DCV noise of Keysight (Agilent) 34970A

[Alex Nikitin]

After buying one 34970A unit here on the forum, and attempting to replace the "jumpy" LM399H reference with a nice low noise ADR1399 (which actually increased the noise at 1NPLC 2-3 times), followed by a sort of an investigation to make some sense of how the DCV noise depends on the reference (mostly unsuccessful, with some possible clues), followed by purchasing two more 34970A (one on eBay, as spares and one again here), I have now compared all three for a DCV noise with a rather strange results. The first unit (Agilent #1) is now very steady after the reference replacement, no jumps and fairly low noise, the second unit (Agilent #2, from eBay) is fully working with a very similar noise levels to the first (only the display is almost dead), however the #3, Keysight, almost new looking (made in 2017, with an original calibration certificate still available from Keysight), exhibits an even stranger noise pattern, with 1NPLC noise similar to other two units (and no jumps) but a considerably higher noise (x1,5 - x2.5) on longer integration periods (10 and 100NPLC). I plan to investigate this at some point (not urgently though) and this thread is to exchange information/ideas/clues on the reasong for such a strange DCV noise behaviour and perhaps how to fix/improve it.

The table shows the DCV noise (SD, in ppm of the range) for 1, 10, 100NPLC and all DCV ranges, all 3 units, 0V input.

The graph is a representative of 10NPLC measurements, 100V Range, 60sec for all 3 units.

Cheers

Alex

[iMo]

..I had 1.39uV average of the sliding STD (over 25k samples) at 10PLC 10V range with shorted input with 1399 in my 34401P (before the OP27 replacement)..
How did you calculate the STD?

https://www.eevblog.com/forum/metrology/hp-34401a-hacks-and-upgrades/msg6262456/#msg6262456

[Alex Nikitin]

..I had 1.39uV average of the sliding STD (over 25k samples) at 10PLC 10V range with shorted input with 1399 in my 34401A (before OP27 replacement)..
How did you calculate the STD?

1) 1.39uV = 0.139ppm, similar to what I’ve measured on units 1 and 2 .

2) I left it to Python to calculate, for the full length of the record (60s, ~140 samples for 10NPLC)

Cheers

Alex

[Zucca]

what should my 34972A then do? No homework done, sorry for the lazy question.

[Kleinstein]

The noise of unit 3 is higher in the 100 mV range and the 10 V range (with 10 PLC).
The 100 mV range is normally largely dominated by amplifier (and protection) noise, not so much the ADC.
The 10 V range (with 10 PLC) on the other hand is more noise (especially 1/f noise) from the ADC and less from the amplifier.

At least for the 34401 there are at least 2 different amplifier versions (one with discrete JFETs and one with a FET op-amp with BS supply). I would still not expect so much difference in the amplifier noise.
Higher noise with the amplifier and ADC may point to something like EMI, mains hum or unstable supply that could effect both.

[bsw_m]

From 3 unit. Input = 0V. 1plc and 10plc one hour for each voltage range.