Advantest R6581 8.5 digit DMM mini teardown/repair

[serg-el]

Since the machine recognition system is still far from perfect, I manually selected data types.  And he corrected the incorrectly recognized hieroglyphs.  Otherwise, the text turned into some kind of nonsense.

[wutieru]

My 6581 has a loud noise in testing now，short circuit with about 0.08ppm。
Where should I start for repairing? 
Any suggestions,Thanks！

[leighcorrigall]

martinr33 pointed out a very interesting observation today:

The Advantest R6581 is rated to a maximum of 1000 V, whereas the R6581T model is rated to 300 V in my variation.

TiN's model is rated to 120 V! https://xdevs.com/doc/Advantest/R6581T/img/adv_front.jpg

According to the dataset that was provided, the calibration lab tested my unit to 1000 V... 

[bdunham7]

martinr33 pointed out a very interesting observation today:

The Advantest R6581 is rated to a maximum of 1000 V, whereas the R6581T model is rated to 300 V in my variation.

TiN's model is rated to 120 V! https://xdevs.com/doc/Advantest/R6581T/img/adv_front.jpg

According to the dataset that was provided, the calibration lab tested my unit to 1000 V... 

Are those voltage markings on the front panel the maximum the meter can read without damage or are they just indicative of the CAT II rating? 

[dietert1]

Essentially R6581T input protection is a passive circuit, with two clipping diodes and a resistor to limit input current. Then there is some extra circuitry to limit transients to +/- 2 V or so from the previous DC level. That cicuit also makes diode leakage less dependent on DC level. Current limiting resistors are 4x 2K2 and good for about 1 W each, so that would mean a continuous overvoltage of up to 47 V on each resistor, 188 V in total plus the +/- 13 V clipping range. The circuit survives short pulses of higher voltage, e.g. during the time to open the relay and disconnect the direct low voltage input in autorange mode.
Don't know whether the relay opens if you input 100 V while in manual range 10 V.
And one would like to know what is the voltage spec of that relay. There may be the origin of a final spec of 600 V or so. In dry and clean conditions the physical limit for sparks will be yet higher, probably above 1 KV.

Regards, Dieter

[dietert1]

Here i have a proposal for solving the R6581T input mux leakage. It's a partial solution supporting DC with Autozero measurements, with clean break before make operation. No more input leakage over the full range from -12 to +12 V, except input amplifier leakage and leakage of protection circuit  (< 1 pA).
The mod also improves the nonlinearity issue discussed above, as it eliminates deviation by voltage drop across protection resistors caused by leakage. The original circuit leaks to the input with negative input voltages.
The idea is routing either the buffered DC input or Gnd or the original guard level to the MUX drive pin 27. In DC with Autozero mode, only the first two are used and the MUX works independent of the slow input amplifier response. The mod is simple  (one chip) and works. The schematic shows some context.
Removing leakage during Autocal requires a little more, as the Autocal negative (!) test voltages come from another MUX input.

Regards, Dieter

[leighcorrigall]

I received my Advantest R6581T back from Transcat today. Sure enough, it has the problem with the high ohms range. They are not willing to diagnose or repair it. 

When I measure an Ohmite RX-1M 1 GΩ 1 % resistor, the reading is within tolerance initially.

After observing the meter readings for about 2 hours, I believe the measurements decay over time until a steady-state value is reached.

My guess is that the instrument reads correctly for a while but slowly drifts off course. The same values appear for both the front and rear terminals. I am going to take some time to measure all my resistance/voltage standards before opening it up for diagnostics.

Does anyone know how to explain this mechanism?

Can you measure the input bias current, preferably directly with a picoammeter, on the DCV range that would correspond to the voltage used in the 1G resistance mode?  If no picoammeter is available, just shunt the input with a stable 1M resistor and a small (1-10nF) mylar or PP capacitor--each microvolt displayed corresponds to 1 pA bias current.  Start with the meter cold and see what the bias current does over a 4 hour period.

The Advantest R6581T was turned OFF at 7:34 AM for 1 hour before turning the instrument ON at 8:36 AM (corresponding to the large spikes in the graph). It took about 1 hour to reach a steady-state value on the 10 V range at 9:49 AM. The Advantest R6581T positive input was connected to the Keithley 6485 positive input of the inner BNC conductor throughout the entire range of measurements.

EDIT:
At 11:20 AM, the Keithley 6485 reads (-5.1 pA ± 0.25 pA).

[leighcorrigall]

The study above was repeated by leaving the Advantest R6581T OFF for about an hour, then turning the instrument ON at 12:31 PM with AZERO ON prior to startup. The noise is certainly a product of the AZERO function. Interestingly, the signal sharply increases at startup then decays to a mean of zero. The instrument AZERO was turned OFF at 1:53 PM which is indicated by the lower noise at (-5.0 pA ± 0.37 pA).

[Kleinstein]

Even the time with auto zero on look still quite good.  The mean near zero is a slight surprize. AFAIK there is no adjustment pot to the bias current so I would consider this a lucky coincidence.

However the point with zero input voltage is also the easy case, as this means switching between 2 essentially equal voltages . Chances are the spikes get larger when there is an input voltage, like some -5 V.

[leighcorrigall]

Even the time with auto zero on look still quite good.  The mean near zero is a slight surprize. AFAIK there is no adjustment pot to the bias current so I would consider this a lucky coincidence.

However the point with zero input voltage is also the easy case, as this means switching between 2 essentially equal voltages . Chances are the spikes get larger when there is an input voltage, like some -5 V.

 

The Keithley 6485 was zeroed with the Advantest OFF. Hopefully, that was alright to do.

I can repeat the experiment with an Advantest R8340 and apply a voltage while measuring the current. If that isn't good enough, I have an Advantest R6245 SMU that I could try. None of them are calibrated or have logging capabilities at the moment.

[Kleinstein]

The zero of the amp meter should normally be done with an open input. I don't think this would make a large difference, as there may be relays off at the input, when the meter is off.
The test of the input current is not so much about accuracy, more about the rough size. So no calibration for the meter would not be a problem.

[leighcorrigall]

The zero of the amp meter should normally be done with an open input. I don't think this would make a large difference, as there may be relays off at the input, when the meter is off.
The test of the input current is not so much about accuracy, more about the rough size. So no calibration for the meter would not be a problem.

I did a quick measurement with the R6245 by inputting 10 DCV to the Advantest R6581T on the 10 V range. The current reading was roughly less than 0.015 nA without zeroing the SMU or using AZERO on the DMM. If this is interesting, I can try more extensive study like I did with the Keithley 6485.

[dietert1]

As far as i understand the studies above were near zero input voltage to the R6581T. I think 5 pA is a lot, a good R6581T has below 1 pA.
The "around zero" result with Autozero On appears strange. You may need to eliminate RF using a bypass capacitor parallel to the R6581T input. I found a 6.8 uF MKS works well. The R6581T input MUX emits RF, originating from the LM339 comparators that drive the FET
switches.
The interesting part is adding a battery to repeat the leakage current measurement at +6 and at -6 V (or whatever battery you have). I would not use a power supply for that as it may add leakage current, too. If you repeat that with Autozero On and Off, you will probably see strong leakage at negative input voltage and with Autozero On. At least that's what i found. The R6581T was open and the scope connected to the DCV MUX input. I used the + and - voltages from zeners D202 and D203 inside the R6581T as test input and there were 1 V peaks on the 8K8 protection resistor when measuring - 5.6 V. Then i started to think about modding the input MUX.

Regards, Dieter

[leighcorrigall]

Out of curiosity, I cracked open the R6581T and measured the R230 network resistor because I have a hunch that this component could be the culprit. The following resistances were roughly measured:
11.08 kΩ, 4.16 kΩ, 495 Ω (when connected the board, measured in both directions)

Does anyone else get these readings from the board? If so, does your Advantest have trouble reading high-ohms?

According to MickleT's schematics ("simplified diagram of the analogue board" and page 3/5 of "ADC"), the readings should be:
45 kΩ, 4.5 kΩ, and 500 Ω (when connected directly to the DMM)

What makes me suspicious of R230 is that the -1 V range was reported to be a problem by the calibration technician. He explained that the instrument needed to be calibrated 3 times before it was within tolerance and part of the reason was that the - 1 DCV was out on the 10 V range.

Using a Fluke 732A, I measured the following voltages with the R6581T a few days ago:
-0.9999998 V on the 10 V range (definitely an incorrect measurement and repeatable)
-1.000009528 on the 1000 mV range

+1.00000960 on the 10 V range
+1.00000985 on the 1000 mV range

Looking now at MickelT's "DCI & Ohms Artifact Calibration" flow diagrams, it appears that the 10 MΩ, 100 MΩ, and 1000 MΩ all require the -0.1 V, -1 V and -10 V reference voltages for calibration. Are these voltages directly related to the R230 voltage divider? If so, I may have found the problem.

Am I going crazy or could I be onto something? 

[leighcorrigall]

Going into the Internal Diagnostic Modes (HOLD POWER + HOME on boot up):

X1ZERO_1 (FAST MODE) = - 0.0165 VDC
X1ZERO_2 (PREC MODE) = +2.95 VDC
X10ZERO = 295.35 mVDC
X100ZERO = 29.52 mVDC
7.2Vref = +10.0448 VDC
-10Vref = -6.8979 VDC
-1Vref = -689.819 mVDC
-0.1Vref = 69.027 mVDC
Int_temp = 63.32 °C

The only serious modifications I made were to resistors:
R200
R234

This makes absolutely no sense to me. I am so lost... 

[Kleinstein]

Measuring R230 in circuit is problematic. So the different number for the 45 K part is likely from the in circuit effect.
The numbers from the diagnistic message look odd. This could still be some oddity in the software.  A nearly 3 V offset for the precision mode is strange.
This more looks like some offset at the ampliers output or ADC, as the offsets go down with gain. This may be something wrong with R200 or another ADC problem.

It is than still odd to see the small error in the fast mode.
Not sure if the -10/-1/-0,1 V ref values include the offsets - this may make up the difference.

Instead of an in circuit resistance measurement, the better way to check would be a voltage measurement on the running meter.
For R230 this would be the -10 / -1 / -0.1 V test votlages.

The amplifier offset should be visible while measuring a short and than use a 2nd meter to measure the internal voltage at a few points, like the amplifier output.

The wrong reading of -1 V in the 10 V range is really odd. With this the meter should not have passed calibration.  Using repeated tries to get it work is not a good practice ! It is more like one should repeat the ACAL part several times and the meter should be in CAL all the time. If it fails once it fails calibration.
Not working reliable at -1 V is kind of a desaster for the ACAL procedure, as many one the readings are around -1 V. The high Ohm problem may still be a separate problem - possibly just too optimistic specs and a weak design in this area.

A -5 pA input current is not so bad. With quite a few FETs in the MUX and the amplifiers this is a realistic number.

[dietert1]

Yes, there is a certain risk with modding to screw up. Anyway with the schematics we got from Mickle T. maintainance of the analog section of R6581T is easy.
You can use another meter to check the reference voltage chain: 7.xxx V from the LTZ module, +17 V, -19 V and the three Autocal test voltages -10, -1, -0.1 V. All these may deviate from their nominal values by some percent, but their ratios are accurate. If there is an error with the references, the diagnostic mode will generate strange results.


Regards, Dieter

[leighcorrigall]

I suppose there is a possibility that I installed the 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)

So far, the voltage magnitudes appear to be reasonable to me. 

The firmware version states: ".A01". Is there a newer version available?

[serg-el]

The firmware version states: ".A01". Is there a newer version available?

http://www.ko4bb.com/getsimple/index.php?id=download&file=Advantest/Advantest_R6581_8.5_Digits_Digital_Multimeter_EPROM_A02.zip

[MegaVolt]

Here's 03.

[Kleinstein]

The R234 and R230 votlage readings look ok.  Looks the large offset could be the problem to get the wrong numbers.

The expected voltages at R200 would be similar to R201.
The interesting point would be the ADC input, so TP200 , with the input shorted and in 10 V and 1 V range with out auto zero and high PLC.  For the fast amplifier one could also test with a very short integration time.

[dietert1]

The -19 V reference voltage check was missing. The offset numbers in diagnostic mode mean that the precision channel of the input amplifier
produces a constant output offset of about 3 V. Depending on gain this output offset gets attributed the numbers shown in diagnostics. It can be a problem of that amplifier or of the -19 V reference. Better fix this before playing with firmware update.

Regards, Dieter

[leighcorrigall]

The -19 V reference voltage check was missing. The offset numbers in diagnostic mode mean that the precision channel of the input amplifier produces a constant output offset of about 3 V. Depending on gain this output offset gets attributed the numbers shown in diagnostics. It can be a problem of that amplifier or of the -19 V reference. Better fix this before playing with firmware update.

Regards, Dieter

How should I test the -19 V reference voltage? Is that associated with R200?

EDIT:
Looking at the schematics, I assume that the -19 V is associated with R232:
16.5 V, 18.685 V

[Kleinstein]

The -19 V is availabe at several points: the output of U211 and at R200 , R201.

The low offset for the high speed range may be just the difference to the slow output.

Though shown as a problem with the amplifier offset, chances are this is a problem with the ADC or reference (the -19 V).

[leighcorrigall]

...

The expected voltages at R200 would be similar to R201.

...

As indicated by the attached tables, the voltages of R200 and R201 seem to match reasonably well. I shorted the DCV input and the range was set to 10 V during the measurements at each resistor.