Nuances repairing metrology grade gear?

[n_haku]

Recently I got R6871E, in a pretty dusty, but seems working condition. Ofc firtst thing was to disassemble and careful wash and clean part of case, clean pcbs with IPA and compressed air. Second, as obvious, thing, was replacing all electrolytic caps to new. But, after all, its not ordinary equipment, it's 7 digit dmm, there should exist much more nuances that I dont know as I'm not used to repair such gear, like, may be, replace film caps, replace eprom to pin compatible eeprom, replace linear regs (like 7815/79/15 etc), or even replace 74ls with 74hc/hct to decrease heat?
Also, may be it's worth to replace 5V reg on digital board with good CCM buck (I tried this in r6144 with mp2307 module) to reduce heat generation?
It would be really helpful to any advice.

[Berni]

You often don't want to change the thermals of an instrument as it might affect the calibration.

The biggest issue is that you need other meteorology gear to cross check the performance against it. Otherwise you don't know if the meter is still in calibration. Some meters also have drift problems over longer periods of time (not familiar with this one tho)

[Andreas]

Hello,

while exchangeing the electrolytics in the power supply may be a good idea especially if it is known that bad capacitors of a certain date code are used,
I would never exchange any film capacitors (except those for the AC mains line filtering if they are degraded).
Switchmode supply is a bad idea near analog cirquitry. This may introduce unwanted noise and offsets on the amplifiers of several 100 uV. The bad thing is that the autozero cycle will supress those errors by a certain amount. So you will not recognize it immediately.

The effect on calibration was already mentioned.

with best regards

Andreas

[ch_scr]

Be careful, "normal" compressed air contains oil mist - for occasional use, a bottle of dry nitrogen is usually cheaper than a specialized "clean air" compressor.

[alm]

Another thing to keep in mind, not specific to metrology-grade equipment, but to any bench DMM with high impedance inputs (~GOhm) is that there are some very high impedance nodes in the input section of the analog board. So you want to try to avoid touching and soldering in this area, and if you have to, clean very throughly.

[Kleinstein]

If not available in the web, one could consider to read out ROMs / EPROMs to have a copy of the code in case they would fail. If possible a copy of the calibration constants may also be worth it, but it can be tricky with some meters. Normally there is no need to change EPROMs - one can still do the change after they fail, if one has a copy.

Replacing the regulators would change the voltages and can this way effect the calibration. One could still check the voltages to check for large errors - especially to high a voltage for some analog parts may get unnoticed. 

After a repair, or as part of it one should also test the performance, at least as far as the usually limited instruments allow. Easy to test points are the zero drift, noise, turn over error and input bias current. Especially the input current can go up with ESD damages or contamination.

[n_haku]

Thanks for quick replies! I'll try answer one by one without an order.
- Actually, though I asked general advise, not tied to particular equipment, there were some issues with my r6871e dmm, thats why I ask about film caps: after cleaning, and before I did any soldering (with except of AC power transformer voltage configuration), I tested some functions, and got a lot of offsets in any func/range:
200mV DC: +16,5uV, 2VDC: +21,4uV, 20VDC: +72uV,
10 Ohm: -0,17894, 100 Ohm: -0,17885, 1k: -0,1789,
2mADC: +44nA, 20mADC: +0,34uA…
Though nulling it and measuring against ad7461a gave pretty accurate values, so I thought may be problem with film caps in input circuits to filtering (ofc I do not want to touch integrator cap, wich may affect factory calibrated linearity). Also those values almost not changed after recap and further cleaning.
- About switching reg, ofc I won't place it in analog board (analog and digital are different board), but just for digital with a lot 74ls and other old thick process IC, I think it would be nice to not add more heat to prolong it's life. Also continious conduction buck converter introduce no wideband noise.
- Yes, I used canned gas for cleaning.
- As for metrology, it come in pretty dusty condition, but with traces of extencive use in past, also with no cal labels. Besides my precious r6871e, I have ad7461a, a pretty modern 6 ½ DMM with usb, with expired cal half year ago, and r6144 voltage-current generator in excelent condition (checked with ad7461a). I planned to cal ad7461a, but seems cal lab in my city can't handle 6 ½ class dmm, so I leave this task for a while.
- I saved both eprom contents after checked some funtions and self-cal procedure work normally, but seems xdevs already have A00 revision dumps  .

[Kleinstein]

Film capacitors are usually aging well, with not much problems. Anyway I would not expect a bad capacitor to cause an offset voltage. If at all the problem would be more with additional leakage current.
The offset errors don't look like they are all at the input or all at the amplifier output. So it looks more like 2 or more sources, or maybe excessive bias current in the feedback path.
Another possible source for the offset could be dirt and thus a need for new zeros after cleaning.

[n_haku]

Forgot to mention, but those offcets shows almost no change during warmup, so AZ seems working normally.
If I understand manual correctrly, full scale cal can be skipped, so I can cal zeros separately. I'll try it after another attept of cleaning analog board, these flying wiring require a lot of attention.

[Kleinstein]

Before another cleaning attempt, I would really measure the input current. The input current / input resistance are among the first things that can change with cleaning. So one could see if there is some change to the better or worst after cleaning.

Chances are the offset was adjustend before with a zero calibration. This could have been in a dirty state with possible old / bad electrolytic caps or maybe with the meter in a hot rack.
The change to the board (e.g. cleaning, possibly an open case) - though ideally not the electrolytic caps could have effcted the offset.
Another point to possibly check is if the offset is different for different integration times, like 1 PLC and 10 PLC. The 1 PLC case would be more senstive to effects from switching effects.

[Conrad Hoffman]

Never change a film cap unless it has actually failed. People are also too quick to change aluminum electrolytics. I'd check the performance of the meter and if it's not broke, don't fix it. Maybe brush out the dust with a clean soft brush, anti-static vacuum if you have one, then leave it alone. If it has any batteries, do check/change those if you can do it without losing calibration.

[n_haku]

Tried indirect measure input current with various resistors in parallel to input:
(az on/az off, 20V, 20 plc, 7 digit)
100meg=4.5mV/0.6mV
1G=22mV/1.2mV
10G=40mV/-13mV
open=226mV/-3V

Direct measure with diy tia (1v/1na, 16Hz bw -3db) almost meaningless, hmm. I didn't thought az would introduce such huge current spikes.

Now I think those offcets are just aging, as it seems last use (and cal) was pretty long time ago.

[bsw_m]

My old research on charge injection (and the current spikes from it) from the operation of Az in the Advantest R6871E.
Az is on, the 20V range is selected.

The names of the attached files indicate the test conditions: the applied voltage to the DMM input, and the current scale.

The large current spike at a negative input voltage is not caused by charge injection. It is caused by the parasitic current through the switch and the settling time of the amplifier output voltage.
I consider this to be a design error.

With Az off, the amplifier of this DMM has an input current of less than 3pA.