quarta-feira, 21 de fevereiro de 2024

 HP-41C Troubleshooting

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

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

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

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.

In one of the moments where it seemed 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.

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 (see below).

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.

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.


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.






For the final job, shorter lengths of wire will be used. to reduce signal cross talk. Despite the low frequencies in use (300 or 400 kHz), their square and pulse waveforms generates lots of RF harmonics. 
I will never use the zebra connector again in this unit because I was lazy and have soldered the wires directly over the PCB pads.

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.










Sem comentários:

Enviar um comentário