HP-41C Troubleshooting

[UraniumNutt]

Hello! I was referred to this forum by a Reddit user. I was having some trouble with my HP-41C as discussed in this Reddit post: https://www.reddit.com/r/hpcalc/comments..._41c_help/

To repeat what I originally posted:

I recently bought an HP 41C from eBay, and have mostly been happy with it. There are just a few buggy / broken behaviors that it has.

The first one I found was that the beep function does not work. Instead, the calculator freezes, and the only way to recover is to remove the batteries. Interestingly, clearing flag 26 (audio enable) allows the function to run without freezing, but obviously without any beeping.

The second problem I found was that the alpha store and recall functions do not work. When the alpha store function is run, the calculator just prints "yes". If you try to recall the string, it just displays zero (while still in alpha mode).

The third problem is that the rad function does not work. If you try to, all it does it print "yes". My calculator also came with a surveying module. If you remove it and run the rad function, then the calculator locks up, like when running the beep function.

I don't know if the behavior that I am encountering is because of something I am doing, or if it is a hardware problem. Does anyone know what is going on?

I was hoping that the folks here would be more familiar with these calculators and would have a better idea of what was going on. For background, I am an electrical / computer engineering student, so if I knew what was going on, I feel reasonably confident I could resolve the issue.

[rprosperi]

I don't have specific answers, but I'll suggest a few things that will help folks to reproduce your symptoms and/or confirm what sounds like a h/w problem.

A. Always remove modules when troubleshooting, to verify the where the problem(s) are from (device or ROM)

B. Give exact examples of your precise steps and results, to insure both that you are using them correctly and also for helpers to ensure they are using identical steps.

Examples:

To test BEEP, I disabled USER mode, then pressed [shift] [4], and the device locked-up.

To test Alpha store, saving "ABC" into register 00, with USER mode disabled, I pressed [Alpha] ABC [shift] [STO] 00, and the 41C responds by displaying "YES". Checking the register to see what was now in there, I pressed [shift] [R/S] 00 (to VIEW reg 00) and the contents were not what I saved.

I don't have a 41C or ROM with me, so can't test these now, but it sounds like it's some kind of h/w issue with the 41C itself. Add some more details as shown above, and other helpers here will confirm their results which should pretty quickly narrow it down.

Good luck, and if you resolve the problem(s), please note the solution in a post below so that future readers may use these results to solve similar problems.

[Diego Diaz]

Hi,

As previously stated, a detailed step by step sequence of the symptoms' appearance is a good starting point to diagnose any faulty device.

In the HP-41C case (apart from removing anything connected to the I/O ports) should begin with a [MEMORY LOST] ;-)

Anyhow, in most cases, with this model (assumed Full-Nut), weird behaviours are related to cracked screw post(s) and/or internal corrosion due to battery leakage.

Pictures are more than welcome in case you decided to open the unit.

Best regards.

Diego.

[UraniumNutt]

Hello, It's been a while. Here are some step by step descriptions of some the the problems my 41C has. All of these examples start from a reset state by holding down [←] while the unit is off, then pressing [ON], and then releasing [←]. I also do not have any modules installed.

Beep lockup:

After the sequence [shift], then [4], the calculator freezes. It does not respond to any input, except the reset sequence. However, what happens after the reset sequence is not consistent. Sometimes it displays zero, other times it displays CLX.

RAD mode lockup:

After the sequence [xeq], [alpha], [7], [Σ+], [log], [alpha], the calculator also freezes. The calculator is mostly unresponsive, unless the [on] button is pressed several times. The number of times it takes is not consistent, but eventually the display will return to zero.

alpha store / recall:

After the sequence, [alpha], [x⇄y], [chs], [chs] (stores the string "foo"), then [shift], [sto], [0], [0], [alpha], the calculator does nothing. If [alpha] is pressed several more times (an inconsistent number of times), the calculator will eventually display zero again. Pressing [alpha] shows the string "foo" still in the alpha register. It is cleared by pressing [←] like expected. If you try to recall foo with [shift], [rcl], [0], [0], then the string "0.0000" is loaded into the alpha register.

More background information:

This calculator came with a custom 3d printed battery compartment. After opening the unit up, there were some signs of corrosion, but nothing looked damaged.

In an attempt to debug before, I removed the beeper. While I was removing it, I noticed that one of the IC sockets was not populated. I thought that could have been missing intentionally, but after looking at the HP 41C service manual, it looks like my revision of the board should have that IC. I am wondering if it was removed by the previous owner, and it is the reason that unit is buggy.

For those interested, here are some pictures of the PCB I took. Images

[Mark H. Shin]

I realize my suggestion might take the "fun" out of self repairing your device...

It does appear to be a hardware problem.

There is a seller on a well know auction site that is selling working CPU boards for 41C/CV for $30.

I'll PM you the link if you should be interested in replacement...

[brouhaha]

(02-03-2023 02:51 AM)UraniumNutt Wrote:  For those interested, here are some pictures of the PCB I took. Images

Looks like the right chips for a 41C. A 41CV would have more RAM chips.

The earliest 41C had five 1LA7 RAM chips, of 16 registers each, with the -01 suffix being the "status registers" (stack, alpha, etc), and the other four (-02 through -05) providing the 64 registers for user program, data, key assignments, and buffers. The 82106A memory module also contained four 1LA7 chips, mounted chip-on-board and encapsulated in epoxy.

The HP-41CV kept the 1LA7-01, but the user memory was provided by five new 1LE7 chips, which provide 64 registers each, so the 41CV has the full complement of "normal" memory (vs. "extended" memory).. The 1LE7-01 provides the first 64 user registers, equivalent to those of the 41C. The 1LE7-02 through -05 provide the remaining 256. The 82170A quad memory module contains four 1LE7 chips. (The 82106A and 82170A modules should never be used with a 41CV or 41CX.)

At some point HP stopped making the 1LA7 variants other than the -01, and started using a common board for the 41C and 41CV. Both had the 1LA7-01 and 1LE7-01, but the 41CV has the additional four 1LE7 chips. To cram six RAM chips in, they used three 8-pin-DIP footprints for RAM, and double-stacked the chips. The 41C only needs two, so it doesn't double-stack them, and one DIP footprint goes unused.

The 1LE9 hips are the ROMs, and the same ones are used for the 41C and 41CV. I think the ones shown are the "GFF" set. There was only one later revision for the 41C and 41CV, which was the "HFF" set, buy I'm not sure whether there was a 1LE9 for ROM0H. That might have only been in the later 120Kbit ROM chip.

Whatever the fault in your 41C is, it's not that the DIP footprint is empty.

It could be a bad chip (almost any of them), a cracked trace, corrosion between the logic board, compression connector, keyboard PCB, or power and ports flex circuit. Unfortunately it's not easy to diagnose.

[jebem]

For those who may be interested.

I had the same type of symptoms on one of my 41CV calculators, and managed to fix it using a workaround.

After analyzing the power supply IC voltages and signals (all OK, despite 140mVpp ripple from the DC-DC converter), power supply IC/Processor IC handshaking (OK), and Processor signals (OK, but lots of noise at the buzzer pin), I came to the conclusion that there was a failure in the Processor IC at the internal driver stage to the piezoelectric buzzer.

Picture: +VCC ripple from the DC-DC converter (Power Supply IC plus external inductor)


It was presenting all sort of garbage at the LCD, processor lockups, effectively preventing a successful initialization.

Picture: Example of one of the random/garbage chars/indicators on LCD

Sometimes it would initialize completely, but it was basically unusable due to instability.

In one of the moments where it seems to be working, I noticed that the execution of BEEP or TONE commands would systematically crash the machine (despite the sound output seemed to be normal), either with a blank LCD, or a "Memory lost" message, or just displayed lots of garbage characters.

Video: Memory lost after executing BEEP and recovery via soft reset

I did the obvious next step, and removed the buzzer.
The calculator issues went away immediately. It never failed again.

So the failure was either due to a defective buzzer or an internal failure of the CMOS driver stage to the buzzer.

From the electronics point of view, the buzzer is essentially a capacitor (around 20nF).
Indeed the buzzer was fine. Infinite resistance, zero leakage DC current under 15VDC (maximum output from my lab power supply).

To confirm the Processor IC buzzer driver stage failure, I loaded it with different resistors (a benign load type as opposed to a capacitor or an inductor).
From 100KR to 1MR would crash the machine, the lower the value the more frequent crashes.
With 3M3 it would work, apparently normal.

I set for a 5M6 precision metal film resistor as a permanent load on the Processor IC buzzer output pin.
The reason for this pull down resistor is that I wanted to have the BEEP/TONE working, so must be sure to not have undefined logic states at this pin (see below).

Also added three 0.1uF bypass capacitors, one at the +VCC pin of the Processor IC, and one in parallel with each electrolytic cap.

With this calculator powered off (sleep mode), the current consumption initially drops to around 3uA, then increases to a maximum of 74uA after a few minutes, and then decreases to around 20uA average.
Comparing this to the published figures in the service manual, I would say this machine is within specification.

Picture: Pull down resistor added, as well as three bypass caps
Picture: Detail of modification. Waiting now for the SMD CD4093 IC

Under rest state, the Processor buzzer pin is at GND level (0 Volt).
Under BEEP/TONE execution, there is a square wave with +VCC peak-to-peak amplitude (6.6V in my unit, 6.5V nominal).
Lots of random noise with around 1Vpp at the GND level was seen as well, indicating the failure.

So I tested a simple CMOS NAND gate (set as inverter) from a CD4093B Schmitt-trigger IC, where the NAND input was connected to the Processor buzzer output having the 5M6 pull down resistor, and the output connected to one of the buzzer leads.
The other buzzer lead connected to +VCC.

Picture: Testing a workaround. Average 7.4mA current consumption during the execution of BEEP command

This was 5 years ago. Recently I found the machine in a box of faulty machines, and tested it again.
As at the time I didn't care to install the workaround components properly, this time I have ordered a new CD4093 SMD type to be small enough to fit under the disc buzzer.

I may publish pictures if there is interest in it.


ADDITIONAL NOTES

Troubleshooting setup

In order to do a proper troubleshooting, I removed the zebra connector at the bottom and the flex power connector at the top.
The lab power supply set at 6VDC was connected to a pair of wires soldered to the KB/Display PCB assembly.
The Processor PCB assembly was connected to the KB/Display PCB by using 30 lengths of wire wrapping (thin teflon isolated wire) directly soldered.

Picture: Wire wrapping first row
Picture: Wire wrapping completed

Soft reset versus hard reset

A soft reset uses the Back + ON keys combination, but it doesn't guarantee to do its job in every situation.

A hard reset guarantees to put the machine in complete absence of electrical potencial (voltage) at the components.
To do it, remove the power supply from the machine (batteries or external), then use a tweezer (or a clip) to firstly short the larger 470uF cap, then short the smaller 100uF one.

By the way, HP have chosen excellent electrolytic capacitors for this 41CV.
I removed them and they measured very well, as new, with very low Vloss (1 to 3%), low ESR (less than 0.5R on both), and the capacitance value was above nominal.

[rprosperi]

Hi Jose, long time no read... hope all has been going well.

Yes, please post photos of the repair process. All rescue/recovery techniques are worth sharing IMHO.

Soldering 30 wires for the temporary connection is (to me, someone who struggles sometimes to do 2 such small wires) pretty bold, I suspect most would not have considered that, but given the nature of the 41's physical architecture, make sense.

Thanks for sharing this.

[jebem]

(02-21-2024 01:11 PM)rprosperi Wrote:  Hi Jose, long time no read... hope all has been going well.

Hi Bob,

True, life was in the way of my past hobby and then I was into radios.
But I never forgot you good people here in the forum, and keep occasionally reading the so many interesting posts.

When I have the opportunity I will upload some pictures of it.

Cheers,
Jose

[jebem]

The 4093 chip SMD version arrived and I could install it and close the calculator.

See my post #8 above for details (Meanwhile I have added some pictures and video there as well).

Ugly cosmetic job but it will work as intended for as long as needed:

Picture: CD4093BM SMD fixed with Silicone adhesive over polyamide tape.
Video: Testing the buzzer fix
Picture: Second layer of polyamide tape to isolate the circuit.
Picture: Processor PCB in final position, external power cabless about to be removed.
Picture: Preparing the rendez vous of the two PCB assemblies
Picture: Cabinet closed, new feet installed, charging complete of my modified 82120A Battery Pack.
Video: Final test. XEQ TONE 0 to 9 on a ISG loop