Author Topic: DVMs aren't benign  (Read 2413 times)

0 Members and 1 Guest are viewing this topic.

Offline Conrad HoffmanTopic starter

  • Super Contributor
  • ***
  • Posts: 2077
  • Country: us
    • The Messy Basement
DVMs aren't benign
« on: February 02, 2019, 06:55:50 pm »
I was recently doing some voltage measurements in the 5-20 mV range and the results seemed a bit off. Just for a reality check, I hooked up a ratio transformer and several different meters, checking from 10 VAC on down. These were an HP 400E, an HP3478 and my ancient HP 3455. All the meters were hooked up in parallel and the signal was 1 kHz. The 400E seemed way off on the lower ranges. The other two were pretty good. When I unhooked the other meters from the 400E, it was dead on, better than the other two. I assume the DVMs put some sort of HF pulses on the line that the fairly high bandwidth 400E can see. Anybody else ever get tripped up by this sort of thing?
 

Online RandallMcRee

  • Frequent Contributor
  • **
  • Posts: 544
  • Country: us
Re: DVMs aren't benign
« Reply #1 on: February 02, 2019, 07:29:48 pm »
I have noticed spurious signals from my DVMs when measuring noise using the 10000x LNA preamplifier.

It clearly shows a lot of noise with most meters. Take the meter off and the noise measurement goes back to normal (in the 1uV range). All of my DVMs cause some sort of problem though I have not tried to characterize it. I'm in the "ok, don't do that camp".

Also, when you turn your meter off, some meters seems to have a fairly low impedance and perturb the circuit. Turn it back on and its back to high impedance.
 
The following users thanked this post: tooki

Offline tooki

  • Super Contributor
  • ***
  • Posts: 13156
  • Country: ch
Re: DVMs aren't benign
« Reply #2 on: February 02, 2019, 07:30:32 pm »
Isn’t it far more likely that you’re dealing with loading effects from the input impedance of the other meters?
 

Offline Conrad HoffmanTopic starter

  • Super Contributor
  • ***
  • Posts: 2077
  • Country: us
    • The Messy Basement
Re: DVMs aren't benign
« Reply #3 on: February 02, 2019, 10:31:01 pm »
My first thought is no, because the ratio transformer output impedance should be fairly low compared to the meters. That was the reason I went with the ratio transformer rather than a KVD. But, I don't have a number for it, so now I have to find out! Other then the 400E, which is 10 megohms input and 25 pF, the other meters are 1 megohm. Many/most DVMs with 10 meg inputs turn out to have 1 meg when used for AC.
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: DVMs aren't benign
« Reply #4 on: February 03, 2019, 03:49:44 am »
It is called charge injection or more often charge pumping in connection with meters and obviously not often discussed or even acknowledged by the meter manufacturers.  (1)  It also shows up when trying to measure the input resistance of one digital multimeter with another digital multimeter.  Some, usually much older, digital multimeters have front end designs which do not produce this problem usually by including a precision analog high impedance buffer stage so there is some virtue to keeping an old meter around which has no charge pumping as a sanity check.

The charge pumping could come from several sources depending on the design.  The ones that immediately come to mind are chopper stabilization, automatic zeroing, sampling kickout.  Some meters allow disabling of automatic zeroing to reduce or eliminate charge pumping.

Chopper stabilized operational amplifiers have the same problem when the inputs of separate devices are connected together.  If they are operating at close to the same chopping frequency, then they synchronously demodulate the charge injection from the other producing a DC offset or low frequency AC signal at their output.

(1) HP/Agilent/Keysight is the only company I know to mention it and today they advertise some of their bench meters as having "low charge injection".  Does that mean that their other meters have high charge injection?
 
The following users thanked this post: Conrad Hoffman, helius

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 15148
  • Country: de
Re: DVMs aren't benign
« Reply #5 on: February 03, 2019, 10:41:39 am »
It is still od to see the effect of the other meters, when measuring AC voltage.  The HP3478 and especially the 3455 are old - before the AZ OPs became common. When in AC mode, there should be no auto zero active at the input. So the effect is kind of surprising.

With increasing use of WLAN and mobile phones, even just a piece of wire can pick up RF junk that can fool some meters, especially older ones not made with these frequencies in mind.

I once had trouble with charge injection on an HP3457 (large brother to the 3478)  - but this was with low voltage DC and only with AZ mode active. An external amplifier was lower noise anyway.
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: DVMs aren't benign
« Reply #6 on: February 03, 2019, 04:28:27 pm »
It is still od to see the effect of the other meters, when measuring AC voltage.  The HP3478 and especially the 3455 are old - before the AZ OPs became common. When in AC mode, there should be no auto zero active at the input. So the effect is kind of surprising.

With increasing use of WLAN and mobile phones, even just a piece of wire can pick up RF junk that can fool some meters, especially older ones not made with these frequencies in mind.

I once had trouble with charge injection on an HP3457 (large brother to the 3478)  - but this was with low voltage DC and only with AZ mode active. An external amplifier was lower noise anyway.

That puzzles me also.  The AC converter in series with the input should suppress any charging pumping and an AC converter does not require automatic zeroing because it is AC coupled.  I do not see anything about the HP3478A or HP3455A designs to indicate otherwise.

I included sampling kickout as a possibility because some meters could use a sampling input stage and then compute the AC digitally.  I believe some handheld multimeters may have started doing this in the 1990s in lieu of analog AC to DC conversion and honestly if I was designing a multimeter from scratch, I would consider this method of operation.
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 15148
  • Country: de
Re: DVMs aren't benign
« Reply #7 on: February 03, 2019, 05:09:18 pm »
I agree that sampling the AC waveform could be a good modern way, but I would expect sampling after an initial buffer / amplifier stage. Depending how the meter is connected common mode ripple from the supply could cause some trouble.

As a first point is might be a good idea to check whether the 3478 or 3455 is injecting extra AC.  Chances are it's only one of them.
 

Online David Hess

  • Super Contributor
  • ***
  • Posts: 17427
  • Country: us
  • DavidH
Re: DVMs aren't benign
« Reply #8 on: February 03, 2019, 05:47:08 pm »
I agree that sampling the AC waveform could be a good modern way, but I would expect sampling after an initial buffer / amplifier stage. Depending how the meter is connected common mode ripple from the supply could cause some trouble.

Since the impedance buffer requires good DC accuracy, it likely uses autozero or chopper stabilization although this is not an absolute requirement for good DC accuracy.

I was thinking about common mode noise also.  The low side inputs are galvanically isolated from earth ground however they still have considerable capacitance to earth ground so there is potential for interference there if the grounding situation is marginal.

It is not strictly connected with the subject but the only difference between the old 4-1/2 digit Siliconix LD120 and LD122 integrating converters was the *lack* of the input buffer in the LD122; it only retains the automatic zero switching.  The LD122 provided better performance because it supported using an external buffer with a higher performance than could be provided with a CMOS process like an AD542.  But this chipset still used automatic zero to correct the offset of the external buffer; the major advantage was improved noise and linearity.
« Last Edit: February 03, 2019, 06:47:51 pm by David Hess »
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 15148
  • Country: de
Re: DVMs aren't benign
« Reply #9 on: February 03, 2019, 07:48:52 pm »
Sampling AC measurement does not need a DC precise buffer - one can use digital correction for any offsets with very little effort.
AFAIK the fast amplifier in the 3458 does not use AZ and I won't expect it with other digital AC implementations. It is more likely to exist with analog RMS. However analog RMS is limited at small amplitude anyway - so DC offset is likely not that critical.

For the common mode it is not only capacitance to ground. There is also common mode capacitive coupling to the driving power.
 

Offline Doctorandus_P

  • Super Contributor
  • ***
  • Posts: 4004
  • Country: nl
Re: DVMs aren't benign
« Reply #10 on: February 03, 2019, 09:16:09 pm »
I think that Dave did a trouble shooting video recently of his own uCurrent.
If I remember well, a (substitute) opamp had stability issues when loaded with cables and meter impedances and started oscillating.

So, even when meauring DC, check for weird stuff with your scope.
 

Offline Dave

  • Super Contributor
  • ***
  • Posts: 1356
  • Country: si
  • I like to measure things.
Re: DVMs aren't benign
« Reply #11 on: February 03, 2019, 09:53:20 pm »
I think that Dave did a trouble shooting video recently of his own uCurrent.
If I remember well, a (substitute) opamp had stability issues when loaded with cables and meter impedances and started oscillating.

So, even when meauring DC, check for weird stuff with your scope.
In that particular case, it was not really a problem of opamp choice, but bad circuit design.
<fellbuendel> it's arduino, you're not supposed to know anything about what you're doing
<fellbuendel> if you knew, you wouldn't be using it
 

Offline Conrad HoffmanTopic starter

  • Super Contributor
  • ***
  • Posts: 2077
  • Country: us
    • The Messy Basement
Re: DVMs aren't benign
« Reply #12 on: February 05, 2019, 12:54:47 pm »
Finally got back to this and concluded that the 3478 doesn't contribute much. The shift is caused by the 3455, but I think it's just hum pickup, as properly connecting the shield reduces the error to near zero.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf