8464
===
Subject: Symbolic Differentiation with 48g
I am really confused. I own a HP48g for two month, but it was not
possible for me to run a symbolic differenation.
The example in the short manual for example:
f(x)=5*x^2-6/x
I started the function differentiate, entered the term 5*x^2-6/x in the
field EXPR:, choose X in the field VAR and selected Symbolic in the
field RESULT. After hitting OK I got an numeric value on the stack:
65.1420118343.
This happens with every function I try to differentiate. Only the number
afterward on the stack changed. I never got the symbolic solution.
Anyone an idea, what I did wrong or why the 48g behaves this way?
Danny
===
Subject: Re: Symbolic Differentiation with 48g
 I am really confused. I own a HP48g for two month, but it was not
> possible for me to run a symbolic differenation.
 The example in the short manual for example:
 f(x)=5*x^2-6/x
 I started the function differentiate, entered the term 5*x^2-6/x in
> the field EXPR:, choose X in the field VAR and selected Symbolic in
> the field RESULT. After hitting OK I got an numeric value on the
> stack: 65.1420118343.
 This happens with every function I try to differentiate. Only the
> number afterward on the stack changed. I never got the symbolic
> solution.
 Anyone an idea, what I did wrong or why the 48g behaves this way?
I have an idea. You already have a variable named X, and it has the
value 6.5.
Scott
--
Scott Hemphill hemphill@alumni.caltech.edu
This isn't flying. This is falling, with style. -- Buzz Lightyear
===
Subject: Re: Symbolic Differentiation with 48g
reply-type=response
 I am really confused. I own a HP48g for two month, but it was not possible 
for me
> to run a symbolic differenation.
 The example in the short manual for example:
 f(x)=5*x^2-6/x
 I started the function differentiate, entered the term 5*x^2-6/x in the 
field
> EXPR:, choose X in the field VAR and selected Symbolic in the field 
RESULT. After
> hitting OK I got an numeric value on the stack: 65.1420118343.
 This happens with every function I try to differentiate. Only the number 
afterward
> on the stack changed. I never got the symbolic solution.
 Anyone an idea, what I did wrong or why the 48g behaves this way?
Does the variable x have a value? If so the calculator may be using that 
rather
then the symbol 'x'.
Tom Lake
===
Subject: Re: Symbolic Differentiation with 48g
> Does the variable x have a value? If so the calculator may be using
> that rather
> then the symbol 'x'.
Hi Tom,
there was a var x in the memory/directory with a numeric value. I did
not assumed that the calculator take this in account, when using a
funtion like the differentiation.
Does it mean, that an expression on the stack is automatically evaluate
with a matching variable in the current dir?
Danny
===
Subject: Re: Symbolic Differentiation with 48g
> Does it mean, that an expression on the stack is automatically evaluated
> with a matching variable in the current [or higher] dir?
That's what the HP49/50 series does,
but if you clear flag -3,
the HP48 will normally ignore things stored in variables,
unless you explicitly EVAL an expression;
just repeat your experiment with that flag cleared
for better results.

===
Subject: Re: Symbolic Differentiation with 48g
 Does the variable x have a value? If so the calculator may be using
> that rather
> then the symbol 'x'.

 there was a var x in the memory/directory with a numeric value. I did
> not assumed that the calculator take this in account, when using a
> funtion like the differentiation.
 Does it mean, that an expression on the stack is automatically evaluate
> with a matching variable in the current dir?

 Danny
Not necessarily always, but very often.
When results not what you expect, it is always a good idea to check for
variables being in use in conflicting ways.
Note that the same kind of thing can occur when any variable in your
current program or algebraic is in use as a name in a higher directory.
I.e., if you have a value saved as 'X' in the HOME directory then every
use of X in a subdirectory will is liable to take the saved value from
the HOME directory.
===
Subject: Re: HP's award Winning Support Team Explains the ONe Year Warranty

>> The one year starts when the ORIGINAL purchaser buys it.
>> That includes resellers like Calculators.com, Office Max and Staples.
>> If you buy a new, sealed calculator off the shelf and it's been
>> on the shelf for three months, you'd get a 9-month warranty.
 Only if you don't have the original purchase receipt...
>
have sold it by now. Or if anyone else has one to sell at a reasonable
price please let me know. Unfortunately, I have become paranoid with
respect to trusting this company to do right by me.
===
Subject: Xtide or similar for HP50G

I was trying to set a Tide estimation program on my HP50, and only
find http://www.hpcalc.org/search.php?query=tide+ an RPL version.
I know that Xtide, is available in C++ and wish to know if someone has
ported it to HP49g+ or HP50g ?
Daniel.
===
Subject: Re: Xtide or similar for HP50G
> I was trying to set a Tide estimation program on my HP50, and only
> find http://www.hpcalc.org/search.php?query=hptide an RPL version.
It comes with both HP48 and HP49/50 version libraries, though, FWIW,
and claims to match the results of Xtide.
Did you try it? (I didn't -- no ocean here :)

===
Subject: New hp calculator?
posting-account=3WY0nwoAAAAJz1UJhuywvjIfR3_WSUMZ
AppleWebKit/525.18 (KHTML, like Gecko) Version/3.1.1 
Safari/525.18,gzip(gfe),gzip(gfe)
Apparently hp is about to release a new financial calculator. Go to
http://www.dynatech-taschenrechner.de/ and search for HP 20b.
===
Subject: New hp calculator?
posting-account=3WY0nwoAAAAJz1UJhuywvjIfR3_WSUMZ
AppleWebKit/525.18 (KHTML, like Gecko) Version/3.1.1 
Safari/525.18,gzip(gfe),gzip(gfe)
Apparently hp is about to release a new financial calculator. Go to
http://www.dynatech-taschenrechner.de/ and search for HP-20b
===
Subject: New hp calculator?
posting-account=3WY0nwoAAAAJz1UJhuywvjIfR3_WSUMZ
AppleWebKit/525.18 (KHTML, like Gecko) Version/3.1.1 
Safari/525.18,gzip(gfe),gzip(gfe)
Apparently hp is about to release a new financial calculator. Go to
http://www.dynatech-taschenrechner.de/ and search for 'HP-20 b'
===
Subject: Re: New hp calculator?
> Apparently hp is about to release a new financial calculator.
Apparently, you're really excited about it if you're posting three times!
8^)
===
Subject: Re: New hp calculator?
<fvutt1$hrj$1@aioe.org>
posting-account=3WY0nwoAAAAJz1UJhuywvjIfR3_WSUMZ
AppleWebKit/525.18 (KHTML, like Gecko) Version/3.1.1 
Safari/525.18,gzip(gfe),gzip(gfe)

> Apparently hp is about to release a new financial calculator.
 Apparently, you're really excited about it if you're posting three times!
 8^)
 Tom Lake
Not excited at all. Just corrected typing mistakes.
===
Subject: Re: Questions about SD cards for 50G

> I was wondering what SD cards (brands and max capacity) work with the
> 50G?
> I tried to use the one that came with my camera but it doesn't work,
> I'm now looking for a new one.
> Is it true that the 50G only recognize upto 1GB cards?
>
Some cameras (also cell phones and mp3 players) format SD cards in
something other than normal FAT16 - those cards that are already
formatted by those devices won't work till you re-format them. This
can be done either in the 50 or on a PC, but i recommend doing it on a
PC. Just make sure you specify FAT16 format. You can use anything up
to 2 gb. I use 2gb Kensington MicroSD cards in an SD adapter in my
50
and they work fine! I haven't tried it, but i think the same holds
true for MiniSD cards in an adapter.
===
Subject: Re: Questions about SD cards for 50G
<op.uasnhfa3nn735j@miu.edu>
posting-account=2G4L-woAAABoWmw_YYmxjw-bOIfoIAqk
MathPlayer 2.10b; .NET CLR 1.1.4322),gzip(gfe),gzip(gfe)
> What SD cards (brands and max capacity) work with the 50G?
> I tried to use the one that came with my camera but it doesn't work,
 Does the camera card type say SD exactly,
> or is there any slight difference?
 What capacity and brand?
 It's been reported that there are junk (and counterfeit) cards,
> but no SD card that I've used have ever not worked.
 
it's a 1 GB generic brand (cheapo that came with the camera)
===
Subject: Re: Questions about SD cards for 50G
posting-account=2G4L-woAAABoWmw_YYmxjw-bOIfoIAqk
MathPlayer 2.10b; .NET CLR 1.1.4322),gzip(gfe),gzip(gfe)
> I have the following SD cards:
 Verbatim 256MB
> Sandisk 1GB
> Kingston 2GB
 All working fine. I had problems with the 2GB card when trying to use
> hpgcc programs, but this has been solved by the HPGCC developers some
> time ago.
so the maximun that the 50G can take is 2GB?
===
Subject: Re: Questions about SD cards for 50G

> so the maximun that the 50G can take is 2GB?
The maximum size for an SD card for any device is 2GB.
Anything beyond that is SDHC, which is a different standard, only
supported by SDHC-ready devices.

===
Subject: Re: Questions about SD cards for 50G
posting-account=2G4L-woAAABoWmw_YYmxjw-bOIfoIAqk
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
 so the maximun that the 50G can take is 2GB?
 The maximum size for an SD card for any device is 2GB.
> Anything beyond that is SDHC, which is a different standard, only
> supported by SDHC-ready devices.
 
I just bought a 2 GB patriot brand, it seems to work. Does the library
supposed show :3 when the SD card is inserted?
It's shown on the filer as SD and it says 916832KB (2GB?) . I
haven't formatted it, but when I tried to run the format command on
the 50G, it says error.
===
Subject: Re: Questions about SD cards for 50G
> I just bought a 2 GB patriot brand, it seems to work.
> Is the library supposed show :3: when the SD card is inserted?
No, the card is an external device,
potentially containing objects entirely foreign to the calculator,
whereas only calculator-specific objects can exist inside
the remaining environment.
Calculator libraries can not be executed from SD card
(must be moved to a lower port, e.g. 2, 1, or 0)
However, :3:name is valid for RCL, STO, EVAL
> It's shown on the filer as SD and it says 916832KB (2GB?)
Yes, that's 2GB, with the initial digit truncated,
due to a display column too narrow for all the digits :)
> I haven't formatted it,
> but when I tried to run the format command on the 50G, it says error.
It's pretty certain to be already formatted,
or you can re-format (with greater control) on a computer.
Is it write-protected? (slider on side)
Try copying variables to it,
or make your first instant backup via :3:BK001 ARCHIVE
(and copy libraries using the Filer)


===
Subject: Re: How to install a library frim SD card

> Ok but it is just the copy of the program from port 1 that
> it is at the same speed. I don't understand shadowed rom techniques 
do
> you mean Bank switching?
ROM shadowing is an old technique used to speed up ROM access, and has
nothing to do with the calculator. It used to be common in PC's, don't
is hidden. Any attempts to read/execute something in ROM will actually

===
Subject: Re: How to install a library frim SD card
> Ok but it is just the copy of the program from port 1 that
>> it is at the same speed. I don't understand shadowed rom techniques 
do
>> you mean Bank switching?
 ROM shadowing is an old technique used to speed up ROM access, and has
> nothing to do with the calculator. It used to be common in PC's, don't
> is hidden. Any attempts to read/execute something in ROM will actually
 
job is made by the OS or by the bios?

===
Subject: Re: How to install a library frim SD card
<4830676d$0$30173$426a74cc@news.free.fr>
posting-account=WngkgAoAAACvmZuu3AbTVFyuYsnc_KRO
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
hi,
i tryied many times but i cant do that.
i can copy the library to 1 or 2 rom...
but then?
how i can i use libraries?
which menu?
===
Subject: Re: How to install a library frim SD card
> Ok but it is just the copy of the program from port 1 that
>> it is at the same speed. I don't understand shadowed rom techniques 
do
>> you mean Bank switching?
 ROM shadowing is an old technique used to speed up ROM access, and has
> nothing to do with the calculator. It used to be common in PC's, don't
> is hidden. Any attempts to read/execute something in ROM will actually
 
===
Subject: Re: New hp calculator?
Apparently hp is about to release a new financial calculator.
 Hm..can I guess? It was necessary to develope a new one because there
> is need for faster subtract of huge numbers for financial user? :-D
 Olaf
>
Maybe they developed the new one of lesser quality to be replaced more
often.
===
Subject: Re: New hp calculator?

> Maybe they developed the new one of lesser quality to be replaced more
> often.
Lesser quality compared to the old HP but still better than the all-
time-low 49g+ in terms of build quality.
S.C.
===
Subject: Re: 33s programming
<o1mj24lj1alaaatmt87spb4ic4s7m690gl@4ax.com>
posting-account=Vpj9oAoAAABGzmpTJJmwiSli6FkGLuf9
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> I had a 42s like most of you. But all good things eventually end, and
> mine did.
 Too bad. But I think you know that a 33s doesn't even come close.
 If I understand the basics behind the 33s the MVAR become INPUT
 I substituted the First letter such as R for RPM
 But yet I receive a select fn when I choose SOLVE R
 Of course you do. How should the 33s know what you want it to solve?
> An equation? If yes: which? Or a program, as in this case? Which?
 You simply tell the 33s what program to solve by using FN= <label>.
> Insert the according label here.
 Example:
 In Program mode, key in:
 LBL S
> Input S
> Input D
> Input R
> RCL S
> 12
> x
> Pi
> x
> RCLx D
> RCL- R
> RTN
 Back in Run mode:
 FN= S
> Solve D
 S?
> 10 R/S
> R?
> 4000 R/S
> SOLVING
> D=10,61
 But honestly: a simple equation is much easier to use.
> In Equation mode (EQN) key in your equation:
 R=12*Pi*S*D
 Select equation in list (not neccessary here, because the display still
> shows the last equation you just entered)
 Solve D
 R?
> 4000 R/S
 S?
> 10 R/S
 D=10,61
 Another advantage of using equations (at least that simple ones) is a
> special feature of the 35s and probably of your 33s as well:
 The calculator is able to solve equations *directly* if the desired
> variable occurs only once, i.e. it transforms the equation so that this
> variable can be evaluated directly. E.g. here it's D = R/(12*Pi*S).
> Cool, huh?
 No need to provide initial guesses, no waiting for a result
> (SOLVING...), you get it immediately after the last R/S.
 Dieter
here is the formula
SFM x 12
---------------- = RPM
Dia x PI
Now I hit the EQN button and it says top of List
when I try to punch in S for SFM it just puts in the number of the key
This can't be this hard and they did nor put in the decoder ring LOL
The 42s I program all day long in RPN
Anyhow you would punch in two varibles and then it would solve for the
third.
Question ------ Should I have bought the 35s ? Is it any closer to the
42s?
Snowolf
===
Subject: Re: 33s programming
> here is the formula
> SFM x 12
> ---------------- = RPM
> Dia x PI
Okay, then it's
R = S x 12 / D / Pi
> Now I hit the EQN button and it says top of List
> when I try to punch in S for SFM it just puts in the number of the key
> This can't be this hard and they did nor put in the decoder ring LOL
Hm... obviously you haven't read the manual.
In equation mode variables are entered by RCL <letter>.
So it this case you press...
RCL R = RCL S x 12 / RCL D / Pi
...which shows up as
R=S*12/D/Pi
Now press e.g. Solve D and you're asked for R and S, the two remainig
variables. The result shows up immediately after the last entry.
> The 42s I program all day long in RPN
> Anyhow you would punch in two varibles and then it would solve for the
> third.
> Question ------ Should I have bought the 35s ? Is it any closer to the
> 42s?
No, not at all. In some ways it's better than the 33s but it doesn't
even come close to a 42s or my good old 41.
Now go ahead, take your manual and read the chapter on using equations.
Equation mode is fast and easy to use. My former calculators didn't
offer this mode either and I wondered if it might be useful in any way.
After a short time I realized its benefits. It's really a fast and smart
way to solve problems like yours.
You can have a whole set of regularly used equations in your 33s, they
all show up in the equation list you already mentioned. Simply select
the desired equation (use the up/down keys for scrolling) and then do
whatever you want: Solve for any variable, integrate, or simply evaluate
the equation for different values.
Dieter
===
Subject: FEM48/49 v5.3 updated to SR4
posting-account=R-8hawoAAAA5CgW3hZ9cXaqaEeJoHVEh
Gecko/2008032620 Firefox/3.0b5,gzip(gfe),gzip(gfe)
Hello all,
After a period of some 4.5 years I have made an update to the FEM48
Structural Analysis program (also for HP49/49+/50 calcs). It is a bug
fix of the Query module, the Huber-Hencky stress was not calculated
(and finding me, after all these years).
The update is already posted on hpcalc.org.
Caspar
Please note that the e-mail adres used for posting this message is NOT
in use by me. Have a look at the included manual in the downloadable
package for a correct e-mail address.
===
Subject: Re: FEM48/49 v5.3 updated to SR4
posting-account=lxdVUwoAAADBT622WmcKOi99Rbk4sNLJ
2.0.50727),gzip(gfe),gzip(gfe)
> After a period of some 4.5 years I have made an update to the FEM48
> Structural Analysis program (also for HP49/49+/50 calcs). It is a bug
> fix of the Query module, the Huber-Hencky stress was not calculated
> (and finding me, after all these years).
> The update is already posted on hpcalc.org.
 Caspar
 Please note that the e-mail adres used for posting this message is NOT
> in use by me. Have a look at the included manual in the downloadable
> package for a correct e-mail address.
Good job for the most advance HP structural analysis program
===
Subject: UserRPL program works on 48GX, doesn't on 49G
Hi:
I have a UserRPL program that I've used on a 48GX for a long time. If I
download the text file onto a 48Gx it loads and operates correctly, but if I 

take the same file and download it to my 49G, i get a syntax error and the
program is sitting on the stack. I've gone through it line by line, and the 

character translation on the 49 version matches the 48 version. The only
weirdness I have detected is that the ->NUM command in the 48 doesn't
appear in the 49 AUG, and appears to have been dropped, or at least not
documented. I tried replacing it with EVAL, just to see if the program
would load, but no joy. Are there any well known issues with 48 UserRPL
Jim
===
Subject: Re: UserRPL program works on 48GX, doesn't on 49G

> I have a UserRPL program that I've used on a 48GX for a long time. If I
> download the text file onto a 48Gx it loads and operates correctly, but if 
I
> take the same file and download it to my 49G, i get a syntax error and 
the
> program is sitting on the stack. I've gone through it line by line, and 
the
->NUM exists on the 49.
Somewhere in your program, you are using a variable name with the same
name as an HP49 commands. The HP49 introduced a lot of new commands.
To find out which one it is, edit the string containing the program,
remove the quotes ( ) at the beginning and end and validate with ENTER.
It will show you which variable is causing the error, and put it in
video inverse
--
They who would give up an essential liberty for temporary security,
deserve neither liberty or security (Benjamin Franklin)
===
Subject: Re: UserRPL program works on 48GX, doesn't on 49G
posting-account=Rr1BpAoAAAC35AO1lUfz6ZegzjxdWL8v
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
Most probably you've got user variable which name coincides with some
of the few new built-in commands; You can find comparison chart by VPN
at hpcalc.org
cheers,
reth
> Hi:
 I have a UserRPL program that I've used on a 48GX for a long time. If I
> download the text file onto a 48Gx it loads and operates correctly, but if 
I
> take the same file and download it to my 49G, i get a syntax error and 
the
> program is sitting on the stack. I've gone through it line by line, and 
the
> character translation on the 49 version matches the 48 version. The only
> weirdness I have detected is that the ->NUM command in the 48 doesn't
> appear in the 49 AUG, and appears to have been dropped, or at least not
> documented. I tried replacing it with EVAL, just to see if the program
> would load, but no joy. Are there any well known issues with 48 UserRPL
 Jim
===
Subject: HP50g -- first foray into assembly lang and sysrpl
Originator: roberts@athlonx2.gville (Robert Swan)
Having become reasonably familiar with my new hp50g, and at least
introduced myself to most of the user RPL commands, I thought I should
finally bite the bullet and have a go at some System RPL.
Oddly enough, I was was already fairly familar with Saturn assembly
language; I have owned an HP-71B since about '85 (IIRC) and enjoyed
adding a few extensions to its Forth language. I thought sysrpl
should be fairly easy in comparison. Not so sure now.
Well, enough of my background. On to the background of my questions.
I thought an interesting first challenge would be the old Mastermind
style game, Cows and Bulls. The vanilla flavour of this game
wouldn't require great speed, but I have a variant where the computer
keeps track of possible candidates. After each of your guesses,
besides telling you so many cows and bulls, it goes through the set
of remaining candidates, eliminating those which wouldn't match this
result and tells you how many candidates remain. For example:
Guesses C B
1274 0 1
3579 1 1
6978 0 1 Possible: 3
(FYI, the remaining candidates are: 1538 3258 6534)
There is clearly quite a bit of processing early on when all 3024
candidates are still valid and a combination of sysrpl and assembly
language seemed to be called for (of course it would be cheating to
use C or ARM!).
At long last, down to my questions:
1. How should I structure shared functions in assembly language?
Having implemented the candidate list as a bitmap (just a grob), I
need one function to clear a bit when a candidate is eliminated
and another to test a bit in order to count candidates after a
guess.

The majority of the code for setting or testing a bit is identical
and I have put that code in a PixPrep function which returns
address and mask in registers. This code is included in the
bmreset function's CODE block. I have another function,
bmtest, which calls PixPrep in the guts of bmreset. This
works -- as long as I store the library in port 0, fails in any
other port -- but it seems dodgy. To make this clearer (maybe),
here is the skeleton:
...
::
CODE
C=C&A P
...
GOVLNG =PUSH#ALOOP
ENDCODE
;
::
CODE
GOSUB PixPrep
C=-C-1 P
...
GOVLNG =GETPTRLOOP
*
* Input grob pix# on stack
* Output
* A[0] Pixel nibble's current value
* C[0] Mask of pixel
* D1 Address of pixel's nibble
* Changed
* B
*
PixPrep
GOSBVL =POP#
GOSBVL =SAVPTR
...
RTN
ENDCODE
;
...
I can see options like passing a function selector on the stack,
but I don't find that particularly appealing. I'd really like
there to be multiple known entry points within one code block.
I've probably missed this somewhere in the documentation.
2. Where is there a clear summary of the syntax differences in the
flavours (HP, MASD, others?) of sysrpl and of assembly language?
I was used to HP's Saturn instruction set and have been using
Debug4x as a development environment for this exercise, but I
would like to get away from the PC (Windows in particular)
and MASD is right there on the calculator.
Besides this, I have been confused from time to time when
tutorials downloaded from hpcalc.org were referring to a
syntax that wasn't available in the other environment. IIRC,
it has one. Also took me a while to work out the need for
FEXTERNAL for entry points introduced by ^.
3. Is there any documentation like the HP-71 IDS for entry points?
I had to poke around with Nosy (which is brilliant) to work out
that POP1% sets decimal mode and calls SAVPTR, both of which need
to be catered for before returning. Something warning me of
significant side-effects would be very welcome.
On the whole, I have been pleased with what I have managed to get to
grips with in the last two weeks. While I've had a 48SX since '92,
I never went beyond user RPL with it; there was paid work to do, and
that was Unix/C, so no time to get into my calculator's intricacies.
Consequently, I have missed out on all the history that has led to the
rather baroque interface that's there today. Pretty daunting to get
a handle on it now, but I don't mind a challenge.
Have fun,
Robert.
--
Robert Swan | No, not the antarctic adventurer.
swan.r.l@gmail.com | No, not the Canberra porn monger.
| Yes, that's right, the boring one.
===
Subject: Re: HP50g -- first foray into assembly lang and sysrpl
hello,
for the syntax, I can not help you much. the only advice I can give you is
49/50...
to create library (unless you make them yourself by hand). we use the CRLIB 

that doc for more info on how this works.
as stated by others, unless you know what you are doing (which is not the
case yet), NEVER jump form one ASM code to another... the usual way to
handle stuff like this is to create 1 single code with an 'entry code' that 

goes to a switch case type statement to do various things....
> Having become reasonably familiar with my new hp50g, and at least
> introduced myself to most of the user RPL commands, I thought I should
> finally bite the bullet and have a go at some System RPL.
 Oddly enough, I was was already fairly familar with Saturn assembly
> language; I have owned an HP-71B since about '85 (IIRC) and enjoyed
> adding a few extensions to its Forth language. I thought sysrpl
> should be fairly easy in comparison. Not so sure now.
 Well, enough of my background. On to the background of my questions.
 I thought an interesting first challenge would be the old Mastermind
> style game, Cows and Bulls. The vanilla flavour of this game
> wouldn't require great speed, but I have a variant where the computer
> keeps track of possible candidates. After each of your guesses,
> besides telling you so many cows and bulls, it goes through the set
> of remaining candidates, eliminating those which wouldn't match this
> result and tells you how many candidates remain. For example:
 Guesses C B
> 1274 0 1
> 3579 1 1
> 6978 0 1 Possible: 3
 (FYI, the remaining candidates are: 1538 3258 6534)
 There is clearly quite a bit of processing early on when all 3024
> candidates are still valid and a combination of sysrpl and assembly
> language seemed to be called for (of course it would be cheating to
> use C or ARM!).
 At long last, down to my questions:
 1. How should I structure shared functions in assembly language?
 Having implemented the candidate list as a bitmap (just a grob), I
> need one function to clear a bit when a candidate is eliminated
> and another to test a bit in order to count candidates after a
> guess.
 The majority of the code for setting or testing a bit is identical
> and I have put that code in a PixPrep function which returns
> address and mask in registers. This code is included in the
> bmreset function's CODE block. I have another function,
> bmtest, which calls PixPrep in the guts of bmreset. This
> works -- as long as I store the library in port 0, fails in any
> other port -- but it seems dodgy. To make this clearer (maybe),
> here is the skeleton:
 ...
> ::
> CODE
> C=C&A P
> ...
> GOVLNG =PUSH#ALOOP
> ENDCODE
> ;

> ::
> CODE
> GOSUB PixPrep
> C=-C-1 P
> ...
> GOVLNG =GETPTRLOOP
> *
> * Input grob pix# on stack
> * Output
> * A[0] Pixel nibble's current value
> * C[0] Mask of pixel
> * D1 Address of pixel's nibble
> * Changed
> * B
> *
> PixPrep
> GOSBVL =POP#
> GOSBVL =SAVPTR
> ...
> RTN
> ENDCODE
> ;
 ...
 I can see options like passing a function selector on the stack,
> but I don't find that particularly appealing. I'd really like
> there to be multiple known entry points within one code block.
> I've probably missed this somewhere in the documentation.
 2. Where is there a clear summary of the syntax differences in the
> flavours (HP, MASD, others?) of sysrpl and of assembly language?
 I was used to HP's Saturn instruction set and have been using
> Debug4x as a development environment for this exercise, but I
> would like to get away from the PC (Windows in particular)
> and MASD is right there on the calculator.
 Besides this, I have been confused from time to time when
> tutorials downloaded from hpcalc.org were referring to a
> syntax that wasn't available in the other environment. IIRC,
> it has one. Also took me a while to work out the need for
> FEXTERNAL for entry points introduced by ^.
 3. Is there any documentation like the HP-71 IDS for entry points?
 I had to poke around with Nosy (which is brilliant) to work out
> that POP1% sets decimal mode and calls SAVPTR, both of which need
> to be catered for before returning. Something warning me of
> significant side-effects would be very welcome.
 On the whole, I have been pleased with what I have managed to get to
> grips with in the last two weeks. While I've had a 48SX since '92,
> I never went beyond user RPL with it; there was paid work to do, and
> that was Unix/C, so no time to get into my calculator's intricacies.
> Consequently, I have missed out on all the history that has led to the
> rather baroque interface that's there today. Pretty daunting to get
> a handle on it now, but I don't mind a challenge.
 Have fun,
 Robert.
> --
> Robert Swan | No, not the antarctic adventurer.
> swan.r.l@gmail.com | No, not the Canberra porn monger.
> | Yes, that's right, the boring one.
===
Subject: Re: HP50g -- first foray into assembly lang and sysrpl
here's a hint for code safety:
Don't use inter-code jumps in relocateable code.
Don't do it, never!
It leads to unsafe code, and is bad programming style.
Your call to your subroutine PixPrep is very likely to fail
if the code doesn't reside in main memory (port 0),
because it will be copied to there in pieces,
each RPL sub when it's used,
so there is a big chance that your target location 'PixPrep'
called from a completely different code slice.
And don't forget the always waiting garbage collection,
which will move yor code in memory.
However, if you urgently need to have various routines
call a common sub, then you should consider combining
all code slices into one single code object,
and introduce a suitable stack parameter so the code
can decide where to jump fro mthe start.
The other way to go would be s.t. like this:
RPL
ASSEMBLE
CON(5) =DOCODE
REL(5) ComCodeEnd
* Actual code
RPL
ASSEMBLE
CON(5) =DOCODE
REL(5) ComCodeEnd
* Code
RPL
ASSEMBLE
CON(5) =DOCODE
REL(5) ComCodeEnd
* Code
PixPrep
ComCodeEnd
RPL
This will also work when the code resides in a port other than 0 .
Please note that inter-code jumps MUST ALWAYS be forward jumps.
If Sub1 Sub2 Sub3 all call PixPrep,
PixPrep has to be in the last pseudo-code object.
If PixPrep would reside in the code slice of Sub1,
Sub3 would not have a safe link to PixPrep,
because if it comes to copying to TEMPOB,
only the Sub3 would have been copied,
not Sub1 or Sub2.

> Having become reasonably familiar with my new hp50g, and at least
> introduced myself to most of the user RPL commands, I thought I should
> finally bite the bullet and have a go at some System RPL.
 Oddly enough, I was was already fairly familar with Saturn assembly
> language; I have owned an HP-71B since about '85 (IIRC) and enjoyed
> adding a few extensions to its Forth language. I thought sysrpl
> should be fairly easy in comparison. Not so sure now.
 Well, enough of my background. On to the background of my questions.
 I thought an interesting first challenge would be the old Mastermind
> style game, Cows and Bulls. The vanilla flavour of this game
> wouldn't require great speed, but I have a variant where the computer
> keeps track of possible candidates. After each of your guesses,
> besides telling you so many cows and bulls, it goes through the set
> of remaining candidates, eliminating those which wouldn't match this
> result and tells you how many candidates remain. For example:
 Guesses C B
> 1274 0 1
> 3579 1 1
> 6978 0 1 Possible: 3
 (FYI, the remaining candidates are: 1538 3258 6534)
 There is clearly quite a bit of processing early on when all 3024
> candidates are still valid and a combination of sysrpl and assembly
> language seemed to be called for (of course it would be cheating to
> use C or ARM!).
 At long last, down to my questions:
 1. How should I structure shared functions in assembly language?
 Having implemented the candidate list as a bitmap (just a grob), I
> need one function to clear a bit when a candidate is eliminated
> and another to test a bit in order to count candidates after a
> guess.
 The majority of the code for setting or testing a bit is identical
> and I have put that code in a PixPrep function which returns
> address and mask in registers. This code is included in the
> bmreset function's CODE block. I have another function,
> bmtest, which calls PixPrep in the guts of bmreset. This
> works -- as long as I store the library in port 0, fails in any
> other port -- but it seems dodgy. To make this clearer (maybe),
> here is the skeleton:
 ...
> ::
> CODE
> C=C&A P
> ...
> GOVLNG =PUSH#ALOOP
> ENDCODE
> ;

> ::
> CODE
> GOSUB PixPrep
> C=-C-1 P
> ...
> GOVLNG =GETPTRLOOP
> *
> * Input grob pix# on stack
> * Output
> * A[0] Pixel nibble's current value
> * C[0] Mask of pixel
> * D1 Address of pixel's nibble
> * Changed
> * B
> *
> PixPrep
> GOSBVL =POP#
> GOSBVL =SAVPTR
> ...
> RTN
> ENDCODE
> ;
 ...
 I can see options like passing a function selector on the stack,
> but I don't find that particularly appealing. I'd really like
> there to be multiple known entry points within one code block.
> I've probably missed this somewhere in the documentation.
 2. Where is there a clear summary of the syntax differences in the
> flavours (HP, MASD, others?) of sysrpl and of assembly language?
 I was used to HP's Saturn instruction set and have been using
> Debug4x as a development environment for this exercise, but I
> would like to get away from the PC (Windows in particular)
> and MASD is right there on the calculator.
 Besides this, I have been confused from time to time when
> tutorials downloaded from hpcalc.org were referring to a
> syntax that wasn't available in the other environment. IIRC,
> it has one. Also took me a while to work out the need for
> FEXTERNAL for entry points introduced by ^.
 3. Is there any documentation like the HP-71 IDS for entry points?
 I had to poke around with Nosy (which is brilliant) to work out
> that POP1% sets decimal mode and calls SAVPTR, both of which need
> to be catered for before returning. Something warning me of
> significant side-effects would be very welcome.
 On the whole, I have been pleased with what I have managed to get to
> grips with in the last two weeks. While I've had a 48SX since '92,
> I never went beyond user RPL with it; there was paid work to do, and
> that was Unix/C, so no time to get into my calculator's intricacies.
> Consequently, I have missed out on all the history that has led to the
> rather baroque interface that's there today. Pretty daunting to get
> a handle on it now, but I don't mind a challenge.
 Have fun,
 Robert.
> --
> Robert Swan | No, not the antarctic adventurer.
> swan.r.l@gmail.com | No, not the Canberra porn monger.
> | Yes, that's right, the boring one.
===
Subject: Xlib error running HPtide
posting-account=QSfVcQoAAADkgn5Et4hWcvOa13b-lJy1
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
Hello !
Last weekend I was running the HPTide program on my HP49g+ and HP50g
and in both got the same error:
Does anyone knows why is this error occurring ? The program comes
from HPCALC.ORG and is in C, I am not a C programer so, going to need
some help.
Daniel.
===
Subject: Re: Xlib error running HPtide

this is just a wrong check inside of HPtide and does not mean that an
external lib is missing. Set the flag for the appropiate date
(European/US format) and retry.

===
Subject: Re: Emu48 for HP50G?

 I think you somehow overrate the power of the 50g...
Overrate? The 50G is *already* running an emulator! And it does so
pretty fast. I was thinking we could replace the standard one with
another one, where we could exchange roms.
 Emu48 relies on a Win32 API so far.
 I think even on my N810,
> which has a much faster ARM than the 50g,
> Emu48 (under wine) would be next to unusable.
OK, but you are running one emulation layer on top of another, doing
screen rescaling in real time, etc. And the multitasking OS is wasting
taime in the background. Really no comparison.
 <Half-OT Apart from that, why should I use a 50g with that poorly
> layouted keyboard for simulating a real HP-48,
> which has a perfect keyboard, both from layout and quality?
That is a good point. However, it's been quite common to get questions
from people who bought a 50G and want to use 48 series software.
Wouldn't it be nice for them to simply 'change mode' to 48 series and
have a fully compatible device?
 My approach is the opposite:
> Making a real HP-48 as fast as possible, using SpeedUI
> </Half-OT>
Fine, but a 50G emulating a 48 can also run SpeedUI and be even faster
than the real 48 (like 2x faster).
Anyway, it appears that such a project would have more resistance than
favor.
That's all I needed to know.

===
Subject: Re: Emu48 for HP50G?

> At least the HP-48G+/GX was produced Euntil 2001/2002,
> the latest incarnations even with the high contrast display
 Raymond
>
Yes but the 50g is much faster at mathematical computations and
supports arbitrary-length integer arithmetic, among other things.
I do agree that the 48's are better machines though, since the less
responsive keys on my 50g are annoying.
S.C.
===
Subject: hpgcc and rpl-stack

I made a program with hpgcc which push and pop objects in/from the rpl
stack. I also
do some sat_poke on rpl objects onto the stack. Is the rpl adress
(SAT_STACK_ELEMENT addr field)
of an rpl object on stack may change after a sat_createtemp or
sat_pushptr?
On other question: does anyone made a sat_stack_rot?

===
Subject: Re: hpgcc and rpl-stack
posting-account=y6ooGwkAAACoZd171Nq6oGIjQC5mB4Q9
.NET CLR 2.0.50727; .NET CLR 3.0.04506; .NET CLR 
3.5.21022),gzip(gfe),gzip(gfe)
> Hello
 I made a program with hpgcc which push and pop objects in/from the rpl
> stack. I also
> do some sat poke on rpl objects onto the stack. Is the rpl adress
> (SAT STACK ELEMENT addr field)
> of an rpl object on stack may change after a sat createtemp or
> sat pushptr?
No, hpgcc does not perform an automatic garbage collection, therefore
an object address remains constant until you exit your C program.
 On other question: does anyone made a sat stack rot?
There's not much use for it, since you can use ...pick(n) to access
any stack level directly. You can quickly write a rot function, by
popping all 3 values off the stack, and then pushptr them in the
proper order.

===
Subject: Re: hpgcc and rpl-stack
> hpgcc does not perform an automatic garbage collection, therefore
> an object address remains constant until you exit your C program.
> You can quickly write a rot function,
> by popping all 3 values off the stack,
> and then pushptr them in the proper order.
Since I know nothing, but just have curiosity,
I can ask real dumb questions from pure ignorance
(or feel free to refer me to fine manuals :)
Is there a GC upon (or before) initial entry?
If not, does pop/push affect RPL stack memory,
which could have been very close to limit
(e.g. one word away from needing GC) to begin with?
In any case, there must always be some limit,
at which points the RPL stack(s) either overflow or underflow?
Is there any access to RPL return stack and environments?
which I always know will be enlightening to read,
whether I know the whole story or not.

===
Subject: Re: hpgcc and rpl-stack

 I made a program with hpgcc which push and pop objects in/from the rpl
> stack. I also
> do some sat_poke on rpl objects onto the stack. Is the rpl adress
> (SAT_STACK_ELEMENT addr field)
> of an rpl object on stack may change after a sat_createtemp or
> sat_pushptr?
 No, hpgcc does not perform an automatic garbage collection, therefore
> an object address remains constant until you exit your C program.

 On other question: does anyone made a sat_stack_rot?
 There's not much use for it, since you can use ...pick(n) to access
> any stack level directly. You can quickly write a rot function, by
> popping all 3 values off the stack, and then pushptr them in the
> proper order.
 
I didn't think about how to do it. Your solution is very simple and I
can write it.

===
Subject: Install libraries

such as: http://www.hpcalc.org/details.php?id=3926
(It contains 16 files?!!)
Ali
===
Subject: Re: Install libraries

> such as:http://www.hpcalc.org/details.php?id=3926
> (It contains 16 files?!!)
 Ali
on, just press the [thermo] menu key to run the program, following the
instructions in TH info.txt.
Hope that helps!
-Joe-
===
Subject: Re: Install libraries
posting-account=bHyPMAoAAADLjbTsg7ST0SYzRljqzqPy
.NET CLR 2.0.50727; Media Center PC 5.0; .NET CLR 
3.0.04506),gzip(gfe),gzip(gfe)
> just press the [thermo] menu key to run the program, following the
> instructions in TH info.txt.
I put all files in same directory, then by pressing [thermo] menu key,
it just appears the program source!
TH info.txt not mentions any futher information too.
Ali
===
Subject: Re: Install libraries
> I put all files in same directory,
> then by pressing [thermo] menu key,
> it just appears the program source!
In the case of binary program object files,
pretty much any means of putting (transfering) them
into a calculator directory would work.
In the case of source text files, however,
if you do not use the computer <-> PC transfer software,
or some program to interpret the files, but instead
transfer via an SD card or drop onto the emulator,
then all you will find is the transferred source text,
in this case including this header in each file:
%%HP: T(1)A(R)F(.);
Unless you want to go back and re-transfer each file,
using the software or importer program, you could instead
first edit each file, to remove its header line,
which isn't part of the actual program,
but instead represents calculator modes
which affect the source text of program files.
Then perform 1 TRANSIO RAD and clear flag -51,
to set the same modes as indicated by the removed headers
(clearing flag -51 makes sure that .
is recognized as the decimal point in numbers).
Then, recall each text file to the stack, perform OBJ->
to compile the program, then store the file back
into the variable it came from.
All files are now converted into actual programs.
It may be helpful, in the pursuit of recalling
and storing the contents of objects in the VAR menu,
to note that when the calculator is in RPN mode,
the Left and Right shift keys act as STO and RCL prefixes,
when used with any label in the VAR menu.
The OBJ-> command is the first thing in the PRG TYPE menu.
An importer program (for source text transferred by SD card
or dropped on an emulator) may be found by following a link
in this recent post:
Or jump directly to:
Ascii Import/Export for SD card and Emulator [for all HP48/49/50]
By the way, not all programs come in libraries, which are
special binary objects which must be installed into a port;
this set of programs runs instead from stored ordinary user variables,
and isn't actually a library.

===
Subject: ROM 2.10-7 Question
What steps are needed to install Mr. B.P.'s ROM version 2.10-7 on my
calculator? I know I have to get a lot pf files from his web site. I
have already done this. I am at a loss as to what to do next.
Harold A. Climer
Dept. Of Physics, Geology, and Astronomy
University of Tennessee at Chattanooga
309 Grote Hall
423-425-4546
Harold-Climer@utc.edu
===
Subject: Re: ROM 2.10-7 Question
posting-account=82unwgoAAABKn0foN-teZr4DLYjqF_TL
Gecko/2008032620 Firefox/3.0b5,gzip(gfe),gzip(gfe)
> What steps are needed to install Mr. B.P.'s ROM version 2.10-7 on my
> calculator? I know I have to get a lot pf files from his web site. I
> have already done this. I am at a loss as to what to do next.
> Harold A. Climer
> Dept. Of Physics, Geology, and Astronomy
> University of Tennessee at Chattanooga
> 309 Grote Hall
> 423-425-4546
> Harold-Cli...@utc.edu
The easiest way to do it is to is to download the rom file from B.P.'s
web site. The file you need should be a .bin file. It may be zipped on
his site. (I cannot remember).
Also download the following file. http://www.hpcalc.org/details.php?id=6484
Place the update.scp onto your SD card. Then copy B.P.'s rom file onto
the card, and rename it 4950 92.bin
Then follow the instructions on http://www.hpcalc.org/hp49/pc/rom/
If you don't have an SD card then you can follow the instructions
found in the readme zip.rtf file file in 
http://www.hpcalc.org/details.php?id=6484
substituting the .bin file from B.P.'s site for 4950 92.bin
===
Subject: Re: ROM 2.10-7 Question
posting-account=2G4L-woAAABoWmw_YYmxjw-bOIfoIAqk
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
On May 3, 12:28 pm, username localhost <username.localh...@gmail.com
> What steps are needed to install Mr. B.P.'s ROM version 2.10-7 on my
> calculator? I know I have to get a lot pf files from his web site. I
> have already done this. I am at a loss as to what to do next.
> Harold A. Climer
> Dept. Of Physics, Geology, and Astronomy
> University of Tennessee at Chattanooga
> 309 Grote Hall
> 423-425-4546
> Harold-Cli...@utc.edu
 The easiest way to do it is to is to download the rom file from B.P.'s
> web site. The file you need should be a .bin file. It may be zipped on
> his site. (I cannot remember).
> Also download the following 
file.http://www.hpcalc.org/details.php?id=6484
> Place the update.scp onto your SD card. Then copy B.P.'s rom file onto
> the card, and rename it 4950_92.bin
 Then follow the instructions onhttp://www.hpcalc.org/hp49/pc/rom/
 If you don't have an SD card then you can follow the instructions
> found in the readme_zip.rtf file file 
inhttp://www.hpcalc.org/details.php?id=6484
> substituting the .bin file from B.P.'s site for 4950_92.bin
is it true that the rom 2.10-7 doesnt work with solvesys?
if it's true, then I don't see any advantage of upgrading.
===
Subject: Re: ROM 2.10-7 Question
posting-account=82unwgoAAABKn0foN-teZr4DLYjqF_TL
Gecko/2008032620 Firefox/3.0b5,gzip(gfe),gzip(gfe)
On May 3, 2:28Epm, username localhost 
<username.localh...@gmail.com
> What steps are needed to install Mr. B.P.'s ROM version 2.10-7 on my
> calculator? I know I have to get a lot pf files from his web site. I
> have already done this. I am at a loss as to what to do next.
> Harold A. Climer
> Dept. Of Physics, Geology, and Astronomy
> University of Tennessee at Chattanooga
> 309 Grote Hall
> 423-425-4546
> Harold-Cli...@utc.edu
 The easiest way to do it is to is to download the rom file from B.P.'s
> web site. The file you need should be a .bin file. It may be zipped on
> his site. (I cannot remember).
> Also download the following 
file.http://www.hpcalc.org/details.php?id=6484
> Place the update.scp onto your SD card. Then copy B.P.'s rom file onto
> the card, and rename it 4950 92.bin
 Then follow the instructions onhttp://www.hpcalc.org/hp49/pc/rom/
 If you don't have an SD card then you can follow the instructions
> found in the readme zip.rtf file file 
inhttp://www.hpcalc.org/details.php?id=6484
> substituting the .bin file from B.P.'s site for 4950 92.bin
BTW.
Questions about 2.10-7 are betting to become FAQs.
Perhaps writing up All the relevant information and posting it
somewhere would be a good idea, allowing answers to the common
questions
be a mere pointer to a document.
Such a FAQ should include things like installation instructions, The
fact that 2.10-7 is unofficial, but HP has an official 2.10 release
ready-to-go that they simply have never published, for no real known
reason, etc.
I don't suppose anybody feels willing to draft such a document.
===
Subject: Re: ROM 2.10-7 Question
> Questions about 2.10-7 are betting to become FAQs.
> Perhaps writing up All the relevant information and posting it
> somewhere would be a good idea, allowing answers to the common
> questions.
> Such a FAQ should include things like installation instructions.
Any normal official ROM comes with installation instructions,
which are the same for any ROM, such as 2.09 from either HP or
http://www.hpcalc.org/details.php?id=6484
When installing from an SD card,
the on-board boot loader first loads text file update.scp
to get the name of the actual ROM image to write into flash,
so all you need do to adjust for another ROM image file name on SD card
is to pre-edit update.scp with any computer text editor
(NOT the calc's text editor).
When installing from computer over USB cable,
the file transfer (or ROM installer) program
lets you choose the desired ROM image file.

===
Subject: Re: solvevx error: non unary operator!

 on my HP50G your problem is solved with SOLVEVX. The
> difference to the solution my voyage 200 is, that on the HP50G
> you have to use CONVERT, REWRITE, EXP2POW
> to simplify the result.
That leaves a log term in. Try EXPAND.
'A*B^K=C*X^K'
[Left shift] S.SLV SOLVEVX
Result: 'X=EXP(-(LN(C/(A*EXP(K*LN(B))))/K))'
[Right shift] ALG EXPAND
Result: 'X=1/XROOT(K,C/(A*B^K))'
===
Subject: Re: A bug and a related mini-challenge re LSQ
> Once more into the breach
Lock and load :)
> I cited that sentence from Golub and Van Loan not for the truth of it...
You quoted a textbook for its untruth?
A sentence may have other qualities than truth or untruth. Just because I 
said
I didn't quote it for its truth, doesn't mean that I did quote it for its
untruth.
Coming from the perspective of solving sets of equations
>(rather than solving minimization problems), what it says is true,
I never said it was untrue.
It is also true from the perspective of solving minimization problems if 
one
understands that some minimization problems have exact solutions, 
meaning that
the minimization goes all the way to zero. That's why I wasn't citing it 
for
its truth, but merely that it shows that one authority may appear to 
contradict
another. The contradiction may be resolved by qualifying the word 
solution,
and realizing that the authority cited may not have done so because they 
thought
it was understood. In an earlier context they used the phrase exact 
solution,
probably because they sought to be perfectly clear, and didn't mind the 
expense
of one additional word to achieve that result.
>and since your adversary was speaking from the very same perspective,
>Golub and Van Loan provided a suitable context for elaboration.
> but because it contradicted what the AUR says about underdetermined 
systems.
In the latter postings (perhaps an earlier one referred to
>even more potentially confusing terms like xxxx-determined,
I never referred to xxxx-determined, whatever that is. I have only 
referred
to overdetermined and underdetermined, and I explained what each 
means.
> from the standpoint of how their common meanings
Common meanings? How often in your everyday life have you ever heard 
anyone
use the words overdetermined or underdetermined? However uncommon they may 
or
may not be in ordinary usage, the mathematical usage conforms quite well to 
what
one might expect their common meanings would be.
>clash with localized technical meanings),
The meanings used by me in this thread aren't localized. They are used 
by the
linear algebraists worldwide, with the conventions given by Golub and Van 
Loan.
In an earlier post in this thread I cited what Golub and Van Loan say is 
meant
by the words underdetermined and overdetermined. This is the 
convention
among linear algebraists, as I confirmed by checking all my texts on the
subject. I said that was the convention and that is what I mean when I use 
the
words.
>the AUR was only credited with already containing a one-sentence statement
>about the result of LSQ, which was then adjusted slightly by the citer
>to make a statement with which you just totally agreed:
I did not say that I totally agreed with it. I agree with the truth of 
it,
but I also said it was too succinct for the beginner. The AUR is a 
reference
which is supposed to provide elaboration. You yourself said that LSQ 
has
interesting properties and asked have those been elaborated?. What 
better
place for some elaboration than the AUR?
> The one sentence The resulting vector has a minimum Euclidean norm
>> over all vectors that minimize the residual Euclidean norm
>> is indeed elegant, beautiful and correct.
Any decision of yours to continue to take on (and argue with)
>the rest of the AUR is now your own private war, on the side.
> ... but rather too pithy to serve as the only description in the AUR.
>> I myself found the 2nd and 3rd paragraphs helpful the first time I read 
it.
But you argued as if the person posting had included the rest of the AUR,
>not merely credited it with supplying the one item he finally took from 
it.
I made no such attribution. My sentence I myself found the 2nd and 3rd
paragraphs helpful the first time I read it. is just a side comment, 
obiter
dictum. I didn't say or suggest that the poster said anything about the 2nd 
and
3rd paragraphs; if you think I did, you have misconstrued. I merely said 
that
I found them helpful.
> I cited it to show that even authorities can make not quite true 
statements
>> due to imprecision of language. And to point out the such 
contradictions
>> can be resolved by simply qualifying the word solution.
Or, by eliminating the word entirely,
That's an unworkable solution. Linear algebra texts are going to continue 
to
use the word solution. And, authorities like Golub and Van Loan will 
qualify
as they did in an example I cited when they think it appropriate. But, 
they
won't eliminate it.
I don't think locutions that eliminate the word solution entirely are
necessarily a good thing in mathematics. The word is used a lot, and it 
would
be better to use it in an unambiguous manner than to eliminate it.
>or, by recognizing the difference in perspective (or goal)
>between two different quests --
>one to solve equations or make things equal,
>another to solve a minimization problem
>or minimize sum of squares of differences
>(and then sum of squares of components of vectors which succeed in that).
I deal with this in another post to you.
>> Your original example
> illustrates the infinite number of solutions case
> Why do you so resist saying infinite number of exact solutions
Infinite number of solutions was in the exact quote
>you had just taken from Golub and Van Loan;
It wasn't the Infinite number of solutions phrase I was criticizing; it 
was
the saying of:
'infinite number of solutions ... (where the residual can be made 
zero),'
rather than just infinite number of exact solutions. I was criticizing 
the
use of the 7-word parenthetical (where the residual can be made zero), 
rather
than just saying the single word exact in front of solutions.
>now you're criticizing someone who elaborated further
>on what they had just said, using the same terminology as they used,
>in which solution (to _equations_) was already declared
>as meaning exact, so how many times over need that be repeated?
I'm only criticizing the use of unnecessarily cumbersome phraseology to say 
that
they mean solutions with zero residual. I'm not critcizing them for their
choice to define the single word solution to mean just that. If they 
want to
do that, they may, although I think it is not a good choice. And, I think 
the
choice to use the single word solution in the linear algebra context can 
lead
to confusion. Any possible confusion can be easily avoided by simply 
qualifying
the word where appropriate, a relatively painless thing to do.
Does approximate solution to an equation (or set of them)
>vs exact solution to an equation (or set of them)
I haven't suggested that anything as cumbersome as approximate solution to 
an
equation (or set of them) vs exact solution to an equation (or set of 
them)
was better terminology. I'm only advocating exact solution and 
approximate
solution rather than just solution. The part about '...to an 
equation (or
set of them)' can be inferred from the context since in linear algebra 
it's
usually an equation (or set of them) that we're considering.
>sound like better, clearer terminology to anyone else?
I don't think 'approximate solution to an equation (or set of them) vs 
exact
solution to an equation (or set of them)' sounds like better, clearer
terminology to anyone, even myself. I haven't advocated any such 
cumbersome
terminology.
Do I think exact solution and approximate solution sound like 
better,
clearer terminology just plain old solution to anyone else? Yes, 
indeed. To
anyone else? Apparently to Golub and Van Loan. I cited where they used 
exact
solution rather than just solution. Why did they do that? I suppose 
they
thought it was better, clearer.
Solutions of Linear Matrix Equations. He went even further and added 
best to
approximate solutions.
As I pointed out in my last post, _any_X_whatsoever_ is an
>[approximate] solution to an equation (or set)
>if the equals sign(s) in that equation (or set) is(are)
>regarded as meaning may or may not be equal,
I gave the mathematics dictionary definition of approximate, and you know 
full
well that it doesn't include _any_X_whatsoever_. If that were true, then
approximate would be useless because then it would mean all X. We 
might
disagree over how equal, but it must be more restricted than the 
entire
universe of vectors.
If I asked you to give me a number approximately equal to Pi, you wouldn't 
tell
me that any number whatsoever is approximately equal to Pi; not with the 
common
English meaning and not with the mathematical usage. Actually, I suppose 
you
might, but if you did, I would know that you weren't engaging in serious
discourse.
>unless you better define how equal, which sounds reminiscent of
>all men are created equal -- but some more equal than others :)
If how equal meant equal enough to minimize the norm of the 
residual,
>then that would start to mean something, but I don't recall having
>seen that defined, and the whole direction of choosing English phrases
>like approximate solution (for equations) strikes me
>as being an impediment to most fluidly
>getting the issue across to the minds of readers.
If you want to get the issue across to the readers with exactitude rather 
than
fluidity, then I think using precise phraseology rather than not is a help
rather than an impediment.
> Both locutions are commendable for their precision.
So is the compacting of long-winded descriptions into brief forms,
Saying infinite number of exact solutions rather than 'infinite number 
of
solutions ... (where the residual can be made zero),' is just such a
compacting.
>(hopefully using words and symbols
>whose ordinary meaning doesn't confuse things),
One way of ensuring that ordinary meaning doesn't confuse things is to 
qualify
ordinary words used by mathematics if the possibility of such confusion 
exists
due to their specialized mathematical usage.
>which are more readily used to proceed to building higher forms,
>given precise initial definitions, which is sometimes called
>Mathematics :)
> Golub and Van Loan don't mind saying exact solution
But they didn't say that in the particular excerpt you quoted,
>to which the response was directed.
They didn't say it there (they said it earlier as I cited) because it was 
clear
from the preceeding material in the book that when they said solution in 
the
paragraph about underdetermined systems, they meant exact solutions.
I've said before that sometimes the word solution doesn't need extra
qualification; the precise meaning is evident from the context. But 
sometimes
it does. The fact that they say it all is what is relevant. It shows that 
they
understand the need and don't mind using it when appropriate.
>> you can't just take any B (over the entire space) and find
> a solution, unless [B] happens to have the same characteristic,
> What characteristic of [B] was being referred to?
Readers (including you) got it from the context;
>it of course meant where the same linear relationship
>of the rows of A persists in the elements of B
>(the combined matrix [AB] thus still having linearly dependent rows)
Maybe we should poll readers and see how many of them got that from the
context. I didn't know for sure what he was referring to. If he wants his
readers to know what he means, he should say it.
What we have here is like the classic conflict between legal documents
>which grow so complex with verbosity in the name of rigor
>that no one can understand them, vs. the movement
>(sometimes backed by laws :) to put them in the simplest terms,
>for people of any level:
Adding the single words exact or approximate to solution hardly 
leads to
such complexity that no one can understand.
<SNIP>
===
Subject: Re: A bug and a related mini-challenge re LSQ
> The only possible solutions are the ones whose residual norms
>> are equal to the minimum possible.
There ya go -- a definition which I don't recall noticing earlier;
This is what is meant in the AUR description of LSQ, second paragraph, when 
they
say that with underdetermined systems ...an infinite number of solutions
exist.. I said this in response to your assertion that ...we would have 
to
say that _any_X_whatsoever_ is a solution, That might be what YOU woud 
say,
but I don't say it and it's not the solutions the AUR means when they 
say
...infinite number of solutions...; they say what they mean-- 
...minimum
norm....
>define approximate solution as this, and call it a draw :)
How about if we define it the way the mathematics dictionary does, nearly 
but
not exactly correct? That's how I've used it all along.
In the particular case of LSQ, it can be further qualified by saying
approximate solution having minimum norm.
> We can speak of approximate solutions quite handily.
>> The mathematics dictionary defines approximate as:
>> 1) A result that is nearly but not exactly correct.
> We will have to use the common understanding
>> for what nearly means, I suppose.
>> But let's say within 10% of the correct value for starters.
Uh-oh, there ya go again, in a different sense (Ronald Reagan's :)
I'm trying to give you some slack here. You seem to be saying that
_any_X_whatsoever_ is a 'special word' solution, where 'special 
word' is a
word that you won't tell me. The word I used is approximate, and I'm 
trying
to be liberal with the meaning of approximate in your favor.
Irreconcilable differences?
You created a straw man by saying:
'If we don't assume the word exact in the word solution
(to a set of _equations_ A*X=B, he defines over and over, again and again,
rather than to a related but not identical minimization problem),
then we would have a situation where we would have to say
that _any_X_whatsoever_ is a solution,
just one in which the norm of the residual 'B-A*X'
doesn't happen to be zero,
which somewhat diminishes the value of bringing in a special word
for something which has no special characteristic at all.'
You attempt to show that if the word exact is not in the word 
solution
(thereby, it seems to me, suggesting by implication that the phrase I have 
used,
approximate solution is what you're talking about), then we must say 
that
_any_X_whatsoever_ is a solution.
Then you refer to an un-named special word (which I have asked you to 
name,
and you haven't given me an answer), Finally, you say this something 
has no
special characteristic at all.
Well, it's certainly true that the way you've chosen to characterize the 
word
solution in the preceeding paragraph, it has no special characteristic 
at
all. You have made it difficult to rebut what you say since you haven't 
told
me what the special word is.
But, let's assume it is in fact, approximate. Then using the definition 
from
the mathematics dictionary:
1) A result that is nearly but not exactly correct.
will give any approximate solution a special characteristic, namely that 
is is
nearly but not exactly correct. And this will clearly not admit
_any_X_whatsoever_ by any reasonable interpretation of what nearly 
means.
Since the use of the qualifier approximate does in fact give a special
characteristic to the solution, the special word to which you refer must 
be
something other than approximate (you still haven't told me what it 
is).
It's a straw man argument. You describe a situation I never put forth and
attempt to show that it diminishes the value of bringing in a special 
word
without saying what that special word is. But, the argument doesn't 
work if
the special word is approximate, because approximate doesn't 
allow
_any_X_whatsoever_ for consideration as a solution.
I cited a seminal paper by Penrose where he used the phrase approximate
solutions in the title. You can be sure he didn't mean that just any 
vector in
the n-dimensional row space should be considered an approximate 
solution.
There ARE special characteristics associated with approximate 
solutions;
they are nearly but not exactly correct, and this is very often useful.

> ...is finally what LSQ delivers (or is that still in dispute?)
> When LSQ works correctly, that is what it delivers.
I had conceded about the bug long ago, and moved on.
Was I suggesting that you hadn't conceded the bug? Does the fact that you 
had
conceded it mean that I can't mention it ever again? If I do mention it 
again,
why do you then feel that you must remind me that you condeded it? Do you 
take
my mention of it to mean that I'm denying that you conceded it? 
Sometimes it's not clear what you're getting at when you say things like 
this.
I was simply reiterating the truth, the whole truth, that sometimes LSQ 
doesn't
deliver the expected result.

>> The definition of solution from the mathematics dictionary makes it
>> clear that you could call these interesting results, solutions, 
because...
Which reminds me of the proof (!)
>that there are no uninteresting numbers, [integers, excuse me]
>which is that if there were any uninteresting numbers, [integers]
>then there would be a smallest uninteresting number [integer]
>(in the Euclidean norm [ABS] sense),
>and that would be interesting, wouldn't it?
> The slight edit job on the first sentence of the LSQ description
>> in the AUR says it all, but doesn't provide much elaboration,
>> as I pointed out: LSQ returns a vector such that
>> The resulting vector has a minimum Euclidean norm
>> over all vectors that minimize the residual['s] Euclidean norm.
Right; too condensed, too succinct, but still correct.
> My long post, to which you are responding here,
>> elaborated on the properties of LSQ for underdetermined systems.
Yes, very good elaboration, illustrating case(s)
>of the general, too succinct statement
>(I thought I also just posted appreciation of both kinds of expositions,
>long and short, providing a total range of perspectives)
Why are you mentioning that you just posted appreciation? Did I suggest 
that
you aren't appreciative? I was responding to your question, LSQ has 
certain
properties that make it interesting (have those been elaborated?),, 
reminding
you that I had provided some elaboration.
I could have been snippy (as it seems to me you are doing to a certain 
extent
here, and as I am now), and wondered why you were asking if LSQ properties 
had
been elaborated in response to a post where I had done just that. 
Perhaps you were wondering if further properties had been elaborated, 
beyond
what I had just done in the previous post. In that case, you might have 
said
have any properties been elaborated beyond those you just did?. Your 
question
just seemed a little inconsistent to me, given what had gone before. 
I take the position, apparently somewhat in opposition to you and Joe, that 
a
little extra qualification of words and phrases is a good thing if it lends
clarity and reduces ambiguity, especially in mathematical discourse.
> Often there are an infinite number of solutions that have the same 
minimum
>> residual norm, and requiring that, additionally, the norm of the vector 
itself
>> be minimized provides a unique result (which happens to have some 
desireable
>> properties), rather than an infinity of results!
Yep, always good to narrow down to one single possible result
>(though no proof was offered that this is what happens :)
Were you expecting that I would offer a proof?
The AUR says that ...LSQ returns THE solution with minimum Euclidean 
length..
They didn't say ...LSQ returns A solution.... The clear implication is 
that
the returned minimum length solution is unique. Clearly, THEY believe 
that
there is a unique minimum norm solution.
I got 95% of the way to a proof by showing how the infinite number (if there 
are
an infinite number) of solutions arise.
I said in a previous post:
'It's easy to see that any linear combination of the null vectors n1, n2 and 
n3
added to any solution of the system yields another solution. 
Therefore
there are an infinite number of solutions to the original system, but 
only one
of the solutions has minimum Euclidean length. The solution with 
minimum
Euclidean length is the one which doesn't contain any additive combinations 
of
the null vectors.'
A proof could be completed along these lines (for a complete, formal proof, 
see
a modern linear algebra text):
Note that the row space and the null space are orthogonal. There is a 
unique
solution contained entirely within the row space, for the same reason that 
an
ordinary square linear system with non-singular A matrix has a unique
solution. See a linear algebra text for a proof of this.
Since all the other infinite number of minimum norm solutions are just the
unique row space solution plus a linear combination of the null vectors, we 
need
only show that adding some combination of the null vectors won't make the 
norm
The fact that the row space and the null space are orthogonal guarantees 
that
adding null space vectors will always make the norm larger; that's what
orthogonal means. Cancellation can't occur. The unique minimum norm 
solution
is the solution contained within the row space only, with no projection 
into
the null space.
That's how LSQ does it. It uses the QR decomposition to separate the row 
space
and the null space, and finds the solution that is entirely within the 
row
space.
Interestingly, the solution which LSQ returns for the original system is 
one
of the NON-minimum norm exact solutions. Its residual as calculated on 
the
HP50 comes out exactly zero. It's just not the minimum norm solution 
we
expected.
It's possible to decompose the bad result returned by LSQ for the original
system and see the null space vector that is added to the row space 
solution.
The (bad) result returned by LSQ for the original system is:
Xb=
[[ 6.99541284404 ]
[ 6.15137614679 ]
[ 2.84403669725 ]
[ 3.00458715596 ]
[ 3.84862385321 ]
[ 9.14678899083 ]]
This is the sum of the minimum norm exact solution (the row space part);
Xe=
[[ 3 ]
[ 4 ]
[ 1 ]
[ 7 ]
[ 6 ]
[ 3 ]]
and the null space part (just subtract Xe from Xb):
Xn=
[[ 3.99541284404 ]
[ 2.15137614679 ]
[ 1.84403669725 ]
[ -3.99541284404 ]
[ -2.15137614679 ]
[ 6.14678899083 ]]
To verify that Xn is in the null space, premultiply it by the original A 
matrix
and get [ 0 0 0 0 ]T.
We have Xb = Xe + Xn. Calculate the norm of Xe and Xn separately, then put
those two scalars in a 2-element vector and calculate the norm. It is the 
same
as the norm of Xb, showing that adding the parts Xe and Xn gives a vector 
with a
norm equal to the square root of the sum of the individual Xe and Xn norms.
This can only happen if Xe and Xn are orthogonal.
Or, combine Xe and Xn as the columns of a 6 x 2 matrix and calculate its 
norm;
it is the same as the norm of Xb.
Calculate the dot product of Xe and Xn and get zero, showing that they are
orthogonal.
Otherwise we'd hear but my TI says... :)
Are there any TI calculators that have the advanced linear algebra 
functions
that the HP48G/HP49/HP50 has? I've been considering buying a TI89 on ebay 
so Icould compare with the HP.

>> LSQ can also be used to compute the pseudoinverse of a matrix.
>> This little program does the job:
> << DUP SIZE OBJ-> DROP DUP2 IF > THEN SWAP DROP IDN
> SWAP TRN LSQ TRN ELSE DROP IDN SWAP LSQ END  But note that this fails when given the original A matrix in this thread
>> due to the bug in LSQ
Good to remember that the LSQ bug will affect programs employing LSQ;
>I might have forgotten that :)
I hope it's like the Linear program solutions which cycle,
>however, in that they are so rare that most people have
>never seen one in practice, yet texts properly cover
>how to make it impossible to cause trouble, even then
>(is there any wrapper program to fix LSQ, like the way
>in which spectacles were added to the Hubble telescope? :)
There's no wrapper program, but anything that LSQ can do can also be 
done with
SVD, probably in several steps instead of just one. It's just that once 
you've
got the SVD, you are then doing 12 digit arithmetic, whereas if LSQ does it 
all
at once, you get the benefits of 15 digit arithmetic.
>> The different participants in this thread
> seem to have set out after differing goals;
> perhaps that's why it keeps going round and round.
>> [...]
> Perhaps the illusion of argument in all these cases
> arises from confounding two entirely different things
> to be solved -- one being a set of equations, without reference
> to minimizing anything, just to making things equal (if possible),
> while the other is the solving of a minimization problem
> (minimizing the sum of squares of the amounts of inequality,
> then the sum of squares of the eligible vectors which do that),
> without reference to making anything equal
No comment?
I suppose one could say that, but if one only wants to talk about solving 
a set
of equations, without reference to minimizing anything, just to making 
things
equal (if possible), then what is there to say when no such solution 
exists?
One must simply say the system can't be solved and then be silent. Of 
course,
making things equal IS minimizing; it's minimizing all the way to zero. 
This
has to be true for the succinct sentence that describes LSQ, The 
resulting
vector has a minimum Euclidean norm over all vectors that minimize the 
residual
Euclidean norm. to be true. LSQ can solve square systems just fine. 
Anytime
you would solve a non-singular linear system by putting the B vector on 
level 2
of the stack, the A matrix on level 1 and pressing the divide key, you 
could
also execute LSQ instead of divide and get the desired exact solution (with
possible tiny rounding error, or course). What's nice about LSQ is that if 
the
A matrix is singular, you will get an approximate solution (rather than an
error; but with less control over the details than with the SVD).
In fact, LSQ can find an exact solution to this system:
A=
[[ 1 1 1 1 ]
[ 2 2 2 2 ]
[ 3 3 3 3 ]
[ 4 4 4 4 ]]
B=
[[ 1 ]
[ 2 ]
[ 3 ]
[ 4 ]]
and it's hard to imagine an A matrix any more singular.
The process of solving square systems for which an exact solution exists 
was
totally worked out a long time ago, and is no longer very interesting. The 
only
thing left to say about them has to do with the sensitivity of systems with
large condition number. And the methods for dealing with that are the same
methods, using modern orthogonal decompositions, that are used to find
approximate solutions.
The case of systems for which no such exact solution exists (square or 
not) is
much more interesting. Techniques for effectively dealing with them are
relatively new, and still a research topic. Sometimes what happens is that 
one
may have a system for which an exact solution exists, but which has a 
large
condition number. Then an approximate solution may have more practical
utility than the exact solution. In particular, this may be the case if 
the
system consists of real, error containing data from real world 
measurements.
Very often in such a case, if the measurements had been made without error, 
the
A matrix would have been singular, and it's the errors that make it
non-singular. In that case, the exact solution may be totally bogus, 
induced
by the errors; an approximate solution will be the one you want.
Here's a real world dataset with that sort of problem:
http://www.itl.nist.gov/div898/strd/lls/data/Longley.shtml
Look at the Data file link for the numbers.
Penrose understood the importance of good methods for dealing with these
systems.
The singular value decomposition provides some of the best, most robust,
methods, but it isn't well known, I was very pleased when the HP48G came 
out
with all those additional linear algebra functions. It provides a handy way 
to
practice with these modern matrix decompositions.
Oh, well.
plus the excuse it offered to take excursions into other places.
I seems to me that we used to have deep discussions about things, but not 
so
much lately. It's been several years since Joe got us all going about the
business of finding best fractions.
There has been little discussion about the more modern aspects of linear 
algebra
in the group. I've tried to stir up things a few times in the past, and 
gotten
a few responses, but not everybody has an HP48G/HP49/HP50, which you need 
to
follow along.
Time's fun when you're having flies [Kermit the frog]
.
===
Subject: Re: A bug and a related mini-challenge re LSQ
Hi -- isn't this banter fun?
I think it may actually soon converge
to some agreement or conclusion
(or fatigue on the part of the participants :)
and meanwhile may have left some useful bits in the archives,
or else just in the mind of one participant (me :)
> The only possible solutions are the ones whose residual norms
> are equal to the minimum possible.
>> There ya go -- a definition which I don't recall noticing earlier;
> This is what is meant in the AUR description of LSQ, second paragraph,
> when they say that with underdetermined systems
> ...an infinite number of solutions exist.
I had been confining my consideration to addressing what is written
in this thread, rather than bringing in things written elsewhere
by inference. For example, someone crediting the AUR
as the source for one adopted statement does not mean endorsing
other AUR statements, any more than Barack Obama
endorses what his old friend and minister says,
just because he attended his church, etc.
Most under-determined systems
actually have an infinite number of exact solutions,
so I have tended to assume that meaning for any such statement,
if no other meaning is made explicit.
At any rate, I have the AUR section on LSQ right in front of me,
and it's rather hard to claim that it ever really defines the word
solution, even within its own scope, but just uses it many times,
forgetting to define it in the first place,
in a manner not likely to be self-evident to anyone,
and thus likely to be taken as whatever is most familiar
to the individual reader, from past experience.
After considerable effort, it might be divined that the only meaning
which could make all the other statements defensible would be
any X for which the residual attains its minimum possible norm,
which makes the phrases solution[s] that minimize[s] the residual
Euclidean norm (occurring three times in this little AUR section)
a bit redundant (or perhaps those phrases were intended
as the definition, stuck under every rock in the middle of the garden,
but never declared to actually be a definition).
I can see where one might try to assign the meaning any X at all
to the word solution, but then if A has less than full row rank,
an infinite number of solutions exist would seem rather silly and
meaningless, because an infinite number of any X always exists,
so I think, to make any sense of the whole, we are compelled
to assign the very definition stated above,
which is what should be made explicit to begin with,
to avoid the standard English meaning of the word
being taken for granted,
leading to the mess we can otherwise get into.
Now back to Roger (repeating a few older lines here),
who evidently was trying to straighten out that same mess:
> First, some terminology.
> The AUR uses the word solution in places where
> it should add an adjective in front to be precise.
> I'll use the terms exact solution, approximate solution,
> minimum norm solution and minimum residual solution.
> By approximate solution,
> I mean a solution x that gives a non-zero residual.
> By minimum residual solution, I mean a solution
> which gives a minimum residual norm.
Just like the AUR, however, you never defined solution at all
(or assumed that the AUR had defined it, which I think it never did),
apparently replacing that undefined word
with a set of distinct, longer phrases, so that,
unless solution already means something precise,
then it is a mathematically meaningless word in all of the above,
simply becoming part of the other phrases which are being defined,
rather than having a known meaning to begin with,
to which further qualifiers approximate etc. can then be applied.
In that light, what you have said
would seem to have to mean that an approximate solution
means any X that gives a non-zero residual,
which (particularly when the equations are inconsistent,
so that there is always a non-zero residual)
necessarily means any X whatsoever,
and that's why I considered it not worth attaching
a special word (approximate) which meant next to nothing,
save for NOT a solution to the equation set,
while incorporating at the same time the word solution,
which conjures up in most people's mind something that IS,
and thus the whole phrase is not what I'd offer
in the cause of explaining this topic to anyone new to it,
and just trying to make sense out of it.
On the other hand, if solution already does mean
precisely what I say we are forced to have it mean in the AUR,
given all its usage of the word,
then it already means minimum residual solution,
for which your definition above is then also redundant;
thus it wouldn't be expected that you did mean the same thing
as it finally seems necessary to conclude that the AUR did.
I offered define 'approximate solution' as this,
and call it a draw (isn't that just like what
Josh Waitzkin said, in Searching for Bobby Fischer?)
I think you should have agreed, rather than coming back with
the move let's say within 10% of the correct value for starters,
which is completely opposed to what you started to mean,
and seems guaranteed to derail discussion
(left side within 10% of right side? other way around?
what if both sides have opposite signs?), and at any rate,
sends the mind off in a direction unrelated to LSQ
(which does not seem assured to return an answer
within any pre-defined approximateness in this sense).
I believe I have finally comprehended
most of what the AUR was trying to say,
which you were trying to re-explain,
but since the never-defined word solution
was sitting in the middle of every sentence,
both in the AUR and in your exposition,
upon which other words were building themselves
an inverted pyramid of definitions,
the entire thing collapses into a total heap of confusion,
unless one gets the initial thing right,
and sticks to that same meaning, all the way through,
properly doing which even eliminates the need
for some of the extra verbiage which had been introduced.
What follows may not be exactly right,
but I think it's on the right track,
as to how it could be more clearly explained,
for everyone's benefit:
Some subset of all possible X produces a residual
whose [Euclidean] norm attains a minimum possible value.
In cases labeled under-determined,
there are infinitely many X in that set,
among which the one having minimum norm itself is LSQ's answer.
In cases labeled over-determined,
there is only one X in that set,
which obviously is LSQ's answer.
When the equations are consistent, the minimum norm is zero,
and LSQ's resulting X must satisfy all the equations.
When the equations are inconsistent, the minimum norm
is not zero, and there is no X satisfying all the equations
(the result of LSQ might even satify none of them,
much as a least squares line might actually pass through
none of the points which it is designed to best represent)
Under-determined does not imply consistency of the equations,
and over-determined does not imply inconsistency
[the precise meanings in relation to rank should be defined]
Thus far, within the limited scope just above,
the word solution does not even need to be mentioned
(consistency of the equations stands in for
existence of solutions of the equations).
The very use of the word solution, in fact,
invites confusion with the meaning expected from common usage,
such as solving a set of equations, whereas LSQ solves
a more general goal, which can always be accomplished,
even when no X solves all the equations.
However, in the greater context of that general goal
which LSQ addresses, a solution can then mean
any X for which the [Euclidean] norm of the residual 'B-A*X'
attains its minimum possible value,
as I think we were compelled to conclude was already meant
between the lines in the AUR.
When the equation set has a solution (or many),
an LSQ solution is also a solution to the equation set;
otherwise, the phrase minimum norm least squares solution...
(the first phrase in the AUR)
isn't a solution to the equations themselves,
which the ongoing wording ... to any system of linear EQUATIONS
where A*X=B (with emphasis on = !!!) completely muddles,
and should be replaced by something unmistakably declaring
that LSQ addresses a general minimization problem,
which only sometimes also solves the equations.
It struck me as counter-productive,
obscuring what could be made crystal clear in fewer words,
to throw more phrases needing further definition into the melee,
trying to make up for the AUR's original muddle,
like approximate solution (to what? to the original equations?
certainly not to the LSQ problem itself).
You could say (and did say) that an approximate solution
is any solution in LSQ sense
which isn't a solution to the equations themselves,
but the phrase approximate solution is not needed
to further define or explain anything else pertinent here,
and in my opinion only reinforces the original AUR muddle.
It reminded me of someone I know,
who addresses a honey jar that has become all sticky,
by putting it inside a plastic bag,
and then, when that bag itself becomes all sticky,
puts all of that inside yet another new plastic bag, etc.,
instead of just washing off the original jar to begin with :)
Extra qualifiers like minimum residual solution
are also not needed, because solution (in the generalized sense)
already means that, unless solution means any X whatsoever,
to which I think you've already objected :)
Anyway, while we're working on a minimization problem,
why don't we also minimize the number of phrases we need
to express it, not wrapping slightly cloudy plastic bag wrappers
around it, but just cleaning up the original description,
to make it like a crystal clear original jar?
Finally (ending the scope of this part), although we (you) have
gone into several variations and details about the possible
characteristics of the set of all [generalized] solutions,
for various categories of inputs, the action of LSQ
can still be completely and rigorously defined in a single sentence
(which I think you admitted as elegant and beautiful :)
common to all of them, which can't be all that bad a thing
to also point out, can it?
Another way to describe LSQ follows (you've addressed it all,
but here I'm only trying to squeeze the same material
into a more condensed form, which simply by this
may be easier for people to comprehend all together,
like stepping back, taking a wider view,
and also minimizing the output of words, as far as possible :)
LSQ generalizes the idea of solving an equation set,
by seeking to minimize the Euclidean norm of 'A*X-B',
which is always possible,
rather than the goal to satisfy 'A*X=B',
which is not always possible.
A solution to the expression 'A*X-B',
in the sense of this wider goal,
is any X which minimizes the Euclidean norm of 'A*X-B'
If (and only if) that minimum is zero,
then X is also a solution to the equation set 'A*X=B'
In those cases where there are multiple solutions
to minimizing the Euclidean norm 'A*X-B'
(known as under-determined systems),
LSQ returns that solution (X)
having the smallest Euclidean norm itself,
which has other properties not here elaborated
(but turn to RR's examples to explore that).
[insert further precise definition/explanation
of under/over-determined, etc., if desired,
and contrast with consistent]
Does this harmonize the clashes
between one person's solution and another's?
Can it help other people's understanding?
All of your elaborations of the properties of these systems
certainly educate by example and deeper exploration,
while the level I sought to contribute was to isolate
what could be said in the most complete generality about LSQ,
applying to all cases.
No doubt there must be authors who have already done all of this
much better, even though their own output doesn't seem to have
penetrated into the AUR (or perhaps even here :)
Wolfram always seems to have something, don't they?
Not all on exactly the same issues, but related,
even further expanding the varied uses for the LSQ function:
http://mathworld.wolfram.com/LeastSquaresFitting.html
http://mathworld.wolfram.com/LeastSquaresFittingPolynomial.html
http://demonstrations.wolfram.com/LeastSquares/
http://documents.wolfram.com/v5/Built-inFunctions/AdvancedDocumentation/Line
arAlgebra/4.3.html
(Golub and van Loan finally make their appearance :)
Searching far and wide, out into life:
http://en.wikipedia.org/wiki/Searching for Bobby Fischer
http://www.imdb.com/title/tt0108065/
> Were you expecting that I would offer a proof?
> [that there is only one right answer for LSQ]
Heavens, no; if you had mentioned Fermat's Last Theorem
some years ago (before an accepted proof was published,
if Joe accepts it :) I might have mentioned
that it hadn't yet been proved,
but I would not have meant that statement
as saying that it wasn't so, nor as demanding a proof from you :)
Something very simple, yet never proved at all
(but the unscaled peak keeps attracting climbers):
http://www.petrospec-technologies.com/Herkommer/goldbach.htm
http://plus.maths.org/issue2/xfile/index.html [with calculator]
http://primes.utm.edu/glossary/page.php?sort=goldbachconjecture
http://mathworld.wolfram.com/GoldbachConjecture.html
http://en.wikipedia.org/wiki/Goldbach%27s conjecture
> I got 95% of the way to a proof...
Wow, someday I might be able to claim having had a small part
in producing a masterpiece, just by inadvertently falling into
another creature's shell, just like a grain of sand into an oyster,
after which the oyster produces a big pearl, all on its own :)
Let me know when it comes to the point where we can shake hands again,
even if only after the bout is over :)

===
Subject: Re: A bug and a related mini-challenge re LSQ [analogy with LR]
My stab at a hasty review of thread
(at least the part following acknowledgment of LSQ bug),
from my own end of the telescope:
> LSQ generalizes the idea of solving an equation set,
> by seeking to minimize the Euclidean norm of 'A*X-B',
> which is always possible,
> rather than the goal to satisfy 'A*X=B',
> which is not always possible.
The analogy to least squares line (LR) comes to mind
(may be special case of LSQ, but no matter,
it has its own command).
We have this set of points. Sometimes they all
lie on the same straight line, but when they don't,
we'd like to find one line which best fits the collection.
The thought to make that line minimize the Euclidean norm
of the deviations occurred to someone. The resulting line
may not pass through any of the points at all,
but it does have some neat properties,
such as passing through a point whose cordinates
are the arithmetic averages of all the individual coordinates.
An input set could exist where every point except one
does lie on one straight line, with one outlier,
but LR will still choose a line which passes through none,
though is close to most.
One may analogously have some set of equations, rather than points.
There may be no X which passes exactly through
the constraints of all the equations,
but one can find an X (sometimes many)
which minimize some similar norm,
and this also may or may not pass through (exactly solve)
any of the individual equations, etc.
The AUR, with reference to LSQ:
It uses the word solution in contradictory ways,
speaking both of minimizing Euclidean norm
and solution to equations where A*X=B on the same page,
as if LR had said a line minimizing Euclidean norm of differences
and passing through all the points at the same time,
which it would seem most propitious to resolve
by striking out the latter erroneously restrictive phrase,
which applies only to a special case,
whereas LR (or LSQ) are meant to address more general problems,
sometimes reducing to the special case.
Other proposals to resolve LSQ contradiction:
Say that the word solution is undefined in LSQ,
instead define four new two-word phrases,
each containing the undefined word solution,
each with its own special definition (trying hard not to
accidentally use undefined original word solution in definitions).
Or, strike out the AUR's erroneously restrictive phrase
solution to equations where A*X=B on LSQ page,
give solution the precise minimize Euclidean norm
definition (analogous to LR) that otherwise appears
three times as a qualifier to the phrase solution.
When the latter is done, the set of four different two-word phrases
that each contained the word solution are no longer necessary,
at least not to define LSQ. I say that the AUR does not need
to prefix solution with different words in different places
(any more than LR needs anything like this hair-splitting)
but just to use the appropriate precise definition
in the first place -- or else don't use the word at all :)
The result delivered by LSQ can be characterized
in one general sentence,
without even using the word solution, perhaps as can LR,
but much more can be explored about the nature of various results
which arise from various different classes of inputs,
which can offer a more complete understanding
of how the minimal norm problem (LSQ)
relates to the more restricted solving equations problem,
much as LR relates to the more restricted draw one 2D line
through all these 1D points (or draw one hyperplane
through all these n-dimensional points etc.)

===
Subject: Re: HP's award Winning Support Team Explains the ONe Year Warranty
Hi
On 2008-05-06 17:13:58 +1000, Tom Lake <tlake@twcny.rr.com> said:
 The one year starts when the ORIGINAL purchaser buys it.
> That includes resellers like Calculators.com, Office Max and Staples.
> If you buy a new, sealed calculator off the shelf and it's been
> on the shelf for three months, you'd get a 9-month warranty.
Only if you don't have the original purchase receipt...
--
They who would give up an essential liberty for temporary security,
deserve neither liberty or security (Benjamin Franklin)
===
Subject: Re: HP's award Winning Support Team Explains the ONe Year Warranty
> The one year starts when the ORIGINAL purchaser buys it.
> That includes resellers like Calculators.com, Office Max and
> Staples.
> If you buy a new, sealed calculator off the shelf and it's been
> on the shelf for three months, you'd get a 9-month warranty.
 Well, this is certainly *not* true for the European Union. The
> warranty starts when the end-customer finally buys the calculator and
> it is 2 years by European law.
> So a calculator that has been sitting on the shelf for a year still
> has a 2 year warranty when it is finaly bought.
 HTH
> Andreas
I think I am the only person in the world who has had this much
difficulty. Most of the early 49g+ guys who frequent this group had four
or 5 replacements all under warranty, me ZERO!!, or how do you say ZilCH!
===
Subject: Re: HP's award Winning Support Team Explains the ONe Year Warranty

> Could you explain exactly what it is you expect? A one year warranty is
>> very fair.
 The one year starts when the ORIGINAL purchaser buys it.
> That includes resellers like Calculators.com, Office Max and Staples.
> If you buy a new, sealed calculator off the shelf and it's been
> on the shelf for three months, you'd get a 9-month warranty.
 Tom Lake
But I was going to buy one used from a guy here in the group. You need
to read my post again. The warranty is non-transferable to the second
owner.
===
Subject: Re: HP's award Winning Support Team Explains the ONe Year Warranty
> net@comcast.baranow says...
>> Here is an excerpt from a private email sent today to a would be HP50
>> seller explaining why, after talking to HP Calculator Support, I am
>> unable to take a chance on the 50g which is being offered for resale.
>> Basically, the logic is that all things considered, and having been
>> burned once, it is too risky to purchase this product knowing that more
>> likely than not it will fail and that the terms and conditions for the
>> warranty are far to harsh to roll the dice.
>> ------------------BEGIN EXCERPT---------------------------------
>> The HP support people told me that the warranty expires one year from
>> the time of purchase or one year from the date of manufacture whichever
>> comes first, but that the warranty is not transferable meaning that for
>> you, the original owner, it will be good for one year from the date
>> purchased, but for the second owner, one who is not registered for the
>> warranty, they would use the first 3 digits of the serial number and add 

>> one year to it. So, for anyone who buys it, they will have until the
>> 20th week of year 2008 (the second and third numbers represent the week
>> of production).
>> So, for a calculator that was manufactured, sat on the shelf, was later
>> purchased, and subsequently sold, the new buyer doesn't get a full year
>> warranty. In other words, the meter was running while the cab was
>> stalled in traffic.
>> Exacerbating all of this is the fact that this unit is more likely than
>> not not fail. Nobody in their right mind is going to take a chance with
>> HP with this attitude from their award winning support team.
>> I plan next to write a letter to Corporate Office while continuing to
>> flame the company on comp.sys.hp48 newsgroup. In fact I will probably
>> post this response there now.
>> ---------------------END EXCERPT-----------------------------------
 Could you explain exactly what it is you expect? A one year warranty is
> very fair. How could you expect HP to extend it from the time someone
> purchases the calculator from someone else? How long should they
> purpetuate the warranty? We're not talking about a car or a house here. 

> We are talking about a graphing calculator. Also, you state that the
> unit is more likely than not not fail. Assuming you meant that it
> would more likely than not fail, on what do you base this statement?
> The few folks who post their negative experiences on this newsgroup?
> What about the many happy users who have great experiences with theirs?
 I am not trying to be abrasive here, but try to put things into
> perspective. Don't lose sleep or waste energy trying to flame HP. You
> will have absolutely no effect on them. And many folks will continue to
> buy and enjoy their calculators without ever knowing you made a post
> anywhere.
 It's just a calculator. Relax. 


===
Subject: Questions about SD cards for 50G
posting-account=2G4L-woAAABoWmw_YYmxjw-bOIfoIAqk
MathPlayer 2.10b; .NET CLR 1.1.4322),gzip(gfe),gzip(gfe)
I was wondering what SD cards (brands and max capacity) work with the
50G?
I tried to use the one that came with my camera but it doesn't work,
I'm now looking for a new one.
Is it true that the 50G only recognize upto 1GB cards?
===
Subject: Re: Questions about SD cards for 50G
posting-account=v6m75AoAAAAgp3vyiLFAfyqmfMcXEPkm
1.1.4322; .NET CLR 2.0.50727; WinuE v6; WinuE 
v6),gzip(gfe),gzip(gfe)
I have the following SD cards:
Verbatim 256MB
Sandisk 1GB
Kingston 2GB
All working fine. I had problems with the 2GB card when trying to use
hpgcc programs, but this has been solved by the HPGCC developers some
time ago.
===
Subject: Re: Questions about SD cards for 50G
> What SD cards (brands and max capacity) work with the 50G?
> I tried to use the one that came with my camera but it doesn't work,
Does the camera card type say SD exactly,
or is there any slight difference?
What capacity and brand?
It's been reported that there are junk (and counterfeit) cards,
but no SD card that I've used have ever not worked.

===
Subject: Re: Questions about SD cards for 50G

> I was wondering what SD cards (brands and max capacity) work with the
> 50G?
> I tried to use the one that came with my camera but it doesn't work,
> I'm now looking for a new one.
> Is it true that the 50G only recognize upto 1GB cards?
>
SDHC cards will not work. Any standard SD card should work (2GB cards
too). I use an old 128MB card, never came close to filling it, and I
do a lot of C development and store all my very large binaries on SD.
If your card is a standard SD card and you still have problems then
read: http://h20331.www2.hp.com/Hpsub/downloads/50gUsing_an_SD_Card.pdf.
===
Subject: Re: hpgcc and rpl-stack (again)

> I made a sat pop grob function which takes a grob (open fire 16 gray
> grob) on stack
> and put it in a ggl surface. Mainly It just copying nibbles: the
> structure are exactly
> the same (at least if length%8=0). ESo I wanted to put the 
grob
> address in the
> ggl surface addr field but according to your answer it seems that it
> is
> not possible.
Now I understand what you are trying to do. And the annswer is not,
this is not possible since the GROB in the saturn world is not
guaranteed to be properly aligned (first pixel must be in the first
nibble pointed to by the addr field of the ggl surface).
If all you want to do is display the grob, then there's one trick you
can use to avoid copying the GROB:
The concept is simple:
* If the GROB is properly aligned (which will happen 50% of the
times), this is when the saturn address of the first pixel is
(addr&1)==0, then you can convert the address using sat s2a(address);
This function will obtain the ARM address (char *) of the saturn
address. You can pass this address directly to the ggl surface.
* If the grob is misaligned (by 1 nibble), you can display it line-by-
line:
1) Get the first line's address using sat s2a().
2) Pass this address to the ggl surface (it will point one nibble
before the GROB, since it's misaligned).
3) Set the width to the GROB width+1.
4) Set the ggl surface.x field to 1 (source coordinate = the second
pixel).
5) Draw to the screen using ggl bitblt(), but only one line at a time
(height=1).
6) Add the width in bytes to the addr field.
7) Repeat 5) and 6) through all the lines in the GROB.
This way you can display/extract all the lines in the GROB without
copying it.
You can apply trivial changes to extract a partial GROB (reading only
the appropriate lines, and also setting the x field and width
parameters in the ggl bitblt() call.
I hope this solves your memory issue.
 So it will be possible wtih hggcc 3? I hope we will have hpgcc 3 soon!
You will be able to use the Saturn library with ARM addresses, but not
the GGL library with Saturn addresses. Only the Saturn API will have
this feature.
> I will have to rewrite the functions which deal with RPL objects on
> stack?
Yes, you'll have to if you want to take advantage of the new API.

===
Subject: Re: hpgcc and rpl-stack (again)
> I made a sat_pop_grob function which takes a grob (open fire 16 gray
>> grob) on stack
>> and put it in a ggl_surface. Mainly It just copying nibbles: the
>> structure are exactly
>> the same (at least if length%8=0). So I wanted to put the grob
>> address in the
>> ggl_surface addr field but according to your answer it seems that it
>> is
>> not possible.
 Now I understand what you are trying to do. And the annswer is not,
> this is not possible since the GROB in the saturn world is not
> guaranteed to be properly aligned (first pixel must be in the first
> nibble pointed to by the addr field of the ggl_surface).
> If all you want to do is display the grob, then there's one trick you
> can use to avoid copying the GROB:
 The concept is simple:
> * If the GROB is properly aligned (which will happen 50% of the
> times), this is when the saturn address of the first pixel is
> (addr&1)==0, then you can convert the address using sat_s2a(address);
> This function will obtain the ARM address (char *) of the saturn
> address. You can pass this address directly to the ggl_surface.
> * If the grob is misaligned (by 1 nibble), you can display it line-by-
> line:
 1) Get the first line's address using sat_s2a().
> 2) Pass this address to the ggl_surface (it will point one nibble
> before the GROB, since it's misaligned).
> 3) Set the width to the GROB width+1.
> 4) Set the ggl_surface.x field to 1 (source coordinate = the second
> pixel).
> 5) Draw to the screen using ggl_bitblt(), but only one line at a time
> (height=1).
> 6) Add the width in bytes to the addr field.
> 7) Repeat 5) and 6) through all the lines in the GROB.
 This way you can display/extract all the lines in the GROB without
> copying it.
> You can apply trivial changes to extract a partial GROB (reading only
> the appropriate lines, and also setting the x field and width
> parameters in the ggl_bitblt() call.
 I hope this solves your memory issue.
>
problem with misalignment between saturn (4 bits...)and ARM. I use ggl
functions so my solution (with sat_peek and sat_poke) is (for me) the
best. It just uses more memory.
> So it will be possible wtih hggcc 3? I hope we will have hpgcc 3 soon!
 You will be able to use the Saturn library with ARM addresses, but not
> the GGL library with Saturn addresses.
If I understand we will be able to do the opposite of what I wanted.
Using ARM addresses for saturn stuff when I wanted to use saturn
addresses for ARM objects.
Of course it is not a critic of your work.
In fact I started programming in C because of hpgcc: what a great
tool! Programs are much faster than an ASM (saturn) program on an
hp48(49). Some people like sudoku or cross-words. I like made useless
programs on an hp calc.
>Only the Saturn API will have
> this feature.
> I will have to rewrite the functions which deal with RPL objects on
>> stack?
 Yes, you'll have to if you want to take advantage of the new API.
>
Ok. But If I don't want to take advantage my program will compile?
Of course I will try to use the advantages of the new version.

===
Subject: Re: hpgcc and rpl-stack (again)
<4826caa2$0$4960$426a74cc@news.free.fr>
posting-account=y6ooGwkAAACoZd171Nq6oGIjQC5mB4Q9
.NET CLR 2.0.50727; .NET CLR 3.0.04506; .NET CLR 
3.5.21022),gzip(gfe),gzip(gfe)
> Ok. But If I don't want to take advantage my program will compile?
No, it won't. We want to force you to use the new API, or stay with
HPGCC 2.0.
This is because the main advantage is that the new API is in ROM, so
all your programs will have a significant reduction in size if you use
the new API, on top of the new features.
The best way for us to discourage the use of old and obsolete code is
to remove it.
> Of course I will try to use the advantages of the new version.
I bet you will, that's where the fun is! 

===
Subject: Re: hpgcc and rpl-stack (again)
> Ok. But If I don't want to take advantage my program will compile?
 No, it won't. We want to force you to use the new API, or stay with
> HPGCC 2.0.
> This is because the main advantage is that the new API is in ROM, so
> all your programs will have a significant reduction in size if you use
> the new API, on top of the new features.
> The best way for us to discourage the use of old and obsolete code is
> to remove it.
>
Well I have just to wait the version of hpgcc. The reduction in size
is a great news. I hope I will be able what I am doing now: pop/push
grob (in black&white or gray) to (or from) a gglsurface.
>> Of course I will try to use the advantages of the new version.
 I bet you will, that's where the fun is! 

y
===
Subject: Re: hpgcc and rpl-stack (again)
<48271c2f$0$15460$426a74cc@news.free.fr>
posting-account=y6ooGwkAAACoZd171Nq6oGIjQC5mB4Q9
.NET CLR 2.0.50727; .NET CLR 3.0.04506; .NET CLR 
3.5.21022),gzip(gfe),gzip(gfe)
 When the fun will be avaible?
We want to add as many features as we can, but right now we are
investigating the possibility of an intermediate release, and add more
features later on a new revision.
We still don't have a date, though.

===
Subject: Re: Xtide or similar for HP50G

The program IS accurate enough for my pourposes.
I constrast it with my Suunto altimeter and its estimations where VERY
accurated until I found a bug that could not repeat now.
It works VERY WELL until noon, after this hour on PREDICT TIDE,
option B) TODAYS HI/LOW TIDE & TIMES after 12 hours the program
crash, and says not a valid time, and let two variables on the
directory.
It did not run again, until I clean up the variables from the HPTIDE
directory (t and Eq if I remember well are the name of the
variables), after that clean I could run all the other options but not
this B Todays Hi / Low option.
This morning I run it again, with the hour set on pm time and got the
crash again, but change the hour to a.m. and now the program is
running properly. Hmmm..... can not reproduce the error.
So, it really IS a good one, and I think have to use it more to find
how and where this crash occurs (if ever).
Daniel.
P.D. I tested it on both HP49G+ and HP50g
===
Subject: Re: Xtide or similar for HP50G
>So, it really IS a good one, and I think have to use it more to find
>how and where this crash occurs (if ever).
Daniel.
This program is pretty deep ... have you checked this link?
http://www.flaterco.com/xtide/bibliography.html
===
Subject: how can i copy stuff from my note pad to my calculator

I just bought this hp50g and i have no idea how to work this but I
NEED HELP. All i want to do is to record some stuff from my note pad
into my calculator and also please tell me how I can recall that again
on my calculator.
===
Subject: Re: how can i copy stuff from my note pad to my calculator
posting-account=QSfVcQoAAADkgn5Et4hWcvOa13b-lJy1
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> I just bought this hp50g and i have no idea how to work this but I
> NEED HELP. All i want to do is to record some stuff from my note pad
> into my calculator and also please tell me how I can recall that again
> on my calculator.
I use the old Scribe Lib, to do so. You just have to make a .df (Data
File) file with the document you want to, and Scribe will show you it,
in a very nice way !.
A good explanatory discution on this library could be find here:
===
Subject: Re: how can i copy stuff from my note pad to my calculator
Basically -you're trying to cheat on exams 
well :
you write your text in notepad, copy to SD card
(text file will be presented as string by default)
the easiest way is to leave it on SD and use filer,
to select it and press RCL (or VIEW)
>I just bought this hp50g and i have no idea how to work this but I
> NEED HELP. All i want to do is to record some stuff from my note pad
> into my calculator and also please tell me how I can recall that again
> on my calculator.
===
Subject: Re: how can i copy stuff from my note pad to my calculator
posting-account=zYTuBQoAAAC_bXzGjGVT5rxv8bOnpefP
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> I just bought this hp50g and i have no idea how to work this but I
> NEED HELP. All i want to do is to record some stuff from my note pad
> into my calculator and also please tell me how I can recall that again
> on my calculator.
You can get the manuals from here:
http://h20000.www2.hp.com/bizsupport/TechSupport/DocumentIndex.jsp?contentTy
pe=SupportManual&lang=en&cc=us&docIndexId=64179&taskId=101&prodTypeId=215348&
prodSeriesId=3235173
If you're in a hurry (from page 3141) flip your 50g over, hold 0
and ., and the LCD will scan in your notes. 
===
Subject: Re: How to install a library from SD card
> How do I install a library on my HP50G from an SD card.
In 49G/49G+/50G, libraries may be installed into
just as in HP48S/G; they do not function on the
external media of an SD card (which can contain
objects completely foreign to the calculator)
Copying may be accomplished using the Filer;
a single port number digit (2, 1, or 0) may be typed
when Pick destination appears in its screen heading.
The on-board lithium coin cell feature is provided
you can copy right back again from your SD card,
even the smallest now sold probably having more capacity
that all the libraries most folks would ever use,
plus room for numerous backups of HOME, manuals, etc.

===
Subject: HP 50g serial cable
I bought a serial cable from hpcalc to use the serial port on the HP50g.
At first we couldn't get it working, then we found that the OPENIO command
was needed and I guess it operates a little differently than on the HP 48.
In the process of troubleshooting, we hooked the HP 50 to my son's
Oscilliscope to observe the output, if any.
On the calculator we ran a simple program:
<< DO M XMIT DROP .01 WAIT UNTIL 0 END >>
This repeatedly sent the bits for a capital M character.
We used the M because is has both two adjacent high bits and alternate
high-low combinations. We also tried a capital U, which nicely has
alternating high and low bits.
Then, we did the same for a 48g and 48gx to see how they compare.
The signal for the 50g does have a higher amplitude (higher voltage)
and while the 48g and gx exhibited some timing jitter, the cable from
HPcalc on the 50g had much more stable timing.
Everything worked fine! We took pictures of the o-scope, but I don't know
how to share them here in a newsgroup.
Now that we have proved that the cable is working good, we will go back to
hooking the 50g to a Zebra 105sl Label Printer.

===
Subject: HP48GX : Problem with latest Comm software.
I was wondering if anyone knows where I can get an older version of
the comm software strictly for the HP48 series.( Before the 49 etc
came out)
I have tried to copy files from my GX to my computer using the latest
version from HP with no luck
I did all the self tests and the calculator checks out OK
I used my multimeter and the serial cable from Calpro is OK too.
Harold A. Climer
Dept. Of Physics, Geology, and Astronomy
University of Tennessee at Chattanooga
309 Grote Hall
423-425-4546
Harold-Climer@utc.edu
===
Subject: Re: HP48GX : Problem with latest Comm software.

> I was wondering if anyone knows where I can get an older version of
> the comm software strictly for the HP48 series.( Before the 49 etc
> came out)
> I have tried to copy files from my GX to my computer using the latest
> version from HP with no luck
> I did all the self tests and the calculator checks out OK
> I used my multimeter and the serial cable from Calpro is OK too.
> Harold A. Climer
> Dept. Of Physics, Geology, and Astronomy
> University of Tennessee at Chattanooga
> 309 Grote Hall
> 423-425-4546
> Harold-Cli...@utc.edu
Most terminal emulation software will work.
When I was in college I used KERMIT
you do not need special HP software I have used Hyperterminal in
windows xp for years.
you must select your COM port
set the bits per second: 9600
data bits: 7
Parity: none
stop bits: 1
flow control: none
set the Hyperterminal program (Transfer menu) for KERMIT (SX only) you
can use xmodem (GX only).
set calc to either bin or asc depends on what you type of file you
send.
I have not done this is in a couple of years so you may have to adjust
the settings till it works.
===
Subject: Re: HP48GX : Problem with latest Comm software.
> I was wondering if anyone knows where I can get an older version
> of the comm software strictly for the HP48 series
> (Before the 49 etc came out)
> I have tried to copy files from my GX to my computer
> using the latest version from HP with no luck
> I did all the self tests and the calculator checks out OK
> I used my multimeter and the serial cable from Calpro is OK too.
Many PC programs here:
http://www.hpcalc.org/hp48/pc/link/
HyperTerminal (built into Windows) also transfers files,
using either Kermit or Xmodem, both of which are built into HP48.
Most problems are not due to which software you use, however,
but rather to the connection itself; perhaps see:
Diagnosing COM Failure (HP 48GX Cable) 1997/02/01

===
Subject: Re: HP48GX : Problem with latest Comm software.
Hi Harold,
I am using Version 2.2 Build 2353 with sucsess.
I have used Version 2.3 Build 2439 (german version) at least sucsessfull
with hp 49g+.
I think, you should better search on Eric's site for versions which are
running instead on
official HP sites 
www.hpcalc.org
Heiko
Harold A Climer <hclimer@prodigy.net> schrieb im Newsbeitrag
> I was wondering if anyone knows where I can get an older version of
> the comm software strictly for the HP48 series.( Before the 49 etc
> came out)
> I have tried to copy files from my GX to my computer using the latest
> version from HP with no luck
> I did all the self tests and the calculator checks out OK
> I used my multimeter and the serial cable from Calpro is OK too.
> Harold A. Climer
> Dept. Of Physics, Geology, and Astronomy
> University of Tennessee at Chattanooga
> 309 Grote Hall
> 423-425-4546
> Harold-Climer@utc.edu
>
===
Subject: Re: 48g and 48gII compatibility

> Can programs for the 48g 32k be used on the 48gII and also
> can the programs be exchanged between the 48gII and the 48g
> with a pc link and/or infrared transfer.
 UserRPL (in text form) is for the most part compatible,
> while binary objects are not (because of the major change
> in ROM addresses between the two series, a 48Gii
> being a 49-series product, like the 49G/49G+/50G);
> a backup of HOME, being a binary object,
> is also not directly transferable.
 If you archive your HOME directory directly to a PC,
> the result can be RESTOREd to Emu48, so that
> all subsequent operations may be carried out on a PC,
> between two emulated calcs, then finally either
> transferred individually or all of emilated 48Gii HOME
> sent at once from the PC to the 48Gii -- presuming that
> you have a serial COM port on your PC (or can create one
> via a USB-to-serial converter).
 IR has been reported to be too weak for direct transfer,
> but serial cable transfer (either direct or via computer)
> should be possible, in ascii (text) mode only.
 
Based on your answer I would like to ask a few more things about the
compatibilty of the 48g 32k and the 48gII
1) I already have my hp 48g home directory backed up to my pc. Does
this mean I can transfer my already backed home directory of the 48g
to a 48gII using ascii.
2)That programs for 49,49g+ and 50g will work correctly on the 48gII.
3) And finally, since I have written some very small short programs
in local variable structure for the 48g is this possible to duplicate
on the 48gII. In other words, does the 48gII program in local variable
structure and algebreic structure the same as the 48g.
Ed
===
Subject: Re: 48g and 48gII compatibility
> I already have my hp 48g home directory backed up to my pc.
> Does this mean I can transfer my already backed up home directory
> of the 48g to a 48gII using ascii.
No; it's already a binary object on the PC,
whereas transfer in ascii means to or from
a plain text object on the PC.
However, you can restore that backup to Emu48 on your PC,
copy or move all UserRPL objects to a new subdirectory,
and then convert that entire all-UserRPL subdirectory
to a single string (and save it on the PC as such),
which can then be transmitted in that mode,
provided that it is 100% free of compilation errors
(for example, if you have a variable on HP48
whose name is a new command in HP49 series,
then that variable name would have to be changed
beforehand, else the full text of the entire directory
would cause a compiler hiccup at that variable name).
To convert to and from strings, you can use ->STR and STR->
or small programs which also represent 8-bit characters
using 7-bit strings that can be viewed and edited on all computers,
e.g. program delimiters as << and >>
and the right arrow character as -> etc
(as on-board HP calc Kermit does automatically
in ascii mode with translation mode 3)
> Will programs for 49, 49g+ and 50g work correctly on the 48gII.
All programs for the entire series are compatible -- in theory :)
But since 49G+/50G have a taller screen, for example,
one doesn't know whether or not this will matter;
it usually doesn't, but I have seen at least one 49G+ SysRPL program
which did direct screen writing (without checking)
in a way which could crash
either an HP49 or HP48Gii (the ones with shorter screen);
it also crashed the emulator itself (while emulating an HP49).
There have also been authors who used unsupported entry points
which have changed between ROM versions -- these could crash
_any_ member of the series having a different ROM version,
without discriminating in particular against the 48Gii :)
Therefore, the only proof of the pudding is to try them and see;
UserRPL programs have the least chance of causing problems,
while authors using SysRPL/ML can range from responsible
to negligent, when it comes to sharing the road with other drivers.
> And finally, since I have written some very small short programs
> in local variable structure for the 48g is this possible to duplicate
> on the 48gII. In other words, does the 48gII program in local variable
> structure and algebraic structure the same as the 48g.
UserRPL syntax is exacty the same.
The HP49 series CAS (computer algebra system), however,
may impose restrictions and quirks on re-written HP48 algebraic
commands, (e.g. the syntax for differentials may differ,
the new ISOL takes 15 times as long as the old one,
and won't work if the named variable happens to exist, etc.)

===
Subject: Re: Xtide or similar for HP50G

> This program is pretty deep ... have you checked this 
link?http://www.flaterco.com/xtide/bibliography.html
 -Dale-
Nop !, and looks very interesting.
I am going to look at some of the publications on it so I may can do
my own program, if I can find enough time to do it !
Muuchas Gracias ! (we say in Spanish)

===
Subject: 33s programming

I had a 42s like most of you. But all good things eventually end, and
mine did.
I recently purchased a 33s in hopes of filling the void.
The programs I had used MVAR and RCL for simple variable equations.
Example:
LBL SFM
MVARSFM
MVARDIA
MVARRPM
Fix 02
RCLSFM
12
x
PI
RCLDIA
x
RCLRPM
-
END
If I understand the basics behind the 33s the MVAR become INPUT
I substituted the First letter such as R for RPM
But yet I receive a select fn when I choose SOLVE R
I am still going through the book , but can someone give me a jump
start?
snowolf
===
Subject: Re: 33s programming
> I had a 42s like most of you. But all good things eventually end, and
> mine did.
Too bad. But I think you know that a 33s doesn't even come close.
> If I understand the basics behind the 33s the MVAR become INPUT
 I substituted the First letter such as R for RPM
 But yet I receive a select fn when I choose SOLVE R
Of course you do. How should the 33s know what you want it to solve?
An equation? If yes: which? Or a program, as in this case? Which?
You simply tell the 33s what program to solve by using FN= <label>.
Insert the according label here.
Example:
In Program mode, key in:
LBL S
Input S
Input D
Input R
RCL S
12
x
Pi
x
RCLx D
RCL- R
RTN
Back in Run mode:
FN= S
Solve D
S?
10 R/S
R?
4000 R/S
SOLVING
D=10,61
But honestly: a simple equation is much easier to use.
In Equation mode (EQN) key in your equation:
R=12*Pi*S*D
Select equation in list (not neccessary here, because the display still
shows the last equation you just entered)
Solve D
R?
4000 R/S
S?
10 R/S
D=10,61
Another advantage of using equations (at least that simple ones) is a
special feature of the 35s and probably of your 33s as well:
The calculator is able to solve equations *directly* if the desired
variable occurs only once, i.e. it transforms the equation so that this
variable can be evaluated directly. E.g. here it's D = R/(12*Pi*S).
Cool, huh?
No need to provide initial guesses, no waiting for a result
(SOLVING...), you get it immediately after the last R/S.
Dieter
===
Subject: Re: Emu48 for HP50G?
X
> Well, to evaluate this poll statistically there were not many
> votes 
 Personally I would believe that it would be an advantage for the 50G
> to run as many software as possible.
 Given the possibility to run HP 48 GX applications directly on a 50G
> would broaden the available software and allow to use software that
> will not be ported to the 50G either because the source is not
> available anymore or because there is not interest from the devolper.
X
I could increase the sales
To my opinion could still restrict the software
as only available to HP calculators,
but otherwise let the interest group do their job
my inflated 3 eurocents, wait, US cents are now very weak
my 1,67 eurocents
Veli-Pekka
===
Subject: Re: Emu48 for HP50G?
 I think you somehow overrate the power of the 50g...
Overrate? The 50G is *already* running an emulator! And it does so
pretty fast. I was thinking we could replace the standard one with
another one, where we could exchange roms.
-------------------------
That would then mean that the EMU48 would be running
INSTEAD of the 49-series emulation
--------------------------
> Emu48 relies on a Win32 API so far.
 I think even on my N810,
> which has a much faster ARM than the 50g,
> Emu48 (under wine) would be next to unusable.
OK, but you are running one emulation layer on top of another, doing
screen rescaling in real time, etc. And the multitasking OS is wasting
taime in the background. Really no comparison.
---------------------------
and I wonder why on earth someone would like to run EMU48
on top of EMU49/50
----------------------------
> <Half-OT Apart from that, why should I use a 50g with that poorly
> layouted keyboard for simulating a real HP-48,
> which has a perfect keyboard, both from layout and quality?
That is a good point. However, it's been quite common to get questions
from people who bought a 50G and want to use 48 series software.
Wouldn't it be nice for them to simply 'change mode' to 48 series and
have a fully compatible device?
_______________________
I would be nice if a true hardware eg. KinpOS keyboard assignments
are available, that is: user could map the keys differently
if the default does not somehow apply to one's opinion
outside the final 48 calculator (in the emu-layer then)
> My approach is the opposite:
> Making a real HP-48 as fast as possible, using SpeedUI
> </Half-OT>
Fine, but a 50G emulating a 48 can also run SpeedUI and be even faster
than the real 48 (like 2x faster).
Anyway, it appears that such a project would have more resistance than
favor.
That's all I needed to know.

____________________
You could arrange a group to do the programming anyway
a few negative comments should not put you down
It's quite natural to be skeptical at 1st
and it's good to hear about the possible downsides of the matter
with respects
Veli-Pekka
===
Subject: How to install a library frim SD card
This might seem like a dumb question, but how do I install a library
on my HP50G from an SD card.
I am assuming that the library has been copied from my computer to the
SD card.
I have done this on my HP48GX using the connectivity program. Is there
a difference?
Harold A. Climer
Dept. Of Physics, Geology, and Astronomy
University of Tennessee at Chattanooga
309 Grote Hall
423-425-4546
Harold-Climer@utc.edu
===
Subject: Re: How to install a library frim SD card
> This might seem like a dumb question, but how do I install a library
> on my HP50G from an SD card.
> I am assuming that the library has been copied from my computer to the
> SD card.
> I have done this on my HP48GX using the connectivity program. Is there
> a difference?
> Harold A. Climer
> Dept. Of Physics, Geology, and Astronomy
> University of Tennessee at Chattanooga
> 309 Grote Hall
> 423-425-4546
> Harold-Climer@utc.edu
Using the filer, copy the library from the SD card to the flash memory
===
Subject: Re: How to install a library frim SD card
posting-account=KsGA2QoAAADjMXoAtcKvhALzbY-wSq4Z
rv:1.9b5) Gecko/2008032619 Firefox/3.0b5,gzip(gfe),gzip(gfe)
> I have done this on my HP48GX using the connectivity program. Is there
> a difference?
Not really. Just recall it to the stack and you can do what you
always did (2 STO). However, I usually just select the libraries I
want in the filer using ENTER to select multiple, press the COPY
softkey, and then select port 2.
TW
===
Subject: Re: How to install a library frim SD card
CLR 1.1.4322),gzip(gfe),gzip(gfe)
> This might seem like a dumb question, but how do I install a library
> on my HP50G from an SD card.
port 1 or port 2. Do an ON-C warmstart and that's it.
Port 1 is faster but can be cleared by power failure, since it's
fast enough - I rarely use port 1).
Bill
===
Subject: Re: How to install a library frim SD card
>> This might seem like a dumb question, but how do I install a library
>> on my HP50G from an SD card.
 port 1 or port 2. Do an ON-C warmstart and that's it.
 Port 1 is faster but can be cleared by power failure, since it's
> fast enough - I rarely use port 1).
 Bill
Not so safe...
When I was doing some saturn ASM coding on the hp49 I had
few problems with port 1 and fewer problems with port 2 (but I had
some...). Of course I was doing things with the I/O Ram so with
a wrong codes anything may happen...
Anyone had the same problem? An ASM prog which purge some programs
on port 2? Of course this prog only deals with main ram (that's
what I thought: but when a very buggy program deals with I/O ram...).

===
Subject: Re: How to install a library frim SD card
>> This might seem like a dumb question, but how do I install a library
>> on my HP50G from an SD card.
 port 1 or port 2. Do an ON-C warmstart and that's it.
 Port 1 is faster but can be cleared by power failure, since it's
> fast enough - I rarely use port 1).
 Bill
Why Port 1 is faster than port 2? In fact when you execute
a program in port 1 or 2 the hp copies it in the main ram (port 0) and
after execute the program. Is this operation faster with port 1 than
with port 2? Of course a lib with a lot of entries will execute
slowly (a lots of programs have to be copied). If the lib have only one
program: I think we have no difference.
I am sure that it works like this since the hp49g. On the 48gx
I think that programs can execute directly from port 1 (because
Jazz and the Metakernel need a 128Ko card in slot 1).
Maybe it is different with hpgcc program? Is it possible to
execute directly a C program on port 2? I did not investigate this
but it is what I understood after a (quick) read of the Armtool
help file.

===
Subject: Re: How to install a library frim SD card
<482a1f0d$0$23037$426a34cc@news.free.fr>
posting-account=y6ooGwkAAACoZd171Nq6oGIjQC5mB4Q9
.NET CLR 2.0.50727; .NET CLR 3.0.04506; .NET CLR 
3.5.21022),gzip(gfe),gzip(gfe)
> Why Port 1 is faster than port 2? In fact when you execute
> a program in port 1 or 2 the hp copies it in the main ram (port 0) and
> after execute the program. Is this operation faster with port 1 than
> with port 2? Of course a lib with a lot of entries will execute
> slowly (a lots of programs have to be copied). If the lib have only one
> program: I think we have no difference.
In the original Saturn hardware, the ram access was much faster than
the flash access, just because typically, rom and flash chips are
slower (therefore the existence of shadowed rom techniques, etc.). But
speed and cached, so even if there's a timing difference, it will be
negligible when seen from the (slower) Saturn world. Also, newer flash
chips are much faster than what they used to be.
So, on an ARM based calc, I'd say port 1 speed = port 2 speed, but his
comment is accurate (negligible but accurate).
<...
> Maybe it is different with hpgcc program? Is it possible to
> execute directly a C program on port 2? I did not investigate this
> but it is what I understood after a (quick) read of the Armtool
> help file.
You understood correctly. A properly aligned C program doesn't need to
be copied in ram. It is executed in-place.

===
Subject: Re: How to install a library frim SD card
> Why Port 1 is faster than port 2? In fact when you execute
>> a program in port 1 or 2 the hp copies it in the main ram (port 0) and
>> after execute the program. Is this operation faster with port 1 than
>> with port 2? Of course a lib with a lot of entries will execute
>> slowly (a lots of programs have to be copied). If the lib have only one
>> program: I think we have no difference.
 In the original Saturn hardware, the ram access was much faster than
> the flash access, just because typically, rom and flash chips are
> slower (therefore the existence of shadowed rom techniques, etc.).
Ok but it is just the copy of the program from port 1 that
it is at the same speed. I don't understand shadowed rom techniques do
you mean Bank switching?
But
> speed and cached, so even if there's a timing difference, it will be
> negligible when seen from the (slower) Saturn world. Also, newer flash
> chips are much faster than what they used to be.
> So, on an ARM based calc, I'd say port 1 speed = port 2 speed, but his
> comment is accurate (negligible but accurate).
 <...
>> Maybe it is different with hpgcc program? Is it possible to
>> execute directly a C program on port 2? I did not investigate this
>> but it is what I understood after a (quick) read of the Armtool
>> help file.
 You understood correctly. A properly aligned C program doesn't need to
> be copied in ram. It is executed in-place.
 
That's a great news! So we can make programs in C that are huge in size
(less than 128kb I guess) and have port 0 and 1 for the variables.

===
Subject: ADDENDUM toHP-48SX Engineering Mathematics Library
Does anyonr have a copy of the addendum to John Holland's book
HP-48SX Engineering Mathematics Library?
It was added to the book in 1993 when the GX came out.
There were several commands on the math library that had the same name
as ones added to the GX.
You had to write three short programs that sent you to the card
versions instaed of the built in ones on the GX.
Harold A. Climer
Dept. Of Physics, Geology, and Astronomy
University of Tennessee at Chattanooga
309 Grote Hall
423-425-4546
Harold-Climer@utc.edu
===
Subject: Re: Menu Label Builder
posting-account=QBubbAoAAABVkhM_r1qx-x7X_DKfLaNb
CLR 1.1.4322; .NET CLR 2.0.50727),gzip(gfe),gzip(gfe)
> I was thinking it might be a complete new menu system that would build
> menus, control they keys and shift functions, etc. EIf so, I 
was
> wondering how such a thing fit into 1Kb (and then I would bow to
> Raymond and worship his amazing skill)
I bow to Raymond, and Detlef, and worship their amazing skill
anyway 
Werner
===
Subject: Re: Menu Label Builder
<mnhollinger@yahoo.com> schrieb im Newsbeitrag
 [..]
> I would never choose an onboard system over a portable one.
> Pretty hard to carry around a total station in your pocket
>wouldn't ya think?
>
That may have been one of the reasons why some companies
made surveying packages for HP calculators, for use with
different surveying equipment.
For example, I made a complete surveying package for the HP-48
back in 1990-91, when I was working for W&W Software Products.
The package was (and still is) completely menu driven,
with built-in graphical help on each topic,
showing at least one picture of the related mathematical context,
and also parameter help screens, showing the used variables.
The package features a dynamic point database,
and can act as a remote control for various theodolites and total stations
from different brands, like Geodimeter, Nikon, Topcon, Leitz/Leica, etc
This means the user can interactively adjust and trigger his
measurement device using an HP-48, w/o even looking at
the measurement device, and send data and commands to the device,
and in return receives the relevant data into his HP-48.
Theoretically it's even possible to make time-controlled measurements,
but this feature wasn't reachable for the user from the VM1D UI.
AFAIK this package was only sold on the German market,
for a (back then) very competitive price of about 900 DEM,
w/o the HP-48 unit.
The buyer got a 128K OTP card, and a finger-thick manual.
However, a real surveyor only needed the first few introductory
pages of the manual, and then he could use the system,
as most of the UI was relativlely self-explanatory.
Raymond
===
Subject: Re: My Next Calculator
> What about battery duration?
 MS
IMHO, even the 50g's battery life is subpar. I need to replace the
batteries every month or so even with light to moderate usage.
S.C.
===
Subject: Re: InForm Builder help

Why dont you make a backup of the HP49g+ (together with 
flags and
keys) move the lists to the new HP50g and restore them into the new
machine.
That worked for me when moving from HP49g+ to 50g !
Daniel.
===
Subject: Re: InForm Builder help
posting-account=4ikQLgoAAAB2vmCjLE2ERFyw6LPpHSV4
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> Hello Scott !
 Why dont you make a backup of the HP49g+ (together with 
flags and
> keys) move the lists to the new HP50g and restore them into the new
> machine.
 Daniel.
Hi Daniel,
Now thats a good way to get most everything over. I was working on a
routine at work ( nothing much else to do) and needed an input form.
So I tried to make it on the 50 and kept getting this error. Strange
since I never had any issues before. InForm Builder has been very
stable for me. I have the minimum ROM of 2.08. Extable is there. So it
is a bit of a mystery to me. If I ever figure it out I'll post here.
Scott H.
===
Subject: Re: UserRPL is really slow on the49G
Hi John:
Switching to approximate mode had no discernable effect on speed. I'm going 

to try some of the other tips in this thread.
Jim
In MODE CAS menu, or Right-shift(hold down) + ENTER
or -105 SF (the top status line contains = in Exact mode,
~ in Approximate mode)
===
Subject: Re: UserRPL is really slow on the49G
On 2008-05-22 12:04:41 +1000, Jim Fekete <fetejim@ameritech.net> said:
> Hi John:
 Switching to approximate mode had no discernable effect on speed. I'm 
going
> to try some of the other tips in this thread.
 Jim
You have to switch to approximate mode *before* you edit your program
and save it again.
If all you do is switch to approximate before running the program, it
isn't going to make any difference
JY
--
They who would give up an essential liberty for temporary security,
deserve neither liberty or security (Benjamin Franklin)
===
Subject: Re: UserRPL is really slow on the49G
the program, but do I also have to add decimal pointst to keyboard inputs? 

I'd REALLY like to not have the extra keystrokes. Can I convert to real
before storing the data input?
Jim
> Compiling programs from text (or transferring from original HP48 series)
> while in Exact mode (rather than Approximate mode,
> which better matches the HP48 series)
> could also slow down some loops, and some re-written
> algebraic functions (ISOL being the worst example I know of)
> have also drastically increased their execution time.
Put decimal points after all of the integers in your program (for
example, if you're multiplying something on the stack by 5, put 5. *
instead of 5 * ). The decimal points force approximate mode instead of
using long ZINTs and this speeds up calculation time significantly.
S.C.
===
Subject: Re: UserRPL is really slow on the49G
On 2008-05-22 12:03:21 +1000, Jim Fekete <fetejim@ameritech.net> said:
> the program, but do I also have to add decimal pointst to keyboard 
inputs?
> I'd REALLY like to not have the extra keystrokes. Can I convert to real
> before storing the data input?
>
Yes, if you're in approximate mode when entering the data...
--
They who would give up an essential liberty for temporary security,
deserve neither liberty or security (Benjamin Franklin)
===
Subject: Re: UserRPL is really slow on the49G
On 2008-05-21 09:32:59 +1000, Jim Fekete <fetejim@ameritech.net> said:
> Now that I have my 48GX UserRPL program running on the 49G, I'm wondering
> why I bothered. It runs SIGNIFICANTLY slower on the 49G than on my 48GX. 
I
> thought this thing was faster than the 48GX, and would be an upgrade, 
even
> with the crappy keyboard. Is this a typical experience? Can I speed it 
up?
> Is the 50G any better? Or is it time to E-bay another 48? (I just had 
to
> put one down.... <snif>)
>
In most cases it will be just as fast, if not faster.
However, by default if you type a number like 1 , the 49G uses a new
data type: integer rather than a real.
And integer has unlimited precision and calculations on those are slower.
So you had a program like:
<< 1 100 FOR I I 1 + 1 DISP NEXT >>
it will be slower than:
<< 1. 100. FOR I I 1. + 1. DISP NEXT >>
As with integer, they will have to be converted into reals each time.
If you transfered your program from an HP48, put your calculator into
approximate mode (check the manual), then edit your program , validate
, save it.
In approximate mode, all numbers will be treated as real.
Then you won't be able to tell the difference in speed between the 48
and the 49
This is a FAQ
--
They who would give up an essential liberty for temporary security,
deserve neither liberty or security (Benjamin Franklin)
===
Subject: Re: HP 48 SX power Problem - Desperate Help
Unfortunately seems not to work.....
Moreover why is the reset procedure working? The reset procedure involves
the ON button. Do we have two different circuits for the ON button??
Stefano
>> I bought my HP48 in 1991 and I drop it in my cupboard in 2001 (I removed 

>> the
>> batteries and it was perfectly working).
>> Now, I get it from my closet, I replaced the batteries but I'n not able
>> to
>> power on / power off the calc!!!
>> The only workaround I found to start the calculator, it's to apply the
>> reset
>> procedure ON+A+B, while to power off I have to remove the batteries...
>> What has it happened?
>> Stefano
 A similar thing occurred with my HP48S. The contact for the ON button
> is dirty or oxidized. You might be able to clear the problem with
> repeated operation of the ON button.
 I had a problem exiting Alarm CAT display with the ATTN button.
> That is when I discovered that holding the calculator and torquing it
> slightly with ON corner and upper right corner would enable the ON
> button to make contact. Hold the calculator with both hands, placing
> left thumb on ON button with right hand on upper right corner and
> torque the unit slightly while pressing the ON button.
 BTW, if you leave the calculator ON, it will turn off after several
> minutes of non use.
===
Subject: Re: HP50g -- first foray into assembly lang and sysrpl
Hi Robert,
here's a short description of what happens,
given my code construct with Sub1 Sub2 Sub3, and PixPrep.
Also assume the whole code resides in a location
If you call Sub3, the code slice between the Sub3 entry
and ComCodeEnd will be copied to TEMPOB
as one monolithic block.
If you call Sub2, the code slice between the Sub2 entry
and ComCodeEnd will be copied to TEMPOB
as one monolithic block.
This includes the Sub3 entry in this case.
If you call Sub1, the code slice between the Sub1 entry
and ComCodeEnd will be copied to TEMPOB
as one monolithic block.
This includes the Sub2 and Sub3 entries in this case.
This works because the appriorate length fields
span from the distinct start to the common end,
or in other words, Sub1 includes Sub2,
which includes Sub3, which includes PixPrep.
> 1. Sub3 invoked, and just it (including PixPrep) loaded into memory
> and executed.
> 2. Sub1 invoked (Sub3 & PixPrep still in memory, I assume),
>
At this point you can simply forget about any previous invocation
of Sub3, as the new TEMPOB copy has no relation to the old one.
Even if the old TEMPOB code is still in memory, consider it gone.
There are many documents about RPL and ML programming,
most of them can be found on www.hpcalc.org .
There are RPL and ML tutorials, RPL and ML entry lists and descriptions,
also some fundamental docs like RPLMAN.doc , and much more
Have fun
Raymond

===
Subject: EE Pro for HP
posting-account=NADa3AoAAABsb_5Gw-DJgHEyd9fPxCdx
SLCC1; .NET CLR 2.0.50727; .NET CLR 3.0.04506; Media Center PC 5.0; 

InfoPath.2; .NET CLR 3.5.21022; Zune 2.0; .NET CLR 
1.1.4322),gzip(gfe),gzip(gfe)
I know I had this for my HP49, does anyone still have this app so I
can install it on my HP50?

===
Subject: Re: EE Pro for HP

> I know I had this for my HP49, does anyone still have this app so I
> can install it on my HP50?

There used to be a 49g+/50g version ported by GaaK, but that has since
been taken down due to copyright issues with Da Vinci Technologies.
S.C.
===
Subject: Re: EE Pro for HP
posting-account=zwF5mgoAAABtN7MLMYmZfSdAoZ2In_RC
rv:1.8.1.14) Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
If you check the archives of this mailing list, you'll see that Brian
Maguire was going to start selling it some time in 2005, but that
never happened. He had a reason for each time it was delayed, but
then he disappeared.
You can see the whole saga if you find the thread that started with
the subject line:
EE?Pro / EEPro / EE Pro: for HP 49
It's a depressing read though, given that Brian never released
anything.
-Jonathan
===
Subject: Sparcom cards
I have several Sparcom cards for my HP48GX; Mechanical engineering
Pac and M.E. Reference. Is it possible to copy these to my HP50G for
my own use only? I ask if it is possible and if it is legal?
Also is it possible to get a copy the manuals somewhere.
I have them around the house somewhere, but have not located them yet.
Harold A. Climer
Dept. Of Physics, Geology, and Astronomy
University of Tennessee at Chattanooga
309 Grote Hall
423-425-4546
Harold-Climer@utc.edu
===
Subject: Re: Sparcom cards
> I have several Sparcom cards for my HP48GX; Mechanical engineering
> Pac and M.E. Reference. Is it possible to copy these to my HP50G for
> my own use only? I ask if it is possible and if it is legal?
Legal? Maybe.
Possible? Most likely.
Will they run? Probably not.
Too many address changes are inbetween the 48GX and 50G. I doubt that the
rom code is 100% relocatable.
Rich
===
Subject: Re: Sparcom cards
I have several Sparcom cards for my HP48GX; Mechanical engineering
> Pac and M.E. Reference. Is it possible to copy these to my HP50G for
> my own use only? I ask if it is possible and if it is legal?
> Also is it possible to get a copy the manuals somewhere.
> I have them around the house somewhere, but have not located them yet.
For manuals you could try Samson Cables - Books & Manuals
http://www.samsoncables.com/catalog/index.cfm
Or see message 4 on the link below:
www.hpmuseum.org/cgi-sys/cgiwrap/hpmuseum/archv013.cgi?read=40080
===
Subject: solvevx error: non unary operator!
hello all, i've just bought a hp 49g+,(ye i hpe ot wit falty kboar)
and i was trying to fast solve(just to spare some seconds) this easy
equation:
- A*(B^(K))=C*(X^(K))
- SOLVEVX
this gives me back an error:
SOLVEVX Error: Non Unary operator
What does it means??
the ti 89 emu that i've just downloaded gives me the correctr answer in just 

one try,
can you help me in reconsider HP??
===
Subject: Re: solvevx error: non unary operator!
>hello all, i've just bought a hp 49g+,(ye i hpe ot wit falty kboar)
>and i was trying to fast solve(just to spare some seconds) this easy
>equation:
- A*(B^(K))=C*(X^(K))
>- SOLVEVX
this gives me back an error:
>SOLVEVX Error: Non Unary operator
What does it means??
the ti 89 emu that i've just downloaded gives me the correctr answer in 
just
>one try,
>can you help me in reconsider HP??

Put the calc from approximate ( ~ ) mode to exact ( = ) mode,
[Right Shift] + [Enter]
Damir
===
Subject: Re: solvevx error: non unary operator!
<mujj14lgs4lmpb2cfal44nc2qm637aqh12@4ax.com>
CLR 1.1.4322),gzip(gfe),gzip(gfe)
>one try,
>can you help me in reconsider HP??
 Put the calc from approximate ( ~ ) mode to exact ( = ) mode,
> E[Right Shift] + [Enter]
That's what I found too. The interesting question: why doesn't the
calc do an automatic mode switch, as it usually does?
Bill
===
Subject: Re: solvevx error: non unary operator!

> That's what I found too. The interesting question: why doesn't the
> calc do an automatic mode switch, as it usually does?
I think, you should look at flag 123: if it is CLEAR, mode switch is
allowed,
for SET, it is not allowed
Peter
===
Subject: Re: solvevx error: non unary operator!

> I think, you should look at flag 123: if it is CLEAR, mode switch is
> allowed,
> for SET, it is not allowed
I tried that, and it doesn't affect it. If mode switching is off, it
Bill
===
Subject: Re: solvevx error: non unary operator!
posting-account=itRlSAoAAADIJvOzSGM8mjsK7gvlk1wg
Gecko/20071127 Firefox/2.0.0.11,gzip(gfe),gzip(gfe)
Hello Hierro
on my HP50G your problem is solved with SOLVEVX. The difference to
the solution
my voyage 200 is, that on the HP50G you have to use CONVERT, REWRITE,
EXP2POW
to simplify the result.
I varied the system flags 1 general solution/principal value and 119
rigorous on/off
but it continued to work.
Have you purged all involved variables before making your test?
On my HP50G I am using ROM-Version 2.10-7. Perhaps you can upgrade to
this.
You should find that ROM on hpcalc.org
Peter
===
Subject: Re: solvevx error: non unary operator!
>Hello Hierro
on my HP50G your problem is solved with SOLVEVX. The difference to
>the solution
>my voyage 200 is, that on the HP50G you have to use CONVERT, REWRITE,
>EXP2POW
>to simplify the result.
I varied the system flags 1 general solution/principal value and 119
>rigorous on/off
>but it continued to work.
Have you purged all involved variables before making your test?
On my HP50G I am using ROM-Version 2.10-7. Perhaps you can upgrade to
>this.
>You should find that ROM on hpcalc.org
Peter
I was wondering where you got a copy of ROM version 2.10-7?
The HP web site does not have it.
I also remember vaguely about either the OS or CAS had to be compiled
from source code before you could install on your calculator because
of copyright or some other reason. Is this true for 2.10-7
Harold A. Climer
Dept. Of Physics, Geology, and Astronomy
University of Tennessee at Chattanooga
309 Grote Hall
423-425-4546
Harold-Climer@utc.edu
===
Subject: Re: solvevx error: non unary operator!
<is6l14l68ve4c31h5in7ednnbt45neiitp@4ax.com>
posting-account=itRlSAoAAADIJvOzSGM8mjsK7gvlk1wg
Gecko/20071127 Firefox/2.0.0.11,gzip(gfe),gzip(gfe)
> I was wondering where you got a copy of ROM version 2.10-7?
> The HP web site does not have it.
Hello Harold,
with the VERSION command my HP50G displays 2.10-7, so
I hope it is 2.10-7 . I found this ROM-Version one or two
years ago on
http://www-fourier.ujf-grenoble.fr/~parisse/english.html
which is the home page of Bernard Parisse. He was one
of the creators of the HP-CAS and then published his
extensions by himself.
Peter
===
Subject: Re: Assembly language splitting the way to communicate ?
hello,
bit 0 and 1 at #0011B are used to control this:
Bit 0 control the type of IR comunication. if set to 0, it's IrDa, at 1is a 

tentative implementation of the old HP Ir protocol (does not realy work).
Bit 1 is for USB/RS232 selection. set to 0, USB, if set to 1: serial.
cyrille
 I've just bought an hp50g calculator, congratulations, that's a great
> model.
 However, adapting my previous programs, just like Navigator and Chat
> HP, I was wondering :
 want to use to communicate ?
 Since now, you have IrDA, USB and serial...
 Before, with the HP48GX, you had the serial RS232C and the IR.
> With the HP49G, you only had wire RS232C...
> The hp49g+ was coming back to the IR, with the IrDA specifications,
> plus the USB.
> But the hp50g has that particulatiry to have 3 communication ports,
> which may be more complicated to select in Saturn assembly language


===
Subject: hp49g+/50g - Del operator only in Cartesian coords?
Hi all,
As far as I can tell on my hp49g+, the vector analysis functions based on 
the
Del operator only work in Cartesian coords. Putting the calc in cylindrical 
or
spherical doesn't change anything.
For instance, in cylindrical coordinates the Laplacian of Rcos(T)+Z = 0. 
But if
I put the calc in polar mode and do:
R * cos(T) + Z
['R' 'T' 'Z']
LAPL
yields
-R*COS(T)
I can write my own functions but just wonder if I'm missing something to 
make
the calculator do vector analysis in other coordinate systems.

===
Subject: Re: hp49g+/50g - Del operator only in Cartesian coords?
> As far as I can tell on my hp49g+,
> vector analysis functions based on the Del operator
> work only in Cartesian coords.
> Putting the calc in cylindrical or spherical doesn't change anything.
By definition, the _internal_ components of vectors
across the entire HP28/48/49/50 series of calculators
are always _rectangular_ coordinates.
RECT/CYLIN/SPHERE modes govern the convenient _display_
of converted _numeric_ coordinates on the screen
(with cute angle symbols attached to angular values),
and for the potential _numeric_ coordinate transformations
between internal (always rectangular) coordinates
and a set of display coordinates,
implied by the numeric-only commands V-> ->V2 and ->V3
V-> ->V2 and ->V3 are those new commands
which replace the R->P and P->R found in older calc series;
those old commands, in turn, had operated on pairs of real numbers
on two stack levels, before the HP28/48/49/50 series
had even introduced the concept of multi-component objects
that occupy only a single stack level, plus other changes
which significantly altered the general paradigm
of how we use our calculators.
As far as I am aware, the CAS never undertook to display expressions
for what would be the transformed _symbolic_ coordinates of a _symbolic_ 
vector,
which is still internally always expressed in rectangular coordinates,
in any sort of analog of what it does with purely numeric vectors
(and complex numbers) -- is that what you miss seeing?
(along with the lack of those same cute angle symbols
in front of symbolic expressions for angular components?)

===
Subject: Re: hp49g+/50g - Del operator only in Cartesian coords?
reread shortly to be sure I've absorbed it all!
> As far as I can tell on my hp49g+,
>> vector analysis functions based on the Del operator
>> work only in Cartesian coords.
 By definition, the _internal_ components of vectors
> across the entire HP28/48/49/50 series of calculators
> are always _rectangular_ coordinates.
 RECT/CYLIN/SPHERE modes govern the convenient _display_
> of converted _numeric_ coordinates on the screen
> (with cute angle symbols attached to angular values),
This is the key bit of info I wasn't aware of - that vectors are always 
rectangular.
> As far as I am aware, the CAS never undertook to display expressions
> for what would be the transformed _symbolic_ coordinates of a _symbolic_ 
vector,
> which is still internally always expressed in rectangular coordinates,
> in any sort of analog of what it does with purely numeric vectors
> (and complex numbers) -- is that what you miss seeing?
I miss a little more than display differences. When problems have certain
symmetries, they're easier to solve in one coordinate system versus another. 

The challenge is that the del vector operator doesn't have a specific form 
in
other coordinate systems. The divergence function, for example, in 
Cartesian is
the good ol':
div D = d/dx Dx + d/dy Dy + d/dz Dz
yet in cylindrical is
div D = (1/p) d/dp (p Dp) + (1/p) d/do Do + d/dz Dz
where p is distance rho, o is angle phi, and z is distance. So the 
calculation
for div itself is different, and of course is different again in spherical. 
I
was hoping, not knowing just what coord mode choice implied, that when I put 
the
hp49 in cylindrical mode that it would perform vector analysis for that 
system.
 Easy enough to write my own functions, but I didn't want to reinvent 
the wheel unnecessarily.

===
Subject: Coordinate modes: RECT, CYLIN, SPHERE, V->, ->V2, ->V3
Elaborating about coordinate modes, with an example to follow:
By definition, the _internal_ components of vectors
across the entire HP28/48/49/50 series of calculators
are always _rectangular_ coordinates.
While the calculator is automatically displaying the stack,
if a complex number or vector of two or three
real-valued components is to be displayed,
the stack displayer considers the implicit
RECT/CYLIN/SPHERE (coordinate display) mode,
and displays a set of coordinates appropriate to that mode,
also prefixing a cute angle symbol
to each angular component displayed.
The ->STR command also does the same thing,
displaying, in the string, the kinds of components called for
by the RECT/CYLIN/SPHERE (coordinate display) mode.
While the calculator is parsing input from the
command line, if a complex number or vector
of two or three real-valued components is encountered,
in which a cute angle symbol prefixes the second
(and possibly third) component, the parser likewise
converts all coordinates into purely rectangular coordinates,
to store in the internal object;
the STR-> command (compile a string) likewise does the same thing.
Note that in the case of command line (or string) parsing,
the presence or absence of angle symbols
explicitly determines the input's coordinate mode,
ignoring any implicit RECT/CYLIN/SPHERE display mode.
The V-> command gives user programs direct access
to the internal coordinate conversion function
which takes a single object's internal rectangular coordinates
as input, and outputs separate individual coordinates,
appropriate to the current RECT/CYLIN/SPHERE display mode.
When string or screen output is created from those coordinates,
cute angle symbols are automatically prefixed as appropriate,
but when you use V-> yourself, it's left to you
to be aware of which values represent angles.
The ->V2 and ->V3 commands give user programs direct access
to the inverse coordinate conversion function,
in which the implicit RECT/CYLIN/SPHERE mode tells these functions
which (if any) of the separate input components are angles;
the output consists of a single object,
whose internal coordinates are always rectangular.
It should be evident that V-> is the inverse of ->V2 or ->V3
It should also be evident that V-> is not a shortcut for
ARRY-> DROP or OBJ-> DROP; if you want to split a vector
into its actual (rectangular) components, use one of the latter,
because the former could produce completely different outputs,
in which some values might actually represent angles.
Likewise, do not use ->V2 or ->V3
as shortcuts for 2 ->ARRY or 3->ARRY
because only in RECT mode will the effect be the same.
Finally, note that the DEG/RAD/GRAD mode affects every operation
mentioned above, where either the input or output involves angles,
whose values can be expressed in either degrees, radians, or grads.
Example:
First set these modes: SPHERE RAD STD
Input [ 2. 3. 6. ] and press ENTER, see screen display:
[ 7. <).982793723247 <).541099525957 ]
which are equivalent spherical polar coordinates.
Note that the vector object was NOT internally converted,
but remains the same; only the _display_ changes in appearance.
Apply V-> to the above, get three separate stack outputs:
7. .982793723247 .541099525957
having the same numeric values as the coordinates just displayed;
this operation actually does convert internal coordinates
to coordinates suitable for display, in the implied mode.
Apply ->V3 to these stack arguments, see screen re-display:
[ 7. <).982793723247 <).541099525957 ]
having re-assembled the separate components into the
single object which would display with those components,
two of which, in this case (SPHERE mode) were interpreted as angles;
the object's real _internal_ components have, however,
been converted to rectangular by the command just performed.
Press cursor-down to invoke the [matrix] editor, verifying
that the object's real _internal_ (rectangular) components
are actually [ 2. 3. 6. ]
as they will also appear again on screen after a RECT command.
To demonstrate parsing (and command line input):
RAD [ 7. <).982793723247 <).541099525957 ] STR->
(or typing the entire vector expression into the command line)
always results in the same actual object,
regardless of the current RECT/CYLIN/SPHERE mode,
because parsing considers only explicit angle symbols
as indicators of which values are angles,
although the current RECT/CYLIN/SPHERE mode
still influences how the object is then re-displayed,
either by automatic stack display or by ->STR
Is the following (on HP49G/50G) a counter-example?
Set Exact mode (-105 CF), then input [ 2 3 6 ] SPHERE
Why does it still display as [ 2 3 6 ] ?
Answer: because the object is currently a _symbolic_ matrix,
and the automatic stack displayer (decompiler) does not attempt
to display more complex _symbolic_ expressions for spherical coordinates,
nor to insert cute angle symbols in front of any such expressions,
neither of which is likely to be either desired or useful to most users.
However, performing V-> on this particular object will convert
its integer components to real, and will then convert
to numeric spherical coordinates, just as above;
you may optionally follow with ->V3
to obtain what you might have been expecting to begin with.
You can also input [ 2 3 6 ] SPHERE in _Approximate_ mode,
to directly produce a _numeric_ vector, which displays as before.
If you want to create expressions for symbolic spherical coordinates,
however, you can perfectly well do so, by applying formulas
to the symbolic original components which OBJ-> will separate for you.

===
Subject: Keystrokes for Spanish initial exclamation -- HP50
posting-account=tcgJPQoAAABxYymug63nNNjPqNtEpLIz
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
I'm trying to decipher the CHARS display for the inverted exclamation
mark (character no. 161). The keystrokes displayed are Alpha Right-
shift and then an indecipherable blob. Does anyone know what the
keyboard-access key stroke sequence is? On the HP-48, it was alpha
right-shift DEL.
Irl
===
Subject: Re: Keystrokes for Spanish initial exclamation -- HP50
> I'm trying to decipher the CHARS display for the inverted exclamation
> mark (character no. 161). The keystrokes displayed are Alpha Right-
> shift and then an indecipherable blob. Does anyone know what the
> keyboard-access key stroke sequence is? On the HP-48, it was alpha
> right-shift DEL.
It appears that alpha, right-shift, 2 produces the regular '!', and alpha,
hold-down-the-right-shift-and-press-2 produces the inverted version. The
blob must mean hold.
--
Dave Boyd
If we hit that bullseye, the rest of the dominoes will fall
like a house of cards. Checkmate. -Capt. Zapp Brannigan, D.O.O.P.
===
Subject: Re: Keystrokes for Spanish initial exclamation -- HP50
<Xns9A918C5555457boyddacomcastnet@64.209.0.94>
posting-account=Q2CEjQoAAACue5ZDDhUeRzv1w0u8hxZE
rv:1.9b5) Gecko/2008032619 Firefox/3.0b5,gzip(gfe),gzip(gfe)
> hold-down-the-right-shift-and-press-2 produces the inverted version. 
EThe
> blob must mean hold.
That blob is actually the incredibly wonderful minifont rendering of
an ampersand (&), which makes sense if you think of it as rightshift
AND 2 versus rightshift then 2.
TW
===
Subject: Re: Keystrokes for Spanish initial exclamation -- HP50
>> hold-down-the-right-shift-and-press-2 produces the inverted version. 
ETh
> e
>> blob must mean hold.
 That blob is actually the incredibly wonderful minifont rendering of
> an ampersand (&), which makes sense if you think of it as rightshift
> AND 2 versus rightshift then 2.
I did figure that out, but I have an aversion to following up on my own
posts just to note something like that.
--
Dave Boyd
If we hit that bullseye, the rest of the dominoes will fall
like a house of cards. Checkmate. -Capt. Zapp Brannigan, D.O.O.P.
===
Subject: Re: Keystrokes for Spanish initial exclamation -- HP50
posting-account=tcgJPQoAAABxYymug63nNNjPqNtEpLIz
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> hold-down-the-right-shift-and-press-2 produces the inverted version. 
The
> blob must mean hold.
 That blob is actually the incredibly wonderful minifont rendering of
> an ampersand (&), which makes sense if you think of it as rightshift
> AND 2 versus rightshift then 2.
 TW
+ Advanced User Reference. ... RTFM, I suppose ... but it was a bit
hard to find.
===
Subject: Emu48 for HP50G?
posting-account=y6ooGwkAAACoZd171Nq6oGIjQC5mB4Q9
.NET CLR 2.0.50727; .NET CLR 3.0.04506; .NET CLR 
3.5.21022),gzip(gfe),gzip(gfe)
I see that Emu48 has been ported to many platforms. I was wondering
how much ram does it need to run, the idea is to perhaps port a
special version that could be stored in Flash memory on a 50G calc
(perhaps replacing the current emulator completely). Then the
calculator would be able to run not only as a standard 50G, but as a
true 48 or any of the other models that Emu48 support. This would open
the door to use old software that is only compatible with the 48
series and is otherwise dead.
A properly adapted Emu48 could be able to hot-swap calculator models
from the SD card.
The big question is: can it run with 512 kb of ram? (assuming the
standard emulator is gone).
Is there any interest in such a port?

===
Subject: Re: Emu48 for HP50G?
I think you somehow overrate the power of the 50g...
Emu48 relies on a Win32 API so far.
I think even on my N810,
which has a much faster ARM than the 50g,
Emu48 (under wine) would be next to unusable.
<Half-OT>
Apart from that, why should I use a 50g with that poorly
layouted keyboard for simulating a real HP-48,
which has a perfect keyboard, both from layout and quality?
My approach is the opposite:
Making a real HP-48 as fast as possible, using SpeedUI
</Half-OT>
Raymond
> I see that Emu48 has been ported to many platforms. I was wondering
> how much ram does it need to run, the idea is to perhaps port a
> special version that could be stored in Flash memory on a 50G calc
> (perhaps replacing the current emulator completely). Then the
> calculator would be able to run not only as a standard 50G, but as a
> true 48 or any of the other models that Emu48 support. This would open
> the door to use old software that is only compatible with the 48
> series and is otherwise dead.
> A properly adapted Emu48 could be able to hot-swap calculator models
> from the SD card.
> The big question is: can it run with 512 kb of ram? (assuming the
> standard emulator is gone).
 Is there any interest in such a port?
> 
===
Subject: Re: Emu48 for HP50G?
<fvdk18$36f$01$1@news.t-online.com>
posting-account=sOAX1QkAAAC-FcySTSbz29Uk8huUtFRz
CLR 1.1.4322; .NET CLR 2.0.50727),gzip(gfe),gzip(gfe)
> <Half-OT Apart from that, why should I use a 50g with that poorly
> layouted keyboard for simulating a real HP-48,
> which has a perfect keyboard, both from layout and quality?
 My approach is the opposite:
> Making a real HP-48 as fast as possible, using SpeedUI
> </Half-OT
> Raymond
>
Well, the real HP48 is getting harder and harder to obtain whereas the
50g is currently being produced.
Plus, the HP48 is old. The SX was released in 1990, so the oldest 48's
are already 18 years old. Granted, the HP48's were built to last, but
all electronics will inevitably fail at some point or another.
S.C.
===
Subject: Re: Emu48 for HP50G?
At least the HP-48G+/GX was produced until 2001/2002,
the latest incarnations even with the high contrast display
Raymond
<sc_usenet@hotmail.com> schrieb im Newsbeitrag
> <Half-OT Apart from that, why should I use a 50g with that poorly
> layouted keyboard for simulating a real HP-48,
> which has a perfect keyboard, both from layout and quality?
 My approach is the opposite:
> Making a real HP-48 as fast as possible, using SpeedUI
> </Half-OT
> Raymond
>
Well, the real HP48 is getting harder and harder to obtain whereas the
50g is currently being produced.
Plus, the HP48 is old. The SX was released in 1990, so the oldest 48's
are already 18 years old. Granted, the HP48's were built to last, but
all electronics will inevitably fail at some point or another.
S.C.
===
Subject: Re: Emu48 for HP50G?
posting-account=sOAX1QkAAAC-FcySTSbz29Uk8huUtFRz
CLR 1.1.4322; .NET CLR 2.0.50727),gzip(gfe),gzip(gfe)
> I see that Emu48 has been ported to many platforms. I was wondering
> how much ram does it need to run, the idea is to perhaps port a
> special version that could be stored in Flash memory on a 50G calc
> (perhaps replacing the current emulator completely). Then the
> calculator would be able to run not only as a standard 50G, but as a
> true 48 or any of the other models that Emu48 support. This would open
> the door to use old software that is only compatible with the 48
> series and is otherwise dead.
> A properly adapted Emu48 could be able to hot-swap calculator models
> from the SD card.
> The big question is: can it run with 512 kb of ram? (assuming the
> standard emulator is gone).
 Is there any interest in such a port?
> 
How would you adapt to the new keyboard? The HP48 has six columns of
keys above the number pad whereas the 50g has five columns. Also, the
key labels would not match up.
S.C.
===
Subject: Up to the minute gizmo and gadget news and 

Amazing website for free up to date information on gizmo and gadget
news and reviews and much much more!http://gizmohacker.com/
===
Subject: Re: A bug and a related mini-challenge re LSQ
Once more into the breach.
>To cut to the chase:
I have, completely consistently in this thread,
>as far as my own comprehension goes,
>used the word solution (to a system of equations)
>to mean only one thing, in the exact same sense
>as Golub and Van Loan, whom you quoted yourself,
>meaning that X which makes A*X=B (which you call exact solution).
You quote Golub and Van Loan as saying
>for an under-determined system [which I think meant
>comparing the number of rows vs. columns] there are
>either no solutions or an infinite number of solutions.
I cited that sentence from Golub and Van Loan not for the truth of it, but
because it contradicted what the AUR says about underdetermined systems. I
cited it to show that even authorities can make not quite true statements 
due to
imprecision of language. And to point out the such contradictions can be
resolved by simply qualifying the word solution.
In an earlier section of their book, titled The Full Rank L(east) 
S(quares)
Problem, they say:
Where there are more equations than unknowns, we say that the system Ax=b 
is
overdetermined. Usually an overdetermined system has no EXACT SOLUTION
(emphasis mine) since b must be an element of Range(A).... They say 
exact
solution where the extra word gives precision and clarity, at the price of 
only
one extra word; what's so bad about that?
Your original example
>illustrates the infinite number of solutions case
Why do you so resist saying infinite number of exact solutions, where to 
do so
only takes one more word, as opposed to putting in parentheses, (where 
the
residual can be made zero). Both locutions are commendable for their
precision.
>(where the residual can be made zero),
>my perturbed examples illustrate the no solutions case
Again, this could be no exact solutions instead of a parenthetical of 8 
words.
Golub and Van Load don't mind saying exact solution, nor do I. Why do 
you?
>(where the residual can not be made zero), and
>LSQ does something that can be described to cover both cases,
>which it has been my aim to try to say, in the simplest possible terms.
Unless great care is exercised, simplest possible terms may be ambiguous due 
to
imprecision. Extra adjectives can add precision. and in mathematics, 
elegance
and beauty are connected with accuracy and precision. For example, just 
below
you refer to some characteristic of [B] without telling us what that
characteristic is. Simplest possible terms aren't necessarily sufficient 
for
the reader to know what is meant.
When [A] has linearly dependent rows, there will not in this sense
You see, here's where I would say things differently. Instead of saying 
in
THIS sense, where the reader has to assume you mean in the sense of a few 
lines
before, You probably mean that, but I would rather not have to make an
assumption about it. I would say in the sense that the residual is zero 
and
then there is no possible misunderstanding.
Of course, another way of dealing with the problem is just to define at the
beginning your use of the word solution to mean with a residual of 
zero
which you have done at the beginning of this post. But then if you want to 
say
approximate solution, since you have defined solution to mean 
exact
solution, you're saying in effect approximate exact solution.
What's so bad about just saying exact solution and approximate 
solution when
that's what you mean?
>be a solution for arbitrary B, meaning that you can't
>just take any B (over the entire space) and find a solution,
>unless [B] happens to have the same characteristic,
What characteristic of [B] are you referring to?
>which it had in your original problem, but not in the perturbations
>which I had explored, to see what could more generally be said
>about the action of LSQ -- the first time I looked, I had indeed
>not found the point, but after you elaborated on the subject,
>revealing its true nature, I returned to it,
>and found a generalized meaning
>which I re-posted to bring out, a meaning which, to me,
>takes a very long-winded treatise, containing all sorts
>of qualified words, and boils it down to one simple sentence,
>which I then thought might be of interest to others
>to see and understand, as much as it was to me.
Regardless of any other sentences in the AUR which might also
>flop around as to the meaning of the word solution,
>I quoted only that one sentence which states precisely
>what LSQ does,
But apparently it doesn't precisely state what LSQ does because you're
advocating changing it. If it's precise, why would it need changing? 
I
suppose you could mean something other that what I do by the word 
precise.
If we do what you do, which is to take the word solution to only mean 
that the
residual is zero, then sometimes there are NO solutions. Then not only is 
the
sentence imprecise, it's not always true. This is because the residual 
isn't
zero in all cases for the vector returned by LSQ. If the word solution 
is
understood to mean more than just residual of zero, then the sentence 
could be
true, but, in my opinion, not precise.
>in the simplest possible terms,
>yet it needs just a slight edit to make it just right.
Just above you say that the one sentence ...states precisely what LSQ 
does...
and now you say that it needs '...just a slight edit to make it just 
right'.
What does just right mean? Something more accurate than precise? 
If it
already ...states precisely what LSQ does..., why does it need a slight 
edit?
The original sentence (needing a slight modification) is:
The resulting vector has a minimum Euclidean norm
>over all vector solutions that minimize the residual Euclidean norm.
Okay, my inserting proposed in front of solutions
>wasn't the best way to modify it;
>I should have instead proposed this alternative,
>eliminating the hot button word solutions
>from being mentioned at all:
It need not be a hot button word if it is properly qualified when 
appropriate.
The resulting vector has a minimum Euclidean norm
>over all vectors that minimize the residual Euclidean norm.
How about that? Is there anything left now, to nit-pick about?
I don't think it's nit-picking to point out imprecise language. The 
sentence is
an exactly correct description of what LSQ does, AFAIK. Had you said it or 
had
the AUR said it, I wouldn't say that it was imprecise or ambiguous, but I 
think
it would be too condensed, too succinct.
As far as what it says about solving non-square linear systems, it's as if 
you
explain number theory.
Or, suppose you were discussing some properties of integers that only 
applied to
positive integers and you didn't say positive integers, assuming that 
your
reader would know that you meant positive integers. Some readers might know 
and
some might not. I'm sure you understand that sometimes you should say 
positive
integers.
After reading the ...long-winded treatise, containing all sorts of 
qualified
words..., you came to this point:
...the first time I looked, I had indeed
not found the point, but after you elaborated on the subject,
revealing its true nature, I returned to it,
and found a generalized meaning...
You're admitting that elaboration of the subject was helpful to you in
understanding the subject, more elaboration in fact than was available in 
what
came after fhe first sentence of the LSQ description in the AUR. But are 
you
now advocating the elimination of any further elaboration beyond your 
revised,
first, simple sentence in the AUR? You seem to be suggesting that.
Maybe I'm misunderstanding you. Maybe you're not advocating simplifying 
what is
in the AUR. Maybe you're just saying that the simple sentence The 
resulting
vector has a minimum Euclidean norm over all vectors that minimize the 
residual
Euclidean norm. says it all, FOR YOU, and you find it elegant and 
beautiful.
If so, then mirabile dictu. But, it sure seems to me that you're 
suggesting
that I and others should also be satisfied with that simple sentence.
You now understand that the resulting vector (from LSQ) SOMETIMES 
minimizes
the residual norm all the way to zero, sometimes not. But a new student of 
the
art, upon reading that single sentence, The resulting vector has a 
minimum
Euclidean norm over all vectors that minimize the residual Euclidean 
norm.
probably wouldn't infer that.
I see no harm in explaining the various cases. If a reader sees The 
resulting
vector has a minimum Euclidean norm over all vectors that minimize the 
residual
Euclidean norm. as the first sentence in the description of LSQ and 
understands
the full implications of it, well and good; he can skip the rest.
But, I think further explanation is desireable and I see no reason to say
something like ...over all vectors that have a residual norm of zero... 
and
...over all vectors that have a minimum (but not zero) residual norm... 
rather
than exact solution (as Golub and Van Loan do) or approximate 
solution.
The one sentence The resulting vector has a minimum Euclidean norm over 
all
vectors that minimize the residual Euclidean norm. is indeed elegant, 
beautiful
and correct, but rather too pithy to serve as the only description in the 
AUR.
I myself found the 2nd and 3rd paragraphs helpful the first time I read it.
Two posts ago, I had put it instead as a two-phase problem,
>also without using the word solution:
 1. First find the minimum-norm possible residual ('B-A*X'), then
> 2. Find the minimum norm X giving that value of the residual.
Was there something wrong with that?
>Does it differ from the single sentence version in the AUR,
>once the word solution is removed?
There's nothing wrong as far as it goes, and I explain in detail what IS 
wrong
(not really wrong, but rather lacking) just above. How would the 
linear
algebra neophyte know that sometimes the minimum residual is zero, and 
sometimes
it isn't? And, that distinction is often critical.
Does anyone find it useful
>to have attempted to distill what LSQ means,
>into the simplest possible (and most general) terms?
Yes, in the same sense that it was useful for Peano to have shown that all 
of
arithmetic can be derived from a few axioms. Elegant and beautiful, but of 
less
didactic utility than a fuller explanation. I think it's worthwhile to
explicitly point out the differences in behavior for the underdetermined 
and
overdetermined cases.
>[to which I gave credit to the AUR, for having already said,
>save for one word that should have been struck out]
One can sympathize with all the world's conflicts and disputes,
>which rattle back and forth like this example,
Churchill did say that it's better to talk, talk, talk than to shoot, 
shoot,
shoot. Too bad everybody doesn't feel that way. :-(
>yet hold out the prospect that there exists a commonality
>which all can perceive, which in due course will resolve them,
>and bring harmony to this world. This ultimate harmony is not,
>however, on the level of words, but resides inside the soul,
>the one thread which unites all things, manifest and unmanifest,
>into the pure being which is found inside the self.
http://www.lynchweekend.org/
---
RR: this is mathematics, not poetry
G. H. Hardy ( http://en.wikipedia.org/wiki/G._H._Hardy )
 A mathematician, like a painter or a poet, is a maker of patterns.
> If his patterns are more permanent than theirs,
> it is because they are made with ideas.
 The mathematician's patterns, like the painter's or the poet's,
> must be beautiful; the ideas, like the colours or the words,
> must fit together in a harmonious way. Beauty is the first test;
> there is no permanent place in the world for ugly mathematics.
I hardly think an adjective in front of the word solution makes 
description of
some of the facts of linear algebra ugly. Poetry has poetic license; 
I've
never heard of mathematical license. Precision, correctness and clarity 
are
sine qua non in mathematics.
I try hard to have those qualities in my writing about mathematical 
subjects.
And I say I try hard because it is hard work and time consuming to do so. I 
can
always tell when I've stayed up too late and written too quickly on this 
group;
those qualities suffer.
http://www.oreilly.com/catalog/9780596510046/
>http://www.amazon.com/Beautiful-Code-Leading-Programmers-Practice/dp/059651
0047
>http://safari.oreilly.com/9780596510046/linpacks_dgefa_subroutine
http://www.ams.org/featurecolumn/archive/kissing.html
>http://www.pballew.net/soddy.html
>http://www.ams.org/mathmedia/mathdigest/199811-temple.html
>http://en.wikipedia.org/wiki/Sangaku
>http://lasi.lynchburg.edu/peterson_km/public/old/projects/problems.htm
http://www.math.nus.edu.sg/aslaksen/teaching/math-art-arch.html
>http://en.wikipedia.org/wiki/Alhambra
>http://twoculturespress.com/a_smoother_pebble.php
>[With Isaac Newton's most famous remark, compare below]
Albert Einstein and Rudolf Steiner - astounding!
>http://home.earthlink.net/~johnrpenner//Articles/Einstein3.html
>http://home.earthlink.net/~johnrpenner//Articles/Einstein2.html
>http://home.earthlink.net/~johnrpenner//Articles/Einstein1.html
 The human mind is not capable of grasping the Universe.
> We are like a little child entering a huge library.
 One cannot help but be in awe when one contemplates the mysteries
> of eternity, of life, of the marvelous structure of reality.
> It is enough if one tries merely to comprehend a little
> of this mystery every day. Never lose a holy curiosity.
 How can cosmic religious feeling be communicated
> from one person to another if it can give rise
> to no definite notion of a God and no theology?
> In my view, it is the most important function of art and science
> to awaken this feeling and keep it alive
> in those who are receptive to it.
 To know that what is impenetrable for us really exists
> and manifests itself as the highest wisdom and the most radiant beauty,
> whose gross forms alone are intelligible to our poor faculties --
> this knowledge, this feeling ... that is the core of the true
> religious sentiment. In this sense, and in this sense alone,
> I rank myself among profoundly religious men.
 Albert Einstein [much more at links above]
-=-=-=-
===
Subject: Re: A bug and a related mini-challenge re LSQ
> Once more into the breach
Lock and load :)
> I cited that sentence from Golub and Van Loan not for the truth of it...
You quoted a textbook for its untruth?
Coming from the perspective of solving sets of equations
(rather than solving minimization problems), what it says is true,
and since your adversary was speaking from the very same perspective,
Golub and Van Loan provided a suitable context for elaboration.
> but because it contradicted what the AUR says about underdetermined 
systems.
In the latter postings (perhaps an earlier one referred to
even more potentially confusing terms like xxxx-determined,
from the standpoint of how their common meanings
clash with localized technical meanings),
the AUR was only credited with already containing a one-sentence statement
about the result of LSQ, which was then adjusted slightly by the citer
to make a statement with which you just totally agreed:
> The one sentence The resulting vector has a minimum Euclidean norm
> over all vectors that minimize the residual Euclidean norm
> is indeed elegant, beautiful and correct.
Any decision of yours to continue to take on (and argue with)
the rest of the AUR is now your own private war, on the side.
> ... but rather too pithy to serve as the only description in the AUR.
> I myself found the 2nd and 3rd paragraphs helpful the first time I read 
it.
But you argued as if the person posting had included the rest of the AUR,
not merely credited it with supplying the one item he finally took from it.
> I cited it to show that even authorities can make not quite true 
statements
> due to imprecision of language. And to point out the such contradictions
> can be resolved by simply qualifying the word solution.
Or, by eliminating the word entirely,
or, by recognizing the difference in perspective (or goal)
between two different quests --
one to solve equations or make things equal,
another to solve a minimization problem
or minimize sum of squares of differences
(and then sum of squares of components of vectors which succeed in that).
>> Your original example
>> illustrates the infinite number of solutions case
> Why do you so resist saying infinite number of exact solutions
Infinite number of solutions was in the exact quote
you had just taken from Golub and Van Loan;
now you're criticizing someone who elaborated further
on what they had just said, using the same terminology as they used,
in which solution (to _equations_) was already declared
as meaning exact, so how many times over need that be repeated?
Does approximate solution to an equation (or set of them)
vs exact solution to an equation (or set of them)
sound like better, clearer terminology to anyone else?
As I pointed out in my last post, _any_X_whatsoever_ is an
[approximate] solution to an equation (or set)
if the equals sign(s) in that equation (or set) is(are)
regarded as meaning may or may not be equal,
unless you better define how equal, which sounds reminiscent of
all men are created equal -- but some more equal than others :)
If how equal meant equal enough to minimize the norm of the 
residual,
then that would start to mean something, but I don't recall having
seen that defined, and the whole direction of choosing English phrases
like approximate solution (for equations) strikes me
as being an impediment to most fluidly
getting the issue across to the minds of readers.
> Both locutions are commendable for their precision.
So is the compacting of long-winded descriptions into brief forms,
(hopefully using words and symbols
whose ordinary meaning doesn't confuse things),
which are more readily used to proceed to building higher forms,
given precise initial definitions, which is sometimes called
Mathematics :)
> Golub and Van Loan don't mind saying exact solution
But they didn't say that in the particular excerpt you quoted,
to which the response was directed.
>> you can't just take any B (over the entire space) and find
>> a solution, unless [B] happens to have the same characteristic,
> What characteristic of [B] was being referred to?
Readers (including you) got it from the context;
it of course meant where the same linear relationship
of the rows of A persists in the elements of B
(the combined matrix [AB] thus still having linearly dependent rows)
What we have here is like the classic conflict between legal documents
which grow so complex with verbosity in the name of rigor
that no one can understand them, vs. the movement
(sometimes backed by laws :) to put them in the simplest terms,
for people of any level:
http://www.plainlanguage.gov/
http://www.plainlanguage.gov/examples/legal/index.cfm
http://www.plainlanguage.gov/examples/award_winning/nogobbledygook.cfm
http://ec.europa.eu/enterprise/regulation/better_regulation/simplification/i
ntro.htm
http://members.aol.com/StatutesPA/73.Cp.37.html
(does this satisfy its own demands?)
Interpretation - This act shall be liberally interpreted to protect 
consumers.
(imprecise, but one gets the gist anyway :)
A plan for the improvement of spelling in the English language
http://www.plainlanguage.gov/examples/humor/marktwain.cfm
(precise, rigorous, but eventually incomprehensible :)
> Simplest possible terms aren't necessarily sufficient
> for the reader to know what is meant.
The unadorned word solution was precisely defined by the other writer,
again and again, and then used in the same sense as defined there;
it just happened to disagree with your way of using it.
Or, to put my current point up again for consideration,
the other writer's perspective (in referring to the word solution)
was consistently that of solving equations,
whereas yours was solving a minimization problem related to those 
equations,
so a neutral way of stating the goal of LSQ
(which does both at the same time, in the case of [exact] solutions)
could be framed in a manner completely free of the word solution,
which, it was hoped, could be agreed upon by both.
did it ever reach a final clear definition of what LSQ does?)
with a vector having the minimum Euclidean norm,
among all vectors that minimize the Euclidean norm of the residual
Of course, if the reader does not know what Euclidean norm means,
then (s)he will not know what _either_ of you was talking about :)
Additional example of the purpose of some [re]writings:
http://www.plainlanguage.gov/examples/before_after/carsafety.cfm
> Of course, another way of dealing with the problem is just to define at 
the
> beginning your use of the word solution to mean with a residual of 
zero
> which you have done at the beginning of this post.
So you noticed that!
And you defined, at your beginning, a set of several longer phrases,
each containing the word solution, in your presentation;
each definition (or set) has as its scope
http://www.bartleby.com/246/1129.html
> But then if you want to say approximate solution,
> since you have defined solution to mean exact solution,
> you're saying in effect approximate exact solution.
But that person did _not_ want to say approximate solution,
and clearly expressed wanting to avoid the terminology clash
between the two different general prespectives mentioned,
finally expressing desire to even drop the term solution entirely,
so that no one could possibly think of an ambiguous meaning for it,
requiring more qualifiers, or use any other phrase containing it.
> What's so bad about just saying exact solution
> and approximate solution when what's what you mean?
How could anyone mean expressions which they had already rejected,
as being confusing and meaningLESS to themselves?
You can translate back into your terminology for yourself;
to try to eliminate conflicting terminology
by eventually dropping all terminology
other than minimizing and Euclidean norm.
> I don't think it's nit-picking to point out imprecise language.
You could even point out mis-spellings or grammatical error,
but these things might not be helpful to the cause
of trying to compress a multi-layered club sandwich
into a size which can be more easily chewed and swallowed.
> The sentence is an exactly correct description of what LSQ does, AFAIK.
> Had you said it or had the AUR said it,
> I wouldn't say that it was imprecise or ambiguous,
> but I think it would be too condensed, too succinct.
In my last post, I congratulated each for achieving some goal,
one for a concise yet fully general expression,
the other for expounding and illustrating cases
with details and examples,
both of which I think have value;
as to their differences in terminology, I suggested
that it was based on differing perspectives, solving equations
vs. solving a minimization problem related to those equations,
and that elimination of the word solution might be useful
in the cause of avoiding that confusion of perspective.
> You're admitting that elaboration of the subject was helpful to you in
> understanding the subject, more elaboration in fact than was available in 
what
> came after fhe first sentence of the LSQ description in the AUR. But are 
you
> now advocating the elimination of any further elaboration beyond your 
revised,
> first, simple sentence in the AUR? You seem to be suggesting that.
No, he is providing another, unified perspective that you didn't,
and you are providing a detailed analysis of cases which he didn't.
And I expect you'll all agree
That he was right to so decree.
And I am right,
And you are right,
And all is right as right can be!
http://math.boisestate.edu/gas/mikado/webopera/mk103.html
http://math.boisestate.edu/gas/mikado/webopera/mk_midi/103.mid
Note that the above work of detailed research, art and love,
is on a beautiful _math_ site :)
> Churchill did say that it's better to talk, talk, talk
> than to shoot, shoot, shoot.
> Too bad everybody doesn't feel that way. :-(
Yeah, like a current U.S. Administration,
which is soon to be replaced by one which may talk differently,
but possibly act the same.
Winston Churchill, however, if I heard correctly,
was then the one who fought to start preparing to be ready to shoot,
before it might be too late, and it appears that he was right, too,
as is yet another approach which neither talks nor shoots,
but opens human awareness to its own fuller, deeper nature,
whereupon it appreciates more and fights less,
identifying and harmonizing itself with a more global self.
Contradictions abound, yet co-exist:
Co-existence, or no existence.
[Piet Hein, http://en.wikiquote.org/wiki/Piet_Hein ]
> I try hard to have those qualities in my writing about mathematical 
subjects.
> And I say I try hard because it is hard work and time consuming to do so.
All careful (and clear) writing takes time; all education takes time;
connecting the dots between all forms of Knowledge,
(including self-Knowledge) takes time.
Time's fun when you're having flies. - Kermit the Frog
Time is a great teacher,
but unfortunately it kills all its pupils. - Hector Berlioz
I wasted time, and now doth time waste me. - Shakespeare
Dost thou love life? Then do not squander time;
for that is the stuff life is made of. - Benjamin Franklin
---
The meanings of words are often the issue
with disagreements or misunderstandings;
I hope that we are coming to a common perspective,
When I use a word, Humpty-Dumpty said,
it means just what I choose it to mean -- neither more nor less
- Lewis Carroll, Through the Looking Glass... (1872)

===
Subject: Re: A bug and a related mini-challenge re LSQ
See my response to John Meyers in another post for some additional 
discussion on
solutions to underdetermined systems.
===
Subject: Re: A bug and a related mini-challenge re LSQ
> First, some terminology. The AUR uses the word solution
>> in places where it should add an adjective in front to be precise.
>> I'll use the terms exact solution, approximate solution,
>> minimum norm solution and minimum residual solution.
> By exact solution, I mean a solution x (a vector)
>> that gives a residual norm, NORM(A*x-b), of exactly zero.
> By approximate solution, I mean a solution x that gives a non-zero
>> residual. There are almost always many approximate solutions,
>> even if there is also an exact solution (he said tautologically).
> By minimum residual solution, I mean a solution which gives a 
minimum
>> residual norm. Typically this term will be used when there are many
>> approximate solutions and no exact solution, but over the set of
>> approximate solutions, only one has the smallest residual norm;
>> this would be the common overdetermined least squares solution.
> By minimum norm solution, I mean a solution vector which has the
>> smallest Euclidean length (norm) compared to all other solutions. In 
some
>> underdetermined systems, there an infinite number of exact solutions,
>> but only one solution vector is of minimum length.
So then, is a minimum norm solution, among all minimum residual 
solutions
>what LSQ does for under-determined systems?
Consider the original system this thread started with, A*X=B, where:
A=
[[ 1 1 0 0 0 -1 ]
[ 1 0 -1 0 1 0 ]
[ 0 1 1 1 0 0 ]
[ 0 0 0 1 1 1 ]]
and B=
[[ 4 ]
[ 8 ]
[ 12 ]
[ 16 ]]
There exists a vector X, [ 3 4 1 7 6 3 ]T such that the residual A*X-B, is 
zero;
that is, all the elements of the vector A*X-B are zero, and therefore its
Euclidean norm, |A*X-B|, is zero.
For any (underdetermined) system such as this where the A matrix has 
dimensions
m x n with m < n, the rank of A can't be larger than the lesser of m and n.
This means that A will have a non-empty null space of dimension at least 
n-m
(I'm not going to consider pathological systems such an A matrix of all 
zeroes).
In this case the rank of A is 3, and its null space has dimension 3. We 
can
find 3 linearly independent vectors that span the null space. Three such
vectors are n1 = [ -1 1 -1 0 0 0 ]T, n2 = [ 1 -1 0 1 -1 0 ]T and n3 = [ 0 -1 
0 1
0 -1 ]T. If you premultiply any linear combination of those vectors by A, 
you
will get a vector of all zeroes.
For example, A*n1 = [ 0 0 0 0 ]T and A*(57*n1 + 33*n2 - 19*n3) = [ 0 0 0 0 
]T.
Therefore if we add any linear combination of n1, n2, n3 to [ 3 4 1 7 6 3 ]T 
and
premultiply by A, we will get [ 4 8 12 16 ]T.
To demonstrate this, let's test whether any of these three vectors is a
solution to the original system:
s1 = [ 3 4 1 7 6 3 ]T
s2 = s1 + 29*n1 = [ -26 33 -28 7 6 3 ]T
s3 = s1 + 17*n1 - 42*n2 + 71*n3 = [ -68 4 -28 36 48 -68 ]T
We find that:
A * [ 3 4 1 7 6 3 ]T = [ 4 8 12 16 ]T
A * [ -26 33 -28 7 6 3 ]T = [ 4 8 12 16 ]T
A * [ -68 4 -28 36 48 -68 ]T =[ 4 8 12 16 ]T
All of them are solutions to the original system A*X=B.
It's easy to see that any linear combination of the null vectors n1, n2 and 
n3
added to any solution of the system yields another solution. 
Therefore
there are an infinite number of solutions to the original system, but 
only one
of the solutions has minimum Euclidean length. The solution with 
minimum
Euclidean length is the one which doesn't contain any additive combinations 
of
the null vectors.
To repeat, any underdetermined system (having dimensions m x n, where m < 
n)
will have a non-empty null space, and there will exist null vectors such as 
the
previously exhibited n1, n2 and n3.
Now consider the system A*X=B', where A is as shown above, and B' is the
original B vector with its last element changed from 16 to 17:
B'=
[[ 4 ]
[ 8 ]
[ 12 ]
[ 17 ]]
There is no 6 x 1 vector X such that the residual norm |A*X-B'| is zero.
However, if we consider all possible 6 x 1 vectors, X, and calculate the
residual norm |A*X-B'| (use the Euclidean norm) for all of them, we will 
find
that there is a minimum residual norm among all of the residual norms thus
obtained. And, there is more than one 6 x 1 vector that gives that minimum
residual norm; in fact, there is an infinite number of them.
Imagine that we have found a vector Z that gives the minimum norm. If nk is 
a
null vector (a vector in the non-empty null space) for the A matrix, then 
the
residual A*(Z + nk)-B' will have the same norm as A*Z-B', because A*nk=0.
If we have a vector Z that gives a minimum norm residual A*Z-B' for the 
system
A*Z=B', the adding any linear conbination of the null vectors of the A 
matrix to
Z will give us another vector that has the exact same residual norm.
If we choose to call the vector Z, a solution to the system A*X=B' 
(even
though the residual norm is not zero), then there are an infinite number of 
such
solutions, which I've just shown how to construct, but there is only one 
such
that its Euclidean length is a minimum.
Saturn Rising said in another post:
>Okay, I've incrementally perturbed until 16 has become 17,
>and the result of B A LSQ is now... still wrong (Rodger's bug),
>but using B B + A A + LSQ gets the right result, which is:
 [[ 3 ]
> [ 4 ]
> [ 1 ]
> [ 7.25 ]
> [ 6.25 ]
> [ 3.25 ]] 'X' STO
In which case 'A*X' is:
 [[ 3.75 ]
> [ 8.25 ]
> [ 12.25 ]
> [ 16.75 ]]
And residual 'B-A*X' is:
 [[ .25 ]
> [ -.25 ]
> [ -.25 ]
> [ .25 ]]
Now that we have the right answers,
>there still remains the question:
What did LSQ minimize?
It did not minimize the length of X
>over all solutions [i.e. where 'A*X=B']
>of the 'under-determined' system,
>because there are no such solutions, not even approximately,
>so what does this result really mean?
I've demonstrated that LSQ did in fact minimize the length of X over 
all
solutions [i.e. where 'A*X=B'] of the 'under-determined' system,, because 
there
are inded such solutions in the sense that the AUR uses the word (which 
is
that A*X-B need not be exactly zero). One might even say that the 
solutions
are approximate.
In the case at hand, the vector B' is not much perturbed from B and the
solution to the system A*X=B' is not very much different from the 
solution
to the unperturbed system A*X=B. It is a reasonable solution to the 
system
A*X=B', because the residual norm is not too large conpared with the 
elements of
the system. It can even be reasonably considered an approximate 
solution to
the original system. The reason the solution is not much perturbed is that 
the
first 3 singular values of A are all 2, and the way we solve systems such 
as
this amounts to deleting the last (nearly zero) singular value, so that the
system has an effective condition number of 1.
The solutions to the system A*X=B' differ in a certain property from 
the
solutions to the system A*X=B. The latter have a minimum residual norm of 
zero
while the former have a minimum residual norm which is not zero. Notice 
that it
takes several words to describe that difference in the preceeding sentence. 
One
might describe the difference with a single aptly chosen word. The 
solutions
to the system A*X=B might be called exact solutions and the 
solutions to the
system A*X=B' might be called approximate solutions. The solutions 
to
systems such as A*X=B' are also called least squares solutions, as are 
the
solutions to underdetermined systems when the residual norm is not zero.
The AUR says that undertermined systems have an infinite number of 
solutions.
What they are referring to is the property of underdetermined systems of 
always
having a non-empty null space, so that given one solution, an infinite 
number
of other solutions may be constructed by adding a linear combination of the 
null
vectors to any given solution.
But Golub and Van Loan say that underdetermined systems have either no 
solution,
or an infinite number of solutions. The apparent contradiction with the AUR 
is
because when Golub and Van Loan say solution, they mean solution 
with a
residual norm of zero, which is not what the AUR means by solution (to 
an
underdetermined system).
The singular value deconposition in its modern generalization was discovered 
by
E. H. Moore in 1935. It was re-discovered by the physicist Roger Penrose 
in
1955. Penrose published a paper in the Proceedings of the Cambridge
Philosophical Society in 1955, titled On Best Approximate Solutions of 
Linear
Matrix Equations, demonstrating the use of the SVD for solving linear 
systems
that don't have a solution with a residual norm of zero.
In 1957 and 1959, T. N. E. Greville published papers wherein he discusses 
the
uses of the pseudoinverse of a matrix. In the 1957 paper, he says 
Bjerhammar
and Penrose have studied the application of matrices to obtain an 
approximate
solution of a system of equations for which no exact solution exists.
The LSQ command is effectively computing the pseudoinverse; that's how it 
does
its magic (when it works right).
There are situations where the use of the terms approximate solution 
and
exact solution are appropriate. Sometimes the reader is not an expert 
on
linear algebra, and what may seem to be a slight redundancy to some may 
help
avoid misunderstanding for others.
>It sounded simpler (and more general)
>where the AUR sums it up in one sentence, though,
>if you completely drop the word solution from it, as someone said,
>but all equivalent things are equal (he said, confusedly :)
> In the early days of linear algebra (Gauss' time), elimination methods
>> were used to solve linear systems. Such methods have a hard time with
>> systems whose A matrix is singular, but such systems can often be solved
>> approximately with modern methods, giving a relatively small residual
>> (we hope). Sometimes systems with a singular A matrix
>> even have an exact solution.
Even in approximate mode? [just a rhetorical, humoretical interjection 
:]
> Modern methods of solution using orthogonal decompositions
>> are able to provide solutions for almost any system.
> The AUR doesn't mention it, but a third party book I have mentions
>> that LSQ uses QR and LQ decompositions to obtain solutions.
Sounds good, and you and Werner seem to have found the bug,
>so congratulations!

===
Subject: Re: A bug and a related mini-challenge re LSQ
(yet another new voice in this crowd :)
One can go eternally round and round,
dwelling (and flipping between)
the variations of possible meaning of the word solution,
even further confounding the issue by bringing in the phrase
least squares solution, which is not even a solution (exact),
in various cases, to the original set of equations
(characterized by the presence of an equals sign,
rather than a may or may not be equal sign :)
but which means the answer to a challenge of minimizing something,
not necessarily of making some things equal --
indeed, it can even end up with NO things being actually equal.
If we don't assume the word exact in the word solution
(to a set of equations A*X=B, he defines over and over, again and again,
rather than to a related but not identical minimization problem),
then we would have a situation where we would have to say
that any X whatsoever is a solution,
just one in which the norm of the residual 'B-A*X'
doesn't happen to be zero,
which somewhat diminishes the value of bringing in a special word
for something which has no special characteristic at all.
Terminology for all X that produce the minimum norm of the residual
begins to narrow the field of all X to a more interesting set,
which it is interesting that you have characterized
in a certain very interesting and satisfying way,
and the selection of an X from that set,
which then minimizes the norm of X,
is finally what LSQ delivers (or is that still in dispute?)
If someone had set after different goals, perhaps
maximize the number of equations actually satisfied,
or minimize the sum of fourth powers, rather than squares,
then other sorts of different interesting result sets might emerge,
but LSQ has certain properties that make it interesting
(have those been elaborated?), as well as the property
that there is an elegant way to get the answer :)
The different participants in this thread
seem to have set out after differing goals;
perhaps that's why it keeps going round and round.
One particpant's goal seems to have been to find
the shortest definition of the most general meaning
of LSQ's result, covering all possible cases,
and a vector having the minimum Euclidean norm,
among all vectors that minimize the Euclidean norm of the residual
seems, to me, to be a satisfactory attainment of that goal,
at the same time managing to completely dispense
with any mention of the word solution,
thus completely obviating its definition.
When the Euclidean norm of the residual can be made zero,
one then also has what that participant
(as well as the other's quoted textbook) calls solution(s),
meaning exact solution(s), meaning where the equals signs
(in that pesky set of original equations )
stand between expressions that are actually equal.
Another participant's goal seems to have been to elaborate
and investigate as deeply as possible into the ramifications
of many different situations -- overdetermined, underdetermined,
exact solution, approximate solution, how to characterize
the entire set of all those which have the minimum norm
of the residual, etc., and that goal, too,
seems to have been well attained,
plus the added bonus of interesting historical notes.
So if those were the goals, then a win-win situation seems to have 
occurred;
everyone attained a goal, and all was posted for all interested parties to 
share.
Perhaps the illusion of argument in all these cases
arises from confounding two entirely different things
to be solved -- one being a set of equations, without reference
to minimizing anything, just to making things equal (if possible),
while the other is the solving of a minimization problem
(minimizing the sum of squares of the amounts of inequality,
then the sum of squares of the eligible vectors which do that),
without reference to making anything equal,
where the two differently stated problems merge
only when both can happen at once.
This seems as likely to be agreed upon, however,
as for the main political parties in any country
to ever agree on anything -- perhaps if only they were
trained as mathematicians, instead of as lawyers and businessmen
(and human beings, with all their inner conflicts raging),
would that bring peace to the world? I think not.

===
Subject: Re: A bug and a related mini-challenge re LSQ
>(yet another new voice in this crowd :)
One can go eternally round and round,
>dwelling (and flipping between)
>the variations of possible meaning of the word solution,
It doesn't take eternal time to list the few adjectives necessary to 
enumerate
the kinds of solutions found in linear algebra; exact and 
approximate cover
most of the cases, with a few terms such as least squares (which is a
particular approximate solution) or least absolute value to take care of 
the
special cases.
>even further confounding the issue by bringing in the phrase
>least squares solution, which is not even a solution (exact),
In that case it's a solution (least squares).
>in various cases, to the original set of _equations_
>(characterized by the presence of an _equals_ sign,
>rather than a may or may not be equal sign :)
There's always the approximately equal sign, ~, which is somewhat more 
useful
than a may or may not be equal sign. 
>but which means the answer to a challenge of minimizing something,
>not necessarily of making some things equal --
>indeed, it can even end up with NO things being actually equal.
One can even solve inequalities, the solution leading to things which 
are not
equal. Mathematics has been able to deal with this.
Not every solution to a mathematical problem ends with things being 
equal.
Sometimes it does, sometimes it doesn't.
My mathematics dictionary has two general definitions for the word 
solution:
1) The process of finding a required result.
2) The result of that process is also spoken of as the solution.
If we don't assume the word exact in the word solution
How about if we DO assume the word approximate, which is what I do when 
the
word exact doesn't apply. If the word exact isn't in the word 
solution, I
don't just leave solution hanging alone; I add approximate.
>(to a set of _equations_ A*X=B, he defines over and over, again and again,
>rather than to a related but not identical minimization problem),
What do you mean, he defines over and over, again and again, rather than 
to a
related but not identical minimization problem? I did define a not 
identical
minimization problem, namely A*X=B', where B' was a perturbed version of 
the
original B.
>then we would have a situation where we would have to say
>that _any_X_whatsoever_ is a solution,
I said to consider all possible 6 x 1 vectors; I didn't say to say that 
they
are solutions. The only possible solutions are the ones whose residual 
norms
are equal to the minimum possible.
>just one in which the norm of the residual 'B-A*X'
>doesn't happen to be zero,
>which somewhat diminishes the value of bringing in a special word
>for something which has no special characteristic at all.
What special word are you referring to here? Could it be 
approximate? We
DON'T have to say that _any_X_whatsoever_ is a solution. We can speak of
approximate solutions quite handily. The mathematics dictionary 
defines
approximate as:
1) A result that is nearly but not exactly correct.
This provides a special characteristic, and its value is that its use 
means
that we don't have to consider _any_X_whatsoever_ to be an approximate
solution.
We will have to use the common understanding for what nearly means, I 
suppose.
But let's say within 10% of the correct value for starters. If you 
consider
_any_X_whatsoever_, you will find values of X that are orders of magnitude
different from the correct solution. We DON'T have to say that those X are
included in the meaning of solutions.just because we don't assume the 
word
exact in the word solution; we can assume (or preferably make 
explicit) the
word approximate. In the linear algebra context, we can start with the
understanding that approximate solution means not exact but nearly 
so and
then further refine it for LSQ to mean minimum norm.
Terminology for all X that produce the minimum norm of the residual
>begins to narrow the field of all X to a more interesting set,
>which it is interesting that you have characterized
>in a certain very interesting and satisfying way,
>and the selection of an X from that set,
>which then minimizes the norm of X,
>is finally what LSQ delivers (or is that still in dispute?)
When LSQ works correctly, that is what it delivers.
If someone had set after different goals, perhaps
>maximize the number of equations actually satisfied,
Some of this fancier sort of thing probably can be done most easily with 
the
singular value decomposition.
>or minimize the sum of fourth powers, rather than squares,
>then other sorts of different interesting result sets might emerge,
Indeed. The definition of solution from the mathematics dictionary 
makes it
clear that you could call these interesting results, solutions, because
setting goals is the same as establishing requirements, and to 
find a
required result is to obtain a solution.
>but LSQ has certain properties that make it interesting
>(have those been elaborated?),
Joe's slight edit job on the first sentence of the LSQ description in the 
AUR
says it all, but doesn't provide much elaboration, as I pointed out: LSQ
returns a vector such that The resulting vector has a minimum Euclidean 
norm
over all vectors that minimize the residual Euclidean norm.
My long post, to which you are responding here, elaborated on the properties 
of
LSQ for underdetermined systems.
Often there are an infinite number of solutions that have the same minimum
residual norm, and requiring that, additionally, the norm of the vector 
itself
be minimized provides a unique result (which happens to have some 
desireable
properties), rather than an infinity of results!
I think I pointed out somewhere else that the system A*X=B, where:
B=
[[ 4 ]
[ 8 ]
[ 12 ]
[ 16 ]]
and A=
[[ 1 1 1 1 ]
[ 2 2 2 2 ]
[ 3 3 3 3 ]
[ 4 4 4 4 ]]
has an exact solution, [ 1 1 1 1 ]T, but normal elimination methods choke on 
it.
LSQ has no problem with it.
LSQ can also be used to compute the pseudoinverse of a matrix. This little
program does the job:
<< DUP SIZE OBJ-> DROP DUP2 IF > THEN SWAP DROP IDN
SWAP TRN LSQ TRN ELSE DROP IDN SWAP LSQ END >>
But, note that this fails when given the original A matrix in this thread 
due to
the bug in LSQ:
[[ 1 1 0 0 0 -1 ]
[ 1 0 -1 0 1 0 ]
[ 0 1 1 1 0 0 ]
[ 0 0 0 1 1 1 ]]
The correct pseudoinverse is:
[[ .25 .25 0 0 ]
[ .25 0 .25 0 ]
[ 0 -.25 .25 0 ]
[ 0 0 .25 .25 ]
[ 0 .25 0 .25 ]
[ -.25 0 0 .25 ]]
>as well as the property
>that there is an elegant way to get the answer :)
The different participants in this thread
>seem to have set out after differing goals;
>perhaps that's why it keeps going round and round.
One particpant's goal seems to have been to find
>the shortest definition of the most general meaning
>of LSQ's result, covering all possible cases,
>and a vector having the minimum Euclidean norm,
>among all vectors that minimize the Euclidean norm of the residual
>seems, to me, to be a satisfactory attainment of that goal,
>at the same time managing to completely dispense
>with any mention of the word solution,
>thus completely obviating its definition.
When the Euclidean norm of the residual can be made zero,
>one then also has what that participant
>(as well as the other's quoted textbook) calls solution(s),
>meaning exact solution(s), meaning where the equals signs
>(in that pesky set of original _equations_)
>stand between expressions that are actually equal.
Another participant's goal seems to have been to elaborate
>and investigate as deeply as possible into the ramifications
>of many different situations -- overdetermined, underdetermined,
>exact solution, approximate solution, how to characterize
>the entire set of all those which have the minimum norm
>of the residual, etc., and that goal, too,
>seems to have been well attained,
>plus the added bonus of interesting historical notes.
So if those were the goals, then a win-win situation seems to have 
occurred;
>everyone attained a goal, and all was posted for all interested parties to 
share.
Perhaps the illusion of argument in all these cases
>arises from confounding two entirely different things
>to be solved -- one being a set of equations, without reference
>to minimizing anything, just to making things equal (if possible),
>while the other is the solving of a minimization problem
>(minimizing the sum of squares of the amounts of inequality,
>then the sum of squares of the eligible vectors which do that),
>without reference to making anything equal,
>where the two differently stated problems merge
>only when both can happen at once.
This seems as likely to be agreed upon, however,
>as for the main political parties in any country
>to ever agree on anything -- perhaps if only they were
>trained as mathematicians, instead of as lawyers and businessmen
>(and human beings, with all their inner conflicts raging),
>would that bring peace to the world? I think not.

===
Subject: Re: A bug and a related mini-challenge re LSQ
> The only possible solutions are the ones whose residual norms
> are equal to the minimum possible.
There ya go -- a definition which I don't recall noticing earlier;
define approximate solution as this, and call it a draw :)
> We can speak of approximate solutions quite handily.
> The mathematics dictionary defines approximate as:
 1) A result that is nearly but not exactly correct.
> We will have to use the common understanding
> for what nearly means, I suppose.
> But let's say within 10% of the correct value for starters.
Uh-oh, there ya go again, in a different sense (Ronald Reagan's :)
Irreconcilable differences?
>> ...is finally what LSQ delivers (or is that still in dispute?)
> When LSQ works correctly, that is what it delivers.
I had conceded about the bug long ago, and moved on.
> The definition of solution from the mathematics dictionary makes it
> clear that you could call these interesting results, solutions, 
because...
Which reminds me of the proof (!)
that there are no uninteresting numbers, [integers, excuse me]
which is that if there were any uninteresting numbers, [integers]
then there would be a smallest uninteresting number [integer]
(in the Euclidean norm [ABS] sense),
and that would be interesting, wouldn't it?
> The slight edit job on the first sentence of the LSQ description
> in the AUR says it all, but doesn't provide much elaboration,
> as I pointed out: LSQ returns a vector such that
> The resulting vector has a minimum Euclidean norm
> over all vectors that minimize the residual['s] Euclidean norm.
Right; too condensed, too succinct, but still correct.
> My long post, to which you are responding here,
> elaborated on the properties of LSQ for underdetermined systems.
Yes, very good elaboration, illustrating case(s)
of the general, too succinct statement
(I thought I also just posted appreciation of both kinds of expositions,
long and short, providing a total range of perspectives).
> Often there are an infinite number of solutions that have the same 
minimum
> residual norm, and requiring that, additionally, the norm of the vector 
itself
> be minimized provides a unique result (which happens to have some 
desireable
> properties), rather than an infinity of results!
Yep, always good to narrow down to one single possible result
(though no proof was offered that this is what happens :)
Otherwise we'd hear but my TI says... :)
> LSQ can also be used to compute the pseudoinverse of a matrix.
> This little program does the job:
>
<< DUP SIZE OBJ-> DROP DUP2 IF > THEN SWAP DROP IDN
SWAP TRN LSQ TRN ELSE DROP IDN SWAP LSQ END > But note that this fails when given the original A matrix in this thread
> due to the bug in LSQ
Good to remember that the LSQ bug will affect programs employing LSQ;
I might have forgotten that :)
I hope it's like the Linear program solutions which cycle,
however, in that they are so rare that most people have
never seen one in practice, yet texts properly cover
how to make it impossible to cause trouble, even then
(is there any wrapper program to fix LSQ, like the way
in which spectacles were added to the Hubble telescope? :)
>> The different participants in this thread
>> seem to have set out after differing goals;
>> perhaps that's why it keeps going round and round.
> [...]
>> Perhaps the illusion of argument in all these cases
>> arises from confounding two entirely different things
>> to be solved -- one being a set of equations, without reference
>> to minimizing anything, just to making things equal (if possible),
>> while the other is the solving of a minimization problem
>> (minimizing the sum of squares of the amounts of inequality,
>> then the sum of squares of the eligible vectors which do that),
>> without reference to making anything equal
No comment?
Oh, well.
plus the excuse it offered to take excursions into other places.
Time's fun when you're having flies [Kermit the frog]
.
===
Subject: Re: A bug and a related mini-challenge re LSQ [what's in a word?]
> [Eliminating the word solution from a discussion is] an unworkable 
solution.
> Linear algebra texts are going to continue to use the word solution.
The AUR and/or these newsgroup discussions are certainly free
to offer discussions that can explain what's going on
without using the word, or which can include,
either before or after such an explanation,
that in connection with LSQ, the word solution is used
to mean any X for which the norm of the residual 'B-A*X'
attains its minimum possible value,
to make completely clear that this is not necessarily
a solution to the equations A*X=B.
What I see as the heart of the differing takes on this subject
is that the word solution was never properly defined anywhere
(when you gave a definition, it was in the form of defining
a longer phrase containing qualifying words in front of
the word solution, which makes it seem that your definition
is about a special kind of solution, which would then
make it seem that solution must be some larger class,
and what larger class is there than all X?).
The AUR also really, IMO, steers the reader (who may not be
an expert in linear algebra, but just an ordinary calculator user)
into believing that solution refers to a solution
of the equations A*X=B -- the AUR actually says that,
even though it also contradicts itself elsewhere.
LSQ can be used for other purposes, even to refine
the accuracy of solutions to the equations, I believe,
so I think it deserves an explanation that clarifies this.
I just made analogy between LSQ and LR, and I like the idea
of expressing LSQ as something looking for a best fit
for a set of equations, just as LR looks for a best fit line,
even to a set of points that don't lie on a line
(LSQ thus also acting upon equation sets that have no
exact-fit solutions) -- this can be pointed out,
whether or not linear algebra texts do so, and I think
it will help to add something for general understanding.
Even the numeric solver, when there is a local minimum
but no exact solution, tries to locate that local minimum instead,
just as one more slightly analogous function that exists in the calculator.
Some of the back and forth in this thread seems to have come from
taking statements to an extreme interpretation; for example,
my saying that there is a correct one-sentence general definition
for the action of LSQ did not mean objection to elaborating
all the particular cases, and all the other material that could
be presented, and my saying that we can also explain LSQ
even without having to use or define the word solution
does not mean that we can't also include a definition --
just that if we are going to explain it using only
the word solution, then we (the AUR, say) had better
actually say what we really mean, as well as not
saying incorrect (and confusing) things.
A rose by any other name would smell as sweet,
but I guess this doesn't work the same for solution :)
Language must be the fundamental sin,
because as soon as human beings got that knowledge,
that's where they've gotten into trouble ever since :)
-- Y
===
Subject: Re: A bug and a related mini-challenge re LSQ [what's in a word?]
> LSQ can be used for other purposes, even to refine
> the accuracy of solutions to the equations, I believe.
My bad memory -- that was RSD (calculates the residual,
B-A*X from B, A, and X, hopefully with increased precision)
..where B and X can be either one vector,
or a matrix of several vectors, just as with LSQ.

===
Subject: Re: Install libraries
<op.ualjjysbnn735j@miu.edu>
posting-account=tK2cUgkAAABPl7vxXuLS8-X63eEhN1C3
1.1.4322; InfoPath.1; .NET CLR 2.0.50727; .NET CLR 3.0.04506.30; 
.NET CLR
3.0.04506.648),gzip(gfe),gzip(gfe)
[how to transfer files from computer to calculator]
-Joe-
===
Subject: Re: Install libraries
Hope you are feeling ever better,
and will never again be run over by cars,
or hacked up by doctors.
I hear that there are acting roles available,
however, which evidently pay well
for just pretending things like that,
which experienced persons could do even better,
and these involve not even memorizing lines --
just getting some fancy make-up,
and lying down while other people talk
(on those CSI shows :)
All best wishes.
.
===
Subject: Re: hpgcc and rpl-stack
<op.uam42lw1nn735j@miu.edu>
posting-account=y6ooGwkAAACoZd171Nq6oGIjQC5mB4Q9
.NET CLR 2.0.50727; .NET CLR 3.0.04506; .NET CLR 
3.5.21022),gzip(gfe),gzip(gfe)
> Since I know nothing, but just have curiosity,
> I can ask real dumb questions from pure ignorance
> (or feel free to refer me to fine manuals :)
Well... we don't have such good documentation, so I can't really use
the usual RTFM. It's not my style anyway.
 Is there a GC upon (or before) initial entry?
Yes, there is a GC if you call it using PrRUN, you can skip it if you
execute it using the other commands (PrLNK and PrJMP), in case you
don't like the speed penalty.
Let me clarify that I'm talking about HPGCC 2.0. The new version
(HPGCC3) behaves completely different.
 If not, does pop/push affect RPL stack memory,
> which could have been very close to limit
> (e.g. one word away from needing GC) to begin with?
You can always run a program under very low memory conditions, with GC
or not. In such case, push will fail (return without pushing the
value) if there's not enough memory.
 In any case, there must always be some limit,
> at which points the RPL stack(s) either overflow or underflow?
Overflow, yes, but it is under control, since the C program will
simply refuse to push any new objects on the stack when memory is
insufficient.
Underflow can't really happen, since you can't have less than 0 items
on the stack . pop will simply fail on an empty stack
 Is there any access to RPL return stack and environments?
No, we have no such functions. But using peek() and poke() you can
push values in the return stack manually. It's not very useful,
though, since you can't touch anything on the RPL stream or your C
program will crash on return.
I hope this satisfies your curiosity,

===
Subject: Re: hpgcc and rpl-stack
Hi
> Overflow, yes, but it is under control, since the C program will
> simply refuse to push any new objects on the stack when memory is
> insufficient.
> Underflow can't really happen, since you can't have less than 0 items
> on the stack . pop will simply fail on an empty stack
Last time I used HPGCC, you couldn't tell unfortunately, weither the
push worked or not. it wouldn't return a value letting you that it
worked. If it couldn't push a value on the stack, it would simply do
nothing
As such, I had to manually test the amount of memory available before I
called push.
JY
--
They who would give up an essential liberty for temporary security,
deserve neither liberty or security (Benjamin Franklin)
===
Subject: Re: hpgcc and rpl-stack
<481fcf48$0$21322$426a74cc@news.free.fr>
posting-account=y6ooGwkAAACoZd171Nq6oGIjQC5mB4Q9
1.1.4322; .NET CLR 2.0.50727),gzip(gfe),gzip(gfe)
> Last time I used HPGCC, you couldn't tell unfortunately, weither the
> push worked or not. it wouldn't return a value letting you that it
> worked. If it couldn't push a value on the stack, it would simply do
> nothing
 As such, I had to manually test the amount of memory available before I
> called push.
And that is why the entire API was rewritten from scratch. The
functions we are discussing don't even exist anymore. The new version
returns BOOL, allowing the proper check.

===
Subject: Re: HP's award Winning Support Team Explains the ONe Year Warranty
net@comcast.baranow says...
> Here is an excerpt from a private email sent today to a would be HP50
> seller explaining why, after talking to HP Calculator Support, I am
> unable to take a chance on the 50g which is being offered for resale.
 Basically, the logic is that all things considered, and having been
> burned once, it is too risky to purchase this product knowing that more
> likely than not it will fail and that the terms and conditions for the
> warranty are far to harsh to roll the dice.
 ------------------BEGIN EXCERPT---------------------------------

> The HP support people told me that the warranty expires one year from
> the time of purchase or one year from the date of manufacture whichever
> comes first, but that the warranty is not transferable meaning that for
> you, the original owner, it will be good for one year from the date
> purchased, but for the second owner, one who is not registered for the
> warranty, they would use the first 3 digits of the serial number and add
> one year to it. So, for anyone who buys it, they will have until the
> 20th week of year 2008 (the second and third numbers represent the week
> of production).
 So, for a calculator that was manufactured, sat on the shelf, was later
> purchased, and subsequently sold, the new buyer doesn't get a full year
> warranty. In other words, the meter was running while the cab was
> stalled in traffic.
 Exacerbating all of this is the fact that this unit is more likely than
> not not fail. Nobody in their right mind is going to take a chance with
> HP with this attitude from their award winning support team.
 I plan next to write a letter to Corporate Office while continuing to
> flame the company on comp.sys.hp48 newsgroup. In fact I will probably
> post this response there now.
 ---------------------END EXCERPT-----------------------------------

Could you explain exactly what it is you expect? A one year warranty is
very fair. How could you expect HP to extend it from the time someone
purchases the calculator from someone else? How long should they
purpetuate the warranty? We're not talking about a car or a house here.
We are talking about a graphing calculator. Also, you state that the
unit is more likely than not not fail. Assuming you meant that it
would more likely than not fail, on what do you base this statement?
The few folks who post their negative experiences on this newsgroup?
What about the many happy users who have great experiences with theirs?
I am not trying to be abrasive here, but try to put things into
perspective. Don't lose sleep or waste energy trying to flame HP. You
will have absolutely no effect on them. And many folks will continue to
buy and enjoy their calculators without ever knowing you made a post
anywhere.
It's just a calculator. Relax. 
-Al
--
~/.signature
===
Subject: Re: HP's award Winning Support Team Explains the ONe Year Warranty
Importance: Normal
> Could you explain exactly what it is you expect? A one year warranty is
> very fair.
The one year starts when the ORIGINAL purchaser buys it.
That includes resellers like Calculators.com, Office Max and Staples.
If you buy a new, sealed calculator off the shelf and it's been
on the shelf for three months, you'd get a 9-month warranty.
Tom Lake
===
Subject: Re: HP's award Winning Support Team Explains the ONe Year Warranty
<MPG.228996eb594a7fd8989680@news.giganews.com> 
<48200534$0$5138$4c368faf@roadrunner.com>
posting-account=57eBeAkAAABkNMhsZnDBzGQsstJRmvKD
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> The one year starts when the ORIGINAL purchaser buys it.
> That includes resellers like Calculators.com, Office Max and
Staples.
> If you buy a new, sealed calculator off the shelf and it's been
> on the shelf for three months, you'd get a 9-month warranty.
Well, this is certainly *not* true for the European Union. The
warranty starts when the end-customer finally buys the calculator and
it is 2 years by European law.
So a calculator that has been sitting on the shelf for a year still
has a 2 year warranty when it is finaly bought.
HTH
Andreas
===
Subject: Re: HP's award Winning Support Team Explains the ONe Year Warranty
Previous FUD:
> The one year starts when the ORIGINAL purchaser buys it.
> That includes resellers like Calculators.com, Office Max and Staples.
> If you buy a new, sealed calculator off the shelf and it's been
> on the shelf for three months, you'd get a 9-month warranty.
> Well, this is certainly *not* true for the European Union. The
> warranty starts when the end-customer finally buys the calculator and
> it is 2 years by European law.
> So a calculator that has been sitting on the shelf for a year still
> has a 2 year warranty when it is finally bought.
And authorized resellers in USA
(which you can find directly from HP site);
no doubt Amazon and other comparable web sellers too
(including commerce.hpcalc.org, I would expect).
-[]-
===
Subject: Re: Emu48 for HP50G?
Hi
> I see that Emu48 has been ported to many platforms. I was wondering
> how much ram does it need to run, the idea is to perhaps port a
> special version that could be stored in Flash memory on a 50G calc
> (perhaps replacing the current emulator completely). Then the
> calculator would be able to run not only as a standard 50G, but as a
> true 48 or any of the other models that Emu48 support. This would open
> the door to use old software that is only compatible with the 48
> series and is otherwise dead.
One issue with Emu48 is that it wouldn't run any of the ARM code. The
new opcode implementation is also not complete. I didn't finish the
Getting the Apple emulator which already emulates the Saturn code in
the HP50 would be more suitable. Who knows, HP may be willing to give
this stuff away.
--
They who would give up an essential liberty for temporary security,
deserve neither liberty or security (Benjamin Franklin)
===
Subject: Re: Emu48 for HP50G?
<481fe0ec$0$9842$426a34cc@news.free.fr>
posting-account=y6ooGwkAAACoZd171Nq6oGIjQC5mB4Q9
1.1.4322; .NET CLR 2.0.50727),gzip(gfe),gzip(gfe)
> One issue with Emu48 is that it wouldn't run any of the ARM code. The
> new opcode implementation is also not complete. I didn't finish the
Well, that's just one more thing to add to the porting effort.
 Getting the Apple emulator which already emulates the Saturn code in
> the HP50 would be more suitable. Who knows, HP may be willing to give
> this stuff away.
I really doubt that HP will release anything.

===
Subject: Re: Emu48 for HP50G?
posting-account=57eBeAkAAABkNMhsZnDBzGQsstJRmvKD
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
Hello ,
> I really doubt that HP will release anything.
I doubt that there is anybody listening at HP 
> Anyway, it appears that such a project would have
> more resistance than favor.
Well, to evaluate this poll statistically there were not many
votes 
Personally I would believe that it would be an advantage for the 50G
to run as many software as possible.
Given the possibility to run HP 48 GX applications directly on a 50G
would broaden the available software and allow to use software that
will not be ported to the 50G either because the source is not
available anymore or because there is not interest from the devolper.
So IMHO it would be a very good thing.
Just my 2 cents.
Best wishes,
Andreas
===
Subject: Re: 10 years ago: Unofficial Emu48 Update
posting-account=WTjC_AoAAAB9ZOTiiFEIrYrm7g9h7BK_
1.1.4322),gzip(gfe),gzip(gfe)
hi i'm abig fan of hp 48 this piece of arte make a great help for me
through colege, now my 15 year old daughter is beginig to use it so i
passed on to her, because now i use this great emulator on my pda
phone. thank you so very much for it.
on another mater i recently aquiere a new wm6 pda phone and im having
problems installing it, i tried to install the emulator several times
but for some reason it just won't work.
please need some help.
===
Subject: Re: HP's award Winning Support Team Explains the ONe Year Warranty

 The one year starts when the ORIGINAL purchaser buys it.
> That includes resellers like Calculators.com, Office Max and Staples.
> If you buy a new, sealed calculator off the shelf and it's been
> on the shelf for three months, you'd get a 9-month warranty.
Only if you don't have the original purchase receipt...
--
They who would give up an essential liberty for temporary security,
deserve neither liberty or security (Benjamin Franklin)
===
Subject: Re: HP's award Winning Support Team Explains the ONe Year Warranty
> The one year starts when the ORIGINAL purchaser buys it.
> That includes resellers like Calculators.com, Office Max and
> Staples.
> If you buy a new, sealed calculator off the shelf and it's been
> on the shelf for three months, you'd get a 9-month warranty.
 Well, this is certainly *not* true for the European Union. The
> warranty starts when the end-customer finally buys the calculator and
> it is 2 years by European law.
> So a calculator that has been sitting on the shelf for a year still
> has a 2 year warranty when it is finaly bought.
 HTH
> Andreas
I think I am the only person in the world who has had this much
difficulty. Most of the early 49g+ guys who frequent this group had four
or 5 replacements all under warranty, me ZERO!!, or how do you say ZilCH!
===
Subject: Re: HP's award Winning Support Team Explains the ONe Year Warranty

> Could you explain exactly what it is you expect? A one year warranty is
>> very fair.
 The one year starts when the ORIGINAL purchaser buys it.
> That includes resellers like Calculators.com, Office Max and Staples.
> If you buy a new, sealed calculator off the shelf and it's been
> on the shelf for three months, you'd get a 9-month warranty.
 Tom Lake
But I was going to buy one used from a guy here in the group. You need
to read my post again. The warranty is non-transferable to the second
owner.
===
Subject: Re: HP's award Winning Support Team Explains the ONe Year Warranty
> net@comcast.baranow says...
>> Here is an excerpt from a private email sent today to a would be HP50
>> seller explaining why, after talking to HP Calculator Support, I am
>> unable to take a chance on the 50g which is being offered for resale.
>> Basically, the logic is that all things considered, and having been
>> burned once, it is too risky to purchase this product knowing that more
>> likely than not it will fail and that the terms and conditions for the
>> warranty are far to harsh to roll the dice.
>> ------------------BEGIN EXCERPT---------------------------------
>> The HP support people told me that the warranty expires one year from
>> the time of purchase or one year from the date of manufacture whichever
>> comes first, but that the warranty is not transferable meaning that for
>> you, the original owner, it will be good for one year from the date
>> purchased, but for the second owner, one who is not registered for the
>> warranty, they would use the first 3 digits of the serial number and add 

>> one year to it. So, for anyone who buys it, they will have until the
>> 20th week of year 2008 (the second and third numbers represent the week
>> of production).
>> So, for a calculator that was manufactured, sat on the shelf, was later
>> purchased, and subsequently sold, the new buyer doesn't get a full year
>> warranty. In other words, the meter was running while the cab was
>> stalled in traffic.
>> Exacerbating all of this is the fact that this unit is more likely than
>> not not fail. Nobody in their right mind is going to take a chance with
>> HP with this attitude from their award winning support team.
>> I plan next to write a letter to Corporate Office while continuing to
>> flame the company on comp.sys.hp48 newsgroup. In fact I will probably
>> post this response there now.
>> ---------------------END EXCERPT-----------------------------------
 Could you explain exactly what it is you expect? A one year warranty is
> very fair. How could you expect HP to extend it from the time someone
> purchases the calculator from someone else? How long should they
> purpetuate the warranty? We're not talking about a car or a house here. 

> We are talking about a graphing calculator. Also, you state that the
> unit is more likely than not not fail. Assuming you meant that it
> would more likely than not fail, on what do you base this statement?
> The few folks who post their negative experiences on this newsgroup?
> What about the many happy users who have great experiences with theirs?
 I am not trying to be abrasive here, but try to put things into
> perspective. Don't lose sleep or waste energy trying to flame HP. You
> will have absolutely no effect on them. And many folks will continue to
> buy and enjoy their calculators without ever knowing you made a post
> anywhere.
 It's just a calculator. Relax. 
 -Al
Unfortunately I am an old guy on a fixed income with no more money to 
waste.
===
Subject: Questions about SD cards for 50G
posting-account=2G4L-woAAABoWmw_YYmxjw-bOIfoIAqk
MathPlayer 2.10b; .NET CLR 1.1.4322),gzip(gfe),gzip(gfe)
I was wondering what SD cards (brands and max capacity) work with the
50G?
I tried to use the one that came with my camera but it doesn't work,
I'm now looking for a new one.
Is it true that the 50G only recognize upto 1GB cards?
===
Subject: Re: Questions about SD cards for 50G
posting-account=v6m75AoAAAAgp3vyiLFAfyqmfMcXEPkm
1.1.4322; .NET CLR 2.0.50727; WinuE v6; WinuE 
v6),gzip(gfe),gzip(gfe)
I have the following SD cards:
Verbatim 256MB
Sandisk 1GB
Kingston 2GB
All working fine. I had problems with the 2GB card when trying to use
hpgcc programs, but this has been solved by the HPGCC developers some
time ago.
===
Subject: Re: Questions about SD cards for 50G
> What SD cards (brands and max capacity) work with the 50G?
> I tried to use the one that came with my camera but it doesn't work,
Does the camera card type say SD exactly,
or is there any slight difference?
What capacity and brand?
It's been reported that there are junk (and counterfeit) cards,
but no SD card that I've used have ever not worked.

===
Subject: Re: Questions about SD cards for 50G
posting-account=zYTuBQoAAAC_bXzGjGVT5rxv8bOnpefP
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> I was wondering what SD cards (brands and max capacity) work with the
> 50G?
> I tried to use the one that came with my camera but it doesn't work,
> I'm now looking for a new one.
> Is it true that the 50G only recognize upto 1GB cards?
>
SDHC cards will not work. Any standard SD card should work (2GB cards
too). I use an old 128MB card, never came close to filling it, and I
do a lot of C development and store all my very large binaries on SD.
If your card is a standard SD card and you still have problems then
read: http://h20331.www2.hp.com/Hpsub/downloads/50gUsing_an_SD_Card.pdf.

===
Subject: Carte math+ pour hp48gx
posting-account=0EN1vwoAAAAmYPG5WRDSqPup8hK_d2A4
Gecko/20080404 Firefox/2.0.0.12;MEGAUPLOAD 1.0,gzip(gfe),gzip(gfe)
Bonjour,
Je suis .88 la recherche de la carte math+ pour hp48gx (impossible de
remettre la main sur celle que je poss.8edais). c'est une carte
possedant des programme pour les etudes de fonctions (recherche de
racines, tableau de variation...). Si quelqu'un sait o.9d je pourrais en
trouver une... merci !
===
Subject: My Next Calculator
Having used HP calculators since 1973, I am now
planning to make my next calculator some kind of
EE pc running *BSD with Maxima as the principal
software used for tensor calculations, my main focus.
Calc from the.com will also be installed.
Dave Feustel
===
Subject: UserRPL is really slow on the49G
Now that I have my 48GX UserRPL program running on the 49G, I'm wondering
why I bothered. It runs SIGNIFICANTLY slower on the 49G than on my 48GX. I 

thought this thing was faster than the 48GX, and would be an upgrade, even
with the crappy keyboard. Is this a typical experience? Can I speed it up? 

Is the 50G any better? Or is it time to E-bay another 48? (I just had to
put one down.... <snif>)
Jim
===
Subject: Re: UserRPL is really slow on the49G
CLR 1.1.4322),gzip(gfe),gzip(gfe)
> Now that I have my 48GX UserRPL program running on the 49G, I'm wondering
> why I bothered. EIt runs SIGNIFICANTLY slower on the 49G than 
on my 48GX. EI
I had that problem. Does your UserRPL contain EVAL commands? This
will slow it down a lot, particularly in loops, since the 49 version
of EVAL is much more complicated than on the 48. The cure is tedious,
but works: replace EVAL with stack manipulations. For instance, if
you have stored values for X and Y, and then 'X*Y' EVAL, it will be
faster to have X Y * on the stack instead of EVAL. The conversion can
be a lot of work for large programs.
Bill
===
Subject: Re: UserRPL is really slow on the49G
Hi Bill:
Nope, no EVAL calls in my program, my programming follows your other
Jim
> Now that I have my 48GX UserRPL program running on the 49G, I'm wondering
> why I bothered. It runs SIGNIFICANTLY slower on the 49G than on my 48GX. 
I
I had that problem. Does your UserRPL contain EVAL commands? This
will slow it down a lot, particularly in loops, since the 49 version
of EVAL is much more complicated than on the 48. The cure is tedious,
but works: replace EVAL with stack manipulations. For instance, if
you have stored values for X and Y, and then 'X*Y' EVAL, it will be
faster to have X Y * on the stack instead of EVAL. The conversion can
be a lot of work for large programs.
Bill
===
Subject: Re: UserRPL is really slow on the49G
> Now that I have my 48GX UserRPL program running on the 49G, I'm wondering
> why I bothered. It runs SIGNIFICANTLY slower on the 49G than on my 48GX.
> I thought this thing was faster than the 48GX, and would be an upgrade,
> even with the crappy keyboard.
Is that the crappy _rubber_ keyboard (on the 49G, rather than 49G+)?
Hmm.. was the original 49G any faster than 48G?
What ROM version (VERSION and VER commands) have you got?
(current 49G+/50G versions are 2.09 and/or 2.10,
depending on whether you get HP's or BP's)
Compiling programs from text (or transferring from original HP48 series)
while in Exact mode (rather than Approximate mode,
which better matches the HP48 series)
could also slow down some loops, and some re-written
algebraic functions (ISOL being the worst example I know of)
have also drastically increased their execution time.
> Is the 50G any better?
The 49G+/50G internal emulation of the old processor is much faster,
and things like USB and SD cards are way cooler.
On the other hand, the new emulations of the clock and alarms
are quite faulty, just in case you're thinking to use it also as a PDA,
which the HP48S/G and 49G were perfectly capable of doing well.

===
Subject: Re: UserRPL is really slow on the49G
posting-account=sOAX1QkAAAC-FcySTSbz29Uk8huUtFRz
CLR 1.1.4322; .NET CLR 2.0.50727),gzip(gfe),gzip(gfe)
> Compiling programs from text (or transferring from original HP48 series)
> while in Exact mode (rather than Approximate mode,
> which better matches the HP48 series)
> could also slow down some loops, and some re-written
> algebraic functions (ISOL being the worst example I know of)
> have also drastically increased their execution time.
Put decimal points after all of the integers in your program (for
example, if you're multiplying something on the stack by 5, put 5. *
instead of 5 * ). The decimal points force approximate mode instead of
using long ZINTs and this speeds up calculation time significantly.
S.C.
===
Subject: Re: UserRPL is really slow on the49G
Hi John:
 Is that the crappy _rubber_ keyboard (on the 49G, rather than 49G+)?
>
Yes, sir. 49G, not G+
> Hmm.. was the original 49G any faster than 48G?
 What ROM version (VERSION and VER commands) have you got?
> (current 49G+/50G versions are 2.09 and/or 2.10,
> depending on whether you get HP's or BP's)
ROM is 1.19-6 Sounds old....
 Compiling programs from text (or transferring from original HP48 series)
> while in Exact mode (rather than Approximate mode,
> which better matches the HP48 series)
> could also slow down some loops, and some re-written
> algebraic functions (ISOL being the worst example I know of)
> have also drastically increased their execution time.
I'll try that. How do I change Exact to Approximate?
> Is the 50G any better?
 The 49G+/50G internal emulation of the old processor is much faster,
> and things like USB and SD cards are way cooler.
 On the other hand, the new emulations of the clock and alarms
> are quite faulty, just in case you're thinking to use it also as a PDA,
> which the HP48S/G and 49G were perfectly capable of doing well.
I'm mainly interested in faster and better screen. My main use for the
calculator is doing calculations in time-speed-distance road rallies. My
program calculates when I should be at a certain point (based on miles
traversed and average speed given by the rallymaster). It then calculates
the difference between the calculated perfect time and the actual clock 

time. I'm measuring differences in 1/100 of a minute differences, and the
beauty of the 48 for this application is the CLK program that HP published, 

that synchronizes the system clock to a time standard (WWV in my case), and 

it REALLY works, I can get the system clock to drift less than a second a
month from WWV. So, now that i got my rally math program working on the 49, 

i'm going to work on getting CLK to run (some kind soul sent me the RPL
code, which also is giving me a syntax error when I compile it, so have 
to
go through it with the fine tooth comb....) If I get the whole mess
running on the 49, i'll upgrade to the 50 in the hope of better screen
display and faster execution. But your mention of faulty emulations of 
the
clock make me a bit nervous. Maybe it's time to find a new platform....
Jim
 
===
Subject: Re: UserRPL is really slow on the49G
> Yes, sir. 49G, not G+
> ROM is 1.19-6 Sounds old....
That was the final (and unofficial, non-HP) ROM version
for 49G; however, it's possible (with the aid of Emu48,
on a PC with a serial port) to ROMUPLOAD 2.09 or 2.10
from a PC to an HP49G, FWIW.
> How do I change Exact to Approximate?
In MODE CAS menu, or Right-shift(hold down) + ENTER
or -105 SF (the top status line contains = in Exact mode,
~ in Approximate mode)
> My main use for the calculator is doing calculations
> in time-speed-distance road rallies.
Want to buy my Curta?
(recommended for road rallies by The Last Whole Earth Catalog :)
http://en.wikipedia.org/wiki/Curta
http://en.wikipedia.org/wiki/Curta#Use in car rallies
http://www.amazon.com/dp/0394704592
http://en.wikipedia.org/wiki/Whole Earth Catalog
> My program calculates when I should be at a certain point (based on miles
> traversed and average speed given by the rallymaster). It then 
calculates
> the difference between the calculated perfect time and the actual 
clock
> time. I'm measuring differences in 1/100 of a minute differences, and 
the
> beauty of the 48 for this application is the CLK program that HP 
published,
> that synchronizes the system clock to a time standard (WWV in my case), 
and
> it REALLY works, I can get the system clock to drift less than a second a
> month from WWV. So, now that i got my rally math program working on the 
49,
> i'm going to work on getting CLK to run (some kind soul sent me the RPL
> code, which also is giving me a syntax error when I compile it, so 
have to
> go through it with the fine tooth comb....) If I get the whole mess
> running on the 49, i'll upgrade to the 50 in the hope of better screen
> display and faster execution. But your mention of faulty emulations of 
the
> clock make me a bit nervous. Maybe it's time to find a new 
platform....
49G should still be okay, but with a 49G+ or 50G, a space rally to Mars
might end up in a roast on Venus :)

===
Subject: Re: UserRPL is really slow on the49G
Hi John:
appreciated. On to the good stuff 
>Want to buy my Curta?
(recommended for road rallies by The Last Whole Earth Catalog :)
http://en.wikipedia.org/wiki/Curta
http://en.wikipedia.org/wiki/Curta#Use_in_car_rallies
http://www.amazon.com/dp/0394704592
http://en.wikipedia.org/wiki/Whole_Earth_Catalog
Whoa, you an old-time rally guy? I've cranked a Curta on rallies before,
back in the 60's it was the tip of the whip technology (or so I've been
told). In fact, the concept for my HP48 rally math program was an
manipulations needed, and I mapped a button to mimic turning the crank
(pumping in a known mileage increment and recalculating perfect time).
Where the 48 left the Curta in the dust was with the syncronizable clock (so 

i can do near real-time splits) and with the ability to remember the
history of the calculations, so you can go backwards if you make a mistake.
>49G should still be okay, but with a 49G+ or 50G, a space rally to Mars
might end up in a roast on Venus :)
Is the clock so bad that the CLK program won't be able to make it behave?
(assuming, of course, I can get it to work at all....)
Jim

===
Subject: Re: UserRPL program works on 48GX, doesn't on 49G
If a string is on the stack and doesn't compile via STR-> or OBJ->
the following simple program can be used to check its syntax,
in a manner similar to what JYA recommends:
<< DUP ->STR { V } +  SWAP INPUT >> 'SYN' STO
This reads a copy of the string into an editor,
at which point you should press ENTER to check it;
the first syntax error, if any, will be highlighted
(a corrected string simply returns to the stack instead).
If you can't correct the string, abort with CANCEL
and drop anything extra that's been returned.
This works on any HP48/49/50,
and also properly handles any embedded quoted strings.
You may use STR-> or OBJ-> to compile a corrected string,
exactly as it would be if typed into the command line
(note that any commands which are not inside of programs or lists
may be executed by doing this); if you want to do this always,
you could insert either of those commands
into the above program, after INPUT.

===
Subject: Re: UserRPL program works on 48GX, doesn't on 49G
fine on 48, horks up on the 49. Program is running.
Jim
> Hi
> On 2008-05-19 10:59:17 +1000, Jim Fekete <fetejim@ameritech.net> 
said:
> I have a UserRPL program that I've used on a 48GX for a long time. If 
I
>> download the text file onto a 48Gx it loads and operates correctly, but
>> if I
>> take the same file and download it to my 49G, i get a syntax error and
>> the
>> program is sitting on the stack. I've gone through it line by line, and 

>> the
 ->NUM exists on the 49.
 Somewhere in your program, you are using a variable name with the same
> name as an HP49 commands. The HP49 introduced a lot of new commands.
 To find out which one it is, edit the string containing the program,
> remove the quotes ( ) at the beginning and end and validate with ENTER.
 It will show you which variable is causing the error, and put it in video 

> inverse
> --
> They who would give up an essential liberty for temporary security,
> deserve neither liberty or security (Benjamin Franklin)
>
===
Subject: Re: UserRPL program works on 48GX, doesn't on 49G
On 2008-05-20 12:37:04 +1000, Jim Fekete <fetejim@ameritech.net> said:
> fine on 48, horks up on the 49. Program is running.
 Jim
Make sur before you compile your program the calculator in in
approximate mode, so whatever is compiled is identical to what the HP48
does and speed is the same
===
Subject: Re: UserRPL program works on 48GX, doesn't on 49G
posting-account=sOAX1QkAAAC-FcySTSbz29Uk8huUtFRz
CLR 1.1.4322; .NET CLR 2.0.50727),gzip(gfe),gzip(gfe)
> Hi:
 I have Ea UserRPL program that I've used on a 48GX for a long 
time. EIf I
> download the text file onto a 48Gx it loads and operates correctly, but if 
I
> take the same file and download it to my 49G, i get a syntax error and 
the
> program is sitting on the stack. EI've gone through it line 
by line, and the
> character translation on the 49 version matches the 48 version. 
EThe only
> weirdness I have detected is that the ->NUM command in the 48 doesn't
> appear in the 49 AUG, and appears to have been dropped, or at least not
> documented. EI tried replacing it with EVAL, just to see if 
the program
> would load, but no joy. EAre there any well known issues with 
48 UserRPL
 Jim
Try this thread where the commands in the 49/50 that are not in the 48
are listed:
S.C.
===
Subject: Re: How to install a library frim SD card
> i can copy the library to 1 or 2 rom...
> but then?
> how i can i use libraries?
> which menu?
The LIB menu (right-shift 2 on 49/50 series),
then press the name of the desired library,
whose command menu will then appear;
if your library's name doesn't appear,
try ON+C (calculator warmstart,
like Ctrl+Alt+Del for computer restart)
If the library comes with instructions for use,
at this point you'll need those instructions,
because any library can be designed to work in any manner
that a programmer wants to build.
If the original object was not actually a library, however
(perhaps a directory or a program instead)
then it belongs elsewhere.

===
Subject: HP 48 SX power Problem - Desperate Help
I bought my HP48 in 1991 and I drop it in my cupboard in 2001 (I removed the 

batteries and it was perfectly working).
Now, I get it from my closet, I replaced the batteries but I'n not able to
power on / power off the calc!!!
The only workaround I found to start the calculator, it's to apply the reset 

procedure ON+A+B, while to power off I have to remove the batteries...
What has it happened?
Stefano
===
Subject: Re: HP 48 SX power Problem - Desperate Help
 I bought my HP48 in 1991 and I drop it in my cupboard in 2001 (I removed 
the
> batteries and it was perfectly working).
 Now, I get it from my closet, I replaced the batteries but I'n not able 
to
> power on / power off the calc!!!
 The only workaround I found to start the calculator, it's to apply the 
reset
> procedure ON+A+B, while to power off I have to remove the batteries...
 What has it happened?
 Stefano
>
A similar thing occurred with my HP48S. The contact for the ON button
is dirty or oxidized. You might be able to clear the problem with
repeated operation of the ON button.
I had a problem exiting Alarm CAT display with the ATTN button.
That is when I discovered that holding the calculator and torquing it
slightly with ON corner and upper right corner would enable the ON
button to make contact. Hold the calculator with both hands, placing
left thumb on ON button with right hand on upper right corner and
torque the unit slightly while pressing the ON button.
BTW, if you leave the calculator ON, it will turn off after several
minutes of non use.
===
Subject: InForm Builder help
posting-account=4ikQLgoAAAB2vmCjLE2ERFyw6LPpHSV4
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
Hi to All,
I have used Steen's InForm Builder for a long time now on a 49g+. I
recently borrowed a 50G and loaded it up with the extable library and
InForm Builder. However when I create a data field and check the 'Real
Number' field property and hit |OK| I error out with Invalid Init-
type!. Now I'm sure there is a flag somewhere I forgot to set or
something. But I just can't track this one down. Wondering if anyone
else has had a similar experience and if you know the solution.
TIA.
Scott H.
===
Subject: Re: HP50g -- first foray into assembly lang and sysrpl
> Having become reasonably familiar with my new hp50g, and at least
> introduced myself to most of the user RPL commands, I thought I should
> finally bite the bullet and have a go at some System RPL.
 Oddly enough, I was was already fairly familar with Saturn assembly
> language; I have owned an HP-71B since about '85 (IIRC) and enjoyed
> adding a few extensions to its Forth language. I thought sysrpl
> should be fairly easy in comparison. Not so sure now.
Do you know this:
http://www.hpcalc.org/details.php?id=5142
 Well, enough of my background. On to the background of my questions.
 I thought an interesting first challenge would be the old Mastermind
> style game, Cows and Bulls. The vanilla flavour of this game
> wouldn't require great speed, but I have a variant where the computer
> keeps track of possible candidates. After each of your guesses,
> besides telling you so many cows and bulls, it goes through the set
> of remaining candidates, eliminating those which wouldn't match this
> result and tells you how many candidates remain. For example:
 Guesses C B
> 1274 0 1
> 3579 1 1
> 6978 0 1 Possible: 3
 (FYI, the remaining candidates are: 1538 3258 6534)
 There is clearly quite a bit of processing early on when all 3024
> candidates are still valid and a combination of sysrpl and assembly
> language seemed to be called for (of course it would be cheating to
> use C or ARM!).
 At long last, down to my questions:
 1. How should I structure shared functions in assembly language?
>
You can't! I think that every objects (from a sys-rpl program) are
pushed on the TEMPOB (temporary objects) area. The solution is:
only one ASM program which pop on the stack a number which says
what the ASM program have to do.
> 2. Where is there a clear summary of the syntax differences in the
> flavours (HP, MASD, others?) of sysrpl and of assembly language?
No I don't think so.
> I was used to HP's Saturn instruction set and have been using
> Debug4x as a development environment for this exercise, but I
> would like to get away from the PC (Windows in particular)
> and MASD is right there on the calculator.
 Besides this, I have been confused from time to time when
> tutorials downloaded from hpcalc.org were referring to a
> syntax that wasn't available in the other environment. IIRC,
> it has one. Also took me a while to work out the need for
> FEXTERNAL for entry points introduced by ^.
Do you know this:
http://staff.science.uva.nl/~dominik/hpcalc/
If you read french http://www.courbis.fr/spip.php?rubrique8
regardez voyage au centre de la hp 48g.
If you can't read french:
http://www.courbis.fr/spip.php?rubrique9
you will find a translation for hp48s(x).
Of course hp48s,49g,50g are not the same but
mostly they are still very intersting books on hp48
models.
===
Subject: Re: HP50g -- first foray into assembly lang and sysrpl
Originator: roberts@athlonx2.gville (Robert Swan)
>here's a hint for code safety:
>Don't use inter-code jumps in relocateable code.
>Don't do it, never!
>It leads to unsafe code, and is bad programming style.
Your advice is welcome, and makes perfect sense. I suspected as much
(I did say it seemed dodgy), but was unsure. That's why I posted.
>Your call to your subroutine PixPrep is very likely to fail
>if the code doesn't reside in main memory (port 0),
>because it will be copied to there in pieces,
>each RPL sub when it's used,
>so there is a big chance that your target location 'PixPrep'
>called from a completely different code slice.
Is there some document where this process is described? There is a
lot of stuff available, but what I've seen has mostly been either
introductory or assumed great knowledge. As I said, something like
the HP-71B's IDS manuals (particularly vol I) that explains the
general workings of the system software would help me a lot.
>However, if you urgently need to have various routines
>call a common sub, then you should consider combining
>all code slices into one single code object,
>and introduce a suitable stack parameter so the code
>can decide where to jump fro mthe start.
Yeah -- what I referred to as passing a function selector
on the stack -- doesn't appeal too much.
>The other way to go would be s.t. like this:
But this way is exactly what I was after. Multiple entry points
to a single block of relocatable code. Worked for me straight away.
>RPL
>ASSEMBLE
> CON(5) =DOCODE
> REL(5) ComCodeEnd
* Actual code

>RPL
>ASSEMBLE
> CON(5) =DOCODE
> REL(5) ComCodeEnd
* Code
RPL
>ASSEMBLE
> CON(5) =DOCODE
> REL(5) ComCodeEnd
* Code
PixPrep
ComCodeEnd
>RPL

>This will also work when the code resides in a port other than 0 .
>Please note that inter-code jumps MUST ALWAYS be forward jumps.
If Sub1 Sub2 Sub3 all call PixPrep,
>PixPrep has to be in the last pseudo-code object.
>If PixPrep would reside in the code slice of Sub1,
>Sub3 would not have a safe link to PixPrep,
>because if it comes to copying to TEMPOB,
>only the Sub3 would have been copied,
>not Sub1 or Sub2.
Hmmm. Again, I'm afraid we're bumping into my ignorance. So Sub1,
Sub2 and Sub3 (and PixPrep) are not all (as a result of the use of
ASSEMBLE and RPL directives) in one atomic block? Let's see if I have
this worked out right then. Structured as per your skeleton code
above, the flow for a call to Sub3 followed by Sub1 might be as
follows:
1. Sub3 invoked, and just it (including PixPrep) loaded into memory
and executed.
2. Sub1 invoked (Sub3 & PixPrep still in memory, I assume), no room
in front of Sub3, so Sub1, Sub2, Sub3 and PixPrep are all loaded
into a new block of memory and Sub1 executed.
Is this something like what happens?
Once again, I suppose, a pointer to system-level documentation
would seem to be the best cure.
>HTH
Have fun,
Robert.
--
Robert Swan | No, not the antarctic adventurer.
swan.r.l@gmail.com | No, not the Canberra porn monger.
| Yes, that's right, the boring one.
===
Subject: Re: My Next Calculator
Importance: Normal
>> What about battery duration?
>> MS
 IMHO, even the 50g's battery life is subpar. I need to replace the
> batteries every month or so even with light to moderate usage.
 S.C.
When I sit at my desk I plug my calc to my notebook and I save battery life. 

But I'm able to use the calc for many days without it. And I carry with me 4 

more batteries.
Anyway I'm worried about a subnotebook PC battery duration. Anyone in this
newsgroup tested it?
MS
===
Subject: Re: HP 48 SX power Problem - Desperate Help
<1211442676.979320@slbhw0>
posting-account=aQItIQoAAADRqM7zx-gk13yzFHjtne3-
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> Unfortunately seems not to work.....
> Moreover why is the reset procedure working? The reset procedure involves
> the ON button. Do we have two different circuits for the ON button??

> Stefano


>> I bought my HP48 in 1991 and I drop it in my cupboard in 2001 (I 
removed
>> the
>> batteries and it was perfectly working).
> Now, I get it from my closet, I replaced the batteries but I'n not 
able
>> to
>> power on / power off the calc!!!
> The only workaround I found to start the calculator, it's to apply the
>> reset
>> procedure ON+A+B, while to power off I have to remove the batteries...
> What has it happened?
> Stefano
 A similar thing occurred with my HP48S. The contact for the ON button
> is dirty or oxidized. You might be able to clear the problem with
> repeated operation of the ON button.
 I had a problem exiting Alarm CAT display with the ATTN button.
> That is when I discovered that holding the calculator and torquing it
> slightly with ON corner and upper right corner would enable the ON
> button to make contact. Hold the calculator with both hands, placing
> left thumb on ON button with right hand on upper right corner and
> torque the unit slightly while pressing the ON button.
 BTW, if you leave the calculator ON, it will turn off after several
> minutes of non use.
I had a similar problem with a 48SX some years ago.
I sprayed the keypad with Radio Shack TV-Tuner/Control Cleaner &
Lubricant
Model: 64-4315 | Catalog #: 64-4315
I sprayed the keyboard in a well ventilated area and let the liquid
drip out of the calculator.
I set the calculator inside so it would completely dry (24 hrs) the
calc worked fine after that.
I do no know if this will fix your problem but it maybe something to
consider.
David
===
Subject: Re: HP 48 SX power Problem - Desperate Help
if you press the cover above B eg, F2
and it then works
the foam or buffering layer inside has lost its flexibility
>> Unfortunately seems not to work.....
>> Moreover why is the reset procedure working? The reset procedure 
involves
>> the ON button. Do we have two different circuits for the ON button??
>> Stefano
> I bought my HP48 in 1991 and I drop it in my cupboard in 2001 (I
> removed
> the
> batteries and it was perfectly working).
> Now, I get it from my closet, I replaced the batteries but I'n not
> able
> to
> power on / power off the calc!!!
> The only workaround I found to start the calculator, it's to apply 
the
> reset
> procedure ON+A+B, while to power off I have to remove the 
batteries...
> What has it happened?
> Stefano
X
===
Subject: Re: UserRPL is really slow on the49G
> Switching to approximate mode had no discernable effect on speed.
As JYA has just said, switching modes after program compilation
is ineffective. But the good news is, that you don't need
to tediously enter decimal points after all whole numbers
to rectify this -- just EDIT the program once,
without making any changes at all,
after having first switched to Approximate mode.
The mere act of on-board editing does two subtle
but very significant things -- first, the binary
(already compiled) program is decompiled
(displayed as source text), then when you press ENTER,
that text is automatically re-compiled.
If the calculator is already in Approximate mode,
that last step automatically makes reals out of all integers,
just as it would have if the original program transfer
had been made with the calculator in Approximate mode to begin with.
As to EVAL taking longer on HP49G series, this applies only
to the evaluation of algebraic expressions (which now internally
invokes CASCOMPEVAL rather than COMPEVAL); in many cases,
particularly when using none of the newer CAS functions in expressions,
this can actually be rectified by using ->LST EVAL or ->PRG EVAL
in place of EVAL (->LST and ->PRG commands are found under
APPS > Development lib, which exists if you keep
flag -86 set whenever restarting the calculator via ON+C),
which pays off when use of the original EVAL more than offsets
the short execution time of the inserted extra command.
Despite all this, however, the time it takes to run a program
inevitably depends on what's going on in that particular program,
because the major reorganization of functions within ROM,
between HP48 series and HP49 series, may have promoted some things
from parts of ROM that needed special mapping to parts that
are always mapped, and may have demoted other parts at the same time;
various older symbolic capabilities were also completely rewritten,
possibly also taking more time (occasionally to an extreme degree),
because of the much larger and more complex CAS.
All the standard numeric functions should take exactly the same time
on 49G (original Saturn CPU) as on 48G; perhaps you can guess
what other functions (and structures, list processing, etc.)
in your programs to consider mentioning,
as possibly responsible for any significant difference between the two.
As Bill Markwick mentioned, use of pure RPL in place of things algebraic
might also turn out to be of benefit.

===
Subject: Restore to EMU48
posting-account=T97vPQoAAAC9kmDX8CtmaQ4OwIkCSTTi
1.0.3705; .NET CLR 1.1.4322; Media Center PC 
4.0),gzip(gfe),gzip(gfe)
Hello once again,
I have backed up my 50g to my PC and now want to use the backup on
EMU48 (packaged with HPUseredit5). I can't do it though. I've tried
dragging and dropping the file to the EMU screen, Opening and
Loading: all to no avail. The Backup menu item is always greyed
out.
Help please.
Incidentally, I have been looking for and English version of
HPUseredit5. Does it exist? If so where?
N
===
Subject: Re: Restore to EMU48
 Incidentally, I have been looking for and English version of
> HPUseredit5. Does it exist? If so where?
Just go to the Options (Opciones) > Language (Idiomas) and select English.
What a great program it is!
===
Subject: Re: Restore to EMU48
<g1a4n9$l23$1@aioe.org>
posting-account=T97vPQoAAAC9kmDX8CtmaQ4OwIkCSTTi
1.0.3705; .NET CLR 1.1.4322; Media Center PC 
4.0),gzip(gfe),gzip(gfe)
On May 24, 11:26Epm, Chris Degnen
 Incidentally, I have been looking for and English version of
> HPUseredit5. Does it exist? If so where?
 Just go to the Options (Opciones) > Language (Idiomas) and select 
English.
 What a great program it is!
Hello Chris,
language options are limited to Italian, Portugese, and two versions
of Spanish: none of which am I fluent in.
N
===
Subject: Re: Restore to EMU48
 language options are limited to Italian, Portugese, and two versions
> of Spanish: none of which am I fluent in.
Same version as mine (which has English).
Did you get it here? : http://www.hpcalc.org/details.php?id=6600
===
Subject: Re: Restore to EMU48
<g1a5os$oq1$1@aioe.org>
posting-account=T97vPQoAAAC9kmDX8CtmaQ4OwIkCSTTi
1.0.3705; .NET CLR 1.1.4322; Media Center PC 
4.0),gzip(gfe),gzip(gfe)
On May 24, 11:44Epm, Chris Degnen
 language options are limited to Italian, Portugese, and two versions
> of Spanish: none of which am I fluent in.
 Same version as mine (which has English).
 Did you get it here? :http://www.hpcalc.org/details.php?id=6600
Can't remember now so I'll try installing from there.
N
===
Subject: Re: Restore to EMU48
posting-account=T97vPQoAAAC9kmDX8CtmaQ4OwIkCSTTi
1.0.3705; .NET CLR 1.1.4322; Media Center PC 
4.0),gzip(gfe),gzip(gfe)
> On May 24, 11:44Epm, Chris Degnen

> language options are limited to Italian, Portugese, and two versions
> of Spanish: none of which am I fluent in.
 Same version as mine (which has English).
 Did you get it here? :http://www.hpcalc.org/details.php?id=6600
 Can't remember now so I'll try installing from there.
 N
about restoring backups from the PC to EMU?
N
===
Subject: Printer for HP 50G
posting-account=IicVKwoAAACbH6mCTTDMOGh_Z0b4FE2T
AppleWebKit/525.18 (KHTML, like Gecko) Version/3.1.1 
Safari/525.18,gzip(gfe),gzip(gfe)
Hello together,
Im a new owner of a HP50G. A calculator dealer in Germany 
said in a e-
mail to me, that the HP 82240B printer doesnt work with the 
HP50G. Is
this true? And when, is there another printer I can use, perhaps an
used one?
Jens
===
Subject: Re: Printer for HP 50G
posting-account=itRlSAoAAADIJvOzSGM8mjsK7gvlk1wg
Gecko/20071127 Firefox/2.0.0.11,gzip(gfe),gzip(gfe)
I have got an HP50g and the HP82240B printer and it
works perfectly. No problem.
Peter
===
===
Subject: Battery Charging
Hi All,
I am wondering if it is possible to charge NIMH batteries in a 50g through 
the usb
socket ?
I am making a small caddy to hold the calculator on the desk and it will 
power the
calculator, just it would be ideal if there was a small resistor isolating 
the
battery from the usb port in the calculator.
For NIMH the charge voltage would not go over 5.6 volts anyway.
I had a 48 and it was stolen and I am still waiting for the 50g so I can't 
do any
tests at present.
best wishes Peter.

===
Subject: broken hp.com links
posting-account=aJhvMAoAAABqICGF80eSUn6d3D9SANQE
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
Is it just me, or are the hp.com United States links to graphing
calculators broken?
http://www.hp.com/calculators/
United States
Graphing Calculators
brings up The page you are looking for is no longer available or has
been moved.
The other countries seem to be okay.
-wes
===
Subject: Re: broken hp.com links
posting-account=sOAX1QkAAAC-FcySTSbz29Uk8huUtFRz
CLR 1.1.4322; .NET CLR 2.0.50727),gzip(gfe),gzip(gfe)
> Is it just me, or are the hp.com United States links to graphing
> calculators broken?
 E Ehttp://www.hp.com/calculators/
> E EUnited States
> E EGraphing Calculators
 brings up The page you are looking for is no longer available or has
> been moved.
> The other countries seem to be okay.
 -wes
HP's website is overcomplicated, bloated (and slow to load), and
unreliable. What's worse is that it doesn't even provide good
information.
S.C.
===
Subject: Re: broken hp.com links
posting-account=QSfVcQoAAADkgn5Et4hWcvOa13b-lJy1
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
I dont have access neither.
Daniel
===
Subject: Re: New hp calculator?
>Apparently hp is about to release a new financial calculator.
Hm..can I guess? It was necessary to develope a new one because there
is need for faster subtract of huge numbers for financial user? :-D
Olaf
===
Subject: Re: New hp calculator?
> Apparently hp is about to release a new financial calculator.
HP-20 B (June 2008)
http://www.dynatech-taschenrechner.de/produkte/produkt.php?prod=4992
Looks like a 10B merged with a 20S.
Dynatech also advertised the 50G in advance, IIRC,
but each is a Summer rerun, not a new show.

===
Subject: Re: New hp calculator?
<op.uaujrn0wnn735j@miu.edu>
posting-account=sOAX1QkAAAC-FcySTSbz29Uk8huUtFRz
CLR 1.1.4322; .NET CLR 2.0.50727),gzip(gfe),gzip(gfe)
> HP-20 B (June 
2008)http://www.dynatech-taschenrechner.de/produkte/produkt.php?prod=4992
 Looks like a 10B merged with a 20S.
 
Looks like it picked up the ALG mode of entry from the 20S also :(
Maybe it inherited the good keys of the Pioneer series? hopefully?
S.C.
===
Subject: Re: New hp calculator?
<sc_usenet@hotmail.com> schrieb im Newsbeitrag
> HP-20 B (June
> 2008)http://www.dynatech-taschenrechner.de/produkte/produkt.php?prod=4992
 Looks like a 10B merged with a 20S.
 
>Looks like it picked up the ALG mode of entry from the 20S also :(
>Maybe it inherited the good keys of the Pioneer series? hopefully?
>
The German description also says the keys don't have the usual resistance
when pressed, but they evaluate the keys as 'better than average' .
More can be said when we actually have a 20b unit to check
Anyone can tell more details on this?
Raymond
===
Subject: Re: New hp calculator?
<fvvq9h$qhu$02$1@news.t-online.com>
posting-account=nKQXlgkAAABFcwCQitpPqM7oKmh-_khh
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
 HP-20 B (June
> 
2008)http://www.dynatech-taschenrechner.de/produkte/produkt.php?prod=4992
 Looks like a 10B merged with a 20S.
 
>Looks like it picked up the ALG mode of entry from the 20S also :(
>Maybe it inherited the good keys of the Pioneer series? hopefully?
 The German description also says the keys don't have the usual resistance
> when pressed, but they evaluate the keys as 'better than average' .
 More can be said when we actually have a 20b unit to check
 Anyone can tell more details on this?
 Raymond
And it has trig functions!!!!
===
Subject: Re: Xtide or similar for HP50G
<op.uatezhxenn735j@miu.edu>
posting-account=QSfVcQoAAADkgn5Et4hWcvOa13b-lJy1
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
Ill try it this weekend, and let you know.
We do have both oceans here (Pacific and Caribbean=Atlantic), so I
really can test it.
Daniel.
===
Subject: hpgcc and rpl-stack (again)
posting-account=6Xk2SwoAAACIm5pYE9m8TqQG94WHZApf
Gecko/20070208 Mandriva/2.0.0.6-12mdv2008.0 (2008.0) 
Firefox/2.0.0.6,gzip(gfe),gzip(gfe)
I want to make somthing like:
SAT_STACK_EELMENT rpl_obj;
sat_get_stack_element(1,rpl_obj);
int *var
var=the adress in c of datas of the rpl_obj (so after the header).
I can do copy (whith sat_peek) of the rpl_obj but si it possible
to get directly into the rpl_obj. Now I make a copy of the rpl_obj
into C world
so I use a lot of memory!

===
Subject: Re: hpgcc and rpl-stack (again)

 I want to make somthing like:
 SAT STACK EELMENT rpl obj;
> sat get stack element(1,rpl obj);
> int *var
> var=the adress in c of datas of the rpl obj (so after the header).
 I can do copy (whith sat peek) of the rpl obj but si it possible
> to get directly into the rpl obj. Now I make a copy of the rpl obj
> into C world
> so I use a lot of memory!
 
If you are using sat peek(), then you already *are* accessing the
object directly. All you need to do is to treat the address you pass
to sat peek as a pointer, and replace all the *var with calls to
sat peek() or sat poke().
The only thing you need to consider is that the saturn adress is a
nibble aligned pointer, so if you are reading bytes, you need to move
the pointer by 2, instead of just one if you use a (char *).
I think you'd be pleased to know that the new hpgcc3 API (besides
being more powerful) works with saturn objects in the C heap or in the
calculator (TEMPOB) without distinction, so you can create,
manipulate, convert, etc. any temporary objects in the C heap, without
messing with the saturn world, and also you can manipulate in the same
way objects directly in the Saturn realm the way you want to do right
now.
But for now you'll have to peek and poke yourself.

===
Subject: Re: hpgcc and rpl-stack (again)

 I want to make somthing like:
 SAT_STACK_EELMENT rpl_obj;
> sat_get_stack_element(1,rpl_obj);
> int *var
> var=the adress in c of datas of the rpl_obj (so after the header).
 I can do copy (whith sat_peek) of the rpl_obj but si it possible
> to get directly into the rpl_obj. Now I make a copy of the rpl_obj
> into C world
> so I use a lot of memory!
 
 If you are using sat_peek(), then you already *are* accessing the
> object directly. All you need to do is to treat the address you pass
> to sat_peek as a pointer, and replace all the *var with calls to
> sat_peek() or sat_poke().
> The only thing you need to consider is that the saturn adress is a
> nibble aligned pointer, so if you are reading bytes, you need to move
> the pointer by 2, instead of just one if you use a (char *).
>
I made a sat_pop_grob function which takes a grob (open fire 16 gray
grob) on stack
and put it in a ggl_surface. Mainly It just copying nibbles: the
structure are exactly
the same (at least if length%8=0). So I wanted to put the grob
address in the
ggl_surface addr field but according to your answer it seems that it
is
not possible.
> I think you'd be pleased to know that the new hpgcc3 API (besides
> being more powerful) works with saturn objects in the C heap or in the
> calculator (TEMPOB) without distinction, so you can create,
> manipulate, convert, etc. any temporary objects in the C heap, without
> messing with the saturn world, and also you can manipulate in the same
> way objects directly in the Saturn realm the way you want to do right
> d now.
> But for now you'll have to peek and poke yourself.
 
So it will be possible wtih hggcc 3? I hope we will have hpgcc 3 soon!
I will have to rewrite the functions which deal with RPL objects on
stack?

.
===
Subject: Re: Questions about SD cards for 50G
>> I was wondering what SD cards (brands and max capacity) work with the
>> 50G?
>> I tried to use the one that came with my camera but it doesn't work,
>> I'm now looking for a new one.
>> Is it true that the 50G only recognize upto 1GB cards?
Some cameras (also cell phones and mp3 players) format SD cards in
>something other than normal FAT16 - those cards that are already
>formatted by those devices won't work till you re-format them. This
>can be done either in the 50 or on a PC, but i recommend doing it on a
>PC. Just make sure you specify FAT16 format. You can use anything up
>to 2 gb. I use 2gb Kensington MicroSD cards in an SD adapter in my
>50 and they work fine! I haven't tried it, but i think the same holds
>true for MiniSD cards in an adapter.
There's no point in using a 2 GB card on the 50, since you cannot store
more on the 2 GB card than on a 1 GB card or on a 512 MB card. How come?
It has to do with formatting the 2 GB as FAT16, which limits the number
of files to a maximum of some 65000. This includes all files in all
subdirectories on the card. It alwo has to do with cluster size,
no matter how small the actual file is, it will still occupy at least
If you switch to a FAT16 formatted 1 GB card, the cluster size will become
it occupied on the 2 GB card. Since almost all files stored by the
will hold no less actual data from the 50 than the 2 GB card: both cards
will store a maximum of some 65000 files.
some 65000 files on the card. A 256 MB card will have a cluster size of
large. So the 256 MB card may not be able to store quite as much data
from the 50 as the 2 GB card.
The bottom line is: it's a waste of card memory space to use a 2 GB
or 1 GB card on the 50. To use the memory on the 2 or 1 GB card
more efficiently, it should be formatted as FAT32, not FAT16. But
formatting them as FAT32 would make them unusable on the 50. So why
not just get an old 128MB or 256 MB card to use with your 50? In
practice, you'll be able to store (almost) as much data from the 50
on it as you could store on a 2 GB card.....
--
----------------------------------------------------------------
Paul Schlyter, Grev Turegatan 40, SE-114 38 Stockholm, SWEDEN
e-mail: pausch at stjarnhimlen dot se
WWW: http://stjarnhimlen.se/
===
Subject: Re: Questions about SD cards for 50G
> There's no point in using a 2 GB card on the 50, since you cannot store
> more on the 2 GB card than on a 1 GB card or on a 512 MB card.
> [if you can still get cards that small :]
You mean more individual files, and then only with FAT16,
but you can store your (large) manuals,
you can store Emu48 (to slip the card into any handy computer),
you can store a zip of everything you've ever downloaded
from www.hpcalc.com, and of course you can store pictures, music and video
(for a card that you share with your phone or PDA :)
Is there yet a book reader program for the HP49G+/50?
Then you can also store the books :)

===
Subject: Re: Questions about SD cards for 50G
<g00qb6$229d$1@merope.saaf.se>
posting-account=y6ooGwkAAACoZd171Nq6oGIjQC5mB4Q9
1.1.4322; .NET CLR 2.0.50727),gzip(gfe),gzip(gfe)
<... The bottom line is: it's a waste of card memory space to use a 2 GB
> or 1 GB card on the 50. ETo use the memory on the 2 or 1 GB 
card
> more efficiently, it should be formatted as FAT32, not FAT16. 
EBut
> formatting them as FAT32 would make them unusable on the 50.
A correction here:
As far as I know, FAT32 is perfectly well supported in both the 49G+
and the 50G. The only thing you can't do with a FAT32 card is to
reflash the rom, but everyday use will work fine.
The only hint is to keep the number of clusters to a manageable
number, or the calculator will be slower to scan the FAT in the card
(the FAT table will become too big). So if you want a 2GB card, use it
with FAT32 but keep a reasonable cluster size that gives you the extra
capacity without too much speed penalty.

===
Subject: Re: Questions about SD cards for 50G
<g00qb6$229d$1@merope.saaf.se>
posting-account=v6m75AoAAAAgp3vyiLFAfyqmfMcXEPkm
1.1.4322; .NET CLR 2.0.50727; WinuE v6; WinuE 
v6),gzip(gfe),gzip(gfe)
> There's no point in using a 2 GB card on the 50, since you cannot store
> more on the 2 GB card than on a 1 GB card or on a 512 MB card. 
EHow come?
 It has to do with formatting the 2 GB as FAT16, which limits the number
> of files to a maximum of some 65000.
Good point. Of course you can always use the extra space on a 1GB or
2GB card if you have lots of 32kB or greater files, but for the
typical files used on HP calculators its a good point.
The max number of files is fixed, for any combination of file sizes.
===
Subject: Re: Questions about SD cards for 50G
posting-account=1ognzAoAAABDku11Zp1mRVw1UdpQAy_1
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
I always get surprised by the greatness of the 50g.
I just wanted to try if my MMC card that came with the camera could
work in the calc and I ended up using it exclusively for the calc!
It's a 16Mb MMC that says CANON on it. Don't know if that's the actual
brand or just a sticker. But my guess is the former.
Can't believe that worked out fine!
And 16Mb seems to me like more than enough!!
 There's no point in using a 2 GB card on the 50, since you cannot store
> more on the 2 GB card than on a 1 GB card or on a 512 MB card. How 
come?
 It has to do with formatting the 2 GB as FAT16, which limits the number
> of files to a maximum of some 65000.
 Good point. Of course you can always use the extra space on a 1GB or
> 2GB card if you have lots of 32kB or greater files, but for the
> typical files used on HP calculators its a good point.
> The max number of files is fixed, for any combination of file sizes.
===
Subject: Re: Questions about SD cards for 50G
> I always get surprised by the greatness of the 50g.
> I just wanted to try if my MMC card that came with the camera could
> work in the calc and I ended up using it exclusively for the calc!
> It's a 16Mb MMC that says CANON on it. Don't know if that's the actual
> brand or just a sticker. But my guess is the former.
> Can't believe that worked out fine!
> And 16Mb seems to me like more than enough!!
Same here, only different sticker Panasonic 
16MB is more than enough (i actualy have have 10 MB free)
but then, for special use dataloging etc... one might need greater card but 

still
let's not forget how much 1 GB realy is, especialy in HP calculator terms.
manjo
===
Subject: Is hpgcc home broken?
posting-account=bfephgoAAABDnwQr7ii9nivg2kXP2z_Z
Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; SV1) ; .NET CLR
1.1.4322),gzip(gfe),gzip(gfe)
I can't connect to www.hpgcc.org
===
Subject: Re: A bug and a related mini-challenge re LSQ [what's in a word?]
> [Eliminating the word solution from a discussion is] an unworkable 
solution.
> Linear algebra texts are going to continue to use the word solution.
The AUR and/or these newsgroup discussions are certainly free
to offer discussions that can explain what's going on
without using the word, or which can include,
either before or after such an explanation,
that in connection with LSQ, the word solution is used
to mean any X for which the norm of the residual 'B-A*X'
attains its minimum possible value,
to make completely clear that this is not necessarily
a solution to the equations A*X=B.
What I see as the heart of the differing takes on this subject
is that the word solution was never properly defined anywhere
(when you gave a definition, it was in the form of defining
a longer phrase containing qualifying words in front of
the word solution, which makes it seem that your definition
is about a special kind of solution, which would then
make it seem that solution must be some larger class,
and what larger class is there than all X?).
The AUR also really, IMO, steers the reader (who may not be
an expert in linear algebra, but just an ordinary calculator user)
into believing that solution refers to a solution
of the equations A*X=B -- the AUR actually says that,
even though it also contradicts itself elsewhere.
LSQ can be used for other purposes, even to refine
the accuracy of solutions to the equations, I believe,
so I think it deserves an explanation that clarifies this.
I just made analogy between LSQ and LR, and I like the idea
of expressing LSQ as something looking for a best fit
for a set of equations, just as LR looks for a best fit line,
even to a set of points that don't lie on a line
(LSQ thus also acting upon equation sets that have no
exact-fit solutions) -- this can be pointed out,
whether or not linear algebra texts do so, and I think
it will help to add something for general understanding.
Even the numeric solver, when there is a local minimum
but no exact solution, tries to locate that local minimum instead,
just as one more slightly analogous function that exists in the calculator.
Some of the back and forth in this thread seems to have come from
taking statements to an extreme interpretation; for example,
my saying that there is a correct one-sentence general definition
for the action of LSQ did not mean objection to elaborating
all the particular cases, and all the other material that could
be presented, and my saying that we can also explain LSQ
even without having to use or define the word solution
does not mean that we can't also include a definition --
just that if we are going to explain it using only
the word solution, then we (the AUR, say) had better
actually say what we really mean, as well as not
saying incorrect (and confusing) things.
A rose by any other name would smell as sweet,
but I guess this doesn't work the same for solution :)
Language must be the fundamental sin,
because as soon as human beings got that knowledge,
that's where they've gotten into trouble ever since :)
-- Y
===
Subject: Re: Install libraries
<op.ualjjysbnn735j@miu.edu>
posting-account=tK2cUgkAAABPl7vxXuLS8-X63eEhN1C3
1.1.4322; InfoPath.1; .NET CLR 2.0.50727; .NET CLR 3.0.04506.30; .NET CLR
3.0.04506.648),gzip(gfe),gzip(gfe)
[how to transfer files from computer to calculator]
-Joe-
===
Subject: Re: hpgcc and rpl-stack
<op.uam42lw1nn735j@miu.edu>
posting-account=y6ooGwkAAACoZd171Nq6oGIjQC5mB4Q9
.NET CLR 2.0.50727; .NET CLR 3.0.04506; .NET CLR 
3.5.21022),gzip(gfe),gzip(gfe)
> Since I know nothing, but just have curiosity,
> I can ask real dumb questions from pure ignorance
> (or feel free to refer me to fine manuals :)
Well... we don't have such good documentation, so I can't really use
the usual RTFM. It's not my style anyway.
 Is there a GC upon (or before) initial entry?
Yes, there is a GC if you call it using PrRUN, you can skip it if you
execute it using the other commands (PrLNK and PrJMP), in case you
don't like the speed penalty.
Let me clarify that I'm talking about HPGCC 2.0. The new version
(HPGCC3) behaves completely different.
 If not, does pop/push affect RPL stack memory,
> which could have been very close to limit
> (e.g. one word away from needing GC) to begin with?
You can always run a program under very low memory conditions, with GC
or not. In such case, push will fail (return without pushing the
value) if there's not enough memory.
 In any case, there must always be some limit,
> at which points the RPL stack(s) either overflow or underflow?
Overflow, yes, but it is under control, since the C program will
simply refuse to push any new objects on the stack when memory is
insufficient.
Underflow can't really happen, since you can't have less than 0 items
on the stack . pop will simply fail on an empty stack
 Is there any access to RPL return stack and environments?
No, we have no such functions. But using peek() and poke() you can
push values in the return stack manually. It's not very useful,
though, since you can't touch anything on the RPL stream or your C
program will crash on return.
I hope this satisfies your curiosity,

===
Subject: Re: Battery Charging
>> I am wondering if it is possible to charge NIMH batteries in a 50g
>> through the usb socket ?
 You won't get much of a charge - the terminal voltage for NiMH cells
> is about 1.48-1.52V, so the 1.25V per cell from the USB will only give
> about 1/2 capacity. Not to mention the difficulty of opening the HP50
> and then putting it back together again.
 Bill
I was going to run it off a constant current source with the current
and voltage limited, that is around 6 volts which is the same voltage
that the standard alkaline batteries provide.
I have the 50g now and it is functioning fine with NIMH cells so I
may not worry as they provide around ten hours of use per charge.
===
Subject: Re: A simple problem with numeric precision
> my good old trusted HP-41 won't do tricks like this.
If you put zero on the stack and then test a flag, etc.
conditionally skipping an EXP (e^x) command,
then this chooses 0/1;
for the opposite, put 1 on the stack
and conditionally skip a LN command.
The 4-level calcs were materially fabulous,
and very brilliant for their time,
yet programming still now seems like bows and arrows
vs. the kind of programming now available.
Come back in another 20 years and see
whether HP50 is also a geologic relic
(with current physical quality,
of course there won't be a working specimen left on Earth,
except in an emulator :)

===
Subject: Re: I'm looking to replace my 49G
posting-account=ujnlqwkAAADa5-KcfiulE2o9CRcf_ASq
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> So I'm looking for the simplest/cheapest HP calculator that has
> that desirable/reliable button press, RPN input only required,
> basic scientific builtin functions like sin, cos, tan,
> exponentials, log functions. I'd prefer the display to show more
> than just the bottom register, maybe 2 at least, but not a
> requirement. Something like the old 42, if memory serves?
The 35S suffers from needing 3 keys to enter hex numbers, really sad.
I have a 50G which has been sitting in my drawer for 6 months, barely
used ( I mean less than 2 hours total).
I'm selling it with a thick hardcover book for the HP49- full of
programs.
===
Subject: Re: I'm looking to replace my 49G
(http://www-dt.e-technik.uni-dortmund.de/~ma/leafnode/beta/)
> <snip /> So I'm looking for the simplest/cheapest HP calculator that has
>> that desirable/reliable button press, RPN input only required,
>> basic scientific builtin functions like sin, cos, tan,
>> exponentials, log functions. I'd prefer the display to show more
>> than just the bottom register, maybe 2 at least, but not a
>> requirement. Something like the old 42, if memory serves?
 The 35S suffers from needing 3 keys to enter hex numbers, really sad.
<snip />
I'm sure it is annoying if you use HEX a lot. I don't.
--
Troy Piggins
===
Subject: Re: I'm looking to replace my 49G
<20080526221136.293@usenet.piggo.com> 
<o8dl34duhi0an0jibe084bpseu5dbmc3u8@4ax.com>
posting-account=ujnlqwkAAADa5-KcfiulE2o9CRcf_ASq
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> If you don't need the graphics, and don't need the advanced programming 
capabilities, consider the hp35S.
I was excited to see HP make a new calculator, but today's HP is a far
cry from the HP that made the 15-C & 16C 9among many other great
calcs)
I like the 35S in general, although the LCD could be a tad brighter.
One of the worst problems with the 35S is- you have to press 3 keys to
enter EACH hex number.
There is not always on hex mode. If you enter in 5 hex numbers, you
need to press 15 keys to designate those number as being he, in
addition to the operations keys.
Really sad
I have a 50G which has barely been used for sale, along with a thick
hardcover book for the HP49 full of programs.
===
Subject: Re: I'm looking to replace my 49G
<20080526221136.293@usenet.piggo.com>
<o8dl34duhi0an0jibe084bpseu5dbmc3u8@4ax.com>
<20080527064621.308@usenet.piggo.com>
(http://www-dt.e-technik.uni-dortmund.de/~ma/leafnode/beta/)
>> If you don't need the graphics, and don't need the advanced
>> programming capabilities, consider the hp35S.
>> It has the nice 'clicky' keys just like the old 41C series.
>> Very nice feel. half the price of a 50G and much thinner. It
>> has it's quirks, but for 'minimal requirements' and RPN input,
>> it's a very nice unit.
>> The 50G, for the high end capabilities, is also a very nice
>> unit. The worst thing I've found with it is the placement of
>> the ENTER key. It's the bottom right key now, and that is a
>> bit of a pain when you've used HP calculators for 25 years with
>> the ENTER key middle left. (the 35s has the ENTER key in the
>> 'right' place.)
 into that one.

===
Subject: Re: Printer for HP 50G

 I think it is the HP 82240A printer which does not work with the HP50.
they all do.
You'll have to run the command OLDPRT first to make it work with the 
82240A.
The 82240B isn't produced anymore and I wonder if anyone still has stock
--
They who would give up an essential liberty for temporary security,
deserve neither liberty or security (Benjamin Franklin)
===
Subject: Re: Printer for HP 50G
<Virgil-71E030.16082225052008@comcast.dca.giganews.com>
posting-account=57eBeAkAAABkNMhsZnDBzGQsstJRmvKD
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
I also confirm that my HP 82240B printer does work perfectly with my
49G+ and my 50G.
Best wishes,
Andreas
http://www.software49g.gmxhome.de
===
Subject: Re: Printer for HP 50G
posting-account=IicVKwoAAACbH6mCTTDMOGh_Z0b4FE2T
AppleWebKit/525.18 (KHTML, like Gecko) Version/3.1.1 
Safari/525.18,gzip(gfe),gzip(gfe)
> I have got an HP50g and the HP82240B printer and it
> works perfectly. No problem.
> Peter
Funny situation here in Germany: Up to now three different HP-dealers
here in Germany gave me the same answer by e-mail: The HP 82240B
printer doesnt work with the 50G. It seems so, that they 
dont want
to earn any money here......
Jens
===
Subject: Re: Printer for HP 50G
Up to now three different HP-dealers
here in Germany gave me the same answer by e-mail: The HP 82240B
printer doesnt work with the 50G.
Culture today is that you never say, I don't know but rather, no.
Apparently being thought of as uninformed is somewhere worse than, on
infrequent occasion, being thought of as wrong?

===
Subject: Re: Battery Charging

> I am wondering if it is possible to charge NIMH batteries in a 50g
> through the usb socket ?
You won't get much of a charge - the terminal voltage for NiMH cells
is about 1.48-1.52V, so the 1.25V per cell from the USB will only give
about 1/2 capacity. Not to mention the difficulty of opening the HP50
and then putting it back together again.
Bill
===
Subject: Re: My Next Calculator

> IMHO, even the 50g's battery life is subpar. I need to replace the
> batteries every month or so even with light to moderate usage.
 S.C.
What ROM do you have ? Because the very first one in the HP49g+ used
to have a very high battery consumtion. (You should wright VERSION
to know it)
With rom 2.10 in HP49g+ and HP50g the battery life goes over 90 days
for me!

===
Subject: Re: InForm Builder help
posting-account=QSfVcQoAAADkgn5Et4hWcvOa13b-lJy1
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> Now thats a good way to get most everything over. I was working on a
> routine at work ( nothing much else to do) and needed an input form.
> So I tried to make it on the 50 and kept getting this error. Strange
> since I never had any issues before. InForm Builder has been very
> stable for me. I have the minimum ROM of 2.08. Extable is there. So it
> is a bit of a mystery to me. If I ever figure it out I'll post here.
 Scott H.
There are too many options in this machines, and even when could be
very good for the deep knowledge of them, have not enough time to go
and find them; so I used to update the ROM to 2.10 in both (HP49g+ and
HP50g) and have the same programs in both of them thru backup and
restore options.
No more troubles.
Daniel.
===
Subject: Re: Printer for HP 50G
> Hello together,
 Im a new owner of a HP50G. A calculator dealer in Germany 
said in a e-
> mail to me, that the HP 82240B printer doesnt work with 
the HP50G. Is
> this true? And when, is there another printer I can use, perhaps an
> used one?
 Jens
I think it is the HP 82240A printer which does not work with the HP50.
===
Subject: I'm looking to replace my 49G
(http://www-dt.e-technik.uni-dortmund.de/~ma/leafnode/beta/)
I upgraded from a 48 to the 49G some time ago. The 48 had a
disagreement with an office fire. It lost.
I've been putting up with the unreliable button pressing issue of
the 49 but can't take it any longer. I loved the feel of the
click of the 48's buttons - you hit a button and you knew it
registered. Also find I'm not using enough of the features of
the 49G to warrant something that high-end.
So I'm looking for the simplest/cheapest HP calculator that has
that desirable/reliable button press, RPN input only required,
basic scientific builtin functions like sin, cos, tan,
exponentials, log functions. I'd prefer the display to show more
than just the bottom register, maybe 2 at least, but not a
requirement. Something like the old 42, if memory serves?
Any recommendations?
--
Troy Piggins
===
Subject: Re: I'm looking to replace my 49G
posting-account=Rr1BpAoAAAC35AO1lUfz6ZegzjxdWL8v
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> I upgraded from a 48 to the 49G some time ago. The 48 had a
> disagreement with an office fire. It lost.
 I've been putting up with the unreliable button pressing issue of
> the 49 but can't take it any longer. I loved the feel of the
> click of the 48's buttons - you hit a button and you knew it
> registered. Also find I'm not using enough of the features of
> the 49G to warrant something that high-end.
 So I'm looking for the simplest/cheapest HP calculator that has
> that desirable/reliable button press, RPN input only required,
> basic scientific builtin functions like sin, cos, tan,
> exponentials, log functions. I'd prefer the display to show more
> than just the bottom register, maybe 2 at least, but not a
> requirement. Something like the old 42, if memory serves?
 Any recommendations?
 --
> Troy Piggins
yea, buy another HP48
===
Subject: Re: I'm looking to replace my 49G
(http://www-dt.e-technik.uni-dortmund.de/~ma/leafnode/beta/)
>> I upgraded from a 48 to the 49G some time ago. The 48 had a
>> disagreement with an office fire. It lost.
>> I've been putting up with the unreliable button pressing issue of
>> the 49 but can't take it any longer. I loved the feel of the
>> click of the 48's buttons - you hit a button and you knew it
>> registered. Also find I'm not using enough of the features of
>> the 49G to warrant something that high-end.
>> So I'm looking for the simplest/cheapest HP calculator that has
>> that desirable/reliable button press, RPN input only required,
>> basic scientific builtin functions like sin, cos, tan,
>> exponentials, log functions. I'd prefer the display to show more
>> than just the bottom register, maybe 2 at least, but not a
>> requirement. Something like the old 42, if memory serves?
>> Any recommendations?
 yea, buy another HP48
Ok. I see the latest is the 48gII. Does that have the nice
clicky keys like the older models, or is it the different ones
that don't feel reliable like the 49G?
I was worried it might be a little much for my minimal
requirements these days. There isn't a cheaper one?
If I'm spending that much, would you recommend the 50g over the
48gII?
--
Troy Piggins
===
Subject: Re: I'm looking to replace my 49G
If you don't need the graphics, and don't need the advanced programming 
capabilities, consider the hp35S.
It has the nice 'clicky' keys just like the old 41C series. Very nice feel. 
half the price of a 50G and much thinner.
It has it's quirks, but for 'minimal requirements' and RPN input, it's a 
very nice unit.
The 50G, for the high end capabilities, is also a very nice unit. The worst 
thing I've found with it is the placement
of the ENTER key. It's the bottom right key now, and that is a bit of a 
pain when you've used HP calculators for 25
years with the ENTER key middle left. (the 35s has the ENTER key in the 
'right' place.)
Ok. I see the latest is the 48gII. Does that have the nice
>clicky keys like the older models, or is it the different ones
>that don't feel reliable like the 49G?
I was worried it might be a little much for my minimal
>requirements these days. There isn't a cheaper one?
If I'm spending that much, would you recommend the 50g over the
>48gII?
--
If you try to fail, and succeed, which have you done?
===
Subject: Re: I'm looking to replace my 49G

(http://www-dt.e-technik.uni-dortmund.de/~ma/leafnode/beta/)
 If you don't need the graphics, and don't need the advanced
> programming capabilities, consider the hp35S.
 It has the nice 'clicky' keys just like the old 41C series.
> Very nice feel. half the price of a 50G and much thinner. It
> has it's quirks, but for 'minimal requirements' and RPN input,
> it's a very nice unit.
 The 50G, for the high end capabilities, is also a very nice
> unit. The worst thing I've found with it is the placement of
> the ENTER key. It's the bottom right key now, and that is a
> bit of a pain when you've used HP calculators for 25 years with
> the ENTER key middle left. (the 35s has the ENTER key in the
> 'right' place.)
into that one.
--
Troy Piggins
===
Subject: Re: I'm looking to replace my 49G
<20080526221136.293@usenet.piggo.com>
posting-account=57eBeAkAAABkNMhsZnDBzGQsstJRmvKD
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> Ok. I see the latest is the 48gII.
The 48gII is not the successor of the 48GX !
Basically the 48gII is is stripped down 49G+ or as called by others a
49G-
Programs for the 48GX series will not work on a 48gII and USER-RPL
programs have to be converted as they have to be converted for the 49G/
49G+/50G series.
> If I'm spending that much, would you recommend the 50g over the
> 48gII?
If you do not want to spend the money for a *used* 48GX get a 50G
which is a vast improvment compared to the 49G series.
HTH
Andreas
===
Subject: Re: I'm looking to replace my 49G
<20080526221136.293@usenet.piggo.com>
(http://www-dt.e-technik.uni-dortmund.de/~ma/leafnode/beta/)
> Ok. I see the latest is the 48gII.
 The 48gII is not the successor of the 48GX !
> Basically the 48gII is is stripped down 49G+ or as called by others a 
49G-
> Programs for the 48GX series will not work on a 48gII and USER-RPL
> programs have to be converted as they have to be converted for the 49G/
> 49G+/50G series.
Oh, I see. Wonderful model naming scheme there then, HP. Very
intuitive :)
>> If I'm spending that much, would you recommend the 50g over the
>> 48gII?
 If you do not want to spend the money for a *used* 48GX get a 50G
> which is a vast improvment compared to the 49G series.
 HTH
> Andreas
--
Troy Piggins
===
Subject: Re: FEM48/49 v5.3 updated to SR4
posting-account=R-8hawoAAAA5CgW3hZ9cXaqaEeJoHVEh
Gecko/2008051206 Firefox/3.0,gzip(gfe),gzip(gfe)
Having just ordered an HP50G... I am looking at some small
improvements for the FEM49 library, like implementing the FAST routine
again.
On the HP48G this turned the display off, but for the HP50G I would
like to use the ARM toolbox with adjustment of the clock speed
programs as distributed by Al Borowski.
The EvalFast routine of FEM48 looks like:
* EvalFast
*
* FAST evaluation (turns LCD off)
::
12 GETcfg (use FAST?)
ITE
::
CODE
GOSBVL =SAVPTR
GOSBVL =DisableIntr
GOSBVL #1D44
GOSBVL =AllowIntr
GOVLNG =GETPTRLOOP
ENDCODE
ERRSET
EVAL
ERRTRAP
::
CODE
GOSBVL =SAVPTR
GOSBVL =DispOn
GOVLNG =GETPTRLOOP
ENDCODE
ERRJMP
;
CODE
GOSBVL =SAVPTR
GOSBVL =DispOn
GOVLNG =GETPTRLOOP
ENDCODE
;
EVAL
;
Could I just replace this by:
* EvalFast
*
* FAST evaluation (turns LCD off)
::
12 GETcfg (use FAST?)
ITE
::
C124s
ERRSET
EVAL
ERRTRAP
::
CFIX
ERRJMP
;
CFIX
;
EVAL
;
===
Subject: Re: FEM48/49 v5.3 updated to SR4
posting-account=57eBeAkAAABkNMhsZnDBzGQsstJRmvKD
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
> Could I just replace this by:
 * EvalFast
> *
> * FAST evaluation (turns LCD off)
> ::
> 12 GETcfg (use FAST?)
> ITE
> ::
> C124s
> ERRSET
> EVAL
> ERRTRAP
> ::
> CFIX
> ERRJMP
> ;
> CFIX
> ;
> EVAL
> ;
I use my code that way:
::
ERRSET
::
SetTo203Mhz
run my code...
SetTo75MHz
;
ERRTRAP
SetTo75MHz
;
HTH
Andreas
===
Subject: Re: FEM48/49 v5.3 updated to SR4

IIRC turning off the display does not improve the speed of the
calculation on an ARM based machine.
This is the code I use inside an ERRSET/ERRTRAP to set the clock speed
to 203 MHz of the ARM processor. It is based on Al Borowskis 
code so
the screen does flicker sometimes.
Please remember that you need an ERRSET/ERRTRAP because if an error
happens the processor will still run with 203 MHz until the next warm
start and at 203 MHz the keyboard does not work, so you need to use
the 75 MHz code twice, once at the end inside the ERRSET/ERRTRAP and
once in case an error happens
This will compile directly in debug4x as the ARM code is encoded as a
binary inside the Saturn ML code.
HTH,
Andreas
This will set the ARM processor to 203 MHz:
CODE
GOSBVL =SAVPTR
GOSBVL PTR IsApple Ticks 3D1 auf simulierten 49G,
Ticks 21 auf simulierten 49G+
GOC weiter203 Ticks 0A falls Carry gesetzt, Ticks
03 falls Carry NICHT gesetzt
GOVLNG =GETPTRLOOP
weiter203
INTOFF INTOFF = disable keyboard
interrupts Ticks 05
LC(5) #000B9 B4h + 5 = B9h
GOSBVL =CREATETEMP erzeuge in TEMPOB einen Platzhalter
f.9fr den Pointer zu unterem String
GOSUB StrEnd203 Der String ist B4 Nibbles lang +5
f.9fr den Prolog -> B4h + 5 = B9h
CON(5) =DOCSTR
REL(5) StrEnd203
NIBHEX
0000000FFF4D29E37610A3EFFC20A3E1202282E0002185E70410A3E4001182E50C20A3E4002
NIBHEX
184E4002184E4002184E4002184E4002184E4002184E27610A3E4001182E1AA20A3E1302282E
0002
NIBHEX 185EFFF8DB8EDDDDDDDD
StrEnd203
C=RSTK % Now you have a pointer to the grob
B=B-CON A,6 Ticks 0A
C=C+B A Ticks 07
D0=C Ticks 08
C=B A Ticks 07 LC(5) #000B9 dauert auch Ticks 07
GOSBVL =MOVEUP Ticks 4C6 hier
D1=D1+ 5
A=DAT1 A
D1=D1+ 12
D0=A
AD1EX von GOSBVL =SAVPTR bis vor LC(5) #FFFF8
dauert dies Ticks AE5
**** Anfang urspr.9fnglicher Original-Code Anfang
* INTOFF % das stimmt nicht: hiermit funktioniert die ON-
Taste bei .86bertaktung nicht Ticks 05
* D0=A % Ticks 08 Pointer zum
String in D0
* D0=D0+ 5 % Ticks 07 .86berspringe
Prolog des String
* A=DAT0 A % Ticks 12 Schreibe die
Gr.9ae des String in A
* D0=D0+ 12 % Ticks 07 Springe an
den Anfang der Stringdaten
* AD0EX % Ticks 08 Tausche A
und D0, D0=Gr.9ae des String A=Anfang der Stringdaten
**** Ende urspr.9fnglicher Original-Code Ende
****************************************************************************
******************************************************
LC(5) #FFFF8 % Ticks 07 Lade #FFFF8
in C
C=C&A A % Ticks 09 Logical AND
also called logical product
D1=C % Ticks 08
RSTK=C % Ticks 08
AD0EX % Ticks 08
A=A-CON A,12 % Ticks 0A
C=A A % Ticks 08
BUSCC
CON(2) #06 % Ticks 08 = MOVEDN $80B60
C=RSTK % Ticks 08
BUSCC
CON(2) #FF % Ticks 06 = ARMSAT $80BFF
GOVLNG =GETPTRLOOP
ENDCODE
********************************************************
And this will set the ARM processor to 75 MHz:
* 13.01.07 ANFANG schalte den Rechnner auf 75MHz zur.9fck ANFANG schalte
den Rechnner auf 75MHz zur.9fck ANFANG
CODE
GOSBVL =SAVPTR
GOSBVL PTR IsApple Ticks 3D1 auf simulierten 49G,
Ticks 21 auf simulierten 49G+
GOC weiter75 Ticks 0A falls Carry gesetzt, Ticks
03 falls Carry NICHT gesetzt
GOVLNG =GETPTRLOOP
weiter75
INTOFF INTOFF = disable keyboard
interrupts Ticks 05
LC(5) #000D1 #CCh + 5 = D1h
GOSBVL =CREATETEMP erzeuge in TEMPOB einen Platzhalter
f.9fr den Pointer zu unterem String
GOSUB StrEnd75 Der String ist CC Nibbles lang +5
f.9fr den Prolog -> #CCh + 5 = D1h
CON(5) =DOCSTR
REL(5) StrEnd75
NIBHEX
0000000FFF4D29E27610A3E4001182E34A20A3E2102282E0002185E37610A3E11B20A3E1202
NIBHEX
282E0002185E70410A3E4001182E20C20A3E4002184E30C20A3E4002184E4002184E4002184E
4002
NIBHEX 184E4002184E50C20A3E4002184E4002184EFFF8DB8E
StrEnd75
C=RSTK % Now you have a pointer to the grob
B=B-CON A,6 Ticks 0A
C=C+B A Ticks 07
D0=C Ticks 08
C=B A Ticks 07 LC(5) #000B9 dauert auch Ticks 07
GOSBVL =MOVEUP Ticks 4C6 hier
D1=D1+ 5
A=DAT1 A
D1=D1+ 12
D0=A
AD1EX von GOSBVL =SAVPTR bis vor LC(5) #FFFF8
dauert dies Ticks AE5
**** Anfang urspr.9fnglicher Original-Code Anfang
* INTOFF % das stimmt nicht: hiermit funktioniert die ON-
Taste bei .86bertaktung nicht Ticks 05
* D0=A % Ticks 08 Pointer zum
String in D0
* D0=D0+ 5 % Ticks 07 .86berspringe
Prolog des String
* A=DAT0 A % Ticks 12 Schreibe die
Gr.9ae des String in A
* D0=D0+ 12 % Ticks 07 Springe an
den Anfang der Stringdaten
* AD0EX % Ticks 08 Tausche A
und D0, D0=Gr.9ae des String A=Anfang der Stringdaten
**** Ende urspr.9fnglicher Original-Code Ende
****************************************************************************
******************************************************
LC(5) #FFFF8 % Ticks 07 Lade #FFFF8
in C
C=C&A A % Ticks 09 Logical AND
also called logical product
D1=C % Ticks 08
RSTK=C % Ticks 08
AD0EX % Ticks 08
A=A-CON A,12 % Ticks 0A
C=A A % Ticks 08
BUSCC
CON(2) #06 % Ticks 08 = MOVEDN $80B60
C=RSTK % Ticks 08
BUSCC
CON(2) #FF % Ticks 06 = ARMSAT $80BFF
GOVLNG =GETPTRLOOP
ENDCODE
* 13.01.07 ENDE schalte den Rechnner auf 75MHz zur.9fck ENDE schalte
den Rechnner auf 75MHz zur.9fck ENDE
===
Subject: Re: FEM48/49 v5.3 updated to SR4
posting-account=57eBeAkAAABkNMhsZnDBzGQsstJRmvKD
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
What I forgot (but you might have discovered this directly in the
code): this is compatible with the old saturn processor as the test at
the beginning checks for an ARM based machine and if it is running on
a Frozen Hampster Butt (FHB) nothing happens. The test is very fast
and allows to have one source code for all machines this way.
=PTR_IsApple_ EQU #3FEF1
* Carry on ARM based machine, first availabe in ROM 1.22
HTH,
Andreas
===
Subject: Re: FEM48/49 v5.3 updated to SR4
posting-account=R-8hawoAAAA5CgW3hZ9cXaqaEeJoHVEh
Gecko/2008051206 Firefox/3.0,gzip(gfe),gzip(gfe)
Hello Andreas,
Since I asked if the routines of Al Borowski could be used (please
read my message again, and nothe the nice error trap...), after
reading your answer I take it that this is possible.
But, is 203 MHz safe to use? In Al's documentation he mentions a safe
option: 124 MHz?
Caspar
===
Subject: Re: FEM48/49 v5.3 updated to SR4
posting-account=57eBeAkAAABkNMhsZnDBzGQsstJRmvKD
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
Hello Caspar
> But, is 203 MHz safe to use? In Al's documentation he mentions a safe
> option: 124 MHz?
I have been using this option on several machines (a lot more than I
own) and never got a complain from anybody nor did someone reported
any trouble so far.
IMHO going from 75 MHz to 124 MHz is hardly noticeable unless we are
talking about a lenghty running program (a minute or more).
Of course the overclocking to 203 MHz is only used rarley at some very
time critical points and never runs longer than one second (in real
time). I do not know what happens if you leave an overclocked machine
that way for a longer time period but if you want to use a different
clockspeed you need to replace the string with the appropriate binary
ARM code and adjust the value for =CREATETEMP as the binary ARM code
is copied into this ram area.
HTH,
Andreas
===
Subject: A simple problem with numeric precision
Hi all,
I'm looking for an elegant solution to the following problem:
Consider the equation
y = ln(4*t*(1-t))
over the range 0 < t < 1.
Even 0 < t < 0.5 will do (=> symmetry).
Now, y shall be evaluated on a 10-digit calculator (here: HP-41 series)
to at least 8, if possible 9-10 significant digits.
Two problems arise:
(1) For t close to 0.5, y approaches zero. Due to its limited precision
the calculator cannot evaluate the equation with sufficient precision:
t = 0.499999 would result in 0.499999 * 0.500001 = 0.24999 99999 99,
which will be rounded to 0.25, thus returning y = ln 1 = 0.
Simple solution: rewrite the equation as
y = ln(1 - 4*(t-0.5)^2)
and get full machine precision by using the 41's ln1+x function,
designed exactly for cases like this:
01 .5
02 -
03 X^2
04 -4
05 *
06 LN1+X
That's it.
(2) On the other hand, for t close to zero this workaround fails:
If t is, for example, somewhere around 1 E-6, the expression 0.5 - t
can not be evaluated to full precision, simply because the calculator
can't distinguish between t = 1 E-6 and, say, t = 1.0000 12345 E-6:
0.5 - 1.0000 12345 E-6 = 0.49999 89999 87655
which becomes 0.49999 90000 in a 10-digit machine,
i.e. we get the same result as for t = 1 E-6.
So a different approach is required, for instance
y = ln(4 * (t - t^2))
which works nicely for small values of t.
01 ENTER
02 X^2
03 -
04 4
05 *
06 LN
But again, this solution does not perform well in the first case,
i.e. for t close to 0.5. Below about t = 0.001 the desired precision can
not be guaranteed for every value of t.
So, at the moment I'm using *both* versions:
the first one for t >= 0.1 and the second one else.
Question: do you know a better way to handle this problem?
Maybe something that uses one *single* approach both for t close to 0.5
and t close to zero.
Dieter
===
Subject: Re: A simple problem with numeric precision
Evaluate an expression, near t=.5 one way, near t=0 another way:
'IFTE(t<.25,expr1,expr2)'
Of course on an HP41 (or in RPL on later series),
just use a logical test within one program,
with different program branches to different program sections
(or internal labels).

===
Subject: Re: A simple problem with numeric precision
> Evaluate an expression, near t=.5 one way, near t=0 another way:
Well, as I said before, that's what I'm doing right now:
| So, at the moment I'm using *both* versions:
| the first one for t >= 0.1 and the second one else.
Even exactly in the way you mentioned (sorry if that wasn't clear):
.1
X<>Y
X<Y?
GTO 01
(routine for t close to 0.5)
GTO 02
LBL 01
(routine for small t)
LBL 02
(...continue..)
The question is: can you do better?
I'm still looking for the ultimately elegant solution. 8-)
Dieter
===
Subject: Re: A simple problem with numeric precision
> | So, at the moment I'm using *both* versions:
> | the first one for t >= 0.1 and the second one else.
 Even exactly in the way you mentioned (sorry if that wasn't clear):
 .1
> X<>Y
> X<Y?
> GTO 01
> (routine for t close to 0.5)
> GTO 02
> LBL 01
> (routine for small t)
> LBL 02
> (...continue..)
 The question is: can you do better?
> I'm still looking for the ultimately elegant solution. 8-)
You can always multiply the numeric value of, say, t < 0.5
(true being 1 and false being 0) by one expression,
then add the logical negative of that times the other expression,
thus having no branches at all :)
I don't get no more elegant myself, but good luck 

===
Subject: Re: A simple problem with numeric precision
> You can always multiply the _numeric_ value of, say, t < 0.5
> (true being 1 and false being 0) by one expression,
Ah - reminds me of a friend's programming technique he used in Gauss
(something like Mathematica for econometrists). Or the good old times
when I used typecasts in Turbo Pascal. 8-)
But well, my good old trusted HP-41 won't do tricks like this.
> then add the logical negative of that times the other expression,
> thus having no branches at all :)
With twice the effort and execution time. 8-)
> I don't get no more elegant myself, but good luck 
Dieter
===
Subject: Re: Restore to EMU48
> The Backup menu item is always greyed out.
On Emu48?
That entire menu on Emu48 refers to saving and restoring
the internal memory state in Emu48 only
not to backing up and restoring the HOME directory of the calculator
to/from files representing calculator objects,
which instead uses the calculator's standard commands,
as emulated faithfully under Emu48.
The internal memory state in Emu48 does include all of HOME
(and stack, flags, current directory, temporary storage,
just like the kind of all-inclusive memory file that Emu48
can save or load for itself), but that's different
from anything that the just calculator alone
(whether real or emulated) handles by itself.

===
Subject: Re: Restore to EMU48

> The Backup menu item is always greyed out.
 On Emu48?
 That entire menu on Emu48 refers to saving and restoring
> the internal memory state in Emu48 only
> not to backing up and restoring the HOME directory of the calculator
> to/from files representing calculator objects,
> which instead uses the calculator's standard commands,
> as emulated faithfully under Emu48.
 The internal memory state in Emu48 does include all of HOME
> (and stack, flags, current directory, temporary storage,
> just like the kind of all-inclusive memory file that Emu48
> can save or load for itself), but that's different
> Efrom anything that the just calculator alone
> (whether real or emulated) handles by itself.
 
Success!
Dragging the Backup into the EMU screen showed Backup HOMEDIR.
Typing RESTORE has given me a clone of the real thing.
N
===
Subject: Re: Restore to EMU48
> I have backed up my 50g to my PC and now want to use the backup on
> EMU48 (packaged with HPUseredit5). I can't do it though. I've tried
> dragging and dropping the file to the EMU screen, Opening and
> Loading: all to no avail. The Backup menu item is always greyed
> out.
To restore a standard backup of your HOME directory to the emulator?
When you drop your backup file onto Emu48
(while it is emulating any HP49/50 series calc),
does the stack show it as either Backup HOMEDIR or as DIR...?
If Backup HOMEDIR then you might try a RESTORE command directly on 
that.
Otherwise store it into any port under any name, e.g. :1:BK001 STO
Then :1:BK001 RESTORE
(or perhaps :1:HOMEDIR RESTORE if it ignored your suggested name
and was stored under its original name)
This should work independently of HPUseredit,
unless that program makes some special kind of its own backup
On-calc backups may also be made to any port or file, e.g.
:3:BK002 ARCHIVE (to an SD card, on a real calculator)


===
Subject: Re: Battery Charging
<483c0e2a$0$21126$5a62ac22@per-qv1-newsreader-01.iinet.net.au>
CLR 1.1.4322),gzip(gfe),gzip(gfe)
> I was going to run it off a constant current source with the current
> and voltage limited, that is around 6 volts which is the same voltage
That should be okay, as long as the current isn't high enough to
overcharge (NiMH can't tolerate overcharging like the old NiCds). C/
10 would be a safe value, which is about 100mA with 1000mAh cells.
According to the label on my Duracell 1000mAh cells, 100mA gives a
full charge in 16 hours.
Bill
===
Subject: 50g/algebraic/solver question
posting-account=0piKEwoAAACqCFJtDqa-a00T5YhKTyux
2.0.50727),gzip(gfe),gzip(gfe)
Have had my 50g for about 2 weeks... trying to figure out how to use
the solver in algebraic mode is like trying to scratch an itch I can't
reach...
Say I have loaded V=IR as the current equation and I see
[V] [I] [R] [EXPR=] [ ] [nXEQ]
at the bottom of the screen.
I do
V STO> 100
I STO> 5
to assign V=100 and I=5. What do I key in to solve for R?
I know there is the NUM.SLV mode where I can move the cursor around in
the SOLVE EQUATION screen to assign the values for relevant
variables and press SOLVE for the variable I want to solve. In
following the manual, I have successfully solved for R in RPN mode,
but there has to be a way to solve for R in algebraic mode? This is
driving me nuts!
===
Subject: Re: 50g/algebraic/solver question
posting-account=itRlSAoAAADIJvOzSGM8mjsK7gvlk1wg
Gecko/20071127 Firefox/2.0.0.11,gzip(gfe),gzip(gfe)
Unfortunately I am not able to do this in algebraic mode,
but in RPN the solution is
V, ENTER, I, ENTER, R, ENTER, *, =
(if there occur any numbers in these operations, you have
to PURGE the appropriate variable)
Now your equation is in the x-register, next steps
R, ENTER, ISOL
Now R = V/I will be in the x-register. Simplest way to
evaluate this is once more NUM.SLV
Peter
===
Subject: Re: Printer for HP 50G
> On 2008-05-26 08:08:22 +1000, Virgil <Virgil@gmale.com> said:
 I think it is the HP 82240A printer which does not work with the
> HP50.
> they all do. You'll have to run the command OLDPRT first to make it
> work with the 82240A.
I have an 82240A printer which works fine with my old 48GX but won't
work at all with my 49G+, even after executing the OLDPRT command.
===
Subject: Re: Printer for HP 50G
> I also confirm that my HP 82240B printer
> does work perfectly with my 49G+ and my 50G.
Does it need to be much closer than HP48 had to be?
There might be less IR transmit power in the 49G+/50G
than the HP48 series, if I'm reading this correctly:
(Cyrille, explaining why IR transfer 48 <-> 49G+/50G may fail).

===
Subject: Re: Printer for HP 50G
On 2008-05-29 07:45:30 +1000, John H Meyers <jhmeyers@nomail.invalid> 
said:
> I also confirm that my HP 82240B printer
>> does work perfectly with my 49G+ and my 50G.
 Does it need to be much closer than HP48 had to be?
>
The IR emitted isn't crippled, only the receiving one is.
However, while I could print over 10 metres away with my 48GX, you
would want to be within the 50cm range with a 50/49g+
--
They who would give up an essential liberty for temporary security,
deserve neither liberty or security (Benjamin Franklin)
===
Subject: Re: Printer for HP 50G
<op.ubvv14u6nn735j@miu.edu>
posting-account=57eBeAkAAABkNMhsZnDBzGQsstJRmvKD
Gecko/20080404 Firefox/2.0.0.14,gzip(gfe),gzip(gfe)
Hello John,
> Does it need to be much closer than HP48 had to be?
I do not remember how far I could place my 48SX to print because it
has been years that my SX stopped working :-(
> There might be less IR transmit power in the 49G+/50G
> than the HP48 series, if I'm reading this correctly:
I placed it about a little less than an inch away from the IR receiver
of the printer and it worked fine. For further distance I have to
recheck.
HTH,
Andreas
http://www.software49g.gmxhome.de
===
Subject: Re: Printer for HP 50G
>Hello together,
Im a new owner of a HP50G. A calculator dealer in Germany 
said in a e-
>mail to me, that the HP 82240B printer doesnt work with the 
HP50G. Is
>this true? And when, is there another printer I can use, perhaps an
>used one?
Jens
You might check Eric's site. There are several programs that enable
the S/SX/G/GX series to print to different kinds of printers. One is
Epson and the other is HP PCL type printers. Since the HP50G has a
serial port again, one of these might work. If not, they might give
you a start in writing a program of your own for the 50G, to do this.
Since most new printers are USB, you might have to get an USB to
serial adapter. There a scads out there. It is also possible to get an
old serial printer to use with your calculator.
Harold A. Climer
Dept. Of Physics, Geology, and Astronomy
University of Tennessee at Chattanooga
309 Grote Hall
423-425-4546
Harold-Climer@utc.edu
===
Subject: Re: Printer for HP 50G
On 2008-05-28 15:52:54 +1000, Virgil <Virgil@gmale.com> said:
 I have an 82240A printer which works fine with my old 48GX but won't
> work at all with my 49G+, even after executing the OLDPRT command.
the only printer I have is a 82240A , and it works fine with my 50...
so it's surprising.
The 49g+/50 support of the 82240 is really poor anyway, while it
officially works it will take several go before I can print anything
--
They who would give up an essential liberty for temporary security,
deserve neither liberty or security (Benjamin Franklin)
===
Subject: Re: Printer for HP 50G
>On 2008-05-26 08:08:22 +1000, Virgil <Virgil@gmale.com> said:
>> I think it is the HP 82240A printer which does not work with the HP50.
they all do.
>You'll have to run the command OLDPRT first to make it work with the 
82240A.
The 82240B isn't produced anymore and I wonder if anyone still has stock
Ebay currently has 14 of them on bid....
--
----------------------------------------------------------------
Paul Schlyter, Grev Turegatan 40, SE-114 38 Stockholm, SWEDEN
e-mail: pausch at stjarnhimlen dot se
WWW: http://stjarnhimlen.se/
===
Subject: Re: Printer for HP 50G
>Up to now three different HP-dealers
>here in Germany gave me the same answer by e-mail: The HP 82240B
>printer doesnt work with the 50G.
Culture today is that you never say, I don't know but rather, no.
>Apparently being thought of as uninformed is somewhere worse than, on
>infrequent occasion, being thought of as wrong?
The culture of the future then might be to never trust someone who
says no. - he's probably just afraid to appear uninformed.... :-/
One could also add a little morsel about the japanese culture, which
is to never say no. because that's not polite. If the correct answer
is no., a japanese would instead say yes, but.... followed by a
polite explanation why it's not possible..... 
--
----------------------------------------------------------------
Paul Schlyter, Grev Turegatan 40, SE-114 38 Stockholm, SWEDEN
e-mail: pausch at stjarnhimlen dot se
WWW: http://stjarnhimlen.se/
===
Subject: Re: HP 48 SX power Problem - Desperate Help
you were totally right!! Pushing above A button it works!!!
2 more questions:
1) Why does the problem only affect ON button?
2) Is it possible to solve it? With a spray?
Stefano
> if you press the cover above B eg, F2
> and it then works
> the foam or buffering layer inside has lost its flexibility
>
===
Subject: Re: HP 48 SX power Problem - Desperate Help
> you were totally right!! Pushing above A button it works!!!
> 2 more questions:
> 1) Why does the problem only affect ON button?
just wait and it will creep to buttons above
> 2) Is it possible to solve it? With a spray?
I simply keep on pressing above button B
as you may guess from answer 1
Other people may help solving the problem permanently
 Stefano

>> if you press the cover above B eg, F2
>> and it then works
>> the foam or buffering layer inside has lost its flexibility

===
Subject: Re: A simple problem with numeric precision
Try rewriting the equation to the following:
y=ln(2*t)+ln(2-2*t)
01 2
02 *
03 LN
04 2
05 LASTX
06 -
07 LN
08 +
This will give good precision over the entire range.
Jens
===
Subject: Re: A simple problem with numeric precision
> Try rewriting the equation to the following:
 y=ln(2*t)+ln(2-2*t)
> ...
> This will give good precision over the entire range.
I think I see the idea behind this (quite clever), but this solution has
the same problem as mine: for t close to 0.5 the result is only okay if
t-0.5 has not too many significant digits:
t = 0.49999 00000 returns -4.0000 00000 E-8 (correct)
t = 0.49999 99000 returns -4.0000 00000 E-12 (correct)
t = 0.49999 99990 returns -4.0000 00000 E-18 (correct)
Fine
(although for t = 0.49999 99999 the result gets wrong)
Problems arise when t-0.5 has more than a few significant digits:
t = 0.49999 10000 returns -3.2400 00000 E-10 (correct)
t = 0.49999 11000 returns -3.1684 00000 E-10 (correct within 1 ULP)
t = 0.49999 11100 returns -3.1613 00000 E-10 (five sign. digits okay)
t = 0.49999 11110 returns -3.1606 00000 E-10 (five sign. digits okay)
Up to here that's at least five sign. digits -
quite nice, although I need at least eight.
t = 0.49999 11111 returns -1.1606 E-10 (wrong)
t = 0.49999 11112 returns +8.3950 E-10 (wrong)
t = 0.49999 11113 returns -7.1603 E-10 (wrong)
t = 0.49999 33333 returns -5.7777 E-10 (wrong)
t = 0.49999 44444 returns -3.2346 E-10 (wrong)
...
But your idea can be improved by replacing ln(2-2t) by ln1+x(1-2t).
This way you get about five digits for t = 0.49999..., or four digits
for t = 0.49999 9..., three for t = 0.49999 99..., and so on.
Anyway - thank you for your time and help. In the meantime I solved the
problem in another way: I used a different approach where t is always
greater than 0.003 so that the equation for t close to 0.5 can be used
with sufficient precision. 8-)
Dieter
===
Subject: Re: A simple problem with numeric precision
X
> Anyway - thank you for your time and help. In the meantime I solved the
> problem in another way: I used a different approach where t is always
> greater than 0.003 so that the equation for t close to 0.5 can be used
> with sufficient precision. 8-)
Care to share?
Please...
> Dieter
>
===
Subject: Re: A simple problem with numeric precision
> Anyway - thank you for your time and help. In the meantime I solved the
> problem in another way: I used a different approach where t is always
> greater than 0.003 so that the equation for t close to 0.5 can be used
> with sufficient precision. 8-)
 Care to share?
> Please...
Okay then - the project behind the original question is an approximation
of the inverse cumulative normal distribution (i.e. enter probability
p = 0.05, get quantile z = -1.64485). The usual (better-than-Hastings)
rational approximations suffer from a limited working range, i.e. they
perform well within the specified error margins only down to a certain
limit, e.g. probabilities down to 1E-10, 1E-20 or 1E-300. Others do not
perform well for results close to zero, i.e. for probabilities close to
0.5 where the result approaches zero. And right here is the link to the
original question.
I'm using two approximations, one for the tail and one for the center.
Both are designed in a way that the error converges against zero the
more you move out in the tail (p -> 0) or the closer you get to p = 0.5.
The switch between both approximations was initially done near p = 1E-5
where the mentioned numeric problems arise.
The equation was a part of the calculation of the independent variable
for the rational center approximation. In the meantime I simplified the
latter and improved the tail approximation, so that now the switch point
is at p = 0.003. This also had some additional benefits.
Dieter
===
Subject: Selling vintage HP calculators
posting-account=3WY0nwoAAAAJz1UJhuywvjIfR3_WSUMZ
AppleWebKit/525.18 (KHTML, like Gecko) Version/3.1.1 
Safari/525.18,gzip(gfe),gzip(gfe)
I am about to sell several of my vintage HP calculators:
- HP 48sx (used, very good condition);
- HP 48g (used, good condition -small scratch on screen-);
- HP 32 SII (used, nearly like new);
- HP 41c tall keys (used, very good condition);
- the never released calculator (X......, new in box).
All are fully functional.
You may email me laurent.joliaferrier(at)free.fr for more information.
===
Subject: HHC 2008: Corvallis Oregon

This year's Annual International Conference of HP Handheld Users (HHC
2008) will be held in Corvallis, Oregon, USA, on the weekend of
September 27-28. Corvallis is where most of the classic HP
calculators were made, and is still the home of many of the big
names in HP history. We're expecting an exceptionally large turnout
this year. It'll be during a busy time if the year (Corvallis is the
home of the state university) so register early before the hotel fills
up.
Further details appear on the ever-expanding official HHC 2008
website:
http://holyjoe.net/hhc2008
-Joe-
===
Subject: Re: HHC 2008: Corvallis Oregon
On 2008-05-29 04:16:44 +1000, Joe Horn <joehorn@holyjoe.net> said:
> This year's Annual International Conference of HP Handheld Users (HHC
> 2008) will be held in Corvallis, Oregon, USA, on the weekend of
> September 27-28. Corvallis is where most of the classic HP
> calculators were made, and is still the home of many of the big
> names in HP history. We're expecting an exceptionally large turnout
> this year. It'll be during a busy time if the year (Corvallis is the
> home of the state university) so register early before the hotel fills
> up.
>
Will Jim Donnelly, Diana Byrne or Bill Wickes attend or anyone of the
old HP team still leaving in Corvallis attend?
--
They who would give up an essential liberty for temporary security,
deserve neither liberty or security (Benjamin Franklin)
===
Subject: Re: HHC 2008: Corvallis Oregon

> This year's Annual International Conference of HP Handheld Users (HHC
> 2008) will be held in Corvallis, Oregon, USA, on the weekend of
> September 27-28. Corvallis is where most of the classic HP
> calculators were made, and is still the home of many of the big
> names in HP history. We're expecting an exceptionally large turnout
> this year. It'll be during a busy time if the year (Corvallis is the
> home of the state university) so register early before the hotel fills
> up.
 Further details appear on the ever-expanding official HHC 2008
> website:http://holyjoe.net/hhc2008
 -Joe-
Yay, I'm moving to Corvallis this weekend to attend Oregon State
University.
===
Subject: Re: I'm looking to replace my 49G
X
> If you try to fail, and succeed, which have you done?
I have failed to succeed...

===
Subject: Re: I'm looking to replace my 49G
One is glad to be of service 
May you and the 35s have a long and happy relationship.
BTW, there's is a lot of info about the 35s on 
http://h20219.www2.hp.com/Hpsub/cache/532210-0-0-225-121.html, HP's