3478A AC Offset

[PA0PBZ]

My 3478A shows a bad offset from zero on AC volts and AC mA. When I short the input and set it to ACV the display shows a steady 0.353 mV. Set to ACA with the input open it shows 0.425 mA, with the input shorted 0.413 mA. Since there is no calibration procedure for zero in the AC ranges I wonder what to do. V and A on DC is ok, but I did the zero calibration anyway. Checked with a DMMCheck it measures OK in DCV and DCA.

[alm]

I have written about this on this forum several times, a Google search for true RMS AC offset on the forum should find it. If you read the specs carefully, you will see ACV is only specced from 5-100% of range or so. Are values within that range correct? You should not expect a shorted true RMS ACV meter to read zero, the RMS converter is non-linear for low values. There is a document on the Keithley website that shows that it doesn't significantly affect the accuracy over the specified range. I don't have a link handy, but I linked to it in earlier posts.

[PA0PBZ]

Its even worse: Accuracy spec is for >10% of full scale, and then still on the 3V range (20-50Hz) 1.14% + 102(!) digits. AC current also >10% of full scale, 1.54% + 163 digits. I guess that makes it worthless for 1mA measurements 

[staxquad]

Its even worse: Accuracy spec is for >10% of full scale, and then still on the 3V range (20-50Hz) 1.14% + 102(!) digits. AC current also >10% of full scale, 1.54% + 163 digits. I guess that makes it worthless for 1mA measurements 

your excuse to buy a 3458A, no dead zone in the mV AC range  (good to 25microV)

(assume 3478A ~ 34401A for mV AC inaccuracies)



In the graph above, it can be seen that compared to the normalized 3458A data (blue), one of the two 34401As (before and after AC re-calibration) goes very non-linear below about 1 mV and somewhere below 100?V, the 34401A displays a solid noise free 0.000 mV on the 100 mV ACV scale. The red curve is one of the 34401A tests before calibration and the green curve is a test after performing the 1/100th scale calibration on page 84 (page 86 on the pdf file) of the 34401A service manual with the same 34401A. (This graph shows the low end 100 mV AC V performance for the same DMM before and after calibration.)

Also, it is not particularly surprising or even necessarily a bad thing that an instrument in this mid level performance class apparently has such limitations. Designing low cost autoranging AC RMS voltage measurement circuits that can perform near zero is an extremely difficult problem.

from http://www.gellerlabs.com/Articles-Notes/34401A%20AC%20zero.pdf

[robrenz]

Exerpt from this related thread.  Notice the shorted ACV value.

8846A AC volts 10mV and below
This is a very crude test but I think it is still valuable.
Rigol 1022 signal generator  with sine wave output at 1KHz feeding a 100 Ohm 10 turn pot as a voltage divider.
Adjusted the Rigol impedance setting so the displayed 2V output equals an actual 2V across the 100 Ohm pot.
Adjusted the voltage divider to get 100:1 attenuation, 2V on Rigol = 0.02V out of the voltage divider. (Rigol only goes to 2mVrms)
Shorted inputs give total min max span of 2.73uV with a SD of .7uV  (noise floor)
These readings are non zeroed but remember I set my voltage divider ratio with the meter.
meter is in 6.5 digit mode with 100NPLC and analog filter on.
Stepped the Rigol down thru the voltages and and gave each reading about 5 seconds of settling time.

The data has readings from 10 mV down to 2mV that are not plotted.  The plotted values are from 0.02mV to 1.0mV
The only thing this shows is the 8846A does not show the nonlinearity that the gellerlabs plots show for the 34401 or 34410 and looks like it would be linear right into the noise floor (Rigol stops at 2mV) No accuracy determination can be made from this crude setup but I think the Rigol and voltage divider are linear enough to see if that graph bulge existed or if the meter had a bogus software zero.

[Wim13]

On my HP 3478 it shows 0.437 mV with shorted inputs.

Indeed according to the service manual, below 10 % full scale, no specification.
But tested with 10 mV AC was correct. 1 mV was 1.12 mV, so thats oke within the specifications.

Testes on a HP 3457, the service manual says below 3 % fulle scale no specification,
but stays oke until 0.1 mV . the HP 3457 gave as shorted reading 0.02235 mV AC

A fluke 175, gave shorted 0.5 mV, and was correct to 1.0 mV AC

[onlooker]

 

Designing low cost autoranging AC RMS voltage measurement circuits that can perform near zero is an extremely difficult problem.

Maybe, someone in the knows can explain this more. I would never have thought non-linearity should be a problem. Is that something a bunch/table of calibration constants can't correct? Unless, the non-linearity is unstable or extremely non-systematic (e.g. strange dependence on freq or dc level (ac+dc)). But judging from the pic with the red and green curves, non of these was happening.

[alm]

They probably do a two point (offset and gain) calibration. Especially for the 3478A, which was designed in early/mid eighties, and based on even older designs. It probably used a very primitive micro-processor that was not able to handle complex calculation (eg. polynomial curves) at high sampling rates. The designs preceding it (eg. 3455A, 3456A) still used manual pots, where complex calibration curves are even more expensive and complex. More calibration points per range would also require more steps during calibration, which is also expensive. I'm guessing automated MET/CAL solutions where not yet common back then, so a tech had to sit down and twiddle knobs on the calibrator to set the correct voltages and frequencies. Even now, more calibration steps means more time, which makes the process more expensive.

For a two point calibration, taking the range below 10% FS into account will degrade accuracy over the rest of the range, so just optimizing the accuracy of the linear portion is the best solution.

[PA0PBZ]

I did some more testing with AC A and AC V using a few DMM's I have here:


Input       Open    Shorted  Open    Shorted   1mA     Shorted  4998mV
            uA AC   uA AC    mA AC   mA AC     AC      AC mV    AC mV
HP3478A                      0.394   0.387     1.060   0.332    4997.5
PM2528      0.0031  0.039    0.046   0.046     1.018   0.46     5016
Fluke87 III 0.44    0.44     0.048   0.048     1.0025  4.32     5002
Metravo 5D  0.03    0.03     0.0002  0.0002    0.9982  3        4979


Some observations:

They all have a significant offset in AC mode.
It looks like shorting the input in AC A makes the offset worse. HP says to leave it open in the zero calibration.
The Fluke 87 III does not have an AC mV range so the comparision is not very fair here.
On the PM2528 the zero offset does not change on the display when changing the range, the decimal point moves but the displayed value stays the same, it looks more like a 'counts' offset. This is more or less true for the other meters.
I should do some more testing with higher AC currents, but at the monent all I have is a DMMCheck plus and that only does 1mA AC.