HP-12C Platinum Version 02 calculator from 2010 week 21?

[jebem]

I saw it advertized in a local Internet site. Asking price was 20 Euros, fixed price. Just offering the calculator and the pouch, both in good working condition.
So yesterday I met the seller and made the trade.

This is my first HP-12C to join my collection, and despite this model belonging to the new manufacturing processes and materials, I'm glad to have acquired it as it looks almost classy as the original models and it works like a charm.

I understand the 12C is the most successful HP calculator of all, and this MoHPC forum has plenty of posts discussing it along the years.
However, for me it is not 100% clear on exactly how many different hardware/firmware versions did HP released along the years.

To increase my doubts, the HP site keeps mentioning that the Platinum uses only one CR2032 3Volt cell (in the specs and on the available user guide).
But mine uses two CR2032 cells in parallel.

The official HP support pages refers to the self test call using the g+ENTER+ON combination where the self test menu has only 3 options: 1.L 2.C 3.H (1.LCD 2.Checksum and Copyright 3. KB test).
But mine has got 4 options in the self test menu: 1.L 2.C 3.H 4.Cpu (1.LCD 2.Checksum, Version and Copyright 3. KB test 4.CPU speed check).
From the option 2 results, I get "Che - - 564Ah", "Ver 02" , "C 2004 hP", so it looks like it is a Version 02 Platinum, ROM checksum 564Ah, Copyright 2004 by HP.

After hours digging the Internet, I managed to get a couple of HP-12C Platinum User Guides from alternative sites, mentioning the two CR2032 cells in the Platinum.
So, it seems that there are at least two different official HP-12C Platinum versions.

Now, what kind of CPU is this model using?
Some sites and posts mentions that if the calculator uses two CR2032 cells it is based on a Atmel AT91SAM7L128 ARM cpu, while others mention that it uses the Sunplus SPLB31A 8502 cpu.
All this information is incomplete, as it omits the Platinum version and manufacturing years on their claim.

Checking the this thread in the HP forum, I have run the described benchmarks used by the poster to explain his issue with his two HP12C calculator generations.

Here are my results:

For Scenario1
Poster HP12C Platinum: 120 n, 1.28 pmt, 103 chs pv, i. ans .7128 @ 22 secs
Poster HP12C: 120 n, 1.28 pmt, 103 chs pv, i. ans: .7128 @ 12 secs
My HP-12C Platinum Version 02: about 2 seconds.

For Scenario2
Poster HP12C Platinum: 10 n, 50 pmt, 1100 chs pv, 1000 FV, i. ans: 3.7805 @ 40 secs
Poster HP12C: 10 n, 50 pmt, 1100 chs pv, 1000 FV, i. ans: 3.7805 @ 12 secs
My HP-12C Platinum Version 02: about 2 seconds.

Correction -Original text:
From the results, I'm inclined to believe that my calculator model version 02 is using a very fast processor, probably the alleged ARM cpu refereed by Tim in the above link.
After some additional testing (see posts below for details), it appears that this machine is not using the ARM processor after all.

Does anyone have more confirmed information on this subject?
I'm not in the mood to open my calculator just to check the processor reference, but eventually I will do it to satisfy my curiosity.




Self Test procedure 1:
Power Off. Press g + ENTER + ON; Release g + ENTER; Release ON: Displays "1.L 2.C 3.H 4.Cpu"



Press 1 (LCD test): Displays all LCD segments ON.



Press 2 (System test): Displays "Che - - 564Ah" (ROM Checksum?)



Press ENTER: Displays "Ver 02" (Platinum Version 02?)



Press ENTER: Displays C 2004 hP" (Copyright 2004 HP?)



Press 3 (Keyboard test): Image not available.
Displays all all LCD segments ON and will shut down them when pressing the corresponding keys)

Press 4 (CPU speed): Displays "3941092 h2" (CPU speed of 3.94 MHz ?)
The value will change in the last 3 digits between test runs.



Self Test procedure 2:
Power Off. Press / + ENTER; Release / ; Release ON (Keyboard test):
Press all the keys from left to right and from top to bottom.
At the end will display "12".



Self Test procedure 3:
Power Off. Press x + ENTER; Release x ; Release ON (System test):
It runs for about 25 seconds and then shows all LCD segments ON.

[jebem]

Additional photos.

[rprosperi]

(07-09-2015 10:26 PM)jebem Wrote:  I understand the 12C is the most successful HP calculator of all, and this MoHPC forum has plenty of posts discussing it along the years.
However, for me it is not 100% clear on exactly how many different hardware/firmware versions did HP released along the years.

This has never been clearly and fully documented, to some extent, because the history of the 12C is far longer than any single person in the HP Calculator group.

There are at least 4 versions of the 12CP (orig silver, late silver, orig black w/1 2032, later black w/2 2032), but I suspect there are many more.

For diags, I think the 15CLE manual (download from HP) has the most accurate description of the diags (in an appendix) in the new Voyagers, but this may be the same as some you've found. It's easy to check.

Also the most recent 12C and 12CP models used the same CPU as other models recently discontinued (e.g. 20b/30b) due to lack of parts from CPU supplier, from which we can infer that another "version" is coming. And while HP's intent is to keep differences invisible to customers, as you've seen with the benchmarks, careful scrutiny by communities like MoHPC will eventually find minor differences. There is MUCH detailed info in postings here (even more in the "old" forum); it would not be a simple task to comb it all and compile a summary, but I, for one, would love to see such a comparison/summary.

[jebem]

(07-09-2015 11:39 PM)rprosperi Wrote:  This has never been clearly and fully documented, to some extent, because the history of the 12C is far longer than any single person in the HP Calculator group.

Also the most recent 12C and 12CP models used the same CPU as other models recently discontinued (e.g. 20b/30b) due to lack of parts from CPU supplier, from which we can infer that another "version" is coming. And while HP's intent is to keep differences invisible to customers, as you've seen with the benchmarks, careful scrutiny by communities like MoHPC will eventually find minor differences.

Thank you, Rob!

I keep reading the old forums and found some additional interesting reports.

In this post (Msg#18), Gene Wright suggests a small looping program to measure how fast the processor is.

Enter this looping program. Press:
shift P/R
shift CLPRGM
+
shift GTO 001
shift P/R
1 ENTER ENTER ENTER ENTER CLX
shift GTO 000
R/S and let it run for 60 seconds. The number in the display is the important one.

My calculator gives me a total count of about 1400.
So it doesn't seems to be using the faster ARM processor, or if it is, it is emulating the original 12C speed, as others reported about the same final count on their machines.
Probably HP used the Sunplus processor in these machines produced in 2010, despite using two CR2032 cells in the battery compartment.

I really have to open this machine one of these days to clarify this issue.




In another post here (Msg#28), Katie has taken current consumption readings on a HP-12C, alleged using the faster ARM processor.

Some of the Katie's readings in April-2009, were:
power off: 4uA
power on, static display, no keystrokes: 45uA
tight program loop: 15mA

On my machine of 2010, the readings are different:
Power off: 0uA
Power on, static display, no keystrokes: 10uA
Tight program loop: 1.4mA

Also on this same post (Msg#22), Gene reports how fast is this new calculator they were testing at the time (April-2009) by using the tight loop test:
Program of
01 + 02 GTO 01
with the stack filled with 0 in X and 1 in Y, Z and T
counts to well over 45,000 in 60 seconds.

This final count of 45000 means that that calculator of 2009 was about 32 times faster than my calculator of 2010, and to do that it was consuming about 11 times more current than mine.


Here are the photos of my HP-12C Platinum calculator of 2010 while taking current readings:

Powered ON, Idle state: 10uA




Powered Off: 0uA




Running the looping program from above: 1.4mA

[rprosperi]

(07-10-2015 06:43 PM)jebem Wrote:  Thank you, Rob!

I keep reading the old forums and found some additional interesting reports.

In this post (Msg#18), Gene Wright suggests a small looping program to measure how fast the processor is.

Enter this looping program. Press:
shift P/R
shift CLPRGM
+
shift GTO 001
shift P/R
1 ENTER ENTER ENTER ENTER CLX
shift GTO 000
R/S and let it run for 60 seconds. The number in the display is the important one.

My calculator gives me a total count of about 1400.

IIRC, the latest 12C (non-platinum) scores about 70,000 on that test, so it seems these results could imply the current 12CP is running older code in an emulator, while the 12C is running code ported to the new processor.

Another interesting data point would be what a 15CLE does for the same benchmark. The 15's better program environment means slight changes to the simple test code, so results will be a bit off, but the gap between 1400 and 70000 is pretty large and it should be obvious which way that went (emulated or ported). Given the amazing speed claims I've read for the 15CLE, I would guess the latter, which also could explain the bugs related to Pausing.

Anyone got a 15CLE that's not still sealed, and willing to run this test?

[BobVA]

(07-10-2015 06:59 PM)rprosperi Wrote:  Anyone got a 15CLE that's not still sealed, and willing to run this test?

Code:

LBL 1
+
GTO 1

Gives 50,590

and

Code:

+
GTO I    (note I = -1)

Interestingly is a bit slower at 49,913. I would have thought the direct line branching would be faster, but perhaps the label search is optimized to ignore "empty" program memory, or the use of indirect addressing has a lot of overhead.

[Dave Frederickson]

(07-10-2015 06:59 PM)rprosperi Wrote:  Anyone got a 15CLE that's not still sealed, and willing to run this test?

12CPS: 1385
12CAE: 75,400
15CLE: 49,300 (Remember when these cost $99! )

[rprosperi]

Thank you Dave and Bob for taking the time to run the test and report results; these all seem consistent by model, but the difference between current 12C's and 15CLE (30% slower) is definitely interesting. I'd speculate (sounds better than guess...) that this is due to the more complex programming model in the 15C design, maybe calculating the offset to jump back each time, while the 12C just jumps to a known offset, or something along those lines.

Would you have guessed the 12C outperforms a 15C?

[Dave - is the 12CPS 1 or 2 2032?]

[Dave Frederickson]

(07-11-2015 02:21 AM)rprosperi Wrote:  Dave - is the 12CPS 1 or 2 2032?

Single 2032.

[BobVA]

(07-11-2015 02:21 AM)rprosperi Wrote:  ... I'd speculate (sounds better than guess...) that this is due to the more complex programming model in the 15C design, maybe calculating the offset to jump back each time, while the 12C just jumps to a known offset, or something along those lines.

I'd speculate your speculation is correct :-)

Either of the ways to do the branch in the 15 has some overhead. Label searches proceed forward in program memory - the wrong way in this case - and indirect branching to a line number involves a register load and whatever complexity is involved in converting that floating point value into a line number. Both seem a lot more complicated than just resetting the program counter in the 12.

[matthiaspaul]

(07-09-2015 11:39 PM)rprosperi Wrote:  Also the most recent 12C and 12CP models used the same CPU as other models recently discontinued (e.g. 20b/30b) due to lack of parts from CPU supplier, from which we can infer that another "version" is coming. And while HP's intent is to keep differences invisible to customers, as you've seen with the benchmarks, careful scrutiny by communities like MoHPC will eventually find minor differences.

While recent revisions of the 12C series (including the 12c 30th Anniversary Edition) since ca. 2008 are based on an Atmel AT91SAM7L128 processor with ARM7TDMI core running the original HP Nut firmware in an emulator, to the best of my information, this does not hold true for the 12C Platinum series (including the 12c Prestige and the HP 12c Platinum 25th Anniversary Edition). AFAIK, recent versions of this calculator are based on a Generalplus GPLB31A processor with 8502 core, a completely different design.

The Atmel AT91SAM7L128 is no longer in production and if HP will continue to offer a HP 12C in the future, they are forced to move the emulator to yet another platform (if they haven't already).

AFAIK, the Generalplus GPLB31A appears to be still in production, therefore the 12c Platinum series is not affected by this. But since the 12c Platinum firmware, which appears to be a functional reimplementation rather than an old firmware running in an emulator, is buggy and apparently it is not under HP's control to fix it, it might be undesirable to continue this series for other reasons.

Some months ago, I tried to compile the information into the corresponding article at Wikipedia:

http://en.wikipedia.org/wiki/HP-12C

If you find omissions or mistakes, I would like to encourage you to fix them in the article.

Greetings,

Matthias

[BartDB]

I remember this discussion when the first 2-cell Platinum appeared just after the faster 12C appeared. As far as I recall the Platinum does not have the faster processor.

I think it just uses a common back for ease and cost of production.

[Tim Wessman]

(08-07-2015 08:53 AM)BartDB Wrote:  I think it just uses a common back for ease and cost of production.

I think that is probably the case as well. While I was not around back then, being able to reduce cycle times from switching molds, reducing maintenance to a single set of tools, and reduction of storage costs when not in use are always goals manufactures have to reduce costs. Moving very similar products to similar production processes and components is always a goal.

[rprosperi]

(08-07-2015 07:45 AM)matthiaspaul Wrote:  AFAIK, recent versions of this calculator are based on a Generalplus GPLB31A processor with 8502 core, a completely different design.

You are quite likely correct and I misspoke by including 12CP in that comment.

Also note that there are significant differences in the performance (and therefore, likely the s/w internals) between the original (no parenthesis) and later versions of the 12CP; the original is notably faster, at least for some benchmarks. Speculation is ports by different subcontractors, when the original CP was "improved" to add the parenthesis and other functions.

Nice job on the Wikipedia page BTW, thanks for doing it and sharing here.

[Gene]

Why is the link to the 12c solutions book on the wiki page going to the RPN only version of the solutions book?

Real question is where is the HP link for the updated version showing solutions in RPN and algebraic that several of us created a few years ago?

That's the version that should be on HP's site and certainly the one linked to on the wiki page.

[Didier Lachieze]

(08-07-2015 03:53 PM)Gene Wrote:  Real question is where is the HP link for the updated version showing solutions in RPN and algebraic that several of us created a few years ago?

it's here: HP 12C Platinum Solutions Handbook

[Gene]

Aha! :-) That's what the link should be to on the wiki page.

tyvm

[matthiaspaul]

The article currently links to

http://www.hp.com/ctg/Manual/c00367123.pdf

which appears to be the Solutions Handbook for the Platinum series.

The link suggested above is:

http://h10032.www1.hp.com/ctg/Manual/c00367123.pdf

Isn't this the same file (both show "rev. 03.04")?

Greetings,

Matthias

[Gene]

The 12c solutions handbook is the one I was referring to. It is unneeded.

The 12c platinum solutions book has RPN solutions in addition to algebraic solutions. It also includes a couple extras that were not in the original 12c version.

I see the platinum version now, but I seem to think it wasn't there when I posted my comment. :-)

However, the original 12c solutions is redundant. Everything it contains is in the platinum solutions book and has been updated with better fonts, formatting, etc. :-)