Old Multimeters with Standard Deviation

[David Hess]

What are some old multimeters that support calculating and displaying the standard deviation of a DC measurement?

I want to make low frequency spot noise measurements.

[macboy]

I have a Keithley 194A which can do this. It takes up to 32k x 16 bit samples at up to 100kHz. It can then calculate various things: average, RMS, StdDev, integral, derivative, peak, and others. Your application may be limited by the 16 bit resolution and inherent noise.

[Vgkid]

Haven't used it in a few years. But I believe my hp 3456A/34401A have it.

[bdunham7]

Haven't used it in a few years. But I believe my hp 3456A/34401A have it.

On the 34401A standard deviation doesn't come standard, but it can be enabled with the following instructions:

DIAG:POKE 28,0,1191 MATH Menu -> MIN-MAX: adds Standard Deviation and Peak-to-Peak

[bdunham7]

What are some old multimeters that support calculating and displaying the standard deviation of a DC measurement?

I want to make low frequency spot noise measurements.

There might a few that do it from the front panel, but AFAIK the typical way was acquire samples as fast as possible via GPIB and do all the calculations later and elsewhere.  As mentioned, the 34401A appears to have this function although I have not tried it.  If you contemplate acquring one I can try it out for you first.

What bandwidth, sample rate and aperture are we talking about here?

[David Hess]

I have a Keithley 194A which can do this. It takes up to 32k x 16 bit samples at up to 100kHz. It can then calculate various things: average, RMS, StdDev, integral, derivative, peak, and others. Your application may be limited by the 16 bit resolution and inherent noise.

I did not get that far going through the old Keithley meters because I recognized it as being the high speed model.  It might be a good default option if I cannot find anything better since it has real bandwidth specifications.  I wonder what its low frequency noise spectrum looks like.

Haven't used it in a few years. But I believe my hp 3456A/34401A have it.

I was surprised that the HP multimeters following the HP 3456A could not do it so I did not search back that far.  The HP 3456A can calculate the variance so standard deviation is one square root away.

I did not find anything about the HP 34401A calculating the standard deviation in the manual, but see below.

On the 34401A standard deviation doesn't come standard, but it can be enabled with the following instructions:

DIAG:POKE 28,0,1191 MATH Menu -> MIN-MAX: adds Standard Deviation and Peak-to-Peak

Why did HP hide this?  I did not find anything about it.

There might a few that do it from the front panel, but AFAIK the typical way was acquire samples as fast as possible via GPIB and do all the calculations later and elsewhere.  As mentioned, the 34401A appears to have this function although I have not tried it.  If you contemplate acquring one I can try it out for you first.

It may come down to transferring the data and doing the calculation on a computer if I have to do it that way.  I am initially looking for something simpler.

What bandwidth, sample rate and aperture are we talking about here?

Sample rate can be largely ignored assuming that it is fast enough because it does not affect standard deviation except in how it limits the aperture (integration time).

Except for the Keithley 194A high speed multimeter discussed above, none of the specifications I have gone through  give bandwidth for DC measurements.  Excluding limits by the signal conditioning, bandwidth is limited by the integration time (aperture) which produces a nonlinear sinc frequency response roughly equivalent to 0.468/t.  For a 1 power line cycle measurement at 60 Hz, that will be 28 Hz.

For a spot noise measurement, the 28 Hz bandwidth limit from a 1 PLC measurement can to be taken into account as sort of an extra pole in the filter response.  If a faster sample rate with a shorter integration time was used, then the increase in bandwidth would make this step unnecessary.

So I want a sample rate of at least 1 PLC, but would not object to something which is faster.

[tggzzz]

HP/Agilent 34410A has it as standard, along with min, max, mean, and count on its small second display. Scroll between them.

I would be miffed if later HPAK equivalents didn't  have it as well.

It might be possible to log the output from any meter, and use a spreadsheet or or stats language to post process results. That wouldn't  preclude other processing, e.g. using a graph to spot trends or popcorn noise, or Allen deviation, or FFT.

[alm]

In addition to the HP 3456A, the 3457A can calculate standard deviation. Might be available cheaper than the 3456A and shipping should certainly be cheaper since it's smaller (less deep, though still 19" wide) and much lighter.

The Prema 6031 (and probably other models from that generation) has about 20 different mathematical functions including standard deviation and variance, though you have to memorize the program number or tape that page of the manual on/near the meter. Prema meters are not for sale very often, though.

[Stray Electron]

What are some old multimeters that support calculating and displaying the standard deviation of a DC measurement?

I want to make low frequency spot noise measurements.

  I think that some of the Keithley meters like the 194, 195 and 196 will do that.  I know that you can also set the sampling rate in the 194 but I'm not sure about the others.  The manuals for all of those are online and are very good. Take a look.

[mawyatt]

Haven't used it in a few years. But I believe my hp 3456A/34401A have it.

On the 34401A standard deviation doesn't come standard, but it can be enabled with the following instructions:

DIAG:POKE 28,0,1191 MATH Menu -> MIN-MAX: adds Standard Deviation and Peak-to-Peak

Yes, these marvelous instruments had many "hidden" features, including an "extra" digit, temperature, extra math functions, and so on. We've enabled these on one of our 34401As, the other is too old and doesn't have the features hidden within.

Best,

[EC8010]

The HP/Agilent 34410A does it, buried in the "Math" menu, but it's a little awkward to use (I've used it). The Agilent/Keysight 34461A does it in its "Math", "Statistics" menu and will display a histogram under its "Display" menu. Having the histogram as well as the standard deviation is extremely useful because it will reveal whether what you are looking at is genuinely a Gaussian distribution or whether it drifts from one to another or hops (popcorn noise) between levels.

[bdunham7]

Sample rate can be largely ignored assuming that it is fast enough because it does not affect standard deviation except in how it limits the aperture (integration time).

For a spot noise measurement, the 28 Hz bandwidth limit from a 1 PLC measurement can to be taken into account as sort of an extra pole in the filter response.  If a faster sample rate with a shorter integration time was used, then the increase in bandwidth would make this step unnecessary.

So I want a sample rate of at least 1 PLC, but would not object to something which is faster.

You can't ignore the sample rate if you are doing math internally on an older unit--the processing time becomes the bottleneck.  I gave it a try on my 34401A and depending on your exact expectations, it sort of works but with some interesting issues.  You'll have to carefully characterize whatever you use unless you find something fully specified.

I started by using an AWG and a calibrator to send 1V_RMS to the meter at 30, 40, 50 and 60Hz.  I used the fast 4.5-digit mode which provides 1000 samples/second internally and over GPIB, but much less over RS232 or when doing the MIN/MAX/AVG/SDV/PP recording internally.  The aperture is 400µS.  For whatever reason, the actual sample rate varied from 50-100 samples per second, and that seemed to depend on both the range selected and the input frequency.  The latter baffles me as no changes were made to the meter settings.

I don't know what voltage levels you want to measure, but 30Hz 5.0mV signal was measured with less than 1% error in 10 seconds with about 1100 readings.  My tests at 1V showed good results at 30Hz, but at 40Hz I had a +2% error which grew to +17% at 50Hz and then back to 2% at 60Hz.  I'm in the US so it's 60Hz all around me.  I tried letting it run for a full minute each time and got the same results with 3000-4000 readings.

I didn't get the custom aperture to work, apparently I need some undocumented commands to set that up.  The next slower built-in aperture is 3ms and the effective sample rate drops to 20 per second or so.  At 30Hz I got a large -40% error, at 10Hz about -5%, at 5Hz -2% and at 2Hz the error was less than 1%.  That was all 5mV_RMS and a full minute of sampling.  So it looks like a custom aperture might actually be quite handy.  I'll have to look into setting that up unless someone here as already figured that out and can tell me.

[iMo]

Yep, thanks a great work of several contributors here there is a long thread on the 34401A hidden menus.
https://www.eevblog.com/forum/testgear/hp-agilent-34401a-hidden-menu/msg3228606/#msg3228606

[J-R]

Maybe not old enough, but Keithley 2015 and 2010...

[alm]

Maybe not old enough, but Keithley 2015 and 2010...

True, and the somewhat more common Keithley 2000. But these aren't that old in my eyes, which probably ages me . My interpretation of the question was something old that can be picked up cheaply, and I think eBay prices for the Keithley 20xx series generally start at a few hundred USD.

[mawyatt]

Yep, thanks a great work of several contributors here there is a long thread on the 34401A hidden menus.
https://www.eevblog.com/forum/testgear/hp-agilent-34401a-hidden-menu/msg3228606/#msg3228606

Exactly, really fun old DMMs that's becoming a collectors item, maybe already is!! Those folks discovered/found out how to unleash these hidden jewels within the 34401As

Best,

[David Hess]

It looks like the HP 3457A is what I should be looking for.  I studied the operating manual and the designers apparently contemplated exactly what I am trying to do measuring low frequency AC.

The problem with the 34401A is that a reasonably priced used one may or may not have the more recent firmware to support the hidden features.  The discussion thread about the hidden menu and bdunham7's report about the sample rate weirdness convinced me that this is not the best option.

The Keithley 2000 only seems to support calculating the standard deviation of the last 10 buffered readings which is not enough, and it is a manual operation.  If these meters can do what I want, then it is not apparent from their poor documentation.

[tggzzz]

It looks like the HP 3457A is what I should be looking for.  I studied the operating manual and the designers apparently contemplated exactly what I am trying to do measuring low frequency AC.

The problem with the 34401A is that a reasonably priced used one may or may not have the more recent firmware to support the hidden features.  The discussion thread about the hidden menu and bdunham7's report about the sample rate weirdness convinced me that this is not the best option.

The Keithley 2000 only seems to support calculating the standard deviation of the last 10 buffered readings which is not enough, and it is a manual operation.  If these meters can do what I want, then it is not apparent from their poor documentation.

Not a 33410a? Those have std dev within the stats functions shown on their smaller/second display, along with min, max, mean and count (i.e. number of readings)

[David Hess]

Not a 33410a? Those have std dev within the stats functions shown on their smaller/second display, along with mim, max, mean and count.

Did you mean 34410a?  It does look like a good solution but for used, they sell for 2 or 3 times the price of the 3457a on Ebay and I think a 34410a would be more difficult to maintain.

[tggzzz]

Not a 33410a? Those have std dev within the stats functions shown on their smaller/second display, along with mim, max, mean and count.

Did you mean 34410a?  It does look like a good solution but for used, they sell for 2 or 3 times the price of the 3457a on Ebay and I think a 34410a would be more difficult to maintain.

Oh, picky picky picky

I have no comment about price, and fortunately I have knowledge of maintainability.

If price is a factor, then I'd look at logging/streaming results over RS232/GPIB/Ethernet into a .txt file, and post process in a spreadsheet.

I've used that trick to check I understood what my 7081 was doing, to spot 1.5ppm zener popcorn noise, and to see the lack of noise in a Weston Standard Cell.

[David Hess]

If price is a factor, then I'd look at logging/streaming results over RS232/GPIB/Ethernet into a .txt file, and post process in a spreadsheet.

I've used that trick to check I understood what my 7081 was doing, to spot 1.5ppm zener popcorn noise, and to see the lack of noise in a Weston Standard Cell.

I do not really object to sending the results directly to a computer, but would prefer a real time readout no matter where this happens.  If I did this, then it would also leave open the possibility of doing a real time low frequency FFT.  However I can settle for a real time spot noise measurement on the multimeter itself which should work on any multimeter which can display the standard deviation.  I would most likely use my laptop which runs Windows 10 if I went that way.

The last time I considered this sort of thing, I was working in Mathematica but never got to the point of trying to interface it, which might not have been possible anyway.  I guess back then Labview was popular but I never got as far as trying it.

[bdunham7]

I do not really object to sending the results directly to a computer, but would prefer a real time readout no matter where this happens.  If I did this, then it would also leave open the possibility of doing a real time low frequency FFT.  However I can settle for a real time spot noise measurement on the multimeter itself which should work on any multimeter which can display the standard deviation.

At least for the 34401A, the MIN/MAX/AVG/SDEV/PP readings are not displayed in real time but have to be manually (or remotely) gated and then read afterwards.  I don't know how it works on the 3457A. 

I did get this whole thing to work pretty well with reasonable-ish gate times.  I set the custom aperture (via remote RS232 connection--there's no way from the front panelsee Hydron's post below) to 1ms [sens:volt:aper .001] and was able to take all those readings at any frequency I was willing to wait long enough for.   The noise floor seems to be ~900nV or so.  This effectively adds an ultra-low-frequency TRMS AC+DC function to the meter since AVG becomes the DC component and SDEV the AC.

Your comments about buying a 34401A for this are probably right--it's an undocumented and unspecified feature that may have unknown pitfalls and good 34401A units are a bit expensive, although perhaps still worth it.  But for those of us that already have a 34401A, it's a nice bonus feature to run across.  I may take some time and try to characterize it better since I haven't seen anyone do a detailed rundown on this feature.  Obviously one big issue is that you lose your normal-mode rejection of the power line field, another is that if the trigger periods are too slow and not randomized there could be an issue with some sort of aliasing. 

I don't have a working complete universal GPIB setup and have to use RS232 for this meter, but if you can get whatever meter you obtain set up and using HKJ's Test Controller software, it has a nice interface and you can do the math there.  I'm pretty sure it can give you a running STDEV as well as a gated one.

[David Hess]

I think the 34410a resolves the problems of using a 34401A, but it looks to be 2 to 3 times more expensive than a 3457a which should be adequate.

If I wanted to spend less on a new instrument, then Siglent makes one, but my experiences with Chinese instruments has been poor.  Their manual made me question their quality.

[mawyatt]

David,

Our experience with Siglent has been just the opposite, and we were reluctant to acquire from them, but decided to get the  SDS2000X+ DSO a few years ago. Out past history has been only to "trust" HP/AG/KS, Fluke and Tektronix, and a few others back then, so forking over our own $ now retired, was well a little risky. Now we have a number of Siglent instruments, DSOs, SA/VNA, AWGs, EL, PSs and a SDM3065X DMM.

All Siglent Instruments have performed well and continue to do so. The SDM3065X is "measured" against the old HP34401A, AG34401A, 3 KS34465As and a Keithley DMM6500. It's a good value, but not a substitute performance-wise for the other mentioned DMMs. Probably the best value overall "new" DMM in this upper end category tho, IMO.

Of course YMMV,

Best,

[Hydron]

RE the custom aperture thing on 34401A, it can be done from the front panel - once custom aperture is enabled via the POKE command from the other thread you need to select "CUSTOM APER" in the resolution menu (the one with slow/fast 4/5/6 digits etc), then a new "APER TIME" submenu appears under the measurement menu to set the aperture time.

Note that it seems that setting the custom aperture decouples the number of displayed digits from the integration time - e.g. you can now have 6 digits with 1 PLC instead of 10 if you want (or even less if you want the last few digits to be a random number generator).

All this works only on FW 7 and above, as mentioned by others, older units won't have the fun hidden goodies.

[iMo]

FYI - NPLC vs Aperture on the 34401A

https://www.keysight.com/us/en/lib/resources/training-materials/adjusting-nplc-and-aperture-to-make-high-speed-measurements.html

Specifying integration time using NPLCs is more commonly practiced because it executes faster and is simpler to configure than using Aperture, which requires careful calculation. Aperture commands allow a 4us resolution, which can help you specify a specific integration time between 300us and 1s. However, unless there is an exact integration time you wish to achieve, the NPLC setting can accomplish your task with greater ease. When using Aperture mode, NPLC mode is disabled and vice versa.

Afaik with NPLCs < 1 the 50/60Hz rejection does not matter.
With 50Hz the "aperture time" should be set always multiple of 20ms (in order to suppress the 50Hz), afaik.
From 300us to <20ms aperture time the 50Hz will be not suppressed.

[Kleinstein]

The HP34401 can do faster than 1 PLC, but this comes with a reduced resolution and of cause no more 50/60Hz suppression.
For the speed one also has to look at the auto zero setting: with AZ active the total conversion takes twice the time (for a input reading and a zero rading). With some other old meters (e.g. Keithley 19x) the AZ mode may take 3 actual conversion and thus 3 x20 ms (for 50 Hz 1 PLC). Chances are one would want the non AZ mode for the fast readings.

The performance in the 100 mV range may not be that good (drift and LF noise) in non AZ mode.

The Sigilent and Rigol meters based on a SD ADC chip tend to not suffer much from non using the AZ mode. For the specific task of fast reading and calculating RMS they can be a good choice (even the SDM3055). The weak point is more with long term drift and some software issues that to a large part seem to be solved for the Sigilent side.

The 34410 is internally quite different from the 34401, it is more like the modern 34461 with an old style display.