Create-a-Calculator Contest
Marino Arturo
anywhere.
Thanks
Marturo
Wayne Brown
> Did HP publish the winners of this contest? I couldn't find the information
> anywhere.
Here's a link to the beginning of a discussion that gives some information
about it:
http://www.hpmuseum.org/cgi-sys/cgiwrap/hpmuseum/archv013.cgi?read=40123
--
Wayne Brown | "When your tail's in a crack, you improvise
fwb...@bellsouth.net | if you're good enough. Otherwise you give
| your pelt to the trapper."
"e^(i*pi) = -1" -- Euler | -- John Myers Myers, "Silverlock"
Joshua J. Belsky
> Did HP publish the winners of this contest? I couldn't find the information
> anywhere.
Bah. Who cares? :)
Not that the c.s.hp48 contest to create a calculator that took place some
years ago was any better. We ended up with an idea for a $50,000 PDA.
--
-Joshua Belsky
jjbe...@yahoo.com
http://belsky.net
Zip Disk
news:vkpo0a6...@corp.supernews.com...
> Marino Arturo <mar...@comcast.net> wrote:
> > Did HP publish the winners of this contest? I couldn't find the
information
> > anywhere.
>
> Bah. Who cares? :)
>
> Not that the c.s.hp48 contest to create a calculator that took place some
> years ago was any better. We ended up with an idea for a $50,000 PDA.
How about some rules. 4-8 line DM HQ display with its own processor & ram.
Processor - open, but a 68K or ARM sounds good. 512KB of STATIC ram above
and beyond anything the OS needs. Get the 41/42S ROMs from you know who.
Beg, whatever it takes. Code for the processor of choice in 'light'
assembler - NO 400 instruction macros!! Make sure it works __exactly__ like
the 41/42S in terms of mathematic algorithm input/output. Do something nice
for user presentation I/O. Dump the classic HP keyboards in favor of one
like the Rockwell 63R - very nice positive action but was easy enough that
allowed touch typing. USB 2 even if it's PIO. Easy Assembler/ML access to
the display controller, KB and USB port. Find a manufacturer who cares about
what they make. Where's not important. Anything I forgot?
Veli-Pekka Nousiainen
news:yrb3b.1725$xD2...@fe01.atl2.webusenet.com...
http://www.hpmuseum.org/software/swcd.htm
"HP-71B Internal Design Spec"
That kind of thing, maybe even:
"HP-71B Hardware Internal Design Specification"
Naturally for *any* new 42S or 49G
VPN
pkhuong
OT comment: By "STATIC RAM", do you mean SRAM or non-volatile memory?
I was reading the winning entry for the design a calc contest, and it
doesn't seem that far-fetched - it's not even a graphing calculator.
How about trying(eh) to make something like it that is somewhat
plausible for a community-driven project? ARM-based boards were pretty
darn expensive last time i checked - can't remember for 68k.
AVRCard:
http://www.avrcard.com/Content/Products/avrcard.htm
EUR 99 each
gcc+libc (and a fp library) were ported to it.
87x54 mm
16 MHz 8-bit RISC (16 MIPS)
32x8 registers
4KB On-Chip SRAM
128 KB On-Chip Flash
4 KB On-Chip EEPROM
32 KB non-volatile RAM (SRAM would have been cool)
Battery power supply(2-5V) on the card
I2C
1 RS-232
1 RS-232/485
Can be programmed through the RS-232
I believe there's a clock (can't remember where i read it)
Not a math machine though, and no FPU(div can take anywhere from 56 to
618 ms fe), but an accessible one.
LCD:
http://www.matrixorbital.com
Has many graphical and character serial(and I2C) LCDs. Many of them
also support connecting a matrix keypad directly through the LCD,
saving a port.
For example:
40x4 characters LCD with interface to a 55 keys matrix
keyboard(http://www.matrixorbital.com/pages/product_view.asp?CatID=1&ProductID=10)
USD125
40x4 LCD with interface to an AT keyboard (i kid you not!
http://www.matrixorbital.com/pages/product_view.asp?CatID=1&ProductID=11)
USD130
Backlight adds ~USD 20. They're both 190mm(L) x 54.0mm(W) x 27.0mm(T).
There are smaller, less expensive models, but they all support ~25
keys keyboards. At the other end of the price spectrum are the graphic
LCDs(122x32 or 240x64, but as i said, the winning entry wasn't
graphing), but they only support 25 keys if any.
So far, there's everything but the batteries and the keypad for ~230
USD, off the shelf, not in bulk :) If i knew where to find an RS-232
or I2C keypad, we could cut that to ~160 USD(a 20x4 LCD, 98x60 mm) +
keypad.
Interesting... I wonder what kind of margin HP had on the 49G :)
Paul Khuong
Veli-Pekka Nousiainen
news:51184814.03082...@posting.google.com...
> "Zip Disk" <zi...@mail1.microsoft.com> wrote in message
news:<yrb3b.1725$xD2...@fe01.atl2.webusenet.com>...
> > "Joshua J. Belsky" <bel...@brionne.cyberverse.com> wrote in message
> > news:vkpo0a6...@corp.supernews.com...
> > > Marino Arturo <mar...@comcast.net> wrote:
> AVRCard:
> http://www.avrcard.com/Content/Products/avrcard.htm
> EUR 99 each
> gcc+libc (and a fp library) were ported to it.
> 87x54 mm
> 16 MHz 8-bit RISC (16 MIPS)
> 32x8 registers
> 4KB On-Chip SRAM
> 128 KB On-Chip Flash
> 4 KB On-Chip EEPROM
> 32 KB non-volatile RAM (SRAM would have been cool)
> Battery power supply(2-5V) on the card
> I2C
> 1 RS-232
> 1 RS-232/485
> Can be programmed through the RS-232
> I believe there's a clock (can't remember where i read it)
>
> Not a math machine though, and no FPU(div can take anywhere from 56 to
> 618 ms fe), but an accessible one.
With 22-bit Program COunter and (MB address space it sounds promising.
The 128KB alternative migth just fit in for a 42SII
with 64KB RAM and 64KB ROM (or rather 32KB+96KB)
but I'm not sure about the 16MHz and short registers
It might be slow...
VPN
Zip Disk
I was referring to good old CMOS Static RAM which is non-volatile for years
with the help of a coin cell. Costs a little more but it's worth every
penny.
I'd guess that it's HUGE. With the expertise on this newsgroup alone, I
think we could easily outdo HP. Most any sensible algorithm questions posted
to math newsgroups return lots of excellent feedback. With flash ROMs we
wouldn't need to get it exactly right the first time. Hell, HP never did.
The trick is to get it manufactured. Guess who's hungry for that sort of
business? China? Before anyone gets all cranked up about Chinese quality,
let me point out that junk does come from China and the USA and Canada and,
well you get the drift. OTH, I have a Chinese made Grundig shortwave radio
that is very state of the art - SMC on both sides of the PCB. Very nicely
done. Much more expensive in parts alone than what we're taking about, yet
the price was about USD$250.00. You've pointed out many examples of what's
available. Anyone interested? BTW, thanks for the time it took to look all
that stuff up!
pkhuong
The main problem is that the software only supports 32 bits floats,
and that the only coprocessors i've seen are also 32 bits; imho,
doubles are absolutely needed for a scientific calculator. As for
speed, well, it can easily be "overclocked" by using another crystal -
someone reported not having any problem at 22 MHz. This
(http://www.forth.com/Content/Products/SwiftX/SXbenchABS.htm) tells us
that, at 16 MHz (the 90S8515 has the same IS, but runs at 8 MIPS max),
the mega128 should be something like ~3-10x slower than a 68332.
Unfortunately, the 68332 boards linked by FORTH Inc are all in the
USD150-250 range. Otherwise, they are interesting: decent size,
memory, etc.
Oh, I found another board in my junk'o links (using the same chip)
http://www.bdmicro.com/mavric/
USD65 (kit). 3x3.5" (75x90 mm), has 128KB external SRAM(2x64 KB
banks), 2 RS-232, etc.
In case it wasn't obvious, i'm very much interested in building my own
"pocket" RPN calculator, even if the interface is 2 4x4 keypads (one
nibble each) and a 2x20 LCD :)
Paul Khuong
Joshua J. Belsky
> The main problem is that the software only supports 32 bits floats,
> and that the only coprocessors i've seen are also 32 bits; imho,
> doubles are absolutely needed for a scientific calculator. As for
Why are doubles needed? Forgive me, but what scientific field needs
that level of precision?
I'm sure there is one, but I don't know of it. Enlighten me.
(This is sarcasm if there aren't any, and genuine curiosity if there are :)
Marino Arturo
"Joshua J. Belsky" <bel...@brionne.cyberverse.com> wrote in message
news:vkpo0a6...@corp.supernews.com...
> > Did HP publish the winners of this contest? I couldn't find the
information
> > anywhere.
>
> Bah. Who cares? :)
>
> Not that the c.s.hp48 contest to create a calculator that took place some
> years ago was any better. We ended up with an idea for a $50,000 PDA.
Well, I care, if you don't have anything to say don't say it...
Jean-Yves Avenard
Belsky" <bel...@brionne.cyberverse.com> wrote:
> Why are doubles needed? Forgive me, but what scientific field needs
> that level of precision?
I don't think that's the main issue.
64 points integers/float can speed up the calculator. All calculators work
with a precision higher than 32 bits.
However, doubling the precision from 32 bits to 64 bits is quite easy.
Jean-Yves
Demeter
> Why are doubles needed? Forgive me, but what scientific field needs
> that level of precision?
>
> I'm sure there is one, but I don't know of it. Enlighten me.
You have got to be kidding, right?
Imagine multiplying a pair of numbers both in the 10 to the
minus 12 range (10e-12)and then take the square root and then
the reciprocal and then repeat for 1000 more pairs
and then add them all up! (cumulative error) or even worse try
an 8th order low pass filter in the several MHz range where the
coefficients are of the order of 10e35 and try and do some more manipulation
on these to find solutions to the nonlinear system that comes
about!
These are all everyday cases in analog electronics!
Take care and be careful.
!Demeter!
pkhuong
> pkhuong <pvk_g...@pvk.ca> wrote:
> > The main problem is that the software only supports 32 bits floats,
> > and that the only coprocessors i've seen are also 32 bits; imho,
> > doubles are absolutely needed for a scientific calculator. As for
>
> Why are doubles needed? Forgive me, but what scientific field needs
> that level of precision?
>
> I'm sure there is one, but I don't know of it. Enlighten me.
>
> (This is sarcasm if there aren't any, and genuine curiosity if there are :)
precision(2^23). I don't know about you, but MY calculator handled
E-99 to E99 externally. Of course, you might say it [10^ą99] is
useless, which brings the question of WHY they ever included this in
all the scientific calculators i used, then :/ Also, keep in mind that
we need higher internal precision than external. Here's a quick
example: I often need to calculate permutations, etc. 60! is already
in the E81 range(and mantissa too small to be useful). Anyway, I'd
obviously rather have a float machine than no machine. Moreover, i
think the TI(nor the HP?) calcs don't have hardware FP either.
* the coprocessors i've seen were all from
http://www.al-williams.com/pak1.htm but i think we'd end up doing a
LOT of software arithmetic even if we used one
BTW, I put a "quick", non-exhaustive list of boards here
http://pvk.ca/personnal/cshp48/boards.htm . The only boards that seem
easily usable are the two ATMEGA128, the uFlashTCP, and the ARM/586
Core (maybe the minipod). All the others don't have RS232 connectors
built-in, though the controllers are (I don't like the idea of
soldering chips+headers). I've seen several cheap 6502(and 8051)-based
boards, but the 6502 seems weak, and i don't know the 8051; i'll have
to ask a friend who's done a lot of 6502 assembly for his opinion.
Paul
Veli-Pekka Nousiainen
news:BB793D5B.12F31%m...@privacy.net...
Hi, Jean-Yves
Maybe it is - for you!
Is it easy to do the Extended Presision Real/Complex repecting DIGITS
and this using native ARM9 instructions to the same level of accuracy
than the BCD algorithms used in P calculators?
I wonder if anybody reading these new could do it.
First to +, - , *, / , then SIN, etc & EXP/LN, etc...
???
VPN - who can program some UserRPL only
pkhuong
something like:
Multiplication/division: trivial :) - just need to make sure to shift
the mantissa so it doesn't overflow or lose precision(i guess it might
be skipped if you don't care too much about internal accuracy)
Addition/substraction: shift the mantissa to put them on the same
exponent, and then add them together.
SIN, EXP, LN (etc), i think, are approximated with series.
For arbitrarily large numbers, it's much slower, but i doubt you'd
have to use BCD if you don't have BCD instructions readily available.