Author Topic: Advantest R6581 8.5 digit DMM mini teardown/repair  (Read 201413 times)

0 Members and 1 Guest are viewing this topic.

Offline szszjdb

  • Regular Contributor
  • *
  • Posts: 208
  • Country: cn
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #75 on: January 13, 2018, 09:14:37 pm »
Hi, Mickle T,
I directly short the GUARD port with the other 4 port together and re-calibrated the zero front , and the zero offset is around 1uv and it is not so stable runing from 100nv to 4.5uv range. The meter has warmed up for about 8 hours.Is it fine?

When I measure the input signal, I just connect the HI and LO port to the 6144 source with the normal test line , the reading is also running in the last 2 digit ,2.0-3.0uv at 10v input.  I know it might causing by the source noise, but how can  I evaluate the noise performance of the meter? In my view , the 8.5 digit meter should keep stable in the uv digit and jump on the nv digit, right?

I have made a LTZ1000 reference board and does it can be used to evaluate the noise performance of the meter?
For running the LTZ1000 board about half an hour, the 10v output has max 2.0uv jumping range in 8.5 digit mode, and the A-ZERO is open and plc=20. Is it aslo fine?

More information, if I go to the diagnostic menu and measure the internal 7.2v  reference , it is also jump about 2-3uv, in the last 2 digit range.  It is fine? The meter  is power on  for 11 hours till now.

Best Regards,
szszjdb
« Last Edit: January 13, 2018, 11:50:13 pm by szszjdb »
 

Online Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #76 on: January 14, 2018, 12:20:46 pm »
... zero offset is around 1uv and it is not so stable runing from 100nv to 4.5uv range. The meter has warmed up for about 8 hours. Is it fine?
It's absolutely not fine :( At least the measurements that I just carried out showed DCV zero about 300 nV with +-250 nV span (100 samples, 100 mV range, 10 PLC, AZERO On).
I think it makes sense to ask questions in that topic: https://www.eevblog.com/forum/metrology/dmm-adc-noise-comparison-testing-project/
 
The following users thanked this post: TiN, szszjdb

Offline szszjdb

  • Regular Contributor
  • *
  • Posts: 208
  • Country: cn
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #77 on: January 14, 2018, 03:31:38 pm »
Hi, Mickle T,
Thanks for your advice.

About my 6581 in diagnostic mode :
1.  the X10Z  is 1.69315mv and X100Z  is 1.8875mv, both seems too huge than the sample 0.06373mv and 0.01811mv in the diag mode document. Is there still have some issue in the circuit?
2. when get into the TRANSFER or A/D check item and choice one to perform, the LCD show the item name, but seems no real action taken because no information show the check finished. How can I find the result of the check?

Does the potentiometer r170 need to be adjusted to get better performance?

Best Regards,
szszjdb

 

Online Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #78 on: January 14, 2018, 04:53:01 pm »
1. "Diag mode document" is not oficial and was compiled by me :) All results, displayed in this mode, are for diagnostic purposes only. No autozero and no cal constants are used in this mode.
2. TRANSFER, A/D and many other checks are instrumental checks, performed with a scope, logical analyzer e.t.c.

R170 potentiometer - InpAmp precision channel coarse balancing (x1Zero_2). I don't think you get a better performance than you have now.
 
The following users thanked this post: TiN, szszjdb

Offline szszjdb

  • Regular Contributor
  • *
  • Posts: 208
  • Country: cn
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #79 on: January 14, 2018, 06:49:52 pm »
Hi, Mickle T,
Thanks a lot.
I found many data of my 6581 worse than yours in diagnostic mode, I just want to find out the reason and make result better. My 6581 has been repaired by somebody and might have some mistake lead to the bad performance. The X10ZERO, X100ZERO, the artifact votage and the current might have chance to improve?

Best Regards,
szszjdb
 

Online Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #80 on: January 14, 2018, 07:05:16 pm »
Diagnostic mode isn't intended for DMM fine tuning. Only a serious malfunctions can be detected in this mode. So, there is absolutely no need (and you can't) to "improve" these data, because they have no direct influence to DMM performance. These zeros, artifact voltages, currents are only raw datas without applying a calibration constants, corrections etc.
 
The following users thanked this post: szszjdb

Offline szszjdb

  • Regular Contributor
  • *
  • Posts: 208
  • Country: cn
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #81 on: January 15, 2018, 03:29:20 am »
Hi, Mr. Mickle.T ,
I see what you mean.
Why I was wondering the hardware abnormal is because the unstable reading of the DCV zero. It is keeping  changing in span of about 4-8uv in 100mv meter range. It is also leading to the bad noise performance in 100mv range ,where the reading of the last 3 digit jumping from time to time.
I have repaired the damaged front port , covered the case and used the short jumper to calibrate , the DCV zero still performed the same.
Is there any posibility that the issue is causing by the amplifier of the 100mv range? How can I check for it?

I will make more test for DCV zero in 10V range to  check the A/D 's noise performance.

Best Regards,
szszjdb
 

Offline szszjdb

  • Regular Contributor
  • *
  • Posts: 208
  • Country: cn
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #82 on: January 17, 2018, 02:49:51 pm »
Hi, Mr. Mickle.T ,
After the test these day, I found the DCV zero of the meter become more stable than before, data as below:
1. DCV ZERO , 10PLC, AZERO ON, N=100, 100mv RANGE, 5 TESTS
        1                2            3              4               5
MAX   -0.00013   -0.00009   -0.0001   -0.0006   -0.00009
MIN   -0.00053   -0.00046   -0.00059   -0.0006   -0.00056
AVG   -0.00035   -0.00029   -0.00033   -0.00035   -0.00037
VPP   0.0004   0.00037   0.00049   0.00054   0.00047
?    8.25E-08   8.18E-08   9.52E-08   8.41E-08   8.06E-08
UCL   -0.0001   -0.00005   -0.00004   -0.0001   -0.00013
LCL   -0.0006   -0.00054   -0.00061   -0.00061   -0.00061

2. EXTERNAL LTZ1000 homebrew REF PCB ,  10PLC, AZERO ON, N=100, 10v RANGE, 5 TESTS
              1                       2                     3                       4                     5
MAX 7.0684339   7.0684347   7.0684342   7.0684347   7.0684352
MIN 7.0684316   7.0684321   7.068432          7.0684324   7.0684329
AVG 7.0684329   7.0684335   7.0684332   7.0684336   7.0684338
VPP 0.0000023   0.0000026   0.0000022   0.0000023   0.0000023
?      4.60E-07      5.91E-07           4.50E-07          4.39E-07         4.98E-07
UCL 7.0684646   7.0684353   7.0684345   7.0684349   7.0684353
LCL  7.0684315   7.0684318   7.0684318   7.0684323   7.0684323

Does the result accecptable?

Many Thanks and Best Regards,
szszjdb
« Last Edit: January 18, 2018, 02:17:00 pm by szszjdb »
 

Online Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #83 on: January 17, 2018, 03:23:34 pm »
Well done! You got excellent results.
 
The following users thanked this post: szszjdb

Offline szszjdb

  • Regular Contributor
  • *
  • Posts: 208
  • Country: cn
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #84 on: January 18, 2018, 06:00:15 pm »
Hi, Mr. Mickle.T ,
Thank a lot!
I was wondering if there have some posible modification for the meter to get higher noise or stability performance?
I have found some modification proving to be effective,  listing below
1. change the 70k RES R2/R3 from reference PCB to metal foil type.
2. change the relay K001 ,K006,K008 to low EMF type.

How about the replace some OPAMP to the newest highend type in the signal chain to reduce noise performance?
Would you please give some advice on this topic?

Best Regards,
szszjdb

 

Online Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #85 on: January 18, 2018, 06:41:51 pm »
All modification you listed (and some others) was well descripted on the bbs.38hot.
You can try and then publish the results, that you received, in another thread.
 
The following users thanked this post: TiN, szszjdb

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14073
  • Country: de
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #86 on: January 18, 2018, 08:56:11 pm »
When looking for points to improve one should first get a good idea of the noise source. There are the input amplifier, the ADC and the reference and reference amplifiers. The ADCs noise also depends on the integration time with the two main parts from general low frequency noise and the second from measuring the residual charge, which gets more and more important for short integration time.

One point where I see a chance would be filtering the reference for the higher frequency (10 s of kHz range) part of the noise. For some reasons many DMMs don't use filtering (or only partial filtering) for this noise contribution. The modulation in the ADC brings back this noise to the near DC range if a small voltage is measured. I don't know, but maybe they did not realize that this higher frequency noise of the reference could matter.

Another obvious noise source are the resistors used for input protection - a super low noise input amplifier is of limited use if there is a 100 K resistor in front.

There is a chance to find better OPs for some places, but changes at the ADC itself are tricky, as fast settling might need an adjustment to the OPs speed. So different OPs would also need changed resistors / caps, which might not be easy.

It looks like the native range of the ADC itself is something like a -1 ... -14 V and the input is initially shifted /divided. This concepts is not really good for very low noise.

 
The following users thanked this post: TiN, szszjdb

Online Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #87 on: January 18, 2018, 09:15:39 pm »
Input protection resistors string is only 4x2.2=8.8 kOms.
The native range of the ADC itself is +-14.5...15V and the input isn't initially shifted /divided.
But there are other big problems, such as no precharge amplifier in the inputs MUX e.t.c. :(
 
The following users thanked this post: TiN, szszjdb

Offline szszjdb

  • Regular Contributor
  • *
  • Posts: 208
  • Country: cn
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #88 on: January 19, 2018, 09:51:00 am »
Hi, Mr. Mickle.T and Mr. Kleinstein ,
My key concern is the short-term stability or the noise performance.
Per your suggestion ,it seems not an easy case to improve the noise performance of my 6581.
So I am planning to change the 70K res of the reference PCB first and left the low EMF relay later when I find them.
All this is for the long-term stability or the temp. drift improvement right now.
Is there any further mods I can try?

Attached is the copied circuit of the reference PCB, FYI.

Best Regards,
szszjdb
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14073
  • Country: de
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #89 on: January 19, 2018, 10:56:57 am »
It is not that easy to improve a high grade instrument, especially not without knowing where the weak spots are.
Form the little circuit diagram parts I saw, I would imagine that filtering the reference could be an option. This is not for the 0.01 Hz -10 Hz range that is essentially impossible to filter, but for the 10 kHz range that might be mixed back. At least filtering should not interfere with the rest of the operation.

Looking at the ADC part (photos) I would expect quite some noise from the AD707's. However changing things here would also change the AC response and would thus need more changes.

I don't think there would be much gained from changing the 70 K resistors in the reference part - the LTZ1000 circuit attenuates changes and noise from the resistors quite a lot.

It also depends on what you are looking for: if it is small signals at a small fraction of the range, it would help to have a special input amplifier before the DMM. If you are looking at a signal near full scale the reference gets important and the existing ADC may be good enough.

Factors like gain drift are more set by the resistors at the ADC. These resistors are much more critical than the resistors at the reference.
 
The following users thanked this post: TiN, Mickle T., szszjdb

Offline szszjdb

  • Regular Contributor
  • *
  • Posts: 208
  • Country: cn
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #90 on: January 19, 2018, 01:32:36 pm »
Hi,Mr. Kleinstein ,
Thanks for your suggestion.
I would add some cap like C2 near the U2A to form a LPF and C9 on the power rail as attached in another circuit. The 70k Res is benefit for the long-term stability of the reference, researched by some DIYer.
The noise of the LTZ1000 reference board is measured to be 1.2-1.4uv and hard to be  improved more .

My main purpose is  to evaluate the performance of the voltage/current source , so the performance of the 10V range is key for that.
I had use the meter to test the DIY LTZ1000 board and you can find that the Vpp is 2.2-2.3uv, in the condition of  10PLC, AZERO ON, N=100, 10v RANGE,
              1                   2                     3                4                  5
MAX 7.0684339   7.0684347   7.0684342   7.0684347   7.0684352
MIN 7.0684316   7.0684321   7.068432      7.0684324   7.0684329
AVG 7.0684329   7.0684335   7.0684332   7.0684336   7.0684338
VPP 0.0000023   0.0000026   0.0000022   0.0000023   0.0000023
?      4.60E-07      5.91E-07       4.50E-07      4.39E-07     4.98E-07

When changed to 50PLC, the Vpp is 1.1-1.5uv.
              1                       2                     3                       4                     5      
MAX 7.068434          7.0684337   7.0684334   7.0684332   7.0684339
MIN 7.0684327   7.0684322   7.0684321   7.0684321   7.0684325
AVG 7.0684334   7.0684331   7.0684329   7.0684327   7.0684334
VPP 0.0000013   0.0000015   0.0000013   0.0000011   0.0000014
?    2.35E-07            3.18E-07            3.32E-07            3.03E-07            2.35E-07

Both the number is 2-3 times larger than the HP3458 ,which is causing by the higher noise floor of 6581, showed in attached. So I was wondering if can modify somewhere to get improvement.
As your said,  it is very difficult to do.

Best Regards,
szszjdb
« Last Edit: January 19, 2018, 01:49:26 pm by szszjdb »
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14073
  • Country: de
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #91 on: January 19, 2018, 02:10:35 pm »
There is not much to improve at the LTZ circuit itself.  One has to be quite careful, as it uses a feedback loop on the reference that is already a little tricky. So it might oscillate (and this way damage the reference) if the modifications are wrong.  I would definitely do a simulation before any change with the reference. The modification might work but I am not so sure it is stable and I am not so sure the noise will actually be lower.

My suggestion would be more at the first amplification stage on the main board. Here it would be just adding a cap at the first OP - so nothing unusual. At least the plan I saw did not have the cap from output to the inverting input I would normally expect.  It could be still a simplification and the cap is already there. The improvement would be more when measuring smaller voltages though, but 7 V is still not all the way to full scale.
Anyway I would not expect a magic large improvement from a small change - maybe a little. If it would be so easy the manufacturer would have done it right from the beginning. Though there are sometimes a few point that deserve a  :palm:. To get a really better reading it would be more like building a new reference with lower noise, like a dual LTZ1000.

Due to the large integration cap the 6581 is not very low noise at short integration times. So it is not so much comparable to the 3458 - more like the Keithley 2002. So it could be normal to get considerably more noise at the 10 PLC setting compared to the 50 PLC setting.

When measuring a low noise external reference one will see at least the combined noise of the external and internal reference. So with a 2 nd LTZ1000 reference it would be normal to see at least 1.4 times the noise of a single reference. So getting something like at least 1.5-2 µV_pp for 100 consecutive readings is kind of normal. So from the data shown I am more surprised to see so little noise in the 50 PLC case.

Anyway for testing the noise one get less scattering values if one looks at the RMS value instead of the peak to peak values. Even better would be Allan variance or similar.
 
The following users thanked this post: Mickle T., szszjdb

Offline Terry01

  • Frequent Contributor
  • **
  • Posts: 907
  • Country: scotland
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #92 on: January 19, 2018, 02:31:05 pm »
Does it means the issue is not so serious and can be fix by  the self of external calibration?

How does that work out so cheap?
The answer is very simple  ;D

Ah I see!

So this is "only" good prices if you have the knowledge to fix them? I do not....but I have a  :-DMM  and a  :-BROKE  and am a bit  :scared:  so you never know I may get one to have a go at!  :box:
Sparks and Smoke means i'm nearly there!
 

Online Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #93 on: January 19, 2018, 02:58:01 pm »
R6581D, R6581T have "good" prices due to lack of huge magical AC board :D
 
The following users thanked this post: szszjdb

Offline szszjdb

  • Regular Contributor
  • *
  • Posts: 208
  • Country: cn
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #94 on: January 20, 2018, 12:13:15 am »
Hi, Mr. Mickle.T and Mr. Kleinstein ,
Thanks for your advice. I got know that the noise performance is hard to be improved by simple mods.

I have not check the DCI performance before and found some stange result when testing today. The DCI zero is not the real 0 and have a large offset, and the number is jumping in the last 3-4 digit when in ua or na range.  Data are as below.

range         100na     1ua        10ua     100ua     1ma     10ma      100ma      1000ma
6581 zero    0.8xxx  0.7xxx   0.00065   0.0065    0.1014   0.000614   0.00602   0.0012

6144 output  100na     1ua           10ua              100ua           1ma        10ma           100ma
34401          310        1.14           10.12           100.16         0.99995    9.99935      99.9925   
6581           12x.xxxx   985.xxxx   10.01xxxx     100.03xxx    1.00023xx    10.003843   100.01975   

As I know , there have not external zero or fullscale calilbration for DCI.  Are there any issues upon my data and which component should I check?
The DCI is abnormal before and repaired by somebody and one of the  relay has been repalced, as attached.

Best Reagrds,
szszjdb
« Last Edit: January 20, 2018, 06:15:42 am by szszjdb »
 

Online Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #95 on: January 20, 2018, 06:45:45 am »
DCI zero calibration is performed during the external zero calibration procedure. But if you have a zero/fullscale calibration problems with DCI, then you have the same problems with Ohms too.
 
The following users thanked this post: TiN, szszjdb

Offline szszjdb

  • Regular Contributor
  • *
  • Posts: 208
  • Country: cn
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #96 on: January 20, 2018, 07:31:30 am »
Hi, Mr. Mickle.T?
Thanks for your help.

I just performed the external zero and DCV calibration and left OHM undo because no accuracy 10K res inhand. Is it the reason for the DCI zero and fullscale issues? If do, I will do the external OHM calibration tonight compared to 34401.
For the noisy data output in the DCI test, is there posibility that the 100mv DCV amplifier has problem ?  I also feel  that the 100mv range DCV performance still have noisy problem. I will confirm it tonight.

Best Regards,
szszjdb

 

Online Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #97 on: January 20, 2018, 08:07:52 am »
I just performed the external zero and DCV calibration and left OHM undo because no accuracy 10K res inhand. Is it the reason for the DCI zero and fullscale issues?
How can you check a DCI fullscale with low-end 6144 and 34401A? So I sees only a large DCI zero offsets and these offsets are not related to external OHM calibration. Some of the possibles sources of the leakage current are U503 MAX327 switches (100 nA and 1000 nA ranges), protection transistors Q502-Q503 and MUX hybrid U001 pin 3,4,5.
 
The following users thanked this post: szszjdb

Offline szszjdb

  • Regular Contributor
  • *
  • Posts: 208
  • Country: cn
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #98 on: January 20, 2018, 08:29:34 am »
Hi, Mr. Mickle.T?
Thanks for your soon reply!

Yes I know that, but just have them inhand. How about use my LTZ1000 with a precision resistor to produce a simply current source to test the noise performance in the lower DCI range ? Just to evaluate ,not calibrate.

How can I test the parts you list? Apply the current in that range and measure the voltage?

Could you give some advice on the detail to test the MUX hybrid U001 pin 3,4,5?  What should be the right voltage? That parts is difficult to find replacement. If DCV 10v work fine, does it mean the U001 is posible ok ?

Best Regards,
szszjdb

« Last Edit: January 20, 2018, 08:44:41 am by szszjdb »
 

Online Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #99 on: January 20, 2018, 08:53:16 am »
It makes sense to short the input for DCV or DCI noise measurement.
MUX hybrid have a complex two-way structure to reduce a stray capacitance and leakage paths. In addition, all of the DCI readings are made in dynamics with auto-zero cycles always ON, i.e. auto-zero mode does not matter. So, it is very difficult to test U001 properly. I'm don't know how.
 
The following users thanked this post: szszjdb


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf