Advantest R6581 8.5 digit DMM mini teardown/repair

[dietert1]

Checking the rundown sources with a multimeter: One could try and measure the nets denoted as +S64 +S4 -S1 -S16 and -S256. They are R213 pins as well, except -S256. -S256 is accessible at the source of Q206. The mosfets drive the current sources either into the integrator input or into ground. All five voltages should be near zero with -S256 the largest.
With DCV, 10 V, 10 PLC and input short i get: +S64  = 5.6 mV, +S4 = 0.1 mV, -S1 = - 0.1 mV, -S16 = -1.7 mV, -S256 = -24.3 mV.

I also probed integrator output with a scope, see above. I could see runup and rundown with different slopes. Guess it requires some hours to find, record and measure each slope.

Regards, Dieter

[leighcorrigall]

Checking the rundown sources with a multimeter: One could try and measure the nets denoted as +S64 +S4 -S1 -S16 and -S256. They are R213 pins as well, except -S256. -S256 is accessible at the source of Q206. The MOSFETs drive the current sources either into the integrator input or into ground. All five voltages should be near zero with -S256 the largest.
With DCV, 10 V, 10 PLC and input short I get: +S64  = 5.6 mV, +S4 = 0.1 mV, -S1 = - 0.1 mV, -S16 = -1.7 mV, -S256 = -24.3 mV.

...

These voltages are with respect to ground and with the R6581T set to VDC mode (10 PLC, 10 V range):

+S64  = 6.25 mV
+S4 = 0.45 mV
-S1 = -0.05 mV
-S16 = -1.5 mV
-S256 = -24.9 mV = Q206 (S)

The +S64 and +S4 have been confirmed a number of times. Are these results caused by the +17 V rail being +16.72 V or is it something else?

[dietert1]

The exact numbers depend on the ADC result and may vary a little. So they agree. I measured +S4 once more, using R213 Pin 7 as precision Gnd and this time i have 0.35 mV.
By the way, when i measure  +S1024 at drain of Q201 i am getting 96 mV, while Sx at the drain of Q200 is -49 mV. A factor two is expected, since the positive source steers 1.5 mA, while the negative one steers 0.75 mA (again DCV, 10V range, 10 PLC, inputs shorted). In this sense one can expect +S4 to be 4x more than -S1.
I think in order to discover the origin of your "stealth offset" you should find a friend with a scope. More subtle errors like integrator input leakage or mosfet switch leakage won't reveal with a multimeter.

Regards, Dieter

[leighcorrigall]

The exact numbers depend on the ADC result and may vary a little. So they agree. I measured +S4 once more, using R213 Pin 7 as precision Gnd and this time I have 0.35 mV.

...

Good to know. Too bad I haven't found anything out of the ordinary except for the 7.2Vref so far.

...

By the way, when I measure  +S1024 at the drain of Q201 I am getting 96 mV, while Sx at the drain of Q200 is -49 mV. A factor of two is expected, since the positive source steers 1.5 mA, while the negative one steers 0.75 mA (again DCV, 10V range, 10 PLC, inputs shorted). In this sense, one can expect +S4 to be 4x more than -S1.

...

+S1024 = 103.8 mV = Q201 (D)
Sx = -49.8 mV = Q200 (D)

...

In this sense, one can expect +S4 to be 4x more than -S1.

...

If I interpret what you said literally, then 4 x +S4 does not equal -S1.

...

I think, in order to discover the origin of your "stealth offset", you should find a friend with a scope. More subtle errors like integrator input leakage or MOSFET switch leakage won't reveal with a multimeter.

Regards, Dieter

That is unfortunate. I do have a loaner. It is a surprisingly expensive computer interface by Analog Arts. 

SL917 http://www.analogarts.com/

I suppose I should begin learning how to use it.

By the way, does anyone know the exact purpose of R170?

[Kleinstein]

R170 is a trimmer at the precision amplifier. It effects the offset and TC of the amplier. As the offset is normally corrected digitally, the drift is the more important parameter. The adjustment range for the offset should be relatively small ( more like <<100 mV).
The adjustment for zero drift could be relatively time consuming, as it includes a change in temperarature.

[serg-el]

1 kOhm

[dietert1]

That trimmer in the precision channel isn't the best design, thanks for the hint. These parts develop contact resistance over the years. In our instrument i am measuring 456 Ohm and 409 Ohm on its two sides while i get 856 Ohm for the total (in circuit). So contact resistance may be 4.5 Ohm with unknown temperature dependence. In our instrument i can replace the potentiometer by one 47 Ohm fixed resistor and a short.
In another instrument the trimmer may be OK or even worse, don't know.

If those mosfet switches SD215DE are all OK, they are similar to each other and the measured voltages proportional to the currents they steer. If +S1024 has 100 mV, then -S256 should have -25 mV, +S64 about 6 mV, -S16 about -1.5 mV, +S4 about 0.4 mV and -S1 about -0.1mV. There is a pattern.
I just happened to notice R236. I think its voltage is proportional to integrator input leakage. So here one wants to see null. One microvolt would indicate one nanoampere, already a lot.

Regards, Dieter

[leighcorrigall]

dietert1, can you please let me know what the voltage output for your R6581 is at U209? I measure 16.72 V.

[dietert1]

I just got 16.66 V at the output of U209. U211 outputs -18.61 V and U210 outputs -9.80 V. Reference voltage is 7.05 V. The LTZ1000 heater gets 9.55 V (emitter of Q1 after about 5 minutes of heating). What else?
By the way i added 100 nF film capacitors to the reference processing stages. Leakage of those caps was checked and below 10 pA. Later i saw 10 nF caps in a HP 3458A schematic.

Regards, Dieter

[leighcorrigall]

I just got 16.66 V at the output of U209. U211 outputs 18.61 V and U210 outputs -9.80 V. Reference voltage is 7.05 V. The LTZ1000 heater gets 9.55 V (emitter of Q1 after about 5 minutes of heating). What else?

...

Thanks, dietert1.

My values are very similar. I was hoping that there was going to be a difference here. The phantom offset is most troubling.

[Kleinstein]

100 nF caps at the reference amplifier make absolute sense. The noise in the 1-5 kHz range can add to the ADC noise, as the modulation during run-up acts like mixer and brings this frequency band to near zero. 2 x the reference noise may be already detectable.

[Mickle T.]

By the way i added 100 nF film capacitors to the reference processing stages.

Without comparative measurements, it looks more like a placebo.

[dietert1]

Did you check the voltage over R236 to be zero? That would be a first test for integrator input offset current.

The idea is like this: Integration time for x1Zero_1 is about 100 usec and for x1Zero_2 is 10 PLC = 200 msec. That gives a factor 2000. Integrator scaling changes by a factor 21.5 nF / 1.5 nF = 14 between the two modes, so x1Zero_2 is affected 2000 / 14 = 140 times  more. Integrator input offset current might produce a large offset for the precision channel without disturbing the fast measurement.

Regards, Dieter

[Kleinstein]

An input current to the integrator would be relative to the current from R200. So the same scale factor would apply for both integration times. It would add some offset, like a costant offset voltage.
Things are a bit different for leakage at the U213/1 at the CAZ capacitor and thus the other input of the integrator. Here the effect would scale with the integration time and integrator capacitance.
This would be something a bit tricky to measure. Indirectly leakage at U213 or bias current of U205 would should up as a current pulse at R236 to0, as the current for the Zero compensation phase.  A measurement with just a DMM is tricky. Even with a scope it is not easy.

Filtering the reference would be mainly because of the 1-5 kHz band of the reference noise (and some OP / resistor noise). For this 100 nF inparallel to some 10 K are sufficient to at least reduce this noise part. The effect is no expected to be large (e.g. - 5-10% for the noise), but the effort is also relatively low.

[dietert1]

My idea about the filter caps is to avoid rectification of pulses at the opamp input transistors. One way to absorb pulses that
may stem from nearby digital circuitry is a capacitive short to the low impedance opamp output. I think this is good design
practice, no need for detailed evaluation. A well designed circuit should have bandwidth limits defined by linear parts.

Concerning integrator offset current: During integrator reset it will flow through R236. In 10 PLC measurement integrator
reset is active for 7 msec. So the test is only 7 msec / 207 msec sensitive. Yet, to produce an offset of 3 V the offset current
would have to be about 0.75 mA * 3 V / 10 V = 0.225 mA, so one would expect 7 mV over R236 . It's a test of Q211,
U213/1, U205 and U206 that one can do with a multimeter. When i look there with a HP 3478A it shows 2 or 3 uV, likely
all thermal EMF.

Regards, Dieter

[leighcorrigall]

...

Concerning integrator offset current: During integrator reset it will flow through R236. In 10 PLC measurement integrator reset is active for 7 msec. So the test is only 7 msec / 207 msec sensitive. Yet, to produce an offset of 3 V the offset current
would have to be about 0.75 mA * 3 V / 10 V = 0.225 mA, so one would expect 7 mV over R236. It's a test of Q211, U213/1, U205 and U206 that one can do with a multimeter. When I look there with an HP 3478A it shows 2 or 3 uV, likely all thermal EMF.

Regards, Dieter

Thanks for the new test, dietert1.

I shorted front inputs in voltage mode and set the integration time to 10 NPLC. The voltage drop across R236 was < 1 uV on a Keithley 2000. 

[leighcorrigall]

Hmmm... Moving onto MickleT's constant current sink schematic.

I am measuring -9.61 between GND4 and TP400 (VS), which I think should be -22 V.

[leighcorrigall]

Hmmm... Moving onto MickleT's constant current sink schematic.

I am measuring -9.61 between GND4 and TP400 (VS), which I think should be -22 V.

There seems to be a lot more problems:

GND4 & TP400 (VS) = -9.61 V (normally -22 V)
GND5 & U107 (pin 4) = -14.9 V (normally -27 V)
GND4 & U403 (pin 5) = -14.9 V (normally +15 V)

[Mickle T.]

The last two isn't a problem but only typos while converting to CAD schematics.
Instead of U107 must be U400, and U403 (pin 5) = -14.9 V is Ok too.

[leighcorrigall]

The last two isn't a problem but only typos while converting to CAD schematics.
Instead of U107 must be U400, and U403 (pin 5) = -14.9 V is Ok too.

Thank you.

Q113 should be renamed to Q400 as well.

These are the values I get:

GND4 & U400 (pin 6) = GND4 & R402 = -10.15 V

GND4 & Q400 (pin 1) = -10.15 V
GND4 & Q400 (pin 2) = -9.61 V
GND4 & Q400 (pin 3) = -27.5 V

Could this transistor be dysfunctional?

[Kleinstein]

From the measurend voltages the transistor is working OK. The question is, why U400 only gives some 10.1 V.
Could it be U401/2 and U401/3 switched wrong for some odd reason.

In the schematis (V5), there seems so be a few more minor mistake with the supply to U410 and likely also U401, U402.  Very likely V+ goes to +5 V and GND to GND of some type.
The voltage noted for pins 20 and 21 of R400 are not compatible with the resistor values given.

The naming for the U406-Q0 and Q1 signal as x1 and x2 is also odd, as the unused range with Q1 active would be more like 1/3 the currents.

[leighcorrigall]

From the measurend voltages the transistor is working OK. The question is, why U400 only gives some 10.1 V.
Could it be U401/2 and U401/3 switched wrong for some odd reason.

In the schematis (V5), there seems so be a few more minor mistake with the supply to U410 and likely also U401, U402.  Very likely V+ goes to +5 V and GND to GND of some type.
The voltage noted for pins 20 and 21 of R400 are not compatible with the resistor values given.

The naming for the U406-Q0 and Q1 signal as x1 and x2 is also odd, as the unused range with Q1 active would be more like 1/3 the currents.

I switched into resistance mode and it reads -21.6 V at TP400. No change occurs when varying the integration time or the ranges.

Maybe this area is working properly.

[Kleinstein]

With -21.6 V at TP400 things are pretty normal considering the slightly lower refvoltage in the meter. When not in Ohms mode (or some ACAL steps) the voltage at TP400 does not really matter. So it may as well enable Q1 insteadt of Q0 and set it to the otherwise unused mode.

So I think at least for the coase main part the Ohms source does woke properly.  This still does not say much about details like leakage, e.g. at Q402-Q405 or D402 that changes with the external voltage the ohms source sees. This is hard to measure and could effect the linearity (and the ACAL results). In this repect I don't like the way the ohms source is build.

[leighcorrigall]

martinr33 and TiN have said that their R6581's have similar diagnostic values as my own unit. This seems to be a common theme with this instrument. 

Might be related to the firmware and never taken care of after a hardware revsion.

...

x1Zero_1 = -0.0167 VDC (10 V range)
x1Zero_2 = +2.9543 VDC (10 V range)
x10Zero = +295.254 mVDC (1000 mV range)
x100Zero = +29.435 mVDC (100 mV range)
7.2Vref = +10.04475 VDC (10 V range), actual: +7.07 V
-10Vref = -6.88804 (10 V range), actual: -9.83 V
-1Vref = -689.901 mVDC (1000 mV range), actual: -0.984 V
-0.1Vref = -69.106 mVDC (100 mV range), actual: -0.098 V
Int_temp = +70.5 °C, actual temperature is much less than this value
10mA = -6.909 mADC (10 mA range)
1000uA = -690.64 uADC (1000 uA range)
100uA = -69.082 uADC (100 uA range)
10uA = -6.905 uADC (10 uA range)
1000nA = +1967.28 nADC (1000 nA range)
100nA = +196.73 nADC (100 nA)
10nA = +285.41 nADC (100 nA)

...

[leighcorrigall]

Does anyone else have a linearity problem with their instrument in the negative 10 V range? A Fluke 732A was used to source 1 V to the R6581T. I know the - 1 V on the 10 V range is problematic because the 1 V measurement on the 1000 mV range matches with the + 1 V on the 10 V range only. Both - 1 V and + 1 V on the 1000 mV range match, which is expected.

R6581T settings:
integration time: 100 NPLC
range: 10 DCV
AZERO: ON/OFF (there was no noticeable difference that I could tell at a given temperature)

I compared my own measurements of the Fluke 732A and an xDevs KX-ref (soaked for 1750 hours) with a calibration laboratory report (before and after calibration). See attached Figure. For the 0 V measurement, I shorted the front input of the R6581T and took measurements (for both AZERO cases) after waiting for a while. After TRANSCAT calibrated the instrument in September, I have not touched the INTCAL feature.

Thanks.