HP3478A “trouble”shooting anomoly.

[jeffjmr]

Rule #1. Don’t fix what ain’t broke.

I am tempted to break rule #1.

My HP3478A is functioning fine, best I can tell with some precision resistors, Chinese DC voltage standard and against my K2015 THD, neither recently calibrated.

But the RIFA caps are cracked, and I believe the electrolytics are original, so replacements are on their way. Since the unit is open, and always eager to learn, I started poking around. 

With Service Manual in hand, I checked all supply voltages are good. All voltages on pins that don’t require jumpering or disconnecting as indicated in the DC troubleshooting section are good. On to the Ohms section.

A 1% 3K resistor reads a stable 3.00817k.

Voltages all check out in Para. 7-C-6.

7-C-8 calls for voltage across R205 to be checked for .47vdc. This when 1ma is supposed to be the source current on the 3K ohms range. I am seeing only 1 or 2 microvolts across R205, while I am measuring 1.015ma on the test leads through the 3k resistor. 

7-C-9 d. (If voltages across R205 are “radically wrong”, which they are) circles me back to 7-C-6 where all voltages check correctly.

What am I missing? Besides breaking rule number 1?

Thanks,
Jeff

[Kleinstein]

At least for the schmematics I have, R205 should not have significant voltage, as it is in series with a small capacitor.
So there is something wrong in the section 7C8:   maybe they mixed up R205 with a different part, like Q205 or R203 ?.

[jeffjmr]

That makes sense. Hard to imagine measurements coming out right with that kind of discrepancy. Although 7-C-18 repeats the .47v check on R205.

But, I have found another discrepancy.

In my haste I missed a bad supply voltage reading. +5, + and -15 all good. But U405 pin 6 is showing -10 when it should be +10. U404 pin 6 is -10 also as it should be, and JM201 is +10 as it should be (All these reference 7-A-3). Additionally, manual says I should see +7 on U461 and I’m seeing -7.

I’m finding the schematics difficult to read on these downloaded manuals, but seems I should be seeing +7 on U405 pin 3 and -7 on pin 2, and I am getting -7 on both. Also getting -10 on both U404 pins 2 and 3 (3 being the source of the -10 on U405 pin 6).

Again, these are pretty significant discrepancies which makes me wonder how the meter can function so well in all modes.

Additional thought?

Thanks,
Jeff

[Kleinstein]

U404 works as a follower. So it should read the same one the input and output. So the -10 V at U405 should also be OK.
The inputs to the OPs should normally see the same voltage at the inputs (pins 2 and 3 with the normal pinout singles). So the readings all look good.

[jeffjmr]

OK, I will rest a bit easier at this point but my confidence is shaken in the service section of the manual. All of section 7-C-18 is predicated on seeing that .47v across R205.

Hoping my preventive maintenance on the caps manages to help keep future errors to a minimum.

Thanks for your generous advice.

Jeff

[MarkL]

...
What am I missing? Besides breaking rule number 1?

Well, I would say only half break rule #1.  The RIFA sounds like it's ready for replacing, but I would check the ripple on the other power rails to decide if it's really needed for the others.

And I would add a rule #0: Backup your calibration constants before you lose them by accident or your battery dies:

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

Fast forward to the last page or two of that thread.  People have written some good utilities to save your calibration constants using the GPIB port.  And some of the utilities also allow you to tweak the values if the need arises (after backup, of course).

[jeffjmr]

Great advice. All caps replaced and all appears well.

Unfortunately, I am not yet set up for GPIB, though I’d like to be.  I’d like an adapter that will work for the 3478A, my Keithley 2015 and my Rigol MSO5000. The Keithley is going in for calibration soon, so would love to get setup before then. 

Months of research have not revealed a clear, affordable path. I haven’t the brainpower, apparently, to discern the practical differences between the $39 ebay versions and the $700 name brands, not to mention the software choices, NI or non-NI, etc.

Any recommendations would be appreciated.

Jeff

[pqass]

Great advice. All caps replaced and all appears well.

Unfortunately, I am not yet set up for GPIB, though I’d like to be.  I’d like an adapter that will work for the 3478A, my Keithley 2015 and my Rigol MSO5000. The Keithley is going in for calibration soon, so would love to get setup before then. 

Months of research have not revealed a clear, affordable path. I haven’t the brainpower, apparently, to discern the practical differences between the $39 ebay versions and the $700 name brands, not to mention the software choices, NI or non-NI, etc.

Any recommendations would be appreciated.

Jeff

Affordability is only an Arduino UNO away.....
https://www.eevblog.com/forum/projects/ar488-arduino-based-gpib-adapter/msg3942883/#msg3942883
https://www.eevblog.com/forum/repair/hp-3478a-how-to-readwrite-cal-sram/msg3210132/#msg3210132
https://www.eevblog.com/forum/testgear/first-bench-multimeter-fluke-8840a-vs-hp-3478a/msg3008694/#msg3008694

[Kleinstein]

Great advice. All caps replaced and all appears well.

Unfortunately, I am not yet set up for GPIB, though I’d like to be.  I’d like an adapter that will work for the 3478A, my Keithley 2015 and my Rigol MSO5000. The Keithley is going in for calibration soon, so would love to get setup before then. 

Months of research have not revealed a clear, affordable path. I haven’t the brainpower, apparently, to discern the practical differences between the $39 ebay versions and the $700 name brands, not to mention the software choices, NI or non-NI, etc.

Any recommendations would be appreciated.

Jeff

For the MSO05000 I would consider using ethernet instead of GPIB (assume it should have LAN): this offers higher speed and additional isolation. It is at least a cheap way to test. With GPIB the cables are also no cheap, so it is not only the adapter.  AFAIK some of the knock off Chinese GPIB adaters also work.

[jeffjmr]

Until I have GPIB/logging capability, I beg your indulgence to look at the "logging" done the old fashioned way.

If looking at this spreadsheet is too tedious, say so and I'll table this issue until I'm setup for GPIB. Otherwise, are there any smoking guns in this data? Offsets alone don't seem to explain some of the numbers.

The convention I used I think is self explanatory; the +/- is jitter of the least significant digit (noise, right?)

Sorry for the attachment confusion. Trying to format for the easiest view.

Thanks for any comments or observations.

Jeff

[Kleinstein]

A noise of +-3 digit for the last digit would be 6 LSB steps peak to peak and this translates to about 1 LSB step of RMS noise. This is a common noise level and sometimes used as the limit to when it is worth to take that last digit serious / count it as real resolution. lower grade meters (e.g. 3 and 4 digits) tend to be a bit lower noise than this.

The ohms resdings seen to be 2 wire ohms and there may be a problem with the zero there.  This could be just a missing zero adjustment call.

[MarkL]

The 3478A can only be calibrated on the ohms ranges for either 2-wire or 4-wire connections.  It may be that your unit has been calibrated for 4-wire measurements.  (And that's really the way you'd want it.)

Try a 4-wire short and see if that gets closer to 0.

[jeffjmr]

They are 2W resistance readings. Unfortunately, the calibration report does not specify 2W or 4W. I hope it was 2W because my shorted 4W numbers are ugly on both meters. More on that later.

I am not concerned about the DC numbers. Both meters appear to be in spec there.

The AC numbers appear high to me but I have nothing to compare them to except each other. After 6 hours of warmup, the HP still reads 406 microvolts on the 300mv range, and I don't like the double 820 on the 3v and 30v ranges, then nearly 10x 76mv on the 300v range. The Keithley is at 730 microvolts on the 100mv scale after 6 hours, which was at 80 microvolts at 3 hours. Curiously the 1v through 750v ranges multiply that 730 microvolts by 10x on each successive range.

Do you agree that the resistance numbers all look like offset/calibration for the HP? But what about the 10x increase with each higher range on the Keithley?

Jeff

P.S. For the benefit of other 3478A owners who might be unaware, apparently there are multiple errors in section 7 of the service manual. Besides the missing voltage across R205, the check for the +10v is also an error, as -10v in both places is the correct value. Disappointing that nobody ever caught and edited that before publication, at lease on SNs in the 2911 range.

[Kleinstein]

The zero reading in the AC ranges going up with the higher ranges is nothing unusual. There are 2 main sources for such an "offset": one is noise / hum / offset at the input and the other is some offset around the RMS converter. The error from the RMS converter would be scaled according to the amplification / attenuation at the input and thus get larger with the rangs.
So it looks the zero effect of your 2015 is more a point from the RMS converter, while the 3478 error looks more like some residual from the input as it does not change much. Are the zero readings with a short, or an open input ?

The 4 wire ohms should be working much better for the low resistors.
Usually there is a seprate zero adjust for 2 wire ohms - this may be a relative measurement by the user just before use.

[jeffjmr]

All of these readings are shorted.

I’m most concerned about the 3478 shorted resistance. The .11 ohms in the 30ohm range is a new issue.  Wasn’t that long ago I checked it against last year’s calibration and it was .0012. Calibration report was .0002 as shown in my earlier post. One year limit is .0041 to -.0041. And of course the offset remains in the higher ranges.

I was hoping a pattern would be revealed in this data that would point to a specific component, or at least an area of the circuitry. And of course dreading the possibility of it being the virtually unobtanium U102.

I don’t want to send either in for a new calibration until I’m sure there are no underlying issues.

Thanks for your help.

Jeff