HP 34401A DMM settling time comparison / questions - Fixed

[radioFlash]

I measured the test points in the AD Converter section as Dr. Frank suggested.

* Measurements were made with Keysight 34465A in High-Z with 10 NPLC. Values shown are average of 5 readings
* Approximate time after turn on for the measurements are shown.
* There is an odd jump in the readings at 40 minutes--this may have been due to my AC unit turning on or off

[radioFlash]

Then check if the input amplifier / multiplexer is ok, you should see exactly these 10V   5V also on pin 4 of U101.

It may be, that this buffered output is also switched on /off, due to AZ measurement, then put your other 34401A in fast DCV mode, so you will see 5.00V briefly flashing, maybe.


You may also measure the 5.000V input voltage before the buffer (maybe relays K103 is defect), that is marked Buffer input, pin 5 of U101.

For that measurement, you have to put your known good 34401A in high Z mode again.

Frank

With 5V on the 34401A's inputs, this is what I measured right after turn on:

U101 / 4 : 5.00578 V (drifting downward)
U101 /5 : 5.00024 (stable, same value as directly measuring the DMMCheck with the 34465A)

[Dr. Frank]

Hi,

Definitely U400-A is defect.
It has 100mV at least of offset, but it should have 100µV maximum offset per datasheet.
Its output at "B" drifts about 500pm, totally uncorrelated to the LM399H reference "C", which drifts 100ppm

These 100ppm drift are also much too high and lasts too long..
I assume that the input transistors of U400-A are broken , drawing several hundreds µA or mA of current, instead of specified 50pA, so affecting the LM399H output.

U400 is an AD706, which should still be available.
Before replacing it, please make sure, that nothing oscillates , neither noise on +15V supply of U400, nor its output..

U101-4 is drifting, because the processor detects some internal offsets and tries to compensate this by adding an offset, or something similar.. this may vanish also, if the reference circuitry is stable again, after replacing U400.

Frank

[radioFlash]

Hi,

Definitely U400-A is defect.

I replaced U400, the Analog Device AD706, and it's not drifting like it was before. I also checked for any oscillations in the supply rails and found none. The fixed unit's behavior is now the same as my other 34401A. Now, I have to get the repaired unit calibrated. Thanks for all your help and suggestions!

[iMo]

Rather older thread but the topic is the same - I've got a 34401A here (1999), where from the cold start till settling (ie against a hot LM399) takes 3.5 hours typically (23.2C ambient, 41.5C inside the box at the end).
Below is the typical warm up curve.
My question is whether to start to mess with the AD706 (for example, when reading this success story), or is it something not worth of the effort..
Thanks.

[Kleinstein]

If using the auto zero mode, there should not be much effect of the OPs. Much of the drift in the warm up period is more due to resistor drift. So there is little hope to improve things.  If really concerned with the warm up, I would consider an extra heater resistor that can be turned on for the first 15 minutes or so to speed up the warming.

[iMo]

Thanks, AZ is on, there is also a thermal isolation from inside the box (shown in an different thread).
PS: added the internal DMM temperature - looks like the temperature needs the same time to settle

[Kleinstein]

Extra thermal insulation tends to slow down thermal equilibrium and make the warm up effect larger.

What is actually shown in the last curve ? 

The small effect of the OPs is valid for OPs that are OK. Broken OPs can have a large effect.

[iMo]

The curve shows the relative voltage diff against an external "hot" LM399 (Vrefext, >1.5khours, on for several days), meter at 10V range, AZ on.
Edit: Display ON.
The ppm is (actual_V - Vrefext)/Vrefext.
The measurement started 15.8secs after switching the meter on.
I've added above the internal DMM temperature graph.
Tamb=23+/-0.2C during the measurement.

[Kleinstein]

The temperature drift does not look excessively large: about some 8 K for the last 5 ppm. This is within the normal range (some +-2 ppm/K) for such an instrument.  The very fast initial part can be higher due to local effects, but I would not worry about the first 10 minutes.

[iMo]

I've calculated the regression (from start+10minutes till the end) of the measured absolute Vrefext voltage against Tdmm and I've got 3.198uV/C.

[Kleinstein]

3.2 µV/K is a little less than 0.5 ppm/K. This is quite good for the gain stability - so nothing to complain about.

[NoisyBoy]

34401A stipulates an one-hour warm-up time to meet the spec. after 60 min of warm-up, it appears the drift is around 2ppm below the stabilized reading in your case.  For a 6.95V reference, you are talking about a deviation of 0.0000139V. 

If that level of accuracy is required, precision bench equipment is constantly left on in most labs to minimize thermal deviation, and cal annually to compensated for aging drift. If you truly need 6.5 digit precision, you should let it warm up for at least an hour before taking measurement.  So if I were you, I would not be messing with it to make it do something it is not designed to do.