Thurlby 1905A DMM ADC repair

[ktulu]

Hello.
My 1905A DMM started to act up on voltage and current measurements when negative (or no) input voltage was applied.
After lots of measurements I figured out, that one of the analog switches inside IC7 (responsible for connecting the reference voltage to the input buffer) is bad.
Since this is part of the ADC circuit, I'm wondering by how much will it affect the accuracy of the unit if I change it?
What concerns me even more is the "selected" written after the part number in the parts listing.
I don't have any precision multi meter that comes even close to this, so no way I could re calibrate it.

[Kleinstein]

IC7 is an old 4053 CMOS switch chip. The selected is probably checking for low leakage. These low cost parts are not checked very much at production. So the maximum leakage current can be quite high though the typical leakage current is rather low (e.g. 10 pA range). The circuit may like to have a reasonable low leakage, like no more than 100 pA. If the leakage is OK, it should not effect the accuracy / calibration very much. The problem is more that the old part may have had leakage during the last calibration. In this case even a perfect replacement can have an effect.

However to me it looks odd to have a problem with negative signals only if IC7 is bad. It would have to be a rather odd failure to cause an error with negative voltages only. There are quite a few other chips / transistors that are more likely to cause such a failure. 

Is the main amplifier / buffer still working ? The signal at the output of IC40 should be a really good test point.

Not very good, but as a try to preserve the calibration, one could try to measure the 6.95 reference directly with the meter in the hope the positive reading is still OK and accurate. The 200 mV signal at R107 (only the positive side) could be another reference point.

The circuit diagram shows a few odd points around LM324 with massive capacitive load - normally a sure reason for trouble  . There may be so extra caps not in the plan.

[ktulu]

Hmm, I'm not sure that you looked at the right service manual? Please check the one I've attached.
IC7 is a CD4016 quad switch used as a H bridge. It switches the 1V reference voltage (which was previously stored in C25) into the negative feedback path of the input buffer. The polarity of this voltage is the reverse of the input voltage, since the unit uses a dual slope ADC.
3 switches are working ok, but the upper right (IC7a) is always closed.
The unit is measuring positive voltages fine. With negative input voltages, the integrator shoots up to the + rail during ramp down because the lack of reverse negative reference voltage. Than starts to be all over the place because of the charge buildup in C27.
I don't have a storage scope, so took some pictures.
Ch1 is the integrator output, Ch2 the buffer output.
Somewhere in the middle of the second picture I've flipped the input voltage from positive to negative. The first division shows the de-integrate phase of the positive input voltage measurement.

[Kleinstein]

Sorry, I looked at the wrong schematics (1906 DMM). For the 1905 circuit it makes totally sense that 1/4 of  IC7 is broken.

With the 4016 switch in the switched capacitor circuit, changing the chip can have an additional effect if the switch capacitance changes.
However C25 is quite large with 1 µF and changes are good that the capacitance would not be much different for the same type of chip.
So I would not expect a major change in the scale factor with a new chip.

With one switch always closed, this can already effect the positive readings too. So there is limited use in doing a reading of the meter internal 7 V reference (only connect the positive side - ground is already connected).

I would not worry much about selected chips - this is likely just checking that leakage is not much higher than typical. Leakage is about the main parameter that really scatters between samples.

[ktulu]

Ok, thank you.

With one switch always closed

Sorry, brain fart. It is always open, not closed.
So I should measure the reference voltage, (with the meter itself) and compare with the new chip installed?
Also I would like to place the new IC in socket.

[Kleinstein]

With the switch open the positive side can be still OK, but zero could already be tricky.

yes measure the internal reference with the meter itself could be a way to preserve / check the scale factor. Chances are the LM399 from the meter may be your most stable reference and reading it's own reference also removes some of the noise.

A socket can be a good idea for tests, but the extra capacitance could give a minute (low ppm range) change in ADC gain.