General Resistance DAS56A

[MosherIV]

Hi

I recently got a General Resistance DAS56A dial a source.



1V range resolution 1uV
10V range resolution 10uV
0.005% accuracy
TC - 2ppm/C
24H stability 10ppm
noise 10ppm or 30uVp-p
Full specification and schematic in the manual :
www.eevblog.com/forum/testgear/i-don't-need-no-stinking-calibration/?action=dlattach;attach=51658

I got it to test the calibration of my DMMs, including Thurlby 1905a (5.5digit) and Solartron 7060 (6.5digit)
Fortuneately the company I work for has allowd me to verify my voltage sources/standatd with a recently calibrated Tektronics DMM4040

There are 3 things I would like clarification/guidance on please?
1. Noise.
The manual has some brief instructions on calibration, where the first thing that needs to be done is to set the zero offset.
Now the instructions say to try and set the offset to somewhere below 5uV
The problem is that the base level noise is 30uV - how do I set the zero offset when the base level noise swamps the readings?

2. Drift
So I did my best to set the zero offset but I noticed that after setting it, after lunch the zero offset has drifted.
Is this normalor is there something wrong with the unit?
I have attached a spreadsheet showing the noise measured by the DMM4040

I have also attached another spreadsheet showing the long term drift over 5 hours with the outpiut set to max - 10V


3. last 2 digits of KVD
It looks like the last 2 digits of the Kelvin Varley Divider are not working properly.
I noticed that in 10V mode, sometimes 100uV range does not work properly.
(the base level noise makes the last digit meaning less - is that what you would expect ?)
When I set any voltage below 1V (in 10V mode) the 100uV range works
When I set the MSD to 9V - the 100uV digit does not match the setting

Does this mean there is something wrong?
Is it the voltage reference?
Or is it the KVD selectors need cleaning?

[MosherIV]

Blast 

How do I attach files like spreadsheets please?
(I tried zipping it and uploading but they do not show ) 

[Vgkid]

A zip file should work. Attachments should accept xls/ods files.

[zlymex]

Blast 

How do I attach files like spreadsheets please?
(I tried zipping it and uploading but they do not show ) 

May be try to save the spreadsheets in .xls extension instead of others suhc as .xlsx

[MosherIV]

OK lets try .xls

Any thoughs on my 3 issues please?
1 noise
2 drift
3 problem with KVD contacts?

[Andreas]

Hello,

1. The noise is given as wideband noise. You are integrating with 10 NPLC. So the noise is largely reduced.
   As long as the noise is gaussian you can average the samples until the noise is small enough.
   I would also consider to use 100 NPLC.

2. is the drift related to the source or to the DMM? the last 10000 measurement values look quite constant.
    Most instruments need at least 2 hrs warm up time.

With best regards

Andreas

[MosherIV]

2. is the drift related to the source or to the DMM? the last 10000 measurement values look quite constant.

Missed the question the first time I read the reply.
I am pretty sure it is the source, but how would I find out for definite?

Yes, it has been overcast these past days and the output was pretty stable after warm up (I give it about 1.5 hours).
I sitting with my back to a window at work and it seems the sun was affecting the readings, cauing what looks like drift. I noticed this in the graph above, I realised it was sunny in the morning and went overcast in the afternoon when the reading stablised.

I think there is something wrong with it. It can set 100uV when the total output is < 4V but as soon as it goes above 4V it reads low by 20/30uV ie
3.00010V is ok get 3.00014 or 3.00015V
4.0001V gives 4.00007V
I think the contacts on the KVD must be dirty. Any suggestions on what to use to clean?
I read about similar Datel calibrator and it said to use IPA and lubricant, what lubricant should I use, eg WD40?

[MosherIV]

It turns out I did not need to worry, access to the contacts for the first KVD decade was very good :




I have done a full tear down here :
https://www.eevblog.com/forum/testgear/general-resistance-das56a-tear-down-and-repair/msg1057652/#msg1057652

It will have to wait till Mon now until I get back into work to see if cleaning the contacts has made a difference.

[GeneM]

I have both a DAS 46A and a TSC 46 and both had minor issues when I bought them-several months apart and from different EBAY sellers.  Before messing with the contacts in the Kelvin-Varley divider, you should take a look at the power supply board and the Chopper-Stabilized Amplifier board.  Towards the left side of the power supply board is a daughter board which is actually the Chopper-Stabilized Amplifier.  It's very easy to miss the fact that it is socketed onto the power supply board.  You can see the CSA daughter board in your tear-down picture https://www.eevblog.com/forum/testgear/general-resistance-das56a-tear-down-and-repair/?action=dlattach;attach=265549 towards the left side of the power supply board.

My DAS 46A had very unstable output readings until I lifted the corners of the CSA board and re-seated it onto the power supply board.   It was not necessary to fully lift it.  After re-seating it, the output was very stable and very easy to zero with the 10-turn pot at the top edge of the power supply board.  Short circuit the meter's leads and make sure it shows '0.'  Then set the Kelvin Varley to zero, connect the meter and it should also show zero-well, adjustable to within 5 micro-volts of zero. 

As an aside, I'm using an HP 3455A bought through EBAY and repaired through several weeks of finding a few minutes at a time to learn how some portions of the circuitry worked.  I've bought four of them (different auctions) in non-working condition and currently have two of them working.

The second thing I did to get my DAS 46A reliable to use was to take advantage of what I learned with the TSC 46.  Although it has additional circuitry for thermocouple calibration, in one mode it fully duplicates the DAS 46A as a voltage reference.  There is very little inside the transformer enclosure-just some bridge rectifiers and 1 Micro-Farad capacitors-so I wouldn't go in there just yet. 

Instead, look at the power supply board.  It uses a clever circuit which only needs only one 1N752A zener diode as a reference for both the + and - 15 volt power supplies that feed the CSA board.  I believe that tries to ensure that both + and - 15 volt outputs drift together, improving stability of the CSA. 

Another voltage regulator circuit provides 1N752 zener stabilized power to the two series-connected 1N827 zeners in the other 'box' located below the Kelvin Varley divider.  That also looks like some gray-beards implemented a very carefully orchestrated circuit with a lot of selected & matched 0.01 % resistors used to partially offset the 1N827 temperature drifts.  The care they took is incredible.  On my TSC 46, I first replaced only the one 50 Micro-Farad capacitor (with a 220 Micro-Farad) which had failed some time before I had bought it.  Replacing the one capacitor brought it back to life.  I used it for many weeks while working on one of the broken 3455A meters, but then one day the TSC 46 became unstable again.  The other 50 Micro-Farad power supply capacitor (-15 volt side) had failed.  So, I replaced all of the 50 Micro-Farad capacitors with 220 Micro-Farad capacitors just to avoid having to do it piece-meal.  They were vintage 1979, so that was not a bad service life. 

I don't have anything good enough to calibrate my 3455A meters.  But I did buy one of those LM584 references on EBAY which had been measured by the seller's 'still in calibration period' Agilent 34401 to give me a ball-park check.  I set the DAS 46 to match the seller's 34401 measurement at each of the four settings and made the following table.

Note the LM584 reference outputs were supposedly 2.49980, 4.99829, 7.50249 and 10.00001.  The third column shows the readings of the LM584 reference from my repaired 3455A. 

DAS 46 Setting   3455A      "LM584" Reference per 3455A:
2.49980              2.49976       2.49977
4.99829              4.99824       4.99832
7.50249              7.50241             7.50252
10.00000            9.99989          10.00010

While I don't know truly how close the DAS 46 is to 'spot-on,' I do have some confidence that it is not terribly far off.  To those who designed and built the General Resistance devices-you should be very, very proud of yourselves. 

So, I'd start by reseating the CSA daughter board and replacing the 50 Micro-Farad power supply board capacitors before I'd take apart the gold-plated Kelvin Varley six-decade divider.  Put an oscilloscope on the voltage regulator outputs and make sure they are not noisy.  The CSA can't give clean output without clean power. 

[MosherIV]

Hi GeneM

I had noticed the PSU board has the Chopper-stabilized amplifier board as a daughter board.
I had not considered this to be the source of the problem. I will definitely, re-seat it to see if this helps.

I have already tried cleaning the first KVD contacts, it made a small difference as far as I can tell.

I will try your suggestions out.

As far as the 'LM584' - did you mean AD584?
The AD584 has far worse noise than the DAS
The data sheet says "NOISE (0.1 Hz TO 10 Hz)" = 50?Vp-p
Where as the DAS56A/47A is 30?Vp-p

[GeneM]

Yes, I'm sorry- it was a AD584KH reference.  Your equipment is far better than what I have.  I had not thought to put an oscilloscope on the output of the DAS 46, though I have looked at it with several multimeters.  The only oscilloscope I have is a 'low mileage' B&K Precision 1476A which I bought new in the mid 1980's.  It's never been calibrated and it only has a 10 MHz bandwidth.  Also, the most sensitive scale it has is 1 millivolt per division.  I had thought it would be useless for this sort of thing and so that may have contributed to my not having looked at the output with it.  It was very useful in showing the DC power supply's capacitors had failed.

The B&K 'scope shows what appears to be about 2 millivolts peak to peak of noise on the DAS 46's output.  But it shows the same on each of the two AD584KH references.  Neither of my 3455A meters gives any indication that the output is noisy.  They bobble by one or two counts at most at any setting I've tried even with 'high resolution' set to give me the full 6.5 digits.  So I'm concluding that my old B&K just isn't up to the task and it's readings are really only telling me that I need a much better scope for this sort of work. 

I've read that Zener diodes are noisy, but the output from the Zeners in the DAS does get filtered through a three-stage RC filter before it gets to the Chopper Amplifier.  Any Zener noise should be very quiet as it enters the CSA.  That's why I was first suspecting the power supply capacitors and the socketed connection to the Chopper Amp rather than the Kelvin-Varley.  It is always possible that your DAS was not well-treated and it could have seen a contaminated environment. 

There is also noise from the 1N752 Zeners in the power supply and there is voltage gain to go from the zener's 5.6 volts to the power supply's 15 volts.  The 50 Micro-Farad capacitors (220 Micro-Farads in each of my DAS / TSC 46) at the input and output of the 15 volt regulators should prevent that from being a noise source.

Since you've got the (far, far) better scope, you could try looking at the five 'stages' of this device for noise and then we'd have a much clearer picture of where the noise is coming from.  It is also possible that your scope's microvolt scale is not really 1000 times better than my B&K 1 millivolt scale and we're really looking at artifacts of that scope's limits.  I'm not saying, implying or even suspecting that is what is happening because I have no experience with really good equipment such as what you have and so I don't know how it behaves when it is being used at microvolt scales.  You can teach me.

Stage 1 would be at the power-line input.  The DAS has an RFI filter which should act to block any noise coming in through the AC power line.  But these machines are 30 plus years old and we have no way to know if they have been over-stressed at some time, so we can't take it for granted that the power filter works-or even that it's not oscillating.  Your scope can tell us.

Stage 2 would be the outputs of each of the +15, -15 for the CSA and +24 for the Zener regulator assembly to ensure there is not significant noise from those sources.  Zeners are noisy and there is voltage gain in those circuits, so a bad zener there could throw noise into the CSA or the Zener regulator assembly.

Stage 3 would be to look at the output of the Zener regulator assembly to see if you have a breaking-down or wildly noisy 1N827A in the voltage regulator assembly.   If the noise is from a noisy 1N827A, then it should be a higher voltage here than anywhere else because that assembly puts out about 12.4 volts which is cut-down before getting into the Kelvin-Varley by 12.4 / 10 (it's much more precise than I'm making sound)

Stage 4 would be to look at the input to the Chopper Stabilized Amplifier from the 1N827A diode pair in the regulator assembly after it has come through the three stage RC filter.  Zener noise should be lower here than at the Zener Regulator assembly output both because it has been reduced to at most 10 volts and because of the RC filter that is supposed to filter the input to the CSA from the Zener assembly.  One or more of those three 1 Micro-Farad capacitors could have broken down or one or more of the resistors opened or nearly opened (or bypassed by some chemical deposits on the board) and so it would not be performing as it should.  There is less shielding from external environment here than in the Zener Regulator assembly or the Kelvin-Varley.  If the Kelvin-Varley was infiltrated by some chemical agent in the environment, then the RC filter would have seen an even greater concentration of whatever agent it was.

Stage 5 is where you've already measured-the output of the CSA.  If the noise is being generated by the Chopper Stabilized Amplifier, then it should be greater at the output than at the input.  The CSA actually is supposed to have a voltage gain of 1.000000, so the output should be exactly like the input except for noise contributed by the CSA.  I have not found a schematic diagram of the CSA, it seems to have been a trade secret of the General Resistance Company.  If yours is bad, then you may want to try replacing it with a modern single-chip CSA.

Alternatively, if you only use it with very high input meters, then you may just take the CSA out of the circuit and run the Kelvin-Varley's output directly to the output jacks.

I don't get to work with my hobbies much, so I may not respond very fast.  I'll try to make a greater effort because I'd really like to know more about these devices and you've got the equipment needed to do that.
 

[MosherIV]

Your equipment is far better than what I have.  I had not thought to put an oscilloscope on the output of the DAS 46

Actually, neither have I (yet). I do mean to put a scope on it at some point.
The figures I quoted for noise are from the specifications.
A 6.5 digit meter should be able to see noise in the 100s uV, 10s uV and possibly 1s uV
The Solartron 7060 (6.5 digit) meter is fast enough to see the noise fluctuations in 10s uV, this is the one I have at home.
The Tektroncis DMM4040 (6.5 digit) meter is fast enough to see the noise fluctuations  in 10s uV, this one I am using at work (where I have taken the picture at the top of this thread).
The Thurlby Thandar 1905a (5.5 digit) meter performs similarly to your HP 3455A - the update rate only sees a small change in lsd count. i have this one at home.
I think the problem is the update rate of the meter. The 6.5 digit meters I have and am using are capable of measuring 100s of times a second - hence can see the noise.

The only oscilloscope I have is a 'low mileage' B&K Precision 1476A which I bought new in the mid 1980's.  It's never been calibrated and it only has a 10 MHz bandwidth.

10MHz bandwidth should be fine, we are just measuring low freq noise. The issue, which you have identified, is the resolution of the lowest range, 1mV is not enough resolution.

I have tried re-seating the CSA board and I think it has helped with the drift.

I have not tried checking the PSU rails and changing the 50uF caps yet. Maybe I will get a chance over the weekend.

Yes, I agree with the 5 steps you have suggested, just need time to do them 

[Andreas]

Hello,

I think the scope is ok.
What you need is a shielded box and shielded wires for the measured device.
Otherwise you get influence by mains line frequency.

Try a cookies box and battery supply for the AD584.

See also AN124
http://cds.linear.com/docs/en/application-note/an124f.pdf

with best regards

Andreas

[villas]

Hi MosherIV,

I got some months ago one DAS-57AL and I have a HP3456a, but no means (GPIB adaptor) to create a long log of measures.

But I can tell my experience with the last digits. My unit agree to HP3456a (uncalibrated) except for the 2 last digits, in both ranges 1V and 10V. In 1V range the last digit (100nV) doesn't seem to work properly, the 6th digit (1uV) works ok but have some offset, this offset looks stable in short term, but varies from one day to another by several micro Volts.

I disassembled my unit, it doesn't have the KVD shielding, and someone had already changed the electrolytics. Now, with all the good information on this post, I will probably open it up again and check on CSA board, also probably try to make a shielding for the KVD.

As soon as I get one GPIB I'll try to make some noise and drift graphs.

[MosherIV]

So I took it apart again over the weekend and gave the end (first KVD) contacts another clean.
Also tried to clean the PCB surfaces of the other KVD dividers as best I could without taking the whole KVD assembly apart.
Seems to have made a difference - I can now set 9.0001V and measure 9.0001V.

While I had it apart I did a few measurements.
I first checked the supply rails for noise (across C4)



It is kind of hard to see but you can just see a bit of ripple on the supply rail.
Bear in mind that is a x1 probe and the scope is on the max gain (2mV/div)

[MosherIV]

At this point I was kind of OK with uV of noise.

Later I came back and checked the output. I had the time base in the usec and the noise looked OK but when I changed the time based to a slower setting, I saw more noise I had not noticed before. 



Sadly, I had put the thing back together at this point and it was late so I did not have time to take it apart again and check the supply rails or even check the rectifier outputs.

Another job for this weekend. 

[Vgkid]

Nice job with those contacts. Always suspect power supply issues  .

[MosherIV]

So I had a chance to have a look at the noise issue.

I started by looking at the PSU, and I was shocked by how much ripple there was around 4V of ripple.
I did as GeneM suggested and replaced the 50uF smoothing caps for 220uF
The ripple is down to less than 1V and the noise seems to have gone, still get some noise.
Will have to wait till Monday when I get it back in the office at work and check with the calibrate 6.4 digit Tektronix DMM4040

Before caps changed



After caps changed

[MosherIV]

I have had a chance to check the DAS56A with the Tektronix DMM4040 - no change   

While it looks like the first stage capacitor replacment got rid of this noise :


Looks like this noise is still there :


The DMM shows there is 20uV of noise

[MosherIV]

So I did some more debugging this weekend and here is what I found :

While probing around for the source of the 20uV noise I figured something out - the source of the noise is not from the Chopper Stabilized Amp, it is completely asynchronous to it




In this picture, I found a nice looking square wave on the CSA board and I am looking at the noise on the +15V line.
You can see there is no correlation between the square wave and the noise

I also found the noise is still there when the power is off ......Hmmmmm

[MosherIV]

During the week, I used a digital scope to look more closely at the noise :



In the second picture, the noise spikes occur at 53KHz freq.



I realized on the weekend talking about it to friends/colleagues that the noise is PC PSU switching noise.
Remeber, I found the noise is still there when the unit is turned off ! - Ahah moment.
Additionally, I was able to see/measure the noise more clearly - 40uV p-p

It could be that I get less noise when using the analogue scope because it is a different location with different set of PCs on the mains 

In any case, I think the source of my inaccuracies from the precision voltage source is being swamped by noise from the mains.

[MosherIV]

The DAS56A already has a mains input filter built into the design. I have circled it in red.



The question is, it is either
a. not working properly - 1000pF caps gone bad? Do inductors go bad?
b. not good enough for modern applications (it is 40 years old and there were no PCs with SMPS back then)?

Looking at the diagram on the top of the IEC filter, it looks like the design is missing the cap between live and neutral.


Can I just add the capacitor in? What value (another 1000pF)?
Why are those ones
(https://www.eevblog.com/forum/testgear/general-resistance-das56a-tear-down-and-repair/msg1075293/#msg1075293)
so big ?
Are modern ones http://cpc.farnell.com/webapp/wcs/stores/servlet/Search?catalogId=15002&langId=69&storeId=10180&categoryId=700000033502&eq=N%3D220842%2B411%26amp%3BNs%3DP_STORE_MARKETING_RANK_FARNELL_CPC%257c0%257c%257cP_MAN_PART_NUM%257c0%26amp%3BNtpc%3D1%26amp%3BNtpr%3D1&pageSize=100&showResults=true&aa=true&pf=110193699&vw= OK ?

OR

Should I fit one of these :
http://cpc.farnell.com/schurter/5500-2077/1-phase-line-filter-medical-m5/dp/FT01659

Any thoughts or advice welcome 

[Gyro]

Across the line capacitors tend to be PP film type in the 100nF range (Class X1 rated). Whether it will help or not depends on whether the noise is common mode or differential. It's pretty unlikely that those 1000pF capacitors have gone bad (or the inductors either). Depending what the caps look like, it might be worth replacing them with modern Y1 rated ceramics anyway (the ones fitted may not have current safety ratings).

It's a bit of a sad lesson of older equipment struggling to survive in our modern SMPS polluted world!

[MosherIV]

I had an old IEC encapsulated filter connector lying around.
As a test, I disconnected the filter in the DAS56A and put the ICE filter in.
I powered it up and checked the +15V rail ...... the noise is still there 

My only other hope now is something like this :
http://cpc.farnell.com/wurth-elektronik/74271221s/ferrite-core-split-10-5mm/dp/FT0165769?ost=FT0165769&selectedCategoryId=&categoryNameResp=All%2BCategories&searchView=table&iscrfnonsku=false
basically a ferrite that clamps around mains cable.

Something else I am going to try is to put together a shielded cable.

The noise is 10 uV (microVolt) p-p on the scope. I would really like to get it down to less than 10uV p-p then I have more confidence in the 5th digit of the unit.
Right now the unit is only good to the 4th digit, ie 1mV

Any thoughts or advice welcome 

[MosherIV]

So I made a bit of progress, I got a couple of ferrite clamps and they seem to make a difference.

Here is a picture of the noise before



Here is a picture of the noise with the clamp on the mains input


Anybody know what the parameter associated with them would make the biggest difference?