[FIXED] HP 3456A DMM fault: displays random numbers

[Kleinstein]

A first point could be checking if there is really no signal to the NOR gate, not even a weak one. This would be just in case the other NOR gate pull the line up from the other input - not very likely, but possible to check with the scope.
Besides actually checking the ROM one could also first check the supply to the ROMs / RAMs, just in case there is a borderline case.
The usual movement of the CPU / RAM / ROM chips in the sockets could be worth a try too.

Checking the ROMs would be with an external programmer, especially if the ROMs are in a socket. The EPROM programmer usually are also able to read a ROM. There are likely ROM files around to compare to and if needed write a new EPROM.

Normally a good firmware should have some kind of  checksum test, but I don't know for the 3456. If it does (the manual may explain a Checksum error) a ROM error is less likely.
It is than still odd why the control line is not working with voltage, but does work in 4 wire Ohms mode.

[enut11]

Before I get into ROM stuff, here is a comparison of what @lowimpedance measured on a good 3456A around U103:

"I see your venerable old meter is misbehaving !. I can do some measurements on a working unit here for comparison if you like, can post results in the thread for others reference.
I did a quick scope measurement around U103 pins 1,2 and 3 and found a low frequency switching signal on pin 3 in DCV, ACV, and OHMS 2W/4W with varying pulse width depending on which function is selected. This signal is related to the auto zero when on, and logic high when off.
 Pin 2 is low in DCV , ACV, and 2W ohms, but high in 4W ohms. Therefore this signal is switched through to pin 1 via a control signal on pin 2.
Hope this helps.
Oh forgot to mention that the auto zero switching signal is blocked from appearing at pin 1 U103 (stays low) in 4W mode ."

My own observations on my faulty 3456A:
The relay clicks on and off continuously while in 2W mode only.
And my measurements around the same chip pins: see pic

[lowimpedance]

Okay to confirm in 2W ohms on my working 3456a there is no relay clicking on/off.
Which relay is switching ?

Setup is Auto Zero - off, Internal trigger -on , Auto range -on, display shows OL.
U103
Pin 1 is logic 0, Pin 2 is logic 0 and Pin 3 is logic 1 with no low frequency square wave signal !.

with Auto zero - on;
Pin 1 has switching signal, Pin 2 is logic 0 and Pin 3 is switching signal

[lowimpedance]

Just a thought, try re seating the front panel connection J4 (a few times to wipe the contact surfaces) also checking the condition of the IDC cable crimp.
If no difference then next pull the in guard and out guard PCBs and re seat to give the connector contacts a good 'wipe'

One other thing make sure there is no supply ripple on the regulated out guard power supply. (A3)
Have you re capped the in guard supply PCB (A10) as many issues are due to the electro's on that module.

[enut11]

Okay to confirm in 2W ohms on my working 3456a there is no relay clicking on/off.
Which relay is switching ?

Setup is Auto Zero - off, Internal trigger -on , Auto range -on, display shows OL.
U103
Pin 1 is logic 0, Pin 2 is logic 0 and Pin 3 is logic 1 with no low frequency square wave signal !.

with Auto zero - on;
Pin 1 has switching signal, Pin 2 is logic 0 and Pin 3 is switching signal

Hard to diagnose as probing around with a DSO stops the relay, but there is a low frequency spike on K102 lower left pinout, if you read the relay label the correct way up.

[lowimpedance]

Is there a corresponding signal on U702 pin 18?.
On my unit that pin on K102 is low in 2W , no 'spikes' and will go high in 4W after a short period.

[enut11]

Okay to confirm in 2W ohms on my working 3456a there is no relay clicking on/off.
Which relay is switching ?

Setup is Auto Zero - off, Internal trigger -on , Auto range -on, display shows OL.
U103
Pin 1 is logic 0, Pin 2 is logic 0 and Pin 3 is logic 1 with no low frequency square wave signal !.

2W selected with AZ OFF
lowimpedance: U103 Pin 1 is logic 0, Pin 2 is logic 0 and Pin 3 is logic 1 with no low frequency square wave signal !. OL on display with no relay switching.
enut11:            U103 Pin 1 is LFSW, Pin 2 is logic 0 and Pin 3 is LFSW (Low Frequency Square Wave).

Well, we certainly have a difference here, and it is, for 2W ohms, a LFSW on U103 pins 1 and 3 for the faulty 3456A.

What does this mean?

[enut11]

Is there a corresponding signal on U702 pin 18?.
On my unit that pin on K102 is low in 2W , no 'spikes' and will go high in 4W after a short period.

Similar signal on U702 pin 18 to K102 so that is another difference on the faulty 3456A.

[lowimpedance]

Sounds like the main controller module is getting confused or sending the wrong setup info to the input switching!. which is why I said to try and eliminate connector and cable interconnect issues.
I seemed to recall reading sometime back that either you or someone had 3456a display issues that was solved by re seating the interconnect cable/connectors  .

[enut11]

Well, I take my hat off to you @lowimpedance.

Pin 15 fell out of connector P5 when I removed the PCB!
Also, pins 12, 13 and 14 are bent and likely to fall out soon.
Wow, the board was incorrectly inserted and instead of the pins sliding in smoothly, the end ones picked up on a weakness in the socket and tried to push the female part out of the white housing.

This has to be the problem with the meter but fixing it now is another challenge. I will try and locate another PCB.
Thanks again
enut11

[lowimpedance]

While that's interesting and good cause to suspect that connector (P5), I had a look at the unit here and it appears that pin is not connected to anything on the PCB so sadly NC and the issue still needs further investigation, (some of the other fingers look 'iffy' too).
But yeah I would certainly give all connectors a critical eye after seeing that as you don't know the units history and if any rough handling of the modules and connectors have taken place.

[enut11]

OK, I will have a look at the supply caps and such. Is it possible to test electros in circuit for ESR?

[lowimpedance]

 The last unit (A10) I fixed all the electro's showed a good capacitance reading but the main 4000uF one was clearly allowing too much ripple into the regulator when scoped, (output showed drop out).
So look at the ESR but dont trust any capacitance measurement. Just noting that in the 5 units I have fixed over the past few years the out guard supply cap has always been good still
and all the failures were on the A10 module.

[enut11]

Re-capped the A10 power supply electros but the problem persists...

The old caps tested ok for ESR and capacitance which is a testament to the quality of the components HP used in this now very old meter.

[enut11]

Well, I take my hat off to you @lowimpedance.

Pin 15 fell out of connector P5 when I removed the PCB!
Also, pins 12, 13 and 14 are bent and likely to fall out soon.
Wow, the board was incorrectly inserted and instead of the pins sliding in smoothly, the end ones picked up on a weakness in the socket and tried to push the female part out of the white housing.

This has to be the problem with the meter but fixing it now is another challenge. I will try and locate another PCB.
Thanks again
enut11

Well, @lowimpedance was right again. I jury rigged the suspect pins on P5 and it did not fix the problems with DCv, ACv and 2Wohms. 4Wohms still works and this is intriguing.

[enut11]

Random numbers on DCv, ACv and 2Wohms persist. 4Wohms still works.

When I remove jumper J29 into the ADC, display goes to all zeros so problem looks to be in the input section.
Further fault finding has not revealed anything amiss in the analog section??

Apart from re-seating connectors, IC chips and the main controller chip in the red socket, I have not done any more tests on the 5v circuitry.
I don't have the knowledge or equipment to test ROM and RAM chips.

Casting out for more ideas that I can follow up on to eliminate any other possible causes.
enut11

[khs]

Some years ago I had a similar problem with one of my HP3456.

The problem was the big relay 01-82.

You may remove and check it.

If it's faulty and if the coil is OK you may try to clean the contacts by connecting the coil to a function generator and run it at the resonance frequency of the relay. Keep the voltage of the function generator as low as possible.

You can 'observe' the contacts with an oscilloscope applying a voltage via a resistor to the contacts and stop 'cleaning' when the contacts are looking fine - not to ruin the relay at the end.

Just my two cents..

[MadTux]

Or just find another instrument for 200$ or so.
That's what I usually do, if I get crazy with locating a fault.
Board swapping usually narrows it down to a board. Comparing signals makes it much easier too.

[Kleinstein]

The µC on the input side seems to gets its commands at a low level. So the problem could be with signal transmission over the isolation barrier. At least there is no accessible ROM at the input side - so nothing to check there. One could check the signal quality at U23.  A signal with boarder line quality could cause trouble with some codes (e.g. to many 1s).

[enut11]

Some years ago I had a similar problem with one of my HP3456.

The problem was the big relay 01-82.

You may remove and check it.

If it's faulty and if the coil is OK you may try to clean the contacts by connecting the coil to a function generator and run it at the resonance frequency of the relay. Keep the voltage of the function generator as low as possible.

You can 'observe' the contacts with an oscilloscope applying a voltage via a resistor to the contacts and stop 'cleaning' when the contacts are looking fine - not to ruin the relay at the end.

Just my two cents..

Thanks @khs. I bypassed the big relay with a croc lead and it made no difference but I will keep this in mind to test again.
enut11

[enut11]

The µC on the input side seems to gets its commands at a low level. So the problem could be with signal transmission over the isolation barrier. At least there is no accessible ROM at the input side - so nothing to check there. One could check the signal quality at U23.  A signal with boarder line quality could cause trouble with some codes (e.g. to many 1s).

@Kleinstein, using a scope, the signal levels at U23 look OK to me

[enut11]

Every now and then the 3456A seems to respond to input.

Eg, inputting 10v on DCv makes the display increment, but it does not last. In less than a minute the display becomes unresponsive to input again.

Currently reading through the manual for some obscure clue as to what is going on.
enut11

[Kleinstein]

From the description so far it looks like the auto zero mode does only work for 4 wire ohms. So chances would be that the non AZ mode could work also for voltage.

The control signal working in 4-wire Ohms mode only is odd, as this means that the control part does not work correct. So far I have not understood if the AZ mode is controlled by the input side µC or the settings send every time from the output side CPU.

In both cases the input side µC could run under borderline conditions, like to low or high supply, high supply ripple or an unstable clock. If in a socket it may help to move the chip a little in the socket, in case there is a bad contact.

[enut11]

Time to finish this story.

As it turned out, I found myself out of depth with this repair but I am glad that I tackled it. I learnt a lot about the venerable old HP 3456A and have even more respect for what it represents to this day.

A younger and wiser techo (@lowimpedance) came to the rescue. Not only did he have to deal with the original fault but also one caused by me, obviously distracted by some domestic duties at the time.

Below is lowimpedance's tale of what he found:
“So the good news is I 'think' I have finally got your 3456 behaving properly now!!. After many red herrings trying to follow the manual service groups descriptions (found quite a few area's that were confusing or had errors in my paper manual copy). Also found some odd differences in some wave forms between your unit and two other working meters which seem to be all red herrings as well, ( ie did not change after repair!).

Okay the issues:
The main one was the plain wrong readings or 'OL' display which very much varied depending on which function and range was selected or whether manual ranging up or going down. That generated more confusion initially.
The clue to the issue area was when probing the Input Amplifier, actual input J19, the meter started to read correctly but only when Auto zero was on. This made sense after finding the issue as the main controller was getting continual updates for 'zero' of the system, and not a one shot measurement at the start when AZ is turned off.

And the culprit..... U104c pin 14 stuck at -18V and not allowing the AZ switching signal present at pin 9 to pass to Q104, so it was held off and thus not providing the system zero ref.

So yes you were on the right path with your initial investigation !. Replacing U104 fixed that problem and the meter measurements started making more sense. .... however...

The next issue was, with the covers back on, Self Test #6 failed after warm up. This was traced to Q116 (input path <10v and 2W ohms) being leaky, getting more so as it warmed up. I replaced it with an identical FET from a DVM plug-in from a scrapped HP 3497a. Virtually the same front end as the 3456A.
 (BIG NOTE here too is the bottom cover is not fully perforated with holes like all the fan-less 3456A's I have seen, 7 or so here! I wonder if its been swapped by accident with an older 3456, and will cause thermal issues!?!)."

Note: Yes, I found the meter would run up to 30C on warm days.

"Now that it passes all Self Tests I measured 10V DC and noted the reading was a bit low. Also, when checking ohms I found the 100 and 1K ohm range current was not 1mA but was 0.92mA.  Measuring a 1K resistor the reading was 950 ohms. Did you ever do Ohms measurements with this meter? "

Note: this turned out to be my fault with a re-wiring error during troubleshooting - should have written down what I did before being called away!

"So after much testing to pin down a leakage path with no success my eye finally caught something that was not right and a quick cross check to a good unit and the schematic proved that a wire was incorrectly terminated to the wrong side of R101 (2K55). Relocating it to the correct position fixed the wrong resistance reading both 2W AND 4W.
As for the DC reading error I think a DC Cal is in order and should correct that issue.

I noted when doing the repair that it has been looked into before your time I guess given the number of cut and re soldered transistor leads and other wires, not to mention the incorrect re connection of the input to the wrong side of R101.

As for your repair on the broken socket pins that is actually quite sound and should not cause any issues, I also could not find any suitable replacement in my junk piles, so best to leave it as is.

Last to do will be a Cal of the 10V range and a test of the Ohms ranges fully to check for further errors. Still might have to do the input switching charge correction/dump adjustments since I replaced Q116 (will think about it!).

I will do a cursory check of the AC function but not any Cal, etc as I suspect you will never use it anyway!?.”

I am indebted to lowimpedance and other Forum members who took the time to contribute to this repair
enut11

[coromonadalix]