HP-32E with errors ...

[ThomasF]

Hi,

I recently got a good looking but dead HP-32E, and after some cleanup and repair of the battery connector, it jumped back into life again!

But, not completely ... there are some strange behaviors that I can't explain.

First of all, when just testing the calculator it seems to work fine, but my "testsuite" with "45 SIN COS TAN ATAN ACOS ASIN" failed, and I realized that ASIN and ACOS doesn't work!
E.g. "45 SIN COS" results in "0.9999" which is ok, but then ACOS gives "Error 0".
And "45 COS" gives "0.7071" (again ok) but then ACOS gives "90.0000" (should have been 45)!
I also noticed that "SINH" always gives "0.0000" as result, but more or less everything else works as expected.

Running the selftest (STO ENTER) gives "Error 9" (leaving the number "1" in x-reg), so something is fishy ...

Anyone have an idea on what is going on here?
If there was a bad connection or shortcut somewhere, I would expect even more to fail, but how could only functions like ASIN, ACOS and SINH fail, but nothing else ... ?
A logical explanation could be that the ROM is broken (i.e. a few words/bits in a common routine for ASIN, ACOS and SINH that fails), if so, I guess that replacing the rom-chip is the only cure ...

Anyone have any thoughts or ideas?

Cheers,
Thomas

[aurelio]

Hi Thomas, actually a strange behaviour..
Is your calculator a "unsoldered chips" one?
Maybe the failure could be due to a bad contact

[ThomasF]

(04-28-2021 10:55 AM)aurelio Wrote:  Is your calculator a "unsoldered chips" one?

Hi Aurelio,

No, it is a soldered device, and it was clean inside.
It was only the battery contacts and flex-cable that was broken due to leaking battery. At least I couldn't see any "residue" on the cpu-board etc.

Cheers
Thomas

[rprosperi]

You'll have to wait for Panamatik to weigh-in for a definitive answer, but it appears the ROM is either damaged or possibly has a bad pin connection?

Is RAM intact? Are values stored in the stack and storage registers preserved when doing various operations?

[ThomasF]

(04-28-2021 01:26 PM)rprosperi Wrote:  Is RAM intact? Are values stored in the stack and storage registers preserved when doing various operations?

Hi Bob!

Yes, as far as I can test, all RAM seems ok (all 9 + 6 registers and the stack)

I've gone through some examples from the manual, and I can see that statistical functions work fine, register arithmetics etc also.
The things that doesn't seems to work are:

ASIN, ACOS, SINH, TANH,
s (standard deviation),
L.R. (and prediction),
and also Q and Q-1 fails.

I.e. "0.1361 Q" should result in "0.6615" but gives "0.5000".
"0.5 Q-1" gives "0.0000" (ok)
"5 Q-1" gives "Error 0" (ok)
"0 Q-1" gives "-9.999999 99" (ok)
"0.25 Q-1" gives after some seconds with a balnk screen "0.0000" (should be "-0.6745")
and last example: "0.01 Q-1" gives after several seconds with a "running-program-blinking-display" "Error 0", but should be "-2.3263"

So, it is really strange - and I'm more and more convinced that something is wrong inside the ROM.

Cheers,
Thomas

[teenix]

(04-28-2021 05:50 PM)ThomasF Wrote:  "0.5 Q-1" gives "0.0000" (ok)
and last example: "0.01 Q-1" gives after several seconds with a "running-program-blinking-display" "Error 0", but should be "-2.3263"

The Error 9 is an internal fault. It is a shame that there was not enough ROM space remaining to add code to display the faulty ROM or RAM, although for an end user it was probably of little use back then, but now it would be.

If it was the ACT I would expect many other faults. As you tested the RAM, it is probably a faulty ROM.

Another way to detect bad chips (not always) is that a faulty chip may be warm after being on for awhile.

The emulator at teenix.org has a ROM trace function. It will tell what ROMs executed for a given keypress when a break point has been set just before the key scan loop starts. Address ($06DB for the HP-32E)

Both of the listed [Q-1] functions execute from the same ROMs so it doesn't really prove which ROM could be in error.

It is time consuming to test code execution for the [Q] functions as many thousands of code lines execute.


Trace example: Q-1

ROM Chip Executions

ROM 0: Executed
ROM 1: Executed
ROM 2: Executed
ROM 3: Executed
ROM 4: Executed
ROM 5: Executed
ROM 6: Executed
ROM 7: Executed
ROM 8:
ROM 9:
ROM A:
ROM B: Executed
ROM C: Executed
ROM D: Executed
ROM E:
ROM F:


cheers

Tony

[PANAMATIK]

(04-28-2021 01:26 PM)rprosperi Wrote:  You'll have to wait for Panamatik to weigh-in for a definitive answer, but it appears the ROM is either damaged or possibly has a bad pin connection?

Is RAM intact? Are values stored in the stack and storage registers preserved when doing various operations?

Obviously I cannot give a definitive answer. But I also think, and you figured it out already, that a defective ROM should be the problem.

I have an HP-25C, which is still calculating, but shows a similar effect, it produces wrong results when using SIN COS TAN etc. I found, by reading the ROMs and comparing, that only one bit of one instruction was flipped.

interestingy I have the HP-32 ROMs here in my spare parts box, because I upgraded my HP-32E to an HP-32E SPICE LP. Possibly your calculator could be repaired by them.

Bernhard

[ThomasF]

(04-29-2021 06:35 PM)PANAMATIK Wrote:  I have an HP-25C, which is still calculating, but shows a similar effect, it produces wrong results when using SIN COS TAN etc. I found, by reading the ROMs and comparing, that only one bit of one instruction was flipped.

Hi Bernhard,

Yes, I can imagine that just one bit could cause behaviours like this. If I had time, I could play with Teenix MultiCalc, and simulate errors in the ROM, but it would not help - just confirm ...

I will send you a PM regarding your offer, resoldering the old ROM's should not be any problems.

BTW, do you know if it is easy to read out the contents from these Spice ROM's? I suspect they are serial ROMs, but checking the schematics, it might work to build a simple reader with a PIC or a Arduino maybe? It would be really satisfying to be able to find the actual fault and repeat the error in an emulator - at least just for fun!

Best regards,
Thomas

[PANAMATIK]

(04-29-2021 10:24 PM)ThomasF Wrote:  BTW, do you know if it is easy to read out the contents from these Spice ROM's? I suspect they are serial ROMs, but checking the schematics, it might work to build a simple reader with a PIC or a Arduino maybe? It would be really satisfying to be able to find the actual fault and repeat the error in an emulator - at least just for fun!

If you know how to do it, it is easy. I did read the ROMs of some SPICE calculators in 2015 with a PIC processor, which sent serial data to a PC while reading the ROMs.

In this thread Post #123 I describe how I read the HP-31E and HP-33E ROMs by invoking STO ENTER ROM test and reading the ISA 56-bit serial addresses and data. But I know from analyzing the STO ENTER function, that the ROM test will be stopped prematurely if the checksum fails. Therefore you have to use another method: disconnect at least the ISA pin from the processor and inject your own address signals.

Bernhard

[ThomasF]

(04-29-2021 06:35 PM)PANAMATIK Wrote:  interestingy I have the HP-32 ROMs here in my spare parts box, because I upgraded my HP-32E to an HP-32E SPICE LP. Possibly your calculator could be repaired by them.

Bernhard

Late update!

I contacted Bernhard who sent me a set of ROMs from a HP-32E, and I managed to fix the faulty calculator into a working one!

When I started, I had a failing 32E (selftest reported Error 9), and this device had two chips installed (there a 3 possible chip positions, and I number them 1-3 with 1 being closest to the display):

1: 1MA1 0209
2: (empty)
3: 1MA 304

   

The chips I got from Bernhard was the following:
(Note that I didn't know the original position of these)

a: 1MA1 0009 (ROM)
b: 1MA1 0010 (ROM)
c: 00031 80008 (RAM)

The donor was a HP-32E which was enhanced with his HP-32E SPICE LP circuit making the ROMs superflous). In a second 32E Bernhard had the following chips (similar to mine):

a: 1MA1 0209 (ROM combined with RAM)
b: 1MA1 0010 (ROM)

So, I first replaced my chip 1MA 304 with the 1MA1 0010, but I couldn't see any changes after that, the calculator worked just as before, with the same faults.
Then I also replaced my 1MA1 0209 with the 1MA1 0009 and that made all the difference (I thought!).
After this step, the calulator now contained the following:

1: 1MA1 0009
2: (empty)
3: 1MA1 0010

The selftest passed, and everything looked ok, but it wasn't ... the RAM was of course missing (so STO/RCL/Sigma+ etc didn't work) as the 0009 is only ROM - no RAM - compared to the 0209 chip.

I then tried the following combination:

1: 1MA1 0009
2: 00031 80008 (RAM)
3: 1MA1 0010

   

Then I was back where I started! Selftest didn't pass, and it worked even worse then before (failing SIN/COS etc which worked before).

So, as a final step I swapped the RAM and ROM chip to the following order (as proposed by Bernhard ):

1: 00031 80008 (RAM)
2: 1MA1 0009 (ROM)
3: 1MA1 0010 (ROM)

And finally - the HP-32E was restored to a completly working condition! Selftest passing, and memory works as expected.

   

My conlusion is that the original 1MA1 0209 in my device was failing, but if it was the RAM or ROM-part I don't know (I suspect the ROM since STO/RCL etc was working ok). A bit strange that the device was failing when the RAM was positioned in port 2, since I was told that the order shouldn't matter - at least not for E-versions (in C-versions the order is important), but apparently the order is important even here.

So again a big thanks to Bernhard at Panamatik to help me (again) restore a classic LED HP calculator back to life!

Hope someone find this information helpful - and that there are help to get when restoring the old beautiful devices!

Cheers,
Thomas

[PANAMATIK]

(05-27-2021 08:45 AM)ThomasF Wrote:  I contacted Bernhard who sent me a set of ROMs from a HP-32E, and I managed to fix the faulty calculator into a working one!