hp 3478A 3V range inaccurate reading

[Warsp]

Hi all,
I recently repaired an hp 3478A I got off of ebay and I have noticed that the 3V range is around 0.4V less than what it should be. I have measured the signal going to the ADC and it appears to have the normal 0-5V (should be 0-10V according to the manual but that is what I measured) signal present like all the other ranges. This makes me think that the input circuitry is fine, but I have no idea how the ADC could be inaccurate on a single range since the same 0-5V signal is going into the ADC on all DC ranges. I personally think it could be a calibration issue, but I am hesitant to erase a possibly perfectly good calibration. Does anyone know what could cause this issue or should I just try to recalibrate it? Also before anyone asks, the power rail voltages are fine and the majority of the supply caps have been replaced.

Thanks!

[Kleinstein]

A point worth checking may be the battery to power the RAM that holds the cal constants. There may be a way to backup the CAL constants via GPIB, but AFAIR it was tricky to write them back if needed.

For the error is this a fixed 0.4 V offset or an error of the scale factor (e.g. all values some 13% too small) ? It may help to get a little more data one the readings: e.g. test something like -3 to 3 V in around 1 V steps.
If the 0.3 V or 30 V range work OK the ADC should be OK. It may also be worth checking the 300 V range - this range uses the same gain setting ( divider 1:100 and than the 3 V range).

[coromonadalix]

yes there is some software who will save the cal constant

https://www.eevblog.com/forum/repair/hp-3478a-how-to-readwrite-cal-sram/


i some case it was possible to edit and correct some values

you have some 3478 threads here

[Warsp]

The RAM battery was replaced when I replaced the power supply capacitors so there should be no problems with it now. I have checked the 300V range and it appears to work correctly. I tested the reading of the multimeter and have provided data below. If it is worth mentioning, there is also a problem with the ohms measurement having an offset of -1ohm, but I believe I can easily fix it in calibration. Does the dc volts setting share the same 0V calibration constant or is there a separate one for each voltage range?

-2.42V -> -3V
0v       -> -0.485V
300mv -> -0.162V
1V       -> 0.568V
2V       -> 1.618V
3V       -> 2.666V
3.32V  -> 3V

Graph:
blue is ideal, red is actual

[Kleinstein]

Essentially all of the signal path for the 3 V range is either used for the 300 mV range or the 300 V range. So a hardware defect should also effect one of the other voltage ranges.
A possible pint to check could be the input bias current. E.g. have a 10 nF low leakage capacitor at at the input in the 3 V range and observe how fast the voltage drifts. Escessive bias could cause an offset, though this should also effect the 300 and 30 mV range if it also happens for low voltage.


From the link below the CAL ram includes offset and gain per range. One could try to read the cal constants and look at the values if they make sense. The 3 V range should not have much offset.
A glitch / flipped bit in the CAL constants (offset) is a possibilty.
It also looks like the slope is also no perfect and a bit off.
Is there a way to check the calibration also for a few more ranges ?  There error there may not be that obvious.

[Warsp]

I put a 10nF film capacitor on the inputs of the meter and it took around 90 seconds for the meter to drift 100mV, but it is probably longer since the meter's "slope" is higher than what it should be. I can not verify if the other ranges are 100% precise since my other meters have drifted a bit, but I believe those ranges are decently accurate. Unfortunately I do not currently have a way to use the GPIB/HPIB, so there is no way to actually verify if the calibration is at fault without recalibrating the 3v range. I may be able to calibrate the 3V range only and use the 30V range as a reference, but if not I will probably end up sending it off to a calibration lab.

[Kleinstein]

A drift of 100 mV in 90 seconds with a 10 nF capacitor are some 11 pA of input current. So this is perfectly fine for the meter. The specs should be something like <50 pA or maybe <20 pA.

It would not make much sense to send a broken meter (if there is more than just a lost cal constant) for calibration / adjustment. Before this step I would check the meter for more time and maybe also try to adjust the 3 V range - the calibration would be lost when letting it be adjusted externally anyway.

It could make sense to get a GPIB connection of some kind, either as a used / clone or possibly build a simple one. Here in the forum is a thread about a relatively simple DIY version that can work with a single instrument at a time.

If the error is in the input part (before then gain step) the 300 mV range should show a comparable offset - so this can be excluded.
If the error is with the gain setting the 300 V range would show a correspondingly 100 times larger error due to the divider in front. No need for high precision to see this. 

[Warsp]

I have recalibrated the 3V range and the problem has appeared to have gone away for now. I don't know what caused the calibration issue in the first place, but I suspect someone might have messed with the calibration before me since the ohms measurement also has some questionable calibration. I will periodically use the meter for the next month and see if the problem comes back, and maybe get a more professional calibration eventually. Thanks for everybody's help.

[fenugrec]

A glitch / flipped bit in the CAL constants (offset) is a possibilty.

Possible, but unlikely - there is a checksum per range (covering offset + gain), so you would need 2 lucky bit flips to maintain a good checksum. With any corruption in the cal constants, the 3478 will tell you right away when powering on.