Fluke 332A adjustment/verification

[VintageNut]

Bought a 332A for keeping a DC standard in my lab. I have use of a Keithley 2002 that is within its one year cal date.

Slid back the cover enough to expose the adjustments. The inner case is a guard and is at the output potential. 1000V at 50mA is potentially lethal. Placed straws in the adjustment holes. Covered the screwdriver blade in shrink wrap for double insulation.

First check is for zero volts on all three ranges 10V, 1000V and 1000V.

10V range error less than1 ppm



100V range error less than 1ppm



1000V range error less than 1 ppm.


No zero adjustment required.

A full-scale check shows 10V range in spec. 1000V range 35 ppm off which is over spec. So, if one range needs adjusted, then they all get the treatment.

Next, set each range for full scale and adjust pots for as close to a full scale reading on the 2002 as possible. The 10V range and 100V range have good resolution turning the pot, single ppm changes are fairly easy. On the 1000V range, any change is going to cause 5ppm movement. It would be a good design change to add another fine tune pot to the circuit.

Once that was finished, I ran through the linearity test with 1.111111, 2.222222, .... 9.999999 on all three ranges. 30+ pictures, just a few here:

444.4444V  error less than 8ppm


77.77777V error less than 4ppm           


The only stinkers in the bunch are 11.11111V and 1.111111V

11.11111V error is 29 ppm. Spec is 32 ppm so no cause for concern but all of the other errors are much less. I am thinking that a string adjustment may fix this but there are no directions for a string adjustment.



1.111111V error is 13ppm. Not bad but larger than almost all of the other errors.


Another test is all 9's. These should be very nearly the same as full scale readings.

999.9999V  error is 1ppm  Nice!


99.99999V error is 2 ppm



9.999999V error is 6ppm


Lest anyone that this is a quick process, it is not. The instrument needs hours to warm up. Measurements take minutes to settle. DMM is set to 10NPLC with 10 length filter. This is pretty much an all day effort.

Now the instrument is closed and back in the equipment rack. Re-check full scale readings.

1000.0000V  less than 2ppm error


100.0000V less than 2ppm error



10.00000V  less than 2ppm error


This was a fun project. If anyone has a procedure for adjusting the string I would appreciate having a copy.

 

[dacman]

I would suggest testing all the electrolytic capacitors.  If you have an LCR meter that uses a bridge circuit, it will give accurate enough results in-circuit at the values of the larger capacitors (i.e., the electrolytics).

For a sample string adjustment procedure, go to www.fluke.com and search for the 332A.

[VintageNut]

Success with string adjustment. Had to print out the string schematic, piece it together and go through it slowly. Read and re-read the US Army instructions. Concluded that it can be done with a DMM.

The schematic is here. It is fuzzy (non-steady hands) but the original is free to download from Fluke even to this day 50 years later. If someone has a paper original I would appreciate a high rez scan.



Here is my procedure:

1. Set range for 1000V.

2. Set decade 1 for 10, all other decades set to 0.

3. Set output on

4. Adjust 1000V range pot for as close to 1000.000 as possible. This pot is difficult to make any better than +/- 5 ppm on my 332A.

5. Set output off

6. Set decade 1 for 1. All other decades are 0.

7. Adjust pot R901 for a reading of 100.0000.

9. Change the first decade dial to 2.  Adjust pot R902 for a reading of 200.0000. This may cause full scale voltage to be over/under 1000.000. Re-adjust 1000V calibration pot for an output of 1000.000 with first decade dial at 10.  Full scale output is the addition of settings 2, 4, 6, 8 and 10 resistors. Adjusting any of the even decade dial setting pots will change the full-scale output.

10. Repeat step 9 until 200V and 1000V are as close to perfect as possible.

11. Set first decade dial to 3. Check that output is 300.0000. You cannot adjust this output. The output is the sum of the 100V setting and the 200V setting. 

12. The dial 1 position and dial 2 position are the foundation of all of the rest of the settings of this decade. You may want to start at the beginning and go over all of the steps to insure that 100.0000V, 200.0000V and 1000.000V are all as close to perfect as you can make them.

13. Change the first decade dial to 4.  Adjust pot R904 for a reading of 400.0000. This may cause full scale voltage to be over/under  1000.000V. Re-adjust 1000V calibration pot for an output of 1000.000 with first decade dial at 10.  Full scale output is the addition of settings 2, 4, 6, 8 and 10 resistors. . Adjusting any of the even decade dial setting pots will change the full-scale output.

Continue through the first decade pot adjustments until you finish up at 1000.000V.

Verify that you have outputs of 100.0000, 200.0000, 300.0000, 400.0000, 500.0000, 600.0000, 700.0000, 800.0000, 900.0000 and 1000.000 for the dial setting 1 through 10 of the first decade dial.

Move on to the second decade dial. Set the first decade to 0 and decades 3 through 7 to zero as well. Use the same process as above to set output voltages of 10.00000, 20.00000, 30.00000, 40.00000, 50.00000, 60.00000, 70.00000, 80.00000, 90.00000 and 100.00000. You will NOT be adjusting the range calibration pots in these steps. 

All of the settings of all of the dials of my 332A are proportional and result in less than 10ppm error. For the first 6 digits, the output is exactly what you set it to. The 7th digit is set to zero and the reading on the DMM for the seventh digit is the uncertainty. ppm error is the 7th digit reading divided by the 332A setting.

Here are two examples

1098.76V, The 0.001 volt digit is the uncertainty. About 7 ppm here


123.456V The 0.0001 volt digit is the ppm error. About 1 ppm here

 
 The 7th digit of the 2002 is stable for this cabling setup. Digits 8 and 9 are pretty much noise.

This process required a bit over two hours to accomplish not including instrument warm up. The 332A seems to require at least two hours to warm up. Very good stability after four hours.

[dacman]

Why didn't you use the procedure in the Fluke manual page 3-2?  It uses the 3rd dial as a reference and steps up from there.

[TiN]

Thanks for posting, enjoy your standard! Also would suggest keeping cover closed for tests, as it may upset temperature compensation of internal parts.

Shipping cost of such beasts make them pretty unaccessible for me :

[VintageNut]

Why didn't you use the procedure in the Fluke manual page 3-2?  It uses the 3rd dial as a reference and steps up from there.

Good question. 00X.0000 on the 1000V range yields 10V. Measured on the 10V range of the 2002 shows 13 ppm high.

[VintageNut]

Thanks for posting, enjoy your standard! Also would suggest keeping cover closed for tests, as it may upset temperature compensation of internal parts.

Shipping cost of such beasts make them pretty unaccessible for me :

It was located in the next state north so the shipping time was just two days. It was open only for adjustments. In fact I got tired of closing and opening it so today I cut slots to allow adjustments with the case closed. Now I just have to make a cover for the slots.

[VintageNut]

I was curious to see if I could find an indication of data of manufacture. So, off came the inside cover for some peeking and poking. Voila! There is a date code on the high voltage transformer.

February 1967! Wow. 48 years young!
 

A look at the high precision resistance string with adjustment pots for the first two decades.


Current reference assembly


Extractable cards in card cage slots

[PaulAm]

I see you have the extender card as well.  That was a  nice touch by Fluke.

I picked up a 332b marked as "for parts or repair" for cheap, and the only thing wrong was operator error - they had the voltage trip control turned all the way counterclockwise.  To be fair, I also had to replace a couple indicator lights.

Shipping can be horrible, but I bought mine from a surplus dealer who does a lot of shipping so it wasn't too bad.  I think it was around $50 US.

Now I'm looking for a Keithley 2001 that I can repair and get calibrated for a reasonable sum so I have something to check it with.   Or maybe I should just mask off the last 4 digits 

[gilbenl]

Very nice! Do you mind sharing how much you paid for it?

[VintageNut]

Very nice! Do you mind sharing how much you paid for it?

$300 including shipping.

[Vgkid]

What a beautiful piece of gear.

[VintageNut]

I see you have the extender card as well.  That was a  nice touch by Fluke.

I picked up a 332b marked as "for parts or repair" for cheap, and the only thing wrong was operator error - they had the voltage trip control turned all the way counterclockwise.  To be fair, I also had to replace a couple indicator lights.

Shipping can be horrible, but I bought mine from a surplus dealer who does a lot of shipping so it wasn't too bad.  I think it was around $50 US.

Now I'm looking for a Keithley 2001 that I can repair and get calibrated for a reasonable sum so I have something to check it with.   Or maybe I should just mask off the last 4 digits 

That is cool that you scored a 332B that only needed a knob turned to get it to operate properly. I would prefer the 332B as it has a trap door to perform adjustments from what I have read.

An interesting outcome from this exercise is that my 2000 DMM DCV measurements are very close to the 2002. My 2000 has had 8 calibrations so my conclusion is that the 2000 has been aged properly and is probably not going to drift significantly until it dies some day. Measuring 10V,  the two DMMs are within less than 5 ppm. They diverge more at 1000V.

I bought a 2001 fixer-upper with the idea that it would be my standard when I return the 2002 loaner.  Now that I have used the 2000 beside the 2002, I am going to use the 2000 as my bench DCV reference.

If you want to discuss purchasing my 2001 fixer-upper, you can send me a pm.

[dacman]

When adjusting the sample string (using the Fluke manual) it doesn't matter that the 3SD, set to X, is 13 PPM off.  (This offset is removed later.)  After the sample string is adjusted, the dials above it will also be 13 PPM off (being referenced to the 3SD), and when the ranges are adjusted at the end, the 3SD, set to X, should then be nominal.

[VintageNut]

Ok. Thanks for the explanation. I will have to read the procedure more carefully and give it a try.

[Dr. Frank]

Ok. Thanks for the explanation. I will have to read the procedure more carefully and give it a try.

Hi, and congratulations on that nice calibrator.

Does it have the mechanical chopper?
And does that still work properly?

Anyhow, you have to linearize the sample string first, starting comparing Deck C vs. Deck B: 00x0000 <=> 0100000 and working upwards.
After that, the 10, 100 and 1000V ranges may be calibrated.
The other way round, you will fail!

I recommend another army document, for the 332B/AF, from 1972, better quality , especially the schematics, which fully covers your instrument (332A),
but also includes the newer versions, with the MOSFET chopper, and the more precise sample string, where DECK C can also be adjusted.

I recommend also to get the 332D, 1975 document, where Fluke improved the linearization process.
They compare consecutive steps on 1.. 10V  voltage  levels, instead of 10..100V, which is much more precise, especially if you use a DMM.
In your 332A case, you would use the 100V range only for the DECK B linearization, and 10V range for the DECK A linearization.

Both documents have around 20MB of size, so I can't upload it here.

Send me a PM, if you'd like to get these.

Frank

Further remarks:
- Linearization at max. 10V level also is more safe, as the inner shield will not rise to more than about 40V.

- The linearization process requires about 0.1ppm resolution and/or linearity of the DMM used.

If you still have access to the 8 1/2 digits Keithley, then it's fine, and you may also adjust the sample string easier, by directly adjusting 00x.00 => 010.00, 01.000, 02.000, 04.000, 06.000, and so forth, instead of comparing two settings like 01.X00 vs. 02.000 and so on.
That's described in the 332D/1972 manual, which uses a 720A KV for linearization.
Latter instrument is specified for 0.1ppm linearity, and would be replaced by the 8 1/2 digit DMM.

- If you have a 6 1/2 digit DMM only, then the linearization is not possible to the required accuracy..

[VintageNut]

Hello Dr. Frank

I believe that I have those documents. It is a bit confusing since I am not using a divider nor a null meter. I have read the documents over and over and tried to perform the steps. I am always left with 20 ppm to 30ppm error with dials set to 00X.0000. I have performed the procedure at least 4 times.

I surmise that I want 00X.0000 to be exactly 10.0000000 volts on the 2002. So I set the 332A range to 1000V and the dials to 00X.0000. The reading on the 2002 is over 20 ppm high. I adjusted the 1000V calibration control so that the 2002 reads 10.0000000V. Then I moved the 332 dials to 010.000000 and adjusted R907 for 10.0000000V on the 2002.

Regularly I go back and check 0X0.0000 and adjust the 1000V range calibration control for 10.0000000V if it has moved.

I worked my way up the dial settings and have all of the readings +/- 1ppm for most and +/-2 ppm for a few of the highest dial settings .

I think this is how it should be done with a DMM. Let me know if you disagree. 

I do not know if it has a mechanical chopper. The serial number is 845 and the date on the transformer is 1967. I do not hear any noise. I have a Keihtley 147 with a mechanical chopper and it can be heard everywhere in the house.

Whatever chopper is used, it is noisy electrically. There is 3-5 ppm noise on most settings.

[Dr. Frank]

Hi,

I quite don't understand your approach, or maybe, I understand your 'thought error':

The linearity adjustment only compares one resistor (or let's say better, the foregoing two of them) to the next one, (with double the value).
These were made equal to about 0.5ppm.

If you proceed further, you end up with a linearity of Deck A and B of about these 0.5ppm of full scale! (Comparable to any modern 6 1/2 digit DMM!!)

If you then adjust the different ranges, i.e, 10V, 100V and 1000V on full scale, i.e  at X000000 settings, this can be done to a few ppm absolute accuracy, initially.

If you output 10V on the 1000V range, the above described linearity really has to be calculated as 0.5ppm (linearity error) of 1000V.

You can't do better than that!!

0.5ppm of 1000V are 500µV.
500µV of 10V are 50ppm error!
Hey, that's 1% of F.S. reading! That's normal!!

Therefore 20ppm are very good, and easily explained:
Always use an instrument on the highest possible setting / readout!

Therefore, for 10V output, use the 10V range, that'll be a few ppm only!


For the chopper: I think, that's inside the double metal box.. see pictures of the A version..

And it's really a wonder, that it's still working..
These mechanical choppers are the first to fail inside these instruments..

The noise is relatively normal, especially, maybe, with a mechanical chopper.

I think I have published noise measurements of my former 332B/AF, having this MOSFET chopper already (I gave it away to Australia, into another 332A).

Please search for "332B .. in the slaughterhouse"

Frank

[guido]

You could put in a modern chopper replacement. See here:

http://friedrich-messtechnik.de/index.php/bauanleitungen/87-...uke335-332

I did in a 335D:






Guess where the connectors on the new pcb came from 

In fact I got tired of closing and opening it so today I cut slots to allow adjustments with the case closed. Now I just have to make a cover for the slots.

A standard feature of the 335D; there is a panel you can remove to access the adjustments. There is pcb material without copper to shield off the inner casing (which could sit at 11kV). Guess there were more complaints.

I see you have the extender card as well.  That was a  nice touch by Fluke.

I picked up a 332b marked as "for parts or repair" for cheap, and the only thing wrong was operator error - they had the voltage trip control turned all the way counterclockwise.  To be fair, I also had to replace a couple indicator lights.

In mine somebody exchanged the main filtercaps. And then switched both fuses at the back. Which resulted in one blown fuse at power on. Easy fix But it was missing a transformer in the null meter section. Despite the fact that i could source it in the US (somebody with a 335 slaughterhause), the null meter still doesn't work (yet).

There was no extender card, so i had to make one myself.

[fmaimon]

- If you have a 6 1/2 digit DMM only, then the linearization is not possible to the required accuracy..

Why? If you use an 6 1/2 DMM with relative mode (8846A, 34401A, etc), you just need to check if every pair of resistors (1, 2, 4 and 8 group) has the same ratio. You should get a good match doing that.

[VintageNut]

Hi,

I quite don't understand your approach, or maybe, I understand your 'thought error':

The linearity adjustment only compares one resistor (or let's say better, the foregoing two of them) to the next one, (with double the value).
These were made equal to about 0.5ppm.

If you proceed further, you end up with a linearity of Deck A and B of about these 0.5ppm of full scale! (Comparable to any modern 6 1/2 digit DMM!!)

If you then adjust the different ranges, i.e, 10V, 100V and 1000V on full scale, i.e  at X000000 settings, this can be done to a few ppm absolute accuracy, initially.

If you output 10V on the 1000V range, the above described linearity really has to be calculated as 0.5ppm (linearity error) of 1000V.

You can't do better than that!!

0.5ppm of 1000V are 500µV.
500µV of 10V are 50ppm error!
Hey, that's 1% of F.S. reading! That's normal!!

Therefore 20ppm are very good, and easily explained:
Always use an instrument on the highest possible setting / readout!

Therefore, for 10V output, use the 10V range, that'll be a few ppm only!


For the chopper: I think, that's inside the double metal box.. see pictures of the A version..

And it's really a wonder, that it's still working..
These mechanical choppers are the first to fail inside these instruments..

The noise is relatively normal, especially, maybe, with a mechanical chopper.

I think I have published noise measurements of my former 332B/AF, having this MOSFET chopper already (I gave it away to Australia, into another 332A).

Please search for "332B .. in the slaughterhouse"

Frank

Hi Dr. Frank

I tried the lower voltage range first and gave up on it. Here is why. The floor is not zero. If you look at the spec, the noise floor (and/or zero offset) on the 100V range can be as much as 20 uV. If you are outputting 1V on the 100V range, 20V of zero offset is 20 ppm error.

The zero offset on the 1000V range is 40 uV maximum which is 4ppm error.

I was able to get all three ranges adjusted below 10uV zero offset. That is still 10 ppm offset at 1V on the 100V range. The initial step in the entire process is to measure 00X.0000 and compare it to 010.0000. I could never get rid of the 10 ppm offset. So, I used the 1000V range where the offset is the lowest ppm. I adjust the 1000V calibration pot to get exactly 10V on the 2002. Now the third decade is exactly correct and the higher decades use the third decade as the reference.

[dacman]

How are the aluminum electrolytic capacitors?  I've replaced lots of them in 332A/B/D and 335D units.  I'd test them all.  Also, the chopper is critical for stability.

[nukie]

Ah ha some discussion about these heavy beasts. I have a 335A and most of the electrolytic caps on the sub boards are toss. I replaced most of them they are getting harder to source. I looked at the main caps at the back covered in cardboard sleeves those are monsters don't know where to find them.

Finally, my relay died one day, I replaced it with a similar coil wattage. I'm surprised the original relay manufacturer is still around and the same model is still being made!!

[VintageNut]

How are the aluminum electrolytic capacitors?  I've replaced lots of them in 332A/B/D and 335D units.  I'd test them all.  Also, the chopper is critical for stability.

Good question about the caps. I have no idea unless I take the boards out and test. That will probably be a winter project. Good idea about checking out the chopper. If that is not stable that would explain some behaviors I am seeing.

Nobody has brought up the pots. At least two of the string pots are dirty/worn. The output makes unexpected jumps while adjusting or the output goes down while you are adjusting up. It would probably make sense to replace all of the pots.

[Dr. Frank]

Hi Dr. Frank

I tried the lower voltage range first and gave up on it. Here is why. The floor is not zero. If you look at the spec, the noise floor (and/or zero offset) on the 100V range can be as much as 20 uV. If you are outputting 1V on the 100V range, 20V of zero offset is 20 ppm error.

The zero offset on the 1000V range is 40 uV maximum which is 4ppm error.

I was able to get all three ranges adjusted below 10uV zero offset. That is still 10 ppm offset at 1V on the 100V range. The initial step in the entire process is to measure 00X.0000 and compare it to 010.0000. I could never get rid of the 10 ppm offset. So, I used the 1000V range where the offset is the lowest ppm. I adjust the 1000V calibration pot to get exactly 10V on the 2002. Now the third decade is exactly correct and the higher decades use the third decade as the reference.

Hi VintageNut,
It's necessary first to separate the parameters zero offset, noise and absolute error.

See 332A-1965 manual, page 1-1, and also the calibration description also.

The zero offset is specified inside the Calibration Accuracy:

10µV  for 10V range,
20µV for 100V range,
40µV for 1000V range

You'll find these values also in the description of the zero offset calibration on page 3-1

noise is specified separately, on page 1-1, 20/30/40µV rms for these three ranges, accordingly.

absolute accuracy is 30ppm for each range, plus the zero offset error.

Linearity error is not specified, but may be estimated to about 0.5ppm for the sample string, following its calibration procedure, and can be ignored in first order.


The zero offset on the order of 10µV that you mentioned,  should be relatively constant, therefore a comparison of two equal values like 002.x000 V vs. 003.0000 V should be possible to 1µV stability / accuracy, as described in the 332D-1972 manual.

In  other word, this constant offset does not affect this adjustment process.
But you did it right, using 10V for equalizing 3rd vs 2nd decade, because that is more stable.


The high noise level of the mechanical chopper on the order of 20..40µV has to be averaged out by a longer time constant of the DMM, (something like NPLC 200, or longer)  to again achieve < 1µV accuracy. That may be a bigger problem, as this slows down the adjustment process.

Things would be a bit better with the MOSFET chopper.
Zero offset is < 2µV on all ranges, and short term noise may be as low as 4µV rms, see my measurements here:

https://www.eevblog.com/forum/testgear/fluke-332baf-in-the-slaughterhouse/msg399310/#msg399310

In the end, I think you succeeded in adjusting your 332A, as far I understand your comment?

Frank