0,2V Offset on DC Measurements with Picotest M3511A DMM

[MaxH]

Hello Everyone,

Im currently starting to debug an M3511A which has a weird constant +0,2V Offset on VDC measurements. This also affects DC and AC current measurements. AC Voltage measurement seem to go to a different ADC.
Could this have something to do with the voltage reference?
I will continue to try to find it out but if somebody might have a hint, it would be usefull.

Thanks in advance,
Max

[David Hess]

A common cause for such an offset is damage to the input protection circuity, or high impedance input buffer.  If input shunt diodes are use which is common, then one of them may have high leakage current which creates an offset though the input protection series resistor which goes to the input.

If you have an electrometer, then you might be able to test this by measuring the current across the input; it should be less than nanoamps.

[MaxH]

Thanks for the answer! Unfortunately I dont have an electrometer.
Here is a picture of the input circuitry, the left relay would be the switch for the AC/DC, down is the Resistornetwork followed by a switch for the ranges and then there is the AD8638 which already has .2V diff at its Input. I will get some better probes to understand the path of those .2 Volts. Might this be a software defined offset?

Greetings,
Max

[Kleinstein]

0.2 V difference at the input of the ad8638 could expain the problem (broken OP). Before replacing it, one should also check the supply to the OP - there is a good chance to have a bootstrapped supply and such a circuit is a little prone to oscillation and there could be supply problems (AC part or DC level) to cause the offset.

There is very likely some offset compensation in software, but this should be more like a few µV as for the offset of the AD8638.
0.2 V of measured offset would be way too much to compensate with software, as the offset would usually not be really stable like a good voltage reference. So I very much doubt it would be a software thing. The analog part should not have that much offset to compensate.


For the high impedance (e.g. > 1 Gohms) ranges one could measure the input bias current by whatching a capacitor (1-10 nF low loss like  PP, PS or C0G) charge / discharge at the input. So there is no need for an electrometer.


Edit:
From the pictures it does not look like a bootstrapped supply, more like a meter with high Z only up to a lower voltage, like 2 V. Checking the supply to the OP is still not a bad idea. I don't see very much protection - so an ESD damage to the input amplifier is a possible cause.

[coromonadalix]

Maybe asking Picotest for an input schematic would help, ask them under an NDA (non disclosure agreement) ??

Have you seen this page  for FW updates   just in case
http://www.picotest.com.tw/h/Data?key=770151152829&set=8&cat=45877

[MaxH]

Thanks for the link, I updated the firmware, unfortunately it didnt change the offset.

The supply to the OP has +8,002V  and -7,946V.

I also realized that I only measure the error if there is a connection between "+ INPUT V" and " - LO / GND" (high ohm or short), its a pretty constant .2V no matter what Input voltage.

I will ask Picotest next for the schematics

[Kleinstein]

The supply to OP looks OK.

If the offset at the OP is measured directly between pins 2 and 3  - this is clearly pointing to a broken chip.  If measured from the input to the OP output, with the OP in a x1 configuration, there could be other error sources, like excessive bias current in combination with some 100 K series resistance from the protection.
Escessive bias current to cause that much error would also be measureable with a normal DMM with 10 M input inpedance. E.g. just connect the other DMM with 10 M input resistance to the inputs and measure in the 200 mV or similar range.

Does the offset ( seen by the DMM itself) change with range. I would expect it to scale with the divider at the input, so much higher offset in the 100 V and 1000 V range, maybe already 10 V.

[MaxH]

Apparently I measured the .2V from the OP Input against the "DMM Input GND" and to the test-pin, between the Input pins of the OP I have have 0V

I looked closer at the Input-Section or the "Input GND" path and there is Q209 which has the .2V Offset. If the "DMM + Input" is left floating I dont see those .2V  (blue badly drawn line)

[coromonadalix]

Just got a feedback from picotest, they wont give any schematics,  even under an NDA, they ask to send the meter for repairs ...

[Kleinstein]

I don't think the signal path would go through Q209, as the JFET switches usually have the gate at the side with 1 pin. So chances are Q209 is more something for protection of similar.

The input shown on the 1 st. picture still looks a little odd. The large ceramic resistor is not a conventional divider, but some 10 M  and than 4 resistors of some 1 M, 100K ,10K , 1K connected to 1 point.  From what one can see, this looks more like an inverting amplifier with gain 0.1 , 0.01,  0.001 for the higher ranges. Still not clear on how the lower ranges for 1 V and 100 mV are made.  It may be through U404 and U402.

I don't think AC volts has a different ADC, it is more like a different amplifier part and having AC coupling. AC amps may be DC coupled and share some of the amplifier with the DC path. The true RMS conveter should than be common for amp and volts and work OK as the AC volts seem to be OK. It may just be that the current is DC coupled all the way and thus also effected by an offset at the input.

The path for the current measurement may be easier to follow than the voltage part.

Some input circuits have the neg side input terminal different from the normal circuit ground. This can be quite confusing than. So one should check that - likely there is a link, but not always.

[MaxH]

The divider network is a CADOCK 1776-C44
The offset error can bee seen until the 100V range, 1000V range has .14V offset (maybe the resistance here is too high). 100mV range has an overload. (I guess it would be also 200mV)

I followed the current path through Q203 and Q202 (high and low current) to R217 (bigger chip under U404) to K201 and K203, after that it also goes to U404 (analog multiplexer)
I didnt manage to find the point where the AC part joins the DC.
Especially since the ground path is indeed a bit confusing.

I will try to draw down some of the paths to better understand where the .2V might occur the first time. (still not 100% sure against what I have to measure them to be sure)

Btw thanks a lot for this amount of help so far. And also for even contacting them.

Attached are photos of the AC part and the "middle" part with the ADC

[MaxH]

Small update:

The 0,2V are only seen on the minus input side or "ground path", not on the + input.
If i measure a voltage with the DMM from + input to the symetric ground (between the +8V and -8V rails) I get correct readings.

Are .2V maybe something that is needed for measuring resistance etc and are now constantly on?

Greetings,
Max

[Kleinstein]

I see 4 possible resons to have the negative side not directly connected to ground. One is the ohms mode. There are different ways to realize excitation for Ohms, especially 4 wire ohms: some keep the common ground and some make the negative side move or have the current source (some Fluke meters). 

Another reason could be getting a larger input range without using high voltage switched. The Sigilent SDM3065 likely uses this to get a high Z 20 V input range with only a +-15 V supply and also switches for more like +-15 V. In this case it would be normal to have a difference there. Not sure if the 3510 has a high Z mode at all, and likely not 10 V, as the input divider seems to use 3 divider ratios. So the 10 V range likely still goes throung the input divider and thus +-1 V (plus some overrange) for the primary range for the ADC.

A 3rd reason may be the ADC, that has a differential input. Many of the SD ADCs get the best INL when the differential input is driven symmetric, without much common mode signal. So the neg side input is no longer at the power supply ground, but movonmg opposite to the positive input. One way to get this is to generate the inverted signal and also use is as the low side input and reference level for much of the input circuit.  It is not a common solution and the divers for the ADC inputs look like 2 differential amplifiers, which is an alternative realization.

Finally it could be simply the supply to the ADC with 0 and 3.3 or 5 V. So an virtual analog ground at some 1.5 V could be a simple idea on how to have one ADC input at this fixed level and use this also for the neg side input. However 0.2 V would be rather low for this. The differential amplifier at the ADC input also doesn't fit this way.

The result from the RMS converter likely joins back at U403 (MUX)

Trouble at the ground side could also explain, why the offset is 0.2 V also for the ranges with gain.

Finding a good ground point is indeed a problem. The middle of the +8 V and -8 V is for the supply, but his may not be the signal ground.
The divider chip R503 (below the LM399 reference) may be an interesting point. This should set the possible gain for amplification (likely x 10) for U501 as kind of the main amplifier for the input. So the pin facing towards the LM399 should be the refrence level for the analog signal. This point should have the same potential as the low side input / common terminal.

[David Hess]

For the high impedance (e.g. > 1 Gohms) ranges one could measure the input bias current by whatching a capacitor (1-10 nF low loss like  PP, PS or C0G) charge / discharge at the input. So there is no need for an electrometer.

If you could watch the capacitor charge or discharge, then the capacitor would not be needed and the current could be measured directly.

[beenai2018]

Edit- Fixed My meter.
Notes here https://www.eevblog.com/forum/testgear/keithley-2110-documenting-fixes/

[MaxH]

Apparently I measured the .2V from the OP Input against the "DMM Input GND" and to the test-pin, between the Input pins of the OP I have have 0V

I looked closer at the Input-Section or the "Input GND" path and there is Q209 which has the .2V Offset. If the "DMM + Input" is left floating I dont see those .2V  (blue badly drawn line)

Late Update: The problem got mostly fixed by replacingh Q209, I also replaced U202 but Im not sure anymore if that was optional.
I seem to get correct measurements now but a slight offset when measuring open leads, I couldnt trace that back yet. Maybe its a wrong transistor type, I used a BC850