3458A Cal Results

[dr.diesel]

A while back I scored a couple 3458As dead on ebay, recapped/replaced all the NVRAMs, and did the R411 60C mod.  I did an initial cal utilizing my unknown eBay 731B and a 10K PWW I already had from Edwin, so the as-found is to be expected.  Both 3458As sat powered for several months before Cal once alive again.

Well last week it turns out I was headed directly past the cal lab that our friend @CalMachine works at, ENI labs in Fort Wayne IN.  I just picked everything back up yesterday and thought I'd share my experience and documentation provided by them.  I also included my 731B and asked them just to compare it to their 732B, no adjustment.

Overall the experience was great, ENI has a really nice facility and tons of equipment, which they are expanding.     ENI is also willing to accept DIYer gear, so if you have fancy homebrew reference and would like it compared to a known 732B, don't hesitate to contact @CalMachine.

Attached are the cal results, I need to further study how the uncertainty values are derived.

[lukier]

Thanks for sharing. Are you going to repeat these calibrations next year (to see the drift)?

A while back I scored a couple 3458As dead on ebay

Wow. Recently I was bidding on a "parts or repair unit" but I was outbid in the end - it went for $2600. IMHO that's too much for an untested unit (as it might have ADC drift or other expensive problems).
It seems it is very hard to get 3458A, even broken one, and then one must add duties, postage and possibly $1300-$2700 for Keysight parts or repair fee.

[dr.diesel]

Thanks for sharing. Are you going to repeat these calibrations next year (to see the drift)?

Yup, annual until I have a reasonable trend, then perhaps every 2 years.

[HighVoltage]

lukier
If you wait long enough, you can get lucky. In my case, I almost waited patiently for 2 years to find a good 3458A

dr.diesel
Thanks for sharing your cal docs.
Do you know what the real 10V value of the 732B was, when your 731B was tested?

[CalMachine]

Do you know what the real 10V value of the 732A was, when your 731B was tested?

The last transfer with Fluke resulted in a value of 9.9999615,  +/- 0.06ppm back in February.  The 732B is back in Fluke's possession for another transfer measurement.  Hopefully, I'll have it back in my hands again sometime next week!

[HighVoltage]

The last transfer with Fluke resulted in a value of 9.9999615,  +/- 0.06ppm back in February.  The 732B is back in Fluke's possession for another transfer measurement.  Hopefully, I'll have it back in my hands again sometime next week!

9.999,961,5 V for your 732B
9.999,803,0 V for D.Diesel' 731B

Not bad for this old 731B

Will your 732B ever be adjusted at the Fluke facility, like closer to 10V exactly or just kept at this value for tracing and comparison reasons?

[CalMachine]

The last transfer with Fluke resulted in a value of 9.9999615,  +/- 0.06ppm back in February.  The 732B is back in Fluke's possession for another transfer measurement.  Hopefully, I'll have it back in my hands again sometime next week!

9.999,961,5 V for your 732B
9.999,803,0 V for D.Diesel' 731B

Not bad for this old 731B

Will your 732B ever be adjusted at the Fluke facility, like closer to 10V exactly or just kept at this value for tracing and comparison reasons?

This unit will, most likely, never be adjusted.  With as many transfers as we've had on this in the past, we develop a linear regression model to better predict the unit's drift and true value over time. 

The only reason why I could see having our 732B adjusted to 10V nominal is if something were to happen to the unit to the point were its history is no longer applicable. 

[ManateeMafia]

Are the 731B measurements made over a period of several days or is it done in a single day?

I think some labs measure over several days but I don't know if they average the readings or if they test for stability and then take a final set of measurements on the last day.

[CalMachine]

Are the 731B measurements made over a period of several days or is it done in a single day?

I think some labs measure over several days but I don't know if they average the readings or if they test for stability and then take a final set of measurements on the last day.

I had let the 731B acclimate to the environment for a good 2 days.  I took multiple sets of 30 measurement samples every day, for 3 days.  Once I felt the mean and std deviation of the measurements were no longer dropping, and the reference had settled to the environment, I took the average of 3 30 measurement sets.

Improvements to my method are being made, as well! 

[Moon Winx]

There is an active debate of whether or not it's good to take several 732 measurements over several days or to take a single measurement the day it comes in and ship it right back out. The output of a 732 over time has a definite sinusoidal pattern laid over a linear drift line. The sinusoidal pattern is a pretty good match to seasonal relative humidity throughout the year. It has been observed that it takes about 40 days for the 732 to acclimate to the local humidity. So there is an argument that the best thing to do is calibrate it in place (these seasonal drifts can be predicted and corrected for), rather than send it anywhere. If you have to send it, get it back to within its environment as soon as possible.

[Moon Winx]

Also, I noticed you are performing these measurements at freezing temperatures. That must be one hell of an air conditioning system in your lab!

[Muxr]

Also, I noticed you are performing these measurements at freezing temperatures. That must be one hell of an air conditioning system in your lab!

hehe, -4 C ouch. That would be cold, below freezing.

[Tony_G]

Out of interest, why do the min, max numbers in the ENI report differ from that in a Keysight one?

For DC volt gain, Keysight:

1 V  min 0.99998942 max 1.00001024

For ENI:

1 V  min 0.99998900 max 1.00001100

Just wondering as I'm a calibration noob.

TonyG

[dr.diesel]

Also, I noticed you are performing these measurements at freezing temperatures. That must be one hell of an air conditioning system in your lab!

Swap F for C, that was actually fixed in the final documentation, sorry, forgot to note that above.

[Moon Winx]

Out of interest, why do the min, max numbers in the ENI report differ from that in a Keysight one?

For DC volt gain, Keysight:

1 V  min 0.99998942 max 1.00001024

For ENI:

1 V  min 0.99998900 max 1.00001100

Just wondering as I'm a calibration noob.

TonyG

I'm sure CalMachine can explain, but until then I'll butt in. Determining limits for a test point is a very complicated process and, in the end, is a mixture of science and art. There are proper universal ways to do this, but there are at least two steps in the accepted process that are completely determined by the lab: guardbanding technique and a subjective confidence coefficient for each component of the measurement uncertainty.

The guardbanding technique changes the test limits based on your measurement uncertainty, and the measurement uncertainty is calculated in real-time and takes into account the stability of the measurement, so it would actually be a red flag to me if you had constant test limits from test-to-test, let alone from lab-to-lab.

[Moon Winx]

...and another reason, the report states the test limits are stated at a 95% confidence level and perhaps Keysight states theirs at a different confidence level. Both sets of limits may be stated at different confidence levels but are actually describing the same data, or the same statistical distribution. Kind of like how you can swap between units degrees C and degrees F, but in the end you are describing the same temperature.

[CalMachine]

There is an active debate of whether or not it's good to take several 732 measurements over several days or to take a single measurement the day it comes in and ship it right back out. The output of a 732 over time has a definite sinusoidal pattern laid over a linear drift line. The sinusoidal pattern is a pretty good match to seasonal relative humidity throughout the year. It has been observed that it takes about 40 days for the 732 to acclimate to the local humidity. So there is an argument that the best thing to do is calibrate it in place (these seasonal drifts can be predicted and corrected for), rather than send it anywhere. If you have to send it, get it back to within its environment as soon as possible.

Noted!  I will look into this further, if possible.

Also, I noticed you are performing these measurements at freezing temperatures. That must be one hell of an air conditioning system in your lab!

 

What can I say...  I like it cold!   Lol!  Our software defaults to ºF and I had forgotten to change it to ºC.

Out of interest, why do the min, max numbers in the ENI report differ from that in a Keysight one?

For DC volt gain, Keysight:

1 V  min 0.99998942 max 1.00001024

For ENI:

1 V  min 0.99998900 max 1.00001100

Just wondering as I'm a calibration noob.

TonyG

Moon Winx explained it very well.   There could be a multitude of reasons why there are differing tolerances.  Confidence levels, Calibration Cycle, Guardbanding, even specific customer's needs.

TonyG were those tolerances stated on a traceable cal?  What's a little funny is, Keysight's stated tolerances are below that of the 3458A's stated specifications.  Interesting!  I'd like to know how they determined that

[Tony_G]

Yes they were. I purchased a 3458A at a great price and wanted to get it under a service contract. To do that it had to be calibrated so I sent it to their Loveland facility.

I have to say that they were absolutely fantastic to work with. Very impressed by them. I'd send you the cert PDF but I'm on vacation and the Keysight infoline site is down now.

TonyG

[PartialDischarge]

My unit arrived yesterday from ap. Attached are the cal results. Played a bit with the unit, after 3h warmup TEMP? reads 43,1º, but today is not that hot in Madrid.
This is what a 0.005% 10k vishay reads after ACAL DCV and ACAL OHM. Adrian told me that he will soon have his heated resistor standard calibrated with less that 1ppm uncertainty.
Will use the 3458 first to calibrate the 335A which is still very accurate but can do better, given that it was probably last calibrated 20 years ago.

[TiN]

What is the magic point in hiding 2823 in S/N number? It's all same for all US-made old HP 3458A's. 
Interesting calibration report, few things cought my eye:
* 90 day spec used as verification criteria. I'd rather expect 24 hour spec, however that is harder to reach for unknown history unit
* ANA mode for ACV instead of higher accuracy SYNC.
* AC HF calibration was not performed, based on equipment list.
* Giggle at using Rb for frequency check of 3458A.

As of 10K reading, that's likely not the true one. Execute OCOMP ON, DELAY 5, OHMF 10E3 , connect 4W, wait for hour or two, and then we see  . Yes, it's slow  .

Thanks for the work, MasterTech and Adrian  .

[Tony_G]

Yes they were. I purchased a 3458A at a great price and wanted to get it under a service contract. To do that it had to be calibrated so I sent it to their Loveland facility.

Here is the official doc from Keysight.

TonyG

[CalMachine]

Yes they were. I purchased a 3458A at a great price and wanted to get it under a service contract. To do that it had to be calibrated so I sent it to their Loveland facility.

Here is the official doc from Keysight.

TonyG

You're lucky!!  N6004465  Is their best metrologist they have.   

Thanks for the cert though.  I wonder what they do for an accredited cal.

[alm]

Yes they were. I purchased a 3458A at a great price and wanted to get it under a service contract. To do that it had to be calibrated so I sent it to their Loveland facility.

Here is the official doc from Keysight.

I assume, from the equipment list, that this is for the Keysight STE9000 'green checkmark' calibration, rather than the much more expensive calibration in their standards lab? Both are done at Loveland, but they do not offer the same performance (or price).

[Tony_G]

Yes, this is the basic Cal (no uncertainties, no guard banding, not in the standards lab). Price was around $600. I needed to do this to get it under the extended warranty but I also wanted to see how it changes from year to year so I'll do it again this year and see.

[alm]

Not sure if you can really rely on the data they provide to track drift. See these posts from the volt-nuts list for more detail about the calibration services. I am sure it will meet its specifications, but it sounds like they are only barely meeting the necessary test uncertainty ratio, so any inference beyond pass/fail is problematic.