Why Not Emulate a New Calculator Design? (Long Again...)



#37

...To continue from my previous posting, I'd like to suggest an idea. Several people have posted here about how neat it would be to create "an HP-41 (or 48GX or 49G+...) for the new millenium," so to speak; Ben Salinas being the most recent example.

I'd like this as much as anyone, but as I have already said, I see these discussion threads as sad pipedreams, at best. The simple fact is that there seems to be no subset of this group sufficiently motivated to undertake such a task AS AN ACTUAL HARDWARE EXERCISE. For those who can remember, the creation of the PPC ROM so many years ago turned out to be a Herculion-seeming task, which involved a significant portion of the PPC Users Group for MONTHS. And they really only did software, relying on an in-place Hewlett Packard program which handled the actual creation of the hardware modules. (Some will also point out, rightly, that like many such projects, when push came to shove, only a small handful of diehards did the key day-to-day coordination and finally tied up all the loose ends. However, at the outset, a large group of PPC'ers was involved...)

Today's product R&D revolves around suites of virtual design tools which create and allow evaluation and vetting of concepts without ever creating actual hardware. Effects-based simulation, systems engineering and cost analyses can all be applied without ever touching a real double-shot-molded key.

If this is the way "adult" industry operates, why not do the same thing? If no one here is willing to tackle real hardware, why not do the next best thing and create a simulation or emulation of a new calculator design which everyone can get involved in by either contributing concepts, software snippets / aplets or simply doing beta testing? With virtual hardware, no one has to be stuck buying anything or worrying about reconfiguring ASIC's...

One seeming advantage of this approach is that the sim can be built one step at a time (e.g., Work out a tentative packaging and display, then implement arithmatic functions on one data type first, then transcentental functions, then more data types, then RPL progamming, then...). Think of it as a GUI where only some of the keys work at first, then more and more. If the imagined physical form becomes unwieldy or obsolete, it can be revised in an instant. New features or approaches can be added (but with increasing difficulty as the design builds up). If a core group of responsible individuals is ambitious enough, the current version can even be "demo'ed" to outsiders (like HP or their Asian OEM) who might be interested enough to transition it to a real product... ***NOTE: Who would profit from this would DEFINITELY be a problem. Maybe Dave Hicks would have to be trusted to work this out.

Finally, if we all run out of steam, the sim could be left here at MoHPC until another group (or a set of diehards) decided to finish it up. Again, the fact that there is no real hardware would liberate the creation process.

So there it is. I don't imagine that this is a perfect idea, but I think it's a workable way to harness all the pent-up creativity that we (apparently) have here.

UNLESS, of course, someone else has a BETTER idea...???


#38

I liked very much your idea, as I'm planning to do it myself as a hobby project (real hardware, that I build). I think that the concept must be tested on an emulator or as algorithms in some high-level language (platform independent). The implementation then can be build in any platform (as my Sharp PC-1260 soon-to-have FLASH ROM).

I as a computer programmer (assembler of several CPUs, pascal) I will support this idea (in my spare time!).

If we let the hardware platform open and focus on the algorithms it will be easy to build the real thing on any hardware we can get.

I'm in a "crusade" of getting the best algorithms for all functions, using reverse engineering and learning on math papers... I just began, I haven't any "solid" yet.

Best regards,

Nelson


#39


#40

Did I mentioned that I only do it as a *hobby* project on my *spare time*? I don't have any commitment with deadlines on this as I'm doing only to myself to learn more. If I can build on this a good project that could be useful to another people, shure I'll release it.

Regards,
Nelson

#41

Fred,

You can check eBay for my HP41 GUI application which represents an HP41-based GUI Windows app calculator that does matrix calculations, support linear, multiple and polynomial regression, offers many additional functions (Bessel, sine integral, cosine integral, and so on), SOLVE, INTEGRATE (and the last two work with user defined functions) AND does graphics plotting. It took me several weeks to build it, but it's a real simulator that kicks a*s. The GUI part replaces a single-line LCD with multiple list boxes that display the stack, commands, memory registers, flags, and programming steps. You can even single-step through a program using a timer, so you can sit back and watch the program execute in that mode. The simulator supports 1000 memory registers and offers many operations that process data in those registers. In addition, the simulator has 27 predefined matrices that you can dynamically resize. You can perform the usualy matrix operations such as transpose, invert, solve equations, and so on.

Namir


#42

Ok Namir,

could you give me the link, please ?


#43

Gladly!

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=3086897530&category=11713&sspagename=STRK%3AMESSE%3AIT&rd=1

#44

Well, Steve...

This has raised its head several times on this Forum. You're looking at the wrong part of the problem.

After the hardware and firmware are done, we have everything we need except money ;) to build a replicatable, *usable* calc. We have all the "fine" skills in hardware, software, etc. but don't have mfg clout to do the plastics work.

And I'm not talking about marginally useless Palm/WinCE emulations. Close, but no cigar.

There's sufficient resources here (firmware, hardware, math skills) to readily take an HP calc and emulate it - even at the instruction level! - and also add improvements, changes, etc. We can do the PCB, have it stuffed w/parts.

But that's not the real work. 30yrs ago, it was - tremendous silicon constraints, yield problems, programming in a restricted environment was difficult, etc. Nowadays, many of us do that kinda stuff everyday and have more headroom for ROM code, faster CPUs, etc.

But what about a case? Keyswitches? Keytops w/injection molded of keytops? Silkscreen overlay for shifted-key legends? These are cheapie items for a large-scale mfgr, but take big NRE/setup costs upfront.

Let's say we wanna "revive" the HP41CX or 42S. (That's what I want.) We can find an LCD unit that'll work. We can use any of a variety of microcontrollers (H8, Z80, ARM, etc.) to emulate the 42S CPU. Writing this emulation layer - esp using C - is not terribily difficult, and writing an I/O abstraction layer to deal with differences in I/O to new LCD, keypad, etc. is not rocket science either. Several months' collaborative work, but eminently doable.

But I don't think anyone has the capability in his garage to build click keyboards domes, injection mold keytops w/legends, etc. or wants to pay $100K setup for molds/jigs for the case.

Bill Wiese

San Jose CA


#45

Bill, thanks for your comments, but I think you missed a key point in my initial argument. I agree with you completely regarding the difficulty of working a COMPLETE hardware project. The cost of creating molds, dies and other tooling, ESPECIALLY for "small" production runs is simply prohibitive.

That's why I suggested that only the concept be developed; really, just enough to show a perspective manufacturer, like HP or KIMPO (is it??), and let THEM take it from there. You'd license the concept or sell it to them outright (really, only the latter is likely plausible). The big problem, of course, is who would get the money... I have no good suggestions on how that would be resolved unless everyone "donated" their work and all the money was funneled back into the MoHPC website for maintenance / improvements...

Finally, I'm both surprised and intrigued by the people who only view this as an opportunity to "reopen" the 41 or 42S (or whatever) assembly line. To me, ALL of these devices have significant limitations and could all stand dramatic impovements. My concept of a "next generation" calculator is radiacally different from anything that I've seen here... (BUT, I haven't read the entire archive...).


#46

Steve,

It is unlikely that HP or Kinpo or TI will listen to us in the detail that we would want. The calc market is unfortunately driven by the so-called 'educational' market, NCTM (Natl Council of Teachers of Mathematics), etc. The biggest bone that has been/will be thrown to us was the new HP33S that has RPN (but no large ENTER key :( )

What an engineer does with a calc at his desk is of little importance to HP now.

And there's really no 'concept' to license for the calcs most of us here want. RPN with a big ENTER key is 31yrs old now. Calcs have been doing exotic math ops for some time...

Perhaps if we offered guidance/firmware services a Kinpo could integrate an "HP devotees" calc into their production, w/real RPN, etc. But that's a longshot esp given they're a bulk mfgr selling millions of calcs a year and surviving on thin margins.

> I'm both surprised and intrigued by the people who
> only view this as an opportunity to "reopen" the 41
> or 42S (or whatever) assembly line. To me, ALL of
> these devices have significant limitations and could
> all stand dramatic impovements.

The problem is, if you're doing math that requires that kinda horsepower, you've got a PC that runs Excel, Matlab or Maple.

The 41/42 calcs were, in many of our humble opinions ;) the epitome of balance between capability, complexity vs. efficiency and "reasonable" computational power. Any more oomph, I'd use a PC. Keystroke programming was useful for intuitively making a "custom calc" for various functions, not replacing a computer (although in early 70s no doubt HP prog calcs did save some mainframe/minicomputer time...)

A decent calc is a good tool to 'play with numbers' in a small intuitive way. To deal with larger problems, data sets, etc. a computer is the way to go. Fancy calcs like the HP48, TI8X, etc. exist because they're at a cheaper price point - and portable - for students.

I had an HP48 once in the mid-90s. Gave it away after a week. Wasn't worth the grief. Too complex to solve little problems (and no 4-level stack) and too simple for big problems, esp when I have a PC at my desk.

My concern now is that calcs are so 'generic', easily obtainable, cheap, etc. that no one thinks about the answers they provide, and folks trust them implicitly - even though errors do exist. I really do wonder if the cheap Asian calc firmware writers really know the fine art of calculating transcendentals (what you learned about 'em in high school is *not* really useful), error propagation, range reduction, etc.


Bill Wiese

San Jose, CA


#47

Bill, no denying that you make some good points, here. If I were just a touch more pessimistic, I'd probably agree with you line-for-line.

In particular,

>> I'm both surprised and intrigued by the people who > only view this as an opportunity to "reopen" the 41 > or 42S (or whatever) assembly line. To me, ALL of > these devices have significant limitations and could > all stand dramatic impovements. <<

>The problem is, if you're doing math that requires that kinda horsepower, you've got a PC that runs Excel, Matlab or Maple.<

...I actually agree with you completely about the "horsepower" issue. NOBODY should be trying to invert 40x40 (much less 20x20) matrices on a hand-held device! However, I think I disagree on another important aspect of computing devices, which is the user (or man-machine) interface. I think I can imagine better things that could be done than have yet been mass-produced. For example, I think that a dedicated device the size of an LX-100, with function keys on one half and a screen the size of a 3x5 card on the other, where calculations accummulate as you do them (exactly as they would if you were using a pencil and paper) would be... interesting.

You may not agree, but the point is things like bigger displays can facilitate increased functionality...


#48

Actually, I didn't use the 49g+ for a 20x20 matrix, but I did use it for a 5x5 matrix with each cell containing a 3rd degree polynomial. It took it a full 2 minutes just to calculate the determinant and then another 2 minutes to calculate the zeroes of the resulting function. (I was performing a complex Sylvester Matrix problem)

I wouldn't try it again.

-Ben


#49

Ben, wasn't talking about you...

#50

>Finally, I'm both surprised and intrigued by the people >who only view this as an opportunity to "reopen" the 41 or >42S (or whatever) assembly line

For me, it would just be a 15C in the case of 12C, with an alphanumeric display and maybe some more memory. I've seen such a beast somewhere on the Web, at the time when speculations about the 33 where up - a bit like the april fool's page here. Great device! Don't need more.


Edited: 24 Mar 2004, 5:30 p.m.

#51

It seems to me the only thing that can't be implemented relatively easily is the keyboard. For that, just utilize one of H-P's grand old designs. I'm playing (or, should I say, briefly played) with an HP-41CV keyboard soldered to a TI-83+SE. Flash-upgradable electronics with plenty of memory, a computer interface and an IDE, tied to one of the all-time favorite keyboards ever manufactured.

If I ever get serious about putting the things together in a nice package, I'll trim the edges of the 41's front case, and epoxy (or otherwise fasten) the 41's keys to another case -- perhaps the TI's.

Another candidate might be the 28s. Not a lusted-after model, generally available, with more double-shot keytops than you can shake a stick at, implemented as modules that should be fairly readily adapted to a new installation . . .


#52

I might get flamed for this, but how about using the 33s's body. The keyboard itself is actually quite decent (as good, if not better than my 32sii). It has a large, good quality (resonably) LCD screen (though the glare is pretty bad) and if that wasn't enough, it could be replaced. It is fairly simple to open (it uses screws I believe). Plus, best of all, it will soon be available to the masses for fairly cheap. Perhaps it would be as simple as stripping teh CPU and mainboard out of it and building another.

Just a thought

-Ben


#53

I suspect that a homebuilt project will likely require a little bit of extra space, as it will not enjoy the benefits of custom VLSI chips and high-density circuit boards typical of a mass-produced piece of consumer electronics.

And, while I agree that the 33s keyboard is more than adequate, die-hard return-to-RPN types will want a wide Enter key in its proper place. Heck, if you're going to work on something in your spare time, it may as well be something you really want!


#54

No, 33S would be fine size-wize - remember, we're just needing one surface mount QFP microcontroller and maybe some small serial RAMs/EEPROMs... all small & thin.

Only prob w/33s case is "chevron" keyboard and lack of large ENTER key above numeric area...

Bill Wiese

San Jose CA


#55

I am just curious, have you actually used a 33s (like really use for more than 10 minutes)? When I got mine, I had a problem with the keyboard for about a day, but then I got used to it and had trouble going back to the traditional enter key (I am used to the large enter key again). The "Cheveron keyboard" (actually a few of my friends noticed that immediately about it) really isn't that bad, it just takes some getting used to.

-Ben

#56

Hey, if it can be made to fit, then the 12c is the prime candidate. There are TONS of 'em!

I suppose it shouldn't make much difference which of the several versions is used, unless they differ in the amount of room available inside.

So, what would a "custom calculator kit" look like?

 * A small PCB with 1-3 chips & an I/O jack, including
* a minimal firmware loader with I/O functions,
* directions for de-populating the original PCB and
wiring the new one into its keyboard matrix, display and
power connections,
* a hole to drill to mount the serial I/O connector, and
* a web site with a shared code base & documentation.

Where do I get one? (Two!!)


#57

Count on me! I only need a *cheap* 12C to start with... The rest I have, but it can be very slow to develop as I need to make a good PCB project and order some samples on a local PCB maker.

My idea is to replace the PCB, if I can make a PCB good enough to mount the key-domes and display. If not, I'll try to make an "add-on" PCB to replace the chips (get rid of them and make wired conenctions - this can be a difficult task).

The "BIOS" of the machine can be a very basic one, just the basic math functions and a program loader from RS-232 (and plenty of room on FLASH).

I have experience with 7-segmented displays showing characters, it isn't so bad.

Best regards,

Nelson

#58

How about using a HP-12C? It stills in production, there is space inside even to put a SD/MMC slot, and need only a PCB recplacement with the key metalic domes (using the original ones - the new 12C still use them??) and the display. The need for a alphanumeric display can be "downsized" to a 7-segment display capable to show "stylized" characters. I have some microcontrollers suitable to the challenge: a Texas MSP430F499 (yes, Texas...), it has LCD driver, 60KB FLASH ROM, 2K RAM, serial, and capable to direct conection to SD/MMC cards.

Best regards,

Nelson


#59

Yep Nelson that's the way!

I'd be happy with a 12C-style device.

All we'd need are *some* keytops. The MSP430 you mentioned is a GREAT little CPU. The display opening in the case would handle a 16-char dotmatrix LCD of appropriate character size.

Bill


#60

I've toyed with this idea before. I think the best way of handling keys (without injection molding) would be to use epoxy and molds. It's a very common practice amongst model builders, especially in small runs. Keys are really the only great barrier I've found to homebrew calculators. PCBs with the lead sizes needed can be made with a laser printer and an iron. LCD's are cheap. SMCs can be installed easily using a toaster oven. Labels to replace the original 12c could likely be made using the same epoxy techniques as the keys, or laser engraving could do a fine job as well.

Sorry for such a rant on my first post here, but this is a very exciting and do-able possibility.

One other idea I've had for the case would be the use of sandwiched layers of sheet metal assembled in much the same way as an old padlock. It's cheap, durable, and will allow the cases to be made thinner than plastic allows.

If anyone is interested in actually organizing and making this happen please reply.


#61

I'm interested, but not in full time ;)

I'm searching on eBay for a "candidate" 12C for the "surgery"...

I have the processor (MSP430F449) with segment-LCD controller, but it has only 60KB of ROM FLASH. I need only to build one prototype PCB to try this on (and exercise my MSP430 assembler). Oh, I'll need to build a JTAG interface for it.

This is very "do-able" indeed, is just a matter of spend some time (and money) on the idea.

If I cannot build myself a prototype PCB and JTAG I'll try to buy them (from Olimex?), up to US$100 (if I can reserve such amount!:).

The operating system of it will be a good team challenge.

Anyway, I'm in!

Best regards,

Nelson

#62

Hey, if the development tools are available at low cost, I'd love to play around in a calculator playground.

If the experts make the crucial component decisions, I'll lend a hand in putting the software together.

#63

Sounds good. You got me in also

-ben


#64

We can start with some hardware-relative issues, can we discuss this here or another place to avoid the "pollution" in this forum?

You guys have my email, I can send my ICQ and MSN contacts by request.

Best regards,

Nelson


#65

It would probably be best to find/setup a place to start development. First, we'll probably want to decide on a format, what hardware to base it on, and any mandatory specifications. I've already begun research into producing epoxy resin based, triple injected keys. There are some alternative epoxy types that may be suitable for fashioning cases.

I have a friend who probably can offer some web space at least while we get started.


#66

We need to exchange some ideas to define the hardware platform. I received an email from Paul Brogger and I don't know if he received the reply due some email server problem (my email is hosted on my own server and is blocked on some ISPs thinking it is spam!).

I have several ideas to start with.

I was thinking in call this project "OpenCalc", but this name is already used in some Linux Calculadors as well... :(

Best regards,

Nelson


#67

Should have a basic website up sometime next week. I think you will all be pleasantly suprised with my solutions for building a case, molded keys, and producing the pcb for very little $$$. (and it looks about like a panel out of an apollo command module) No point in spending $100 on a pcb prototype when you can make it for under $10 yourself.

In the meantime, everyone start brainstorming about what you want in a calc.

-Hugh

#68

Hi,

I'd vote for using a Pioneer case and keyboard for such a project.

They have some advantages:
They are still easily available in huge amounts, have a very aestethical and classical design, have the ENTER key where it belongs, are of the right quality of keyboard & case (even the ID models...), have the right form factor (the 33S is too BIG), and maybe some more I don't recall at the moment.
And generally: Why should someone want to design a new calc if it's not like a classic HP calc with the ENTER bar? 'Normal' calculators are available in nearly any shape (see 33S) and for various tastes.

Raymond

#69

Steve said, "I'd like this as much as anyone, but as I have already said, I see these discussion threads as sad pipedreams, at best. The simple fact is that there seems to be no subset of this group sufficiently motivated to undertake such a task AS AN ACTUAL HARDWARE EXERCISE."

That kind of hardware is probably too big an undertaking for a single person or small group of enthusiasts to do-- _IF_ you want it to look like the HP's of 20 years ago.

It wouldn't have to look like HP did it though. The size of the job drops dramatically if you use off-the-shelf LCDs
available from any number of manufacturers, off-the-shelf hand-held cases like Pac-Tec and others make (and offer custom milling on), and so on. We don't need to spend $100,000 of tooling to make a funcional unit that offers the possibilities desired for the next couple of decades.

For myself, a laptop is much too big. It has to fit in a corner of the atache case. And if it doesn't have HPIL or USB or something like it with available interface converters to interface to a lot of lab equipment at once, I'm not interested. Emulators don't cut it. I have no use for a supercalculator. I personally want a controller.

I develop products for our company all the time, so I have an idea of what's involved. We definitely are not at the HP
product development budget level though. I haven't done much digital design either. I have made small custom workbench computers and used them heavily, and am seriously thinking about making a hand-held, nearly-pocket-sized one as a product. It's definitely for a different market, although partly inspired by the HP-41 and 71.

I initiall posted this yesterday and the system dropped it again. Fortunately I learned from previous experience and kept a copy. Since then someone else said "Why don't we just ask HP to make it again?" That probably makes the most sense. It appeared to me that HP management started to change a lot in the late 80's, and they made some really stupid market decisions. It might be good to rub their noses in it and bring them back to their senses.


#70

Garth...

It's not a worthwhile exercise at this point because the hardware is simple (a cheap microcontroller, PCB, LCD) but the final realization thereof wouldn't have a good KB or case that fits the hand. The software is easy to do, even emulating an existing HP architecture.

I agree that a little prog calc w/decent I/O could be quite useful in a lab setting.


Bill


#71

The type of keyboards HP had in those years truly are expensive to tool up for. Other comparatively low-NRE options might include rubber keypads (a couple grand NRE, and $.25/ea after that) and membrane switches with tactile feel and raised edges (unlike the flat one of the Timex/Sinclair ZX-1000).

The off-the-shelf case would not look nearly as nice as say an HP-41, but there's hardly any reason it really has to be any bigger although the battery compartment might make it slightly thicker. The only thing I'm not sure how you'd implement would be the ports and the covers for ports not being used.

To answer someone else's post, a home-made calculator with something like wire-wrap would indeed be very thick; but done correctly, I was surprised to find how really easy it is to solder SOIC's and QFP's with really fine pitches, even using a soldering iron that covers several pins at once. My first try netted a result that looked like it was done by machine.


Original message: -----------------

It's not a worthwhile exercise at this point because the hardware is simple (a cheap microcontroller, PCB, LCD) but the final realization thereof wouldn't have a good KB or case that fits the hand. The software is easy to do, even emulating an existing HP architecture.

I agree that a little prog calc w/decent I/O could be quite useful in a lab setting.


#72

Garth,

It's not too hard soldering fine-pitch surface mount ICs on to a PCB using a heat gun and alum. foil/sheetmetal "horn" air director, along w/appropriate SMT paste.

Remember this calc is gonna only require one large-pin-count chip, maybe (not necessarily) one or two 8-pin SMT "microDIPs" for serial EEPROM/SRAM.

Bill Wiese

San Jose CA


Possibly Related Threads...
Thread Author Replies Views Last Post
  HP-80 History, Design and Interesting Facts BShoring 1 509 11-30-2013, 08:50 AM
Last Post: Xavier A. (Brazil)
  HP Prime: Long integers (continued) Helge Gabert 2 598 11-07-2013, 11:24 AM
Last Post: Helge Gabert
  HP Prime: Pass "Long" Integers to a Program Helge Gabert 6 1,030 11-03-2013, 01:12 PM
Last Post: Helge Gabert
  HP Prime polynomial long division bluesun08 13 1,509 10-30-2013, 03:29 AM
Last Post: parisse
  A very long HP-17BII equation Gerson W. Barbosa 22 2,372 04-19-2013, 12:37 AM
Last Post: Gerson W. Barbosa
  A long WP-34S night Siegfried (Austria) 10 1,307 04-16-2013, 02:11 AM
Last Post: Siegfried (Austria)
  HP-25 left on for a long, long, while Matt Agajanian 12 1,366 04-10-2013, 11:33 PM
Last Post: Steve Leibson
  AME Design Port Extender for the HP41 Dane Spillers 2 562 03-14-2013, 07:58 PM
Last Post: dane spillers
  RPL long vs. short names peacecalc 5 825 10-30-2012, 01:25 PM
Last Post: peacecalc
  HP41 Long term storage Steve Hunt 6 819 09-27-2012, 11:08 PM
Last Post: db (martinez, ca.)

Forum Jump: