Fluke 732B repair.

[Bill158]

I shipped mine to Tektronix with a new battery and the in-cal light out. They adjusted the output as close to 10V as possible, or the output just happened to be within .2 ppm of nominal. They noted the light was out but they never noted it was adjusted.

I didn't think about the "in-cal" light being on with my 732B.  I would bet that Fluke made the assumption that I had a "cal" from somewhere and didn't want to adjust it further so that I could keep a record of "drift".  I would guess that I should have asked to have the 732B be adjusted to exactly 10.000000 as Fluke perceived 10 volts with their local standards which was a bank of 4 732Bs which seems to be "certified"  every 3 months.  This is simply a "philosophy" in how to calibrate or certify.  Since I have no requirements to maintain the "cal" seals I have opened it up and looked at the adjustment switch settings and they were set to 2755 which is exactly the same as when I shipped it to Fluke.  Thanks for your experience.
Bill

[dacman]

Fluke has a document (on the Fluke website in the 732B documents section) on 0.3 PPM accuracy for the 732B.  It basically states that it takes 7 calibrations to reach this.  These would be without adjustments.  These calibrations would be tracked and trended.  If the unit were to be adjusted, it would be back at calibration number one.  With one calibration, the accuracy at 1 year would be 2 PPM (from the manual) combined with the uncertainty of the calibration.  If the standard is tracked, then it could be shown to an auditor what the stability actually is, what the trend is, and statistical analysis could be employed to calculate a predicted value and the uncertainty of that value.  We could claim 0.3 PPM uncertainty where I work.  If I were to send a unit off, and it were to arrive dead, they'd better not adjust it.  We have data on units dating to the 1990's.  (The last time I checked, Fluke with the JVA was NVLAP accredited and Fluke with the bank of 4 732Bs was A2LA accredited.  They are not going to adjust your zener, and it has nothing to do with the JVA.)

[plesa]

Fluke has a document (on the Fluke website in the 732B documents section) on 0.3 PPM accuracy for the 732B.  It basically states that it takes 7 calibrations to reach this.  These would be without adjustments.  These calibrations would be tracked and trended.  If the unit were to be adjusted, it would be back at calibration number one.  With one calibration, the accuracy at 1 year would be 2 PPM (from the manual) combined with the uncertainty of the calibration.  If the standard is tracked, then it could be shown to an auditor what the stability actually is, what the trend is, and statistical analysis could be employed to calculate a predicted value and the uncertainty of that value.  We could claim 0.3 PPM uncertainty where I work.  If I were to send a unit off, and it were to arrive dead, they'd better not adjust it.  We have data on units dating to the 1990's.  (The last time I checked, Fluke with the JVA was NVLAP accredited and Fluke with the bank of 4 732Bs was A2LA accredited.  They are not going to adjust your zener, and it has nothing to do with the JVA.)

Interesting reading. I did not found how long and how many calibration Fluke is performing for example on their 732B/H or on 732B/C where drift rate should be provided.
Based on my last few month investigation it is quite difficult to find used 732B on market.
Fluke also did not respond to quotation request, but I will expect something 4-6 kUSD (depends on the version).
What was your cost of new or used unit?

[Bill158]

Fluke has a document (on the Fluke website in the 732B documents section) on 0.3 PPM accuracy for the 732B.  It basically states that it takes 7 calibrations to reach this.  These would be without adjustments.  These calibrations would be tracked and trended.  If the unit were to be adjusted, it would be back at calibration number one.  With one calibration, the accuracy at 1 year would be 2 PPM (from the manual) combined with the uncertainty of the calibration.  If the standard is tracked, then it could be shown to an auditor what the stability actually is, what the trend is, and statistical analysis could be employed to calculate a predicted value and the uncertainty of that value.  We could claim 0.3 PPM uncertainty where I work.  If I were to send a unit off, and it were to arrive dead, they'd better not adjust it.  We have data on units dating to the 1990's.  (The last time I checked, Fluke with the JVA was NVLAP accredited and Fluke with the bank of 4 732Bs was A2LA accredited.  They are not going to adjust your zener, and it has nothing to do with the JVA.)

Interesting reading. I did not found how long and how many calibration Fluke is performing for example on their 732B/H or on 732B/C where drift rate should be provided.
Based on my last few month investigation it is quite difficult to find used 732B on market.
Fluke also did not respond to quotation request, but I will expect something 4-6 kUSD (depends on the version).
What was your cost of new or used unit?

I have read and reread that document many times.  Ray Kletke really gets into environmental conditions that you could only detect changes to the 732B output by having a JJVS sitting in the same room and being able to access the 10 volt output whenever you wanted.  But his basic information about having 4 or more 732B/A units to inter-compare can give you the overall drift slope of each unit.  Repeated "calibrations" at Fluke then can also give you the drift of the single 732B that you return at regular intervals.  I have used his basic ideas, without a regular calibration at Fluke for over 10 years, and managed to be only 2.3 ppm high according to the value of my 732B when returned last August from Fluke.  Since up to this time I only had 732A references I could not return one to Fluke WA. because of the lack of battery time involved (12 - 15 hours).  By using the NBS/NIST technical note 430 and the technique described in "Calibration-Philosophy in Practice Volume 1 (Fluke)" pg. 34 I was able to predict changes during those 10 years.  My offset from my 732B after cal could also be error introduced by using the local Fluke Cal Lab, when it was local, and the 732A could be transported by hand to and from the lab and battery life was not a consideration.  However that local Lab was one more 732B/A cal away from Fluke WA. because their units were sent in on a regular basis.  So all in all I was very happy with my results.  I have 10 years worth of delta readings between all units and have a very good regression slope for all units.  But I agree that I lack the absolute knowledge of which unit was drifting other than making an intelligent interpretation of the data, as described in the Philosophy in Cal book.  It should be interesting to see where my 732B is next time I send it in for Cal.
I got my 732B off of ebay for US$2,000.  I was very happy that it worked, almost, correctly.  There was a problem with isolation from guard/ground which turned out to be a missing thermal insulator pad under the voltage regulator U504 which allowed chassis ground to be connected to the Guard/Low of the 732B.  This was probably someone's failed attempt to remove the battery by trying to take the A5 Power Supply PCA off and getting to the battery that way.  They obviously didn't have the 732B manual!  The insulator was just laying on the A5 PCA next to U504.  But I did consider a lot of other 732B units that were on ebay before buying this one.  I sure haven't seen any come up on ebay for a while.

Bill

[dacman]

I don’t know how much our 732Bs cost (they were there when I got there).

NBS TN 430 shows how to do the intercomparisons.  The document from Fluke shows how to do the trend uncertainty calculations.  Over time, the calculation for uncertainty of the projected value of the 732B will get larger and larger, and once the uncertainty gets to a certain value, the unit can be recalibrated.

Intercomparisons can not only detect anomalies with the units, but it is also used to lower the uncertainty using pooled uncertainty.  The 732B values used in calculations are the predicted values (today's values) from the trend charts (with charted calibrations of the units).  After intercomparisons, pooled values are assigned.

For calibrating a Fluke 5720A or 5730A (or even 5700A), the 10 V output of the 732B needs to be known to within ± 1.5 PPM.  For calibrating a Keysight 3458A, it needs to be known to within 2 PPM.  Where I work, we have to know what the 10 V value is and what its uncertainty is, which is statistically calculated.

[Bill158]

Interesting reading. I did not found how long and how many calibration Fluke is performing for example on their 732B/H or on 732B/C where drift rate should be provided.
Based on my last few month investigation it is quite difficult to find used 732B on market.
Fluke also did not respond to quotation request, but I will expect something 4-6 kUSD (depends on the version).
What was your cost of new or used unit?

The last price list I can find for a new Fluke 732B was 4/2001. It shows the 732B/C at US$5,990 and the 732B/H at US$5,330.  You can bet that these prices would be around US$9k to $10k now.
Bill

[plesa]

Thanks Bill and Dacman for information. I will post actual price when Fluke send me quote. Today I also found lab close to me with JJA, what a surprise.

[dacman]

If you want to know how accurate the 732B can be, NIST is using them to certify some (what NIST calls) CJVAs and PJVAs (Conventional Josephson Voltage Array or Programmable Josephson Voltage Array).  In a North American Interlaboratory Comparison run by NIST, by using a bank of four 732Bs as the traveling standard, agreement to within +0.022 / -0.018 ppm was achieved.  (Using a PJVA as the traveling standard, agreement within 0.0005 ppm was achieved.)  Some labs do not need the accuracy, or do not have the space for another JVA to compare against.  It was stated that this approximately 0.02 ppm comparison was achieved by developing coefficients for the 732Bs, such as for pressure.

Where I work, we get our 732Bs calibrated using the Fluke DVMP program.  They send us one 732B, we make the measurements using our meters, and Fluke gives us a NVLAP certified report.  The last report we got had a stated uncertainty of the measurement of 0.06 ppm.

If you want to know how those seemingly continuous comparisons are made on lab 732As or 732Bs, a system already exists that does that (you don't need to build one yourself).  We have a 32 channel scanner from Data Proof and VoltRef software from Data Proof.  (We got the guarded scanner and I'm glad we did.)

[Bill158]

They send us one 732B, we make the measurements using our meters, and Fluke gives us a NVLAP certified report.  The last report we got had a stated uncertainty of the measurement of 0.06 ppm.

Very impressive indeed.  I was wondering how good this method was.  Fluke must have the 732B that they send you characterized very closely.
Question?  What make and model of "our meters" do you use to make the measurement between the 732B outputs?  I have tried several different methods to make these uV measurements but the random noise from both units makes it difficult to get a measurement much below approximately 0.1 ppm repeatability.  Right now I am using a 3458A set to 100 PLC, 40 measurements with the MATH turned on and then take the "mean" value from the 3458A stats.  But my "low" to "high" readings are around 0.8 uV apart, with the "mean" being somewhat in the middle of those.
Bill

[plesa]

Quite simple method!! Thanks for posting. The uncertainty you mentioned  is amazing ( compare to the 732B 30day spec.). What is annular fee to be in DVMP?

[dacman]

Very impressive indeed.  I was wondering how good this method was.  Fluke must have the 732B that they send you characterized very closely.
Question?  What make and model of "our meters" do you use to make the measurement between the 732B outputs?  I have tried several different methods to make these uV measurements but the random noise from both units makes it difficult to get a measurement much below approximately 0.1 ppm repeatability.  Right now I am using a 3458A set to 100 PLC, 40 measurements with the MATH turned on and then take the "mean" value from the 3458A stats.  But my "low" to "high" readings are around 0.8 uV apart, with the "mean" being somewhat in the middle of those.
Bill

About any high end meter should so.  We presently use a Data Proof scanner and VoltRef software and a Keysight 34420A to get the measurements.  VoltRef is designed to follow NBS TN 430 (the document is on the Data Proof website).  If you get two or more Zeners calibrated, then there will be three or more in the system, and the measurements that Fluke wants will follow NBS TN 430.  (Fluke also wants negative readings, which also follow TN 430, but are from the Lo terminals vice Hi.)  The instruction from Fluke is to take 36 (or is it 72) measurements over a three day period, which should help cancel out noise and short term drift of the Zeners.  (If I were calibrating the 1 V tap using the 10 V tap, I would want to use a 3458A.)

The uncertainty of the 3458A at zero is 1 µV (if not nulled) and it is not important to null the 3458A, because this will fall out due to checking each Zener on both channels A and B.  Zeners are also noisy.  That is one reason why so many measurements are required.  (Although they can be too noisy.)  (VoltRef averages about 20 readings for one measurement.)

Here is the Fluke NVLAP scope of accreditation.  At the bottom of page 21 is the uncertainty of the on-site calibration, which they just send a 732B.
http://assets.fluke.com/download/nvlap_certs/NVLAP_scope.pdf

[dacman]

Quite simple method!! Thanks for posting. The uncertainty you mentioned  is amazing ( compare to the 732B 30day spec.). What is annular fee to be in DVMP?

I think the cost is for a single on-site calibration, and I think it is about $1k for just the 10 V tap, and we just order one as required (and it may only be available in the USA).  Fluke will give a prediction formula after three calibrations, although we calculate this ourselves.

[Bill158]

The uncertainty of the 3458A at zero is 1 µV (if not nulled) and it is not important to null the 3458A, because this will fall out due to checking each Zener on both channels A and B.  Zeners are also noisy.  That is one reason why so many measurements are required.  (Although they can be too noisy.)  (VoltRef averages about 20 readings for one measurement.)

I do reverse the connections from each 732B ( REF HI  and  UUT LO) into the 3458A so that any offsets from the connections or the 3458A should be nulled out.  I then take the sum of the two readings, observing signs, and divide by 2 to get the final reading.  This follows what FLUKE says to do when "standardizing" one 732B to the "certified" 732B which is detailed in the manual paragraph 5-9 and replacing fig. 5.1 with fig 5.1 which is in Change #1 at the end of the manual.  I can see why the DATA PROOF and the software makes this MUCH easier to accomplish, especially when you have to take 36 ( or 72 ) readings and then average those readings.  Then when you have to apply TN 430 and do all inter comparisons between 4, or more, units in a bank it could take a day or more, when waiting for thermals and other issues to settle down.  So I guess now I need a DATA PROOF to add to my collection.  Or just ignore the small variations and work with that and use the long term drift trends over years of observation.  My guess is that the DATA PROOF is expensive so I think I will just use my method, which has shown me very good drift trends over 10 years.
At least you answered my question about what meters to use and the "noise" of the zeners.  I had tried the analog low pass filter in the 3456A and the MATH software low pass filter of the 3458A and didn't get significant improvement in stability.  Therefore I was pretty sure that multiple readings was the only way to accomplish all of this.
Thanks,
Bill

[ManateeMafia]

Bill158

Data Proof quoted about $2K for the software. I am guessing that for most people it is easier to do manually.

[dacman]

Bill,
You did not mention the LFILTER function of the 3458A.  Is that the one you meant?  Also, have you read change 3 for the 732B?  It basically states that for the 732B to meet stability specifications that either line power conditioning needs to be used or it needs to be run on the battery.  And I hope you were talking generally about the reversal formula.  The reverse reading is supposed to be subtracted from the forward reading (or the sign reversed and then added) before dividing by 2.

[Bill158]

Bill,
You did not mention the LFILTER function of the 3458A.  Is that the one you meant?  Also, have you read change 3 for the 732B?  It basically states that for the 732B to meet stability specifications that either line power conditioning needs to be used or it needs to be run on the battery.  And I hope you were talking generally about the reversal formula.  The reverse reading is supposed to be subtracted from the forward reading (or the sign reversed and then added) before dividing by 2.

I was using the MATH operation FILTER.  My interpretation of the LFILTER was for filtering high frequency noise from the change of level for triggering a sample to be taken.  I tried several constants for DEGREE for the FILTER operation.  Results were mixed when trying the FILTER operation, but not much improvement as the first reading seems to set the initial value of the difference voltage between units.  The FILTER does decrease the total difference between RMATH 2 (LOWER) and RMATH 13 (UPPER) values but the RMATH 4 (MEAN) still is not that stable.  I hope I have made myself clear on this.
Yes, I was reversing the sign on the reverse measurement, "adding" the two values and dividing by 2 as you were talking about in the last sentence.  So if the first reading was -1.024 uv and the reverse reading was +0.635uv the total is -1.659 uv and then dividing by 2 the result is -0.8295 uv which I round up to -0.830 uv.  I have seen Change #3 and read it but I have never tried taking readings on battery power when doing measurements.  I will have to try this to see if I get any short term improvement between readings.  My knowledge of "power line conditioners" is that they modify the input sine wave by distorting the peak to keep the RMS value under the sine wave constant and then they also have a Faraday shield in their transformer to filter out any high frequency noise coming in from the power line.  The schematic of A4 shows something in the transformer TX1 primary hooked to the power line GND.  I assumed that this is a Faraday copper shield, which is what I would expect to be designed into the transformer.  Then there is something shown on the secondary which has a connection to VGUARD.  But without have the drawing and build components for the transformer it is not clear what is happening here.  But since I am not operating in an industrial environment (i.e. at home) I would not expect much improvement, but I will try this.  Thanks for the suggestion.
Bill

[plesa]

Someone win in lottery
http://www.ebay.com/itm/Fluke-732B-DC-Standard-T70017-/361390435649?hash=item54248c6d41:g:WJIAAOSwWnFV-vgZ
http://www.ebay.com/itm/Fluke-732B-DC-Standard-T70016-/351518993668?hash=item51d82a2d04:g:qg0AAOSwQTVV-voA

I needs to check ebay more often

[Vgkid]

Damn, even I could afford those.

[eas]

I'm guessing they sold less than 60 minutes after being listing. Much like a few 3458a DMMs that sold for $1000-1500 in the last six months.

[Awesome14]

Interesting reading. I did not found how long and how many calibration Fluke is performing for example on their 732B/H or on 732B/C where drift rate should be provided.
Based on my last few month investigation it is quite difficult to find used 732B on market.
Fluke also did not respond to quotation request, but I will expect something 4-6 kUSD (depends on the version).
What was your cost of new or used unit?

The last price list I can find for a new Fluke 732B was 4/2001. It shows the 732B/C at US$5,990 and the 732B/H at US$5,330.  You can bet that these prices would be around US$9k to $10k now.
Bill

I received a quote on a new 732B in early 2015 for 9,100.00USD. I saw 2 on eBay from the same  seller  go for around 200.00USD each. The seller obviously didn't know what he had!