Sick Keithley 197A -- 30 uV instability

[rastro]

It's time to reconsider using contact cleaner on all the switches.   Hank's story as well as the apparent intermittent nature of the 30uV noise issue again points to poor contacts. 
I have fixed at least 6-8 of these series of meters just using contact cleaner on the switches - including the range switches.  The only significant electronic issues I've encountered were on two 197's which had the same defective JFET controlling the high setting for AC volts thus creating problems for all other ranges and functions.

You can argue that the signal does not go through the range switch's but that doesn't mean these can't cause trouble by leaving the system in an partially invalid functional state (and yes even when it's auto ranging).

I don't understand why the resistance towards using contact cleaner on the switches.  It's seems like easy low-hanging fruit.

[technogeeky]

It's time to reconsider using contact cleaner on all the switches.   Hank's story as well as the apparent intermittent nature of the 30uV noise issue again points to poor contacts. 
I have fixed at least 6-8 of these series of meters just using contact cleaner on the switches - including the range switches.  The only significant electronic issues I've encountered were on two 197's which had the same defective JFET controlling the high setting for AC volts thus creating problems for all other ranges and functions.

You can argue that the signal does not go through the range switch's but that doesn't mean these can't cause trouble by leaving the system in an partially invalid functional state (and yes even when it's auto ranging).

I don't understand why the resistance towards using contact cleaner on the switches.  It's seems like easy low-hanging fruit.

Please be careful and slow and think about this. You meter may be calibrated with the contact resistance degradation in there!

[rastro]

Please be careful and slow and think about this. You meter may be calibrated with the contact resistance degradation in there!

You're joking - right???  ...lets not fix the instrument because it might NOT be calibrated any more...

[technogeeky]

Please be careful and slow and think about this. You meter may be calibrated with the contact resistance degradation in there!

You're joking - right???  ...lets not fix the instrument because it might NOT be calibrated any more...

I still haven't read every post in this thread carefully, but no, I'm not joking. If the instrument is truly broken, then by all means fix it. But if you are just worried about some very intermittent problem, know that by fixing the contacts you may go farther out of calibration.

[rastro]

More to the point - if the instrument calibration relies on a baseline of poor connections then it's clearly a house built on shifting sand. 

Most of the systems on the secondary market have seen many years since their last calibration.  That's not to say that the systems aren't stable or reading correct - most will probably still pass baseline performance.  However it's more likely that the meter was last checked/calibrated with the contacts working properly and if anything cleaning would help bring the system back to a valid baseline state.

[Kleinstein]

Poor switch contacts should not have a reproducible influence of the readings. So even if the calibration was done with dirty contacts, cleaning the contacts would not make things worse. The main danger i using a contact cleaner is, that it could contaminate the board and this way increase surface leakage. So one might have to really clean the critical part of the board in case the input bias gets to high - in a not so clean environment, this might be the case anyway.

One point to test is the input bias / input resistance on the 0.2 and 2 V ranges. This is done with measuring a high value resistor or a low leakage capacitor.

One factor could already be running the instrument for some time. This helps to reduce surface humidity. So a meter like this that was in storage for a long time might need several hours, maybe days to get back to a more normal state.

The specs call for a <=2 µV offset. So it could be something is not that good. This could be leakage on the board or of some of the FETs. It is still a question on what is good enough.

[tggzzz]

One factor could already be running the instrument for some time. This helps to reduce surface humidity. So a meter like this that was in storage for a long time might need several hours, maybe days to get back to a more normal state.

My Solartron 7081 states has a procedure after being kept in low temperatures (e.g. 5C). It tells you how to pack it so that the temperature reaches "around 35C" when turned on, and you leave it there for 24 hours.

[rastro]

... The main danger i using a contact cleaner is, that it could contaminate the board and this way increase surface leakage. So one might have to really clean the critical part of the board in case the input bias gets to high - in a not so clean environment, this might be the case anyway.

This is a great point .  Be careful when using contact cleaner as not to let it drip and run across the PCB.  When I clean the switches on the 19X series I hold the switches/front panel facing the floor so that any dripping moves away from the mother board.  Allow it to drain for some time in this position.  Try to use only enough to cover the contact surfaces.  You may want to swab the surrounding area with IPA. 

This kind of potential surface contamination could certainly impact high range ohm readings.   However, if care is taken, cleaning the switches can clear up a lot erratic behavior on this series of meters.

I would avoid disassembling the switch at this point and just try the spray cleaning first.

[technogeeky]

2. Contrary to popular belief, it is possible to disassemble the switches and clean them manually. It involves removing components behind the switches, blocking them from sliding backwards. Then you have to put pressure on the switch shaft downward, so it can squeeze out the back side. Each switch has 4 metal springs in it, which can easily be cleaned using an ultrasonic cleaner and alcohol. The inside of the switch (which just consists of two arrays of vertical pins) can also easily be cleaned thanks to the fact that a Q-Tip cotton swab covered with alcohol is exactly the same size as the channel, so it fits in there, cleans perfectly, and doesn't leave any cotton behind.

technogeeky, that's a good find.  Do you have any pictures to post showing the disassembly process?

No, but I actually need to do one switch again: one of two of my meters came with a busted power on/off switch, solved (by them) by soldering pairs of pins together for that switch.

I replaced the switch with one from a dirt cheap 179, but there is some subtle difference causing the switch to block the locking of [ volts , amps, ohms ]. So I'll document the process.


As for the calibration thing, yeah. As someone here has said, there are two kinds of failures I'm aware of regarding the switches:

1) Them being broken (like when the toggle metal pin digs out of whatever that thing that's in the top of the button shafts, like the power button; the mode select buttons). The springs being broken and jammed. Whatever. The important thing is, these can do anything from totally ruin the usability of the device all the way to having no effect at all. But I class these kind of switch breaking that will not effect calibration (e.g. they do not show up as a difference in resistance or if they do, it turns out not to matter).

2) [Note: I am an amateur, but...] In nearly every case where a corrosion-induced change in resistance of a switch contact can make a difference, it seems like it actually does make a difference. To see this, you can just consider all combinations of switches. But first, you simplify:

• You can ignore all of the range switches in the Volts and Ohms ranges; here they are just selectors fed to the mcu.
• The power switch either works, or it doesn't. In any case, resistances involved can get absorbed into the corrective abilities of the power supplies. So we can ignore these..

However, in nearly every remaining case, the switch contact itself somewhere in the signal path:

• In amps mode, the range switches will be combined into the shunt resistance itself, in series-parallel.
• The Amps switch contacts themselves are also in the signal path, for the same reason.
• Every two-position triplet of the AC/DC switch is in the signal path (and even the guarded path)
• The Volts, Ohms, and Amps switches all have places where one of their contacts is involved in a crucial circuit: connecting the 2V ref to the circuit, connecting signal ground to the circuit, connecting in with the -6.4v ref usage on the 2nd page of the schematic (ohms mode), even the ohms switch triplet [7 8 9] which seems like it can't matter (just a 5v rail). But even that switch contact's resistance is thrown in among 1500  ohms to ground. And then it gets sent to every FET driver opamp on the negative rail.

So unless all of this is totally a wrong way to look at this circuit, and for some reason I don't understand none of these things can matter... then as long as I had a meter which gave calibrated results when it was obviously working, but had intermittent errors due to switch problems of type one... I'll take the problems of type 1, and keep the (probably) calibrated meter. I can't afford calibration, and for the reasons that everyone has (very correctly) harped on, it is very risky working around the many high-impedance sections.

So that is essentially the argument for living with some types of switch problems. It's not perfect, and if I could afford actual calibration, it would be invalid reasoning I suppose.

There are also ton of selected parts in this meter, and while I have guessed at the selection reasons, I don't know any of them for sure. But it seems to me like there is some real FET greybeard work going on. And even some pretty clever selected BJT uses too.

[Alex Nikitin]

And finally, the 197A doesn't zero on DCV with the inputs shorted. Cold it was measuring about 16uV, now that it's warmed up, around 7uV.

Hmm, I would start with the U102 IC and if it is not a problem, then with JFETs around it.

Cheers

Alex

[tooki]

And finally, the 197A doesn't zero on DCV with the inputs shorted. Cold it was measuring about 16uV, now that it's warmed up, around 7uV.

Hmm, I would start with the U102 IC and if it is not a problem, then with JFETs around it.

What exactly would I be looking for?

[Alex Nikitin]

U102 is the input buffer/amplifier, I would just replace it and see if that helps. However, re-reading your description it almost feels like a bad earth, a dry joint or a bad switch contact somewhere. Do these voltage variations react on a mechanical stress (tapping/pushing the board)?

Cheers

Alex

[Kleinstein]

The meter seems to use a kind of automatic zero scheme. So measuring the input and zero and than display the difference. So a bad input amplifier is unlikely to cause an offset. A poor (e.g. damaged from ESD) U102 would be more something causing extra noise.

I see 3 possible sources for an offset:
1) thermal EMF due to temperature gradients at the FET switches or 330 K resistors used for protection - this is given by design, not much one can do about it.
2) RF signals from outside get into the circuit and get demodulated different, depending on the FETs switching. One could test if additional shielding on the outside has an influence and maybe repeat the test in a very quiet environment.
3) Leakage currents (e.g. from the JFETs, transistors used a zener or just dirt) cause a variable offset.

For the last part, one could at least measure the input current in the 200 mV or 2 V range. Ideally it should be rather low (e.g. < 100 pA). If the input bias is high this would point to a leakage problem.

[hendorog]

Just got a 'parts' 197A and it was very intermittent on ohms and pretty sketchy on DCV too.
Traced the ohms issue to the 7,8,9 connections on the ohms switch.

Surprisingly working the switches (many) times made no difference. So I've sprayed IPA into the holes at back of the switches with the board tilted forward.
That has fixed the reliability and resistance of the 8,9 connection, now waiting for the IPA to evaporate to see if the thing works properly now...

[hendorog]

Just got a 'parts' 197A and it was very intermittent on ohms and pretty sketchy on DCV too.
Traced the ohms issue to the 7,8,9 connections on the ohms switch.

Surprisingly working the switches (many) times made no difference. So I've sprayed IPA into the holes at back of the switches with the board tilted forward.
That has fixed the reliability and resistance of the 8,9 connection, now waiting for the IPA to evaporate to see if the thing works properly now...

Update on this. It worked and now the 197A seems to be fine. It matches my other meters very well on DCV at least.
The backlight isn't working - of course I forgot to spray that switch so it could be the same issue.