Author Topic: Advantest R6581 8.5 digit DMM mini teardown/repair  (Read 201420 times)

0 Members and 2 Guests are viewing this topic.

Offline leighcorrigall

  • Frequent Contributor
  • **
  • Posts: 452
  • Country: ca
  • Nuclear Materials Scientist
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #900 on: October 22, 2021, 08:21:17 pm »
I suppose there is a possibility that I installed the R230 and R234 backwards, but I was careful to check the other people's reference pictures before installation.
...


@branadic

The R200 and R234 resistors have been properly oriented. This was confirmed by measuring the original blue resistors that I kept and comparing their resistances to the labels found on the Vishay replacements. Photos were taken before and after the replacement process for reference. The replaced resistors are not causing performance issues with my unit. At this point, I wish the problem was simple and that I made a really easy mistake with the resistor reversal, but there is something else causing the anomaly.  :palm:
MASc, EIT, PhD Candidate
 

Online dietert1

  • Super Contributor
  • ***
  • Posts: 2020
  • Country: br
    • CADT Homepage
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #901 on: October 23, 2021, 12:57:44 pm »
Meanwhile the R6581T with the modded MUX received a fresh VFD.

Here are the internal voltages from diagnostic mode:
x1Zero_1 -0.0305 V
x1Zero_2 -0.00258 V
x10Zero -0.293 mV
x100Zero -0.0629 mV
7VRef 7.062... V
-10V -9.813399 V
-1V -980.702 mV
100 mV -98.178 mV
Int. Temp 35.32 °C (still heating slowly)

I also wired up a setup with a HP3478A and an E3630A power supply to measure input leakage current as voltage along a 10 MOhm resistor. As expected leakage is small without Autozero and much larger with Autozero on. Leakage current is the same with 10 PLC and with 100 PLC. With Autozero off this R6581T input behaves resistive, and a line fit estimates about 2 TOhm.
With Autozero on i got a latch-up when trying -11 V. This means even with the correct control signal the input MUX does not break before make at input voltages below -10 V. Break-before-make is based on the open-collector outputs of the comparators, in conjunction with the different RC delays formed by the 8K4 (break) and the 100K (make) resistors in the MUX hybrid and JFET gate capacitance. But that doesn't work properly if the gate thresholds are at very different voltages. In addition there is this strange patch R083/C026.

Regards, Dieter

Edit: As the MAX328 charge injection is specified as 5 pC, times 2.5 gives a max leakage of 12.5 pA at 10 PLC with Autozero. About a factor 15 lower than now. Maybe that's the way to go.
« Last Edit: October 23, 2021, 01:06:51 pm by dietert1 »
 
The following users thanked this post: leighcorrigall

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #902 on: October 23, 2021, 01:16:50 pm »
C026 slows down some of the switches. As the charging part is though 100 K and discharge though 10 K the effect is stronger for the turn on part than for the turn off part. To some degree this avoids a make before break and reduces the effect of forward biased gate.  R83 is likely mainly there to limit the current spike for the comparator. I don't know if the SW adds some extra phase with no input connected.


Except for the -11 V point the curve with AZ mode does not look so bad. It is still odd to also see a negative current for positive voltages.
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #903 on: October 23, 2021, 02:25:43 pm »
Added the additional data to the table.
For whatever reason the values of your unit are odd, leighcorrigall. I would try to find out why.

-branadic-
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 

Offline leighcorrigall

  • Frequent Contributor
  • **
  • Posts: 452
  • Country: ca
  • Nuclear Materials Scientist
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #904 on: October 23, 2021, 02:48:01 pm »
Added the additional data to the table.
For whatever reason the values of your unit are odd, leighcorrigall. I would try to find out why.

-branadic-

I think this might be the heart of the problem. The actual measurements are not at all close to the results presented in the diagnostic menu. Could this be something wrong with the digital board?  :-//
MASc, EIT, PhD Candidate
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #905 on: October 23, 2021, 03:03:19 pm »
It would be unusual that a problem with the digital board would only effect the diagnostic mode in a away to give wrong values. A defect in the firmware in such a way would be unlikely.

Are the wrong readings at least changing to the degree of normal noise, when a new measurement / update is called for ?

A agree that the odd values there are close to the heart of the problem. The esepcially odd point is the very different offset for the fast and slow mode, unless the fast amplifier value is only the difference to the slow one.
 

Offline Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #906 on: October 23, 2021, 03:25:46 pm »
Some more statistics.
 
The following users thanked this post: leighcorrigall

Offline leighcorrigall

  • Frequent Contributor
  • **
  • Posts: 452
  • Country: ca
  • Nuclear Materials Scientist
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #907 on: October 23, 2021, 03:29:03 pm »
...

Are the wrong readings at least changing to the degree of normal noise, when a new measurement/update is called for?

...

The values are fairly stable. These measurements were taken now:

X1ZERO_1 = -0.0167 DCV
X1ZERO_2 = +2.9543 DCV
X10ZERO = +0.29526 DCV
X100ZERO = +0.0294 DCV
7.2Vref = +10.04475 DCV
-10Vref = -6.89803 DCV
-1Vref = -0.68989 DCV
-0.1Vref = -0.069098 DCV
INT_TEMP = +70.3 °C
MASc, EIT, PhD Candidate
 

Offline leighcorrigall

  • Frequent Contributor
  • **
  • Posts: 452
  • Country: ca
  • Nuclear Materials Scientist
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #908 on: October 23, 2021, 03:44:23 pm »
...

Are the wrong readings at least changing to the degree of normal noise, when a new measurement/update is called for?

...

The values are fairly stable. These measurements were taken now:

X1ZERO_1 = -0.0167 DCV
X1ZERO_2 = +2.9543 DCV
X10ZERO = +0.29526 DCV
X100ZERO = +0.0294 DCV
7.2Vref = +10.04475 DCV
-10Vref = -6.89803 DCV
-1Vref = -0.68989 DCV
-0.1Vref = -0.069098 DCV
INT_TEMP = +70.3 °C

I suppose there is a possibility that I installed the R230 and R234 backwards, but I was careful to check the other people's reference pictures before installation. Tonight, I will check the voltages.

EDIT: Couldn't wait!

R230: 98.5 mV, 0.985 V, 9.85 V
R234: 9.659 V, 7.090 V (across the network: 16.75 V)

It seems that this pattern could all lead back to R234:

7.2Vref - X1ZERO_2 = +7.090 V (R234 voltage)
-10Vref - X1ZERO_2 = -9.85 V (R234 voltage)
-1Vref  - X10ZERO = -0.985 V
-0.1Vref - X100ZERO = -0.0985 V

As a sanity check, I have attached the R234 before and after photos. Note that the camera was pointed in two separate directions.
 Could the Vishay resistor have incorrect labelling? I thought I checked the resistances before installation.
MASc, EIT, PhD Candidate
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #909 on: October 23, 2021, 04:17:10 pm »
The readings are fairly stable, but at least they do change a bit and are thus actually measured and not just old data from memory.

I don't think that R234 is wrong - there were some separately measured voltages at R200 R201 that support that R234 is working OK. With a relacement resistor the numbers may be off a little, like maybe 100 mV, but not so much.
The numbers are more like a common offset of some 3 V for the raw ADC readings. We don't know exactly how the other test voltages are measurend, but this way it would make sense.

The question is a litlle if the offset does include a cal constant (maybe some factory cal) of some kind to effect the x1zero_2 number.

I tend to think there should be no such constant, as the other meters also show an offset of a few mV. That is more than the expected drift even over a long time for the amplifier. So if there would be a cal constant involved the zeros should be much better. For diagnostics it also makes sense to give raw reading with as little cal constants included as possible.

It is still a quite good values for the nominal values of the resistors - so the resistors seem to be rather high accuracy (e.g. better than 0.1%), but still not impossible.
So the offset is likely the reading of the ADC without any individually measured offset constant, just a common constant somewhere in the firmware.

The fast channel offset may be actually the difference to the precision channel. It would be really odd to have an offset at the amplifiers ouput only for the precision channel and not for the fast channel. A more ouput side offset would be more an ADC problem than an amplifier problem.
The R200/R201 measurements also supported that the amplifier was near zero with a shorted input.

It would than still be odd to give the other voltage readings also as raw values without offset correction, but who knows.
So the 10 V instead of 7 V would be because of the 3 V offset added and not a factor 1.4 in the scale factor.

A large offset at the ADC may also effect the linearity (also visible as DNL) of the ADC. For a quite check of the linearity ( more local, DNL like)  one could use a slowly drifting votlage (like a large capacitor slowly discharged through a large resistor) and see if the measurend voltage gives a really smooth curve. The 1PLC data would be the more sensitive to the rundown part than higher PLC.

A 3 V offest at the ADC would also reduce the measurement range somewhat. So does the meter read OK to some +12 V or where even the upper limit is ?
 

Offline leighcorrigall

  • Frequent Contributor
  • **
  • Posts: 452
  • Country: ca
  • Nuclear Materials Scientist
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #910 on: October 23, 2021, 04:40:36 pm »
I just took another set of measurements on the R234. The attached picture indicates where the R234 voltages are coming from.

- the 7k3400 labelled side of the R234 measures a voltage of -7.09 V when the COM is applied to the middle lead
- the 10k000 labelled side of the R234 measures a voltage of +9.65 V when the COM is applied to the middle lead

With the shield and the cover removed, the following information is presented in the diagnostic mode:  :scared:

X1ZERO_1 = -16.668 mVDC
X1ZERO_2 = +0.2954 mVDC
X10ZERO = +0.0295 mVDC
X100ZERO = +29.519 mVDC
7.2Vref = +1.0044 mVDC
-10Vref = -0.6897 mVDC
-1Vref = -0.0689 mVDC
-0.1Vref = -69.01 mVDC
INT_TEMP = +64.05 °C

This is insane.  :-BROKE

MASc, EIT, PhD Candidate
 

Offline leighcorrigall

  • Frequent Contributor
  • **
  • Posts: 452
  • Country: ca
  • Nuclear Materials Scientist
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #911 on: October 23, 2021, 04:53:51 pm »

...

X1ZERO_1 = -16.668 mVDC
X1ZERO_2 = +0.2954 mVDC
X10ZERO = +0.0295 mVDC
X100ZERO = +29.519 mVDC
7.2Vref = +1.0044 mVDC
-10Vref = -0.6897 mVDC
-1Vref = -0.0689 mVDC
-0.1Vref = -69.01 mVDC
INT_TEMP = +64.05 °C

...


I restarted the unit and placed the shield on and the values/units have returned back to what they were before. The process was repeated several times. The values/units are not the same each time, especially for the 7.2Vref (see attached). The problem could be a loose connection near the voltage reference.

EDIT:
The unit seems to be indecisive about whether to use mVDC or VDC units. Notice that two zeros are missing from 7.2Vref when the units are in mVDC? This happens at random...
« Last Edit: October 23, 2021, 05:30:13 pm by leighcorrigall »
MASc, EIT, PhD Candidate
 

Offline serg-el

  • Regular Contributor
  • *
  • Posts: 142
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #912 on: October 23, 2021, 06:05:06 pm »
Is it possible to bring the voltage map U001 in IDLE mode?
« Last Edit: October 23, 2021, 06:07:17 pm by serg-el »
 

Online dietert1

  • Super Contributor
  • ***
  • Posts: 2020
  • Country: br
    • CADT Homepage
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #913 on: October 23, 2021, 06:17:26 pm »
C026 slows down some of the switches. ...
If you look at the schematic again, the patch slows down E_DCV only. Out of curiosity i removed the patch replacing R083 by a short. Now even with Autozero on the positive side looks clean. What a mess.

Concerning the bad display of internal voltages in diagnostic mode, for my instrument that issue disappeared after using the Reset command (first in diagnostic mode menu).

The MUX is in idle mode if you connect the control pin 27 to -18V. Before you need to separate it from the Guard buffer. Not easy unless you cut the pin. I pulled the pin from its hole to apply my mod.

Regards, Dieter
« Last Edit: October 23, 2021, 06:38:08 pm by dietert1 »
 

Online dietert1

  • Super Contributor
  • ***
  • Posts: 2020
  • Country: br
    • CADT Homepage
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #914 on: October 24, 2021, 02:36:32 pm »
Meanwhile i measured the channel resistance of two U001 JFETs and found 28 KOhm both for the DCV input and the 7.2V input. For comparison the MAX328 spec is 1 to 3.5 KOhm.
I think the MCU U300 outputs all necessary signals to replace U001 and U012 to U017 by 2x MAX328 without any glue logic. The channel select address bits are A=U300P26 B=U300P27 and C=U300P30. The two active high enable signals are E1=U300P24 and E2=U300P25. So it is about wiring in the two MAX328 correctly. For a test one can leave the old circuit in place and idle U001. Once the mod has been verified, the old circuit can go and the mod gets mounted in place of U001, with bodge wires to bring in the five MCU signals instead of the 14 enable signals of the hybrid. The footprint of the hybrid allows for DIL16 muxes with guard traces between their pins.

Regards, Dieter

Edit: Removed typo.
« Last Edit: October 24, 2021, 08:56:30 pm by dietert1 »
 

Offline Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #915 on: October 24, 2021, 05:48:12 pm »
Quote
I think the MCU U300 outputs all necessary signals to replace U001 and U012 to U017 by 2x MAX308 without any glue logic.
I don't quite understand why you try to replace a good hybrid MUX with a general purpose MAX308 with huge leakage current.
 

Online dietert1

  • Super Contributor
  • ***
  • Posts: 2020
  • Country: br
    • CADT Homepage
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #916 on: October 24, 2021, 08:18:17 pm »
Believe me, i also thought a lot about this question.
My answer is that all analog muxes nowadays use pairs of complementary mosfet channels. No more JFET muxes anymore. Also i think it will be a lot of work to educate that mux circuit into switching in an orderly fashion. But you can try, it is certainly possible. For example you can increase the break-before-make time interval by shifting the 2.5 V reference level of the comparators down to 1 V. Then you only need 16x RC delays in front of each comparator, e.g. 4K7 with 560 pF. And then you need to cure the control signal on pin 27, using another ten low leakage buffers.
With 2x MUX328 it becomes so easy. The MUX328 isn't general purpose, but especially designed for this application "metering front end". From it's datasheet i read that it's leakage starts at 1 pA at room temperature. That will increase somewhat at 38 or 39 °C inside the R6581T. As soon as i have a running prototype i will measure leakage current using the same method as above and report.

Regards, Dieter

Edit: Removed typo.
« Last Edit: October 24, 2021, 08:55:23 pm by dietert1 »
 

Offline Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #917 on: October 24, 2021, 08:34:34 pm »
Quote
The MUX308 isn't general purpose, but especially designed for this application "metering front end". From it's datasheet i read that it's leakage starts at 1 pA at room temperature...
No, it's one of the average performance MUX. MAXIM has another series of low leakage multiplexers: 326/327/328. I widely used them in impedance synthesizer, voltmeter - voltage calibrator, etc. The main problem is to minimize their parasitic thermoEMF.
 

Online dietert1

  • Super Contributor
  • ***
  • Posts: 2020
  • Country: br
    • CADT Homepage
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #918 on: October 24, 2021, 08:54:13 pm »
Sorry, that was a typo, will edit that. I ordered some MAX328 already a week ago and my prototype has those.
Here i have the existing logic. All this and the hybrid can be replaced by the MUX chips.

Regards, Dieter
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #919 on: October 24, 2021, 08:57:36 pm »
Replacing the JFET based MUX with MOFET based one is a rather big change. AFAIK there are a few meters with a broken MUX, that would warrant such an effort.

The JFET switching has its problems in the control and they are not easy to fix. Additional low leakage buffers for the other channels would be quite some effort - though only 3 (the divider, ohms sense hi and the smalles currents) would need to be really low leakage and GND would not need a buffer at all.

The low leakage version is the max328 (maxim) or MUX36S08 (Ti)

As there are only 11 signals needed (with no AC) one could consider 1 low leakage 4:1 mux for the 3 critical ones and one 8:1 for the less critical ones.
 

Online dietert1

  • Super Contributor
  • ***
  • Posts: 2020
  • Country: br
    • CADT Homepage
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #920 on: October 25, 2021, 06:39:53 am »
Channel resistance 28K + protection resistor 9K = 37 KOhm times 200 pA of switching errors gives 7.4 uV, in other words 0.74 ppm of 10 V. In an 8.5 meter that's a serious problem. Nonlinearities of 0.5 ppm or so have been one of the major concerns with the R6581T. But i am not trying to sell anything, except i may have some spare hybrids soon..

Regards, Dieter

Edit: With the MAX328 mux the error calculation should be about 3.5K + 9K = 12.5 KOhm * 20 pA = 250 nV equivalent to 0.025 ppm of 10 V.
« Last Edit: October 25, 2021, 08:02:26 am by dietert1 »
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #921 on: October 25, 2021, 08:01:50 am »
Is the channel resistance really that higher. The 2N4117 is the highest resistance JFET I know of and this is more like 5 KOhms and thus 10 K for 2 in series. The FET resistance is linear and may show up higher if measured with a higher current.

The resistance at the input in combination with quite some input current can in deed be a problem. The current in AZ mode is not constant, but more like some spikes every 200 ms.  So part of the transients can decay before the actual measurement and the zero reading phase see's another current spike. Worst case the input current effect would be nearly 4 times higher (the pulse effecting more the actual conversion and a similar effect for the zero reading) - at best, much of the switching spike has decayed before the conversion.  There is quite some delay in the AZ mode and thus a chance that much of the input current spike as decayed, at least with a lower impedance signal source.
 

Offline Mickle T.

  • Frequent Contributor
  • **
  • Posts: 463
  • Country: ru
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #922 on: October 25, 2021, 07:56:41 pm »
With Keithley 2100 I got a MUX ON-resistance of about 18 kOhm.
As is well known, the R6581 is a successor to the older (T)R687X DMMs in which the multiplexer consisted of discrete sub-pA leakage 2N4117A FETs on a ceramic base. R6581 is not the only one that uses a hybrid chip. R6561 6.5-d nanovolt-DMM uses the same THD series MUX (Takeda Hybrid Device?), and the newest 7480T uses the similar MUX topology with SMD FETs on a ceramic base.
 
The following users thanked this post: branadic

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #923 on: October 25, 2021, 08:38:09 pm »
So the MUX hybrid can be respinned by using MMBF4117, which is still available, is it that what you're saying? Or is the SOT23 version using something different?

-branadic-
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 

Online dietert1

  • Super Contributor
  • ***
  • Posts: 2020
  • Country: br
    • CADT Homepage
Re: Advantest R6581 8.5 digit DMM mini teardown/repair
« Reply #924 on: October 26, 2021, 08:42:41 am »
Meanwhile i replaced the MUX U001 and its helpers U012 to U017 by 2x MAX328. The mod  worked right from the start.
As expected it behaves much more like an ideal switch. Leakage with "Autozero off" remains within -10 pA to 0 pA over the whole input voltage range -12 V to +12 V. Now the instrument input exhibits a resistive behaviour of about 3.5 TOhm.
With "Autozero on" the behaviour is also resistive and i can now test at -11 and -12 V input voltage. The measurement can be explained by charge pumping the input voltage into a capacitive load of 18 pF  Now effective input resistance is 21 GOhm. If one can precharge the amplifier input to compensate consumption, the residual nonlinear leakage current will be less than 10 or 20 pA once more (except below -10 V). I will try and run the MAX328 from -18 V instead of -15 V to improve that.

The 2x MAX328 will remain. After heat-up i ran an external calibration "zero front" to remove a 3 uV offset. The instrument runs about 3 °C cooler now. There is a bodge wire to let the LM35 temperature sensor run from  5V instead of 15 V.

Regards, Dieter
« Last Edit: October 26, 2021, 09:13:02 am by dietert1 »
 
The following users thanked this post: ch_scr


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf