EEVBLOG #210 Krohn-Hite Voltage Standard

[gordo51]

Hey what do you know, I have the same unit on my bench except according to the date stamp on the LM741 it was built in 1974 !
Pretty much exactly the same layout of components except I have a Teledyne Philbrick 1703 chopper stabilized op amp module in place of the plug-in board.  Beside the reference zener I have a small paper label stuck on with hand written voltage and current values.  Thanks for showing the schematic.  I was able to adjust pots R1, R3 and R4 so now I have 3 zeros after the decimal point on the 10V range and the other two ranges are spot on.  Of course this is according to my uncalibrated HP3456A voltmeter.  As they say "other voltmeters may give you different mileage..."
Keep up the great work.
Gordo51

PS: I didn't realize it at first but my unit has an unlabelled knob to the left of the range switch that works a potentiometer. I didn't think to check out its effect when I was making my adjustments.  You wouldn't think this would affect the calibration settings?  Wonder what it is for?

[Nermash]

It is a nice catch for 20ish bucks, if you are lucky enough:)

BTW, what do they use to calibrate that 3458A?

[SgtRock]

Dear Gordo51:

--Why is it I cannot find EEVBLOG #210. Do you need a DJ decoder ring, or what?

"The biggest competitive advantage is to do the right thing at the worst time." Bill & Dave

Best Regards
Clear Ether

[McMonster]

You need to look on EEVblog's Youtube channel, there's usually a delay between uploading a new video and posting it on the site.

[SgtRock]

Dear McMonster:

--Good to know. Thanks.

"The telegraph is a kind of a very long cat. You pull his tail in New York and his head is meowing in Los Angeles. Radio operates the same way, except there is no cat." Albert Einstein

Best Regards
Clear Ether

[Kiriakos-GR]

My own voltage - current source, looks to use similar technology.

But those English guys had cover with black hard glue the PCB with the IC and the diode, and so I did not take any pictures.
I did open it up yesterday, so to check it out.

And I managed to tweak it, by adding on the main potentiometer , my decade resistor box in parallel,
and I created one high quality resistor divider, that helps out to set voltages with so great detail speaking about bringing to zero all the digits,
at any voltage between  mv to 8V.

https://www.eevblog.com/forum/index.php?topic=1032.0

I will add pictures about it soon, in the main thread about it. 


 

[Kozmyk]

What about the Mystery Device?
Did you get any closer to identifying it?

[EEVblog]

What about the Mystery Device?
Did you get any closer to identifying it?

Haven't really looked into it, sorry.

Dave.

[amspire]

I think the mystery device is a precision 3 ohm resistor. 4 terminal so two terminals can be used for the series current through the resistor, and two outputs to provide the precise 3 ohm output without any lead resistance error.

[Kozmyk]

Well ferreted that man. 8^)

[saturation]

Nice review!  I didn't know about this vblog until today.  I was bidding on the same model as EDC, last month but I was away on vacation.  Perchance were be bidding on the same device, Dave?  Did you get yours from the USA?

Anyway, I love how its made and Dave's video was immensely helpful.  I didn't know it was so simple inside, its very KISS so its likely to last forever.  Its tolerances are huge, from the casing to the PCB: large and crude but solid as a rock.  Those old IC have such clearly made labels and nice trimming, compare to any LM741 today.   Comparing it the 3458a is the ultimate, so its just fantastic this simple device is so accurate and precise.

I managed to snag one of those today so I'll post how I fair with with.

@Gordo51, I have the same set up as yours for testing the MV106.  But I maintain 4 3456a to cross reference each other for accuracy, its a volt-nut fever I picked up last year. 

[saturation]

Calibrating the 3458a is very difficult, its the ultimate DMM.  Only a Josephson Junction does it justice, so comparing it to a metcal labs 3458a is a very authoritative measurement, assuming their 3458a is properly calibrated & maintained, which on the video appears to be.

Calibrating the 3458a is one reason I don't own one, even if you could find an eBay version of it and patch it up to factory condition, one can't get its famed accuracy without shelling the $1200 to get it calibrated, annually.  Thereafter, it has to be auto-cal everyday and kept in a temp and preferably humidity controlled room to insure its 0.05 ppm dcV transfer accuracy. and 4-8ppm yearly accuracy is maintained.

It is a nice catch for 20ish bucks, if you are lucky enough:)

BTW, what do they use to calibrate that 3458A?

[saturation]

I've just had a short exchange with Joe Inglis at Krohn-Hite about the MV106, and thank them for their assistance!  Anyone looking to service any EDC or KH device would do well to contact Joe.

I've been performance testing this device and its stability is very impressive. 

After 24/7 days of monitoring fluctuations of 1.000 000 VDC at 21^oC,  my 3456a is registering variance of .005 0.0005 x 10^-9 using a 3456a.  This device performs similar to what Dave showed in the video but I've stretched it to 168+ hours and had the 3456a analyze variations continuously using its statistical functions.

This is far better than the spec sheet 1-8 h stability of 5ppm-10ppm + 2uV.   

But based on the above data it could be possible to use the calibrator to its lowest resolution of 10nV.

If you can find this model or its variants on eBay it well worth it to own if you need a voltage reference; components are easily obtained if repair were needed, the device is very easy to service and inspect, its a tribute to the EE who designed it that such stability and accuracy can be made this simply. 

Its a lot more accurate and stable than Geller or voltagestandard.com offerings [ which can still be used as transfer references via snail mail] and a lot easier to maintain than a Fluke 732A series if < 30ppm is adequate.  The manual states its operating environment is 20^oC-30^oC so it was designed to be a field instrument, chances are in a more temperature controlled setting it will be far stabler over time.

[EEVblog]

I've just had a short exchange with Joe Inglis at Krohn-Hite about the MV106, and thank them for their assistance!  Anyone looking to service any EDC or KH device would do well to contact Joe.

Yes indeed, Joe was very helpful in getting me the schematic.

I've been performance testing this device and its stability is very impressive. 

After 24/7 days of monitoring fluctuations of 1.000 000 VDC at 21^oC,  my 3456a is registering variance of .005 x 10^-9 using a 3456a.  This device performs similar to what Dave showed in the video but I've stretched it to 168+ hours and had the 3456a analyze variations continuously using its statistical functions.

Wow, nice to know, thanks.
I presume that others would have similar order stability as well?

If you can find this model or its variants on eBay it well worth it to own if you need a voltage reference; components are easily obtained if repair were needed, the device is very easy to service and inspect, its a tribute to the EE who designed it that such stability and accuracy can be made this simply. 

I concur, an amazing device it seems, belying it's design simplicity.

Dave.

[alm]

After 24/7 days of monitoring fluctuations of 1.000 000 VDC at 21^oC,  my 3456a is registering variance of .005 x 10^-9 using a 3456a.  This device performs similar to what Dave showed in the video but I've stretched it to 168+ hours and had the 3456a analyze variations continuously using its statistical functions.

How much does the 3456A contribute to this? In the past, I've measured short term variance (n=1000) to be about 5 * 10^-15 V² on the lowest DCV range with 100 NPLC and shorted inputs (three orders of magnitude less than your measurement), but I'm not sure how much sample variation there is between multimeters. I would expect long term variance to be larger, since I'd expect there to be some slight drift in the gain, eg. due to tempco.

[amspire]

After 24/7 days of monitoring fluctuations of 1.000 000 VDC at 21^oC,  my 3456a is registering variance of .005 x 10^-9 using a 3456a.

Is that meant to be 0.005 x 10^-3?

[alm]

A variance of 0.005 * 10^-9 corresponds to a standard deviation of about 2ppm, not unlikely given a stable source.

[amspire]

Ah! Variance is a squared function.

I knew that. Once upon a time. 

[alm]

I'm guessing that saturation quoted variance because the HP 3456A can only calculate variance, not standard deviation. A seemingly strange omission. Maybe they didn't have the spare CPU cycles to calculate the square root so many times per second? It's not like today when you might stick an embedded ARM system in it which is idle 90% of the time.

[saturation]

Thanks alm.  I made an order of magnitude error: I missed a zero, the variance on the measurement is 0.0005 x 10^-9 so its more like 1ppm at this point.  Apologies to all for the missing zero; have made corrections.

For comparison, the $1000 eBay Fluke 732A voltage reference has a 10V spec'd stability of 0.3ppm at 30 days and 1ppm at 1 year

http://www.ebay.com/sch/i.html?_trkparms=65%253A12%257C66%253A4%257C39%253A1%257C72%253A5841&rt=nc&_nkw=Fluke+732A&_sticky=1&_trksid=p3286.c0.m14&_sop=15&_sc=1

http://us.flukecal.com/products/electrical-calibration/electrical-standards/734a-reference-transfer-standards?quicktabs_product_details=2

The MV 106 manual specs its best accuracy at 5ppm at 1 hour but rated at wide operating temperatures, so the accuracy could be far better if kept in a stabler test temperature.  Note, the Fluke standard is likewise rated for a wide operating temperature.

http://www.krohn-hite.com/htm/ServiceSupport/PDF/Manuals/MV106J-116J%20Manual.pdf

As for the contribution of the 3456a, its a probabilistic guess, I haven't done the output short on the 3456a for a while to see how my 3456a drifts, but with your data as comparison, I'll let you know in 24h.  So far the measured specs are in line with your measurement [ the actual 3456a contribution is negligible] as the measured 1ppm variation at 1 week is below the 1 day spec best case stability for either the 3456a at 8ppm or the MV106 at 5ppm [not counting the LSD drifts and uV offset on the MV106].

After 24/7 days of monitoring fluctuations of 1.000 000 VDC at 21^oC,  my 3456a is registering variance of .005 x 10^-9 using a 3456a.  This device performs similar to what Dave showed in the video but I've stretched it to 168+ hours and had the 3456a analyze variations continuously using its statistical functions.

How much does the 3456A contribute to this? In the past, I've measured short term variance (n=1000) to be about 5 * 10^-15 V² on the lowest DCV range with 100 NPLC and shorted inputs (three orders of magnitude less than your measurement), but I'm not sure how much sample variation there is between multimeters. I would expect long term variance to be larger, since I'd expect there to be some slight drift in the gain, eg. due to tempco.

[saturation]

For a company its size and age, Krohn-Hite has survived much upheaval in the industry to suggest their portfolio of gear is well respected. 

http://www.manta.com/c/mm71qq0/krohn-hite-corp

The manuals I received from KH suggests the guts of the MV106 is a module for a series of instruments: variants exist with a voltmeter, strip chart recorder outputs, mA calibrator, 1V in lieu of 100mV, nanvolt variant with a 1mV scale instead of 10V ... again a testament to their ingenuity of taking a basic design and likely re-purposed to other arenas.  So, its very likely the accuracy is similar if not better in other variants.

100mV variant


10mV variant


1V variant


Amp calibrator variant:



etc.,

There are hardly any reviews of EDC products or discussions of them [ at least via google].  So, very likely this thread will pop up in the near future when EDC or Krohn Hite are searched for.  This device and its derivatives are worthy of more interest from anyone looking a well made, portable affordable voltage standards that can be priced right for small labs.

I've just had a short exchange with Joe Inglis at Krohn-Hite about the MV106, and thank them for their assistance!  Anyone looking to service any EDC or KH device would do well to contact Joe.

Yes indeed, Joe was very helpful in getting me the schematic.

I've been performance testing this device and its stability is very impressive. 

After 24/7 days of monitoring fluctuations of 1.000 000 VDC at 21^oC,  my 3456a is registering variance of .005 x 10^-9 using a 3456a.  This device performs similar to what Dave showed in the video but I've stretched it to 168+ hours and had the 3456a analyze variations continuously using its statistical functions.

Wow, nice to know, thanks.
I presume that others would have similar order stability as well?

If you can find this model or its variants on eBay it well worth it to own if you need a voltage reference; components are easily obtained if repair were needed, the device is very easy to service and inspect, its a tribute to the EE who designed it that such stability and accuracy can be made this simply. 

I concur, an amazing device it seems, belying it's design simplicity.

Dave.

[saturation]

At this post, with n ~> 12000, the variance of a shorted 3456a input is 2.9 x 10^-14 using PLC 100. [ I have 3, the other two are doing long term measurements of the MV106A and Geller reference.]

It contributes about 0.17 ppm to the variation of the measurement.
 
As aside, I measured the power consumption of the MV106 at 5W using a killawatt.

After 24/7 days of monitoring fluctuations of 1.000 000 VDC at 21^oC,  my 3456a is registering variance of .005 x 10^-9 using a 3456a.  This device performs similar to what Dave showed in the video but I've stretched it to 168+ hours and had the 3456a analyze variations continuously using its statistical functions.

How much does the 3456A contribute to this? In the past, I've measured short term variance (n=1000) to be about 5 * 10^-15 V² on the lowest DCV range with 100 NPLC and shorted inputs (three orders of magnitude less than your measurement), but I'm not sure how much sample variation there is between multimeters. I would expect long term variance to be larger, since I'd expect there to be some slight drift in the gain, eg. due to tempco.

[alm]

What about the 24h drift of the gain? For short term (transfer) measurements you can usually ignore drift and expect noise to limit the accuracy, but I wouldn't expect this to be the case over a 24h period. Any estimates from measuring the references?

[saturation]

I checked the shorted output of the 3456a at n~ 1000 and it read like your ~ 5 x10^-15.  Hours later change in variance suggests the 3456a does drift [ as expected] but its tiny,  and still underneath its rated 10ppm at 24h.   

I think I can separate the noise from the drift component.  Noise causes continuous measurement fluctuations but the net variance mean is ~ 0 [unless its not white noise!], while the drifting shifts the baseline and affects the variance; as analogy noise is to a sine wave riding a DC offset, which represents drift.

I find the biggest cause of drift despite a temperature controlled room remains diurnal shifts caused by day and night in the 24 h period, that's when my measures deteriorated to 2.9 x 10^-14 at 24h, and assuming a stable climate [ i.e., no storms to drastically affect humidity or pressure .]  If you measure n=1000 at night for a few hours or in the day for a few hours assuming a stable period where ambient conditions wont' change drastically, the variances are identical, but if you string the two diurnal changes together,  which includes all the ambient condition shifts, the drift component shows itself. 

What about the 24h drift of the gain? For short term (transfer) measurements you can usually ignore drift and expect noise to limit the accuracy, but I wouldn't expect this to be the case over a 24h period. Any estimates from measuring the references?

[alm]

I think I can separate the noise from the drift component.  Noise causes continuous measurement fluctuations but the net variance is ~ 0 [unless its not white noise!], while the drifting shifts the baseline and affects the variance; as analogy noise is to a sine wave riding a DC offset, which represents drift.

I believe you're using variance in a non-technical sense, this is confusing in this context. Since variance is a function of the square of the deviation from the mean, it will be non-zero for anything but a constant signal (pure DC). The mean of a white noise process is zero, however, and the variance is finite.

I find the biggest cause of drift despite a temperature controlled room remains diurnal shifts caused by day and night in the 24 h period, that's when my measures deteriorated to 2.9 x 10^-14 at 24h, and assuming a stable climate [ i.e., no storms to drastically affect humidity or pressure .]  If you measure n=1000 at night for a few hours or in the day for a few hours assuming a stable period where ambient conditions wont' change drastically, the variances are identical, but if you string the two diurnal changes together,  which includes all the ambient condition shifts, the drift component shows itself.

Of course both the DMM and the voltage reference will be affected by this. So taking samples at 24h intervals would give a lower variance? The drift appears to be dominated by temperature coefficient? If you could measure just the effect of the DMM, you could correlate the ambient temperature with the drift of the DMM, and attempt to correct for this. This is cheaper than to get a proper climate-controlled room .

Guess this is why 24h specs are for 23°C ± 1°C.