Author Topic: Advantest R6581 8.5 digit DMM mini teardown/repair (Read 203591 times)

branadic · « **Reply #600 on:** August 27, 2019, 11:54:05 am »

Don't get me wrong Kleinstein, but beside the turn over issue there is no INL issue. It's more like there is a gain error present, that results in a negative slope over the entire 10V range, thus +10V /= -10V.

-branadic-

Dr. Frank · « **Reply #601 on:** August 27, 2019, 12:01:44 pm »

Hi branadic,
haven't you calibrated the 4000As gain yet?
Frank

branadic · « **Reply #602 on:** August 27, 2019, 04:02:25 pm »

Frank, it's not a gain issue of the Datron 4000A, but a gain issue of the R6581D, if gain is the correct word for it. You can clearly see that, as K3458A shows that Datron 4000A produces the correct output voltage. Both meters were in parallel to the Datron 4000A.

-branadic-

Kleinstein · « **Reply #603 on:** August 27, 2019, 04:48:04 pm »

The calibration of the 6581 seems to be off a little (e.g. some 5 ppm) compared to the other instruments.

For most of the part the linearity looks good. The range from 0 to about -2 V and especially the jump near zero is probably the worst part of it. For the U³ part it looks like that much of the error is due to the calibrator, as the curves look similar for both meters. Still there is a little difference.
However the curve does not look like there should be a lot of turn over error.

branadic · « **Reply #604 on:** August 27, 2019, 05:29:51 pm »

Kleinstein, no calibration is not off, it was freshley calibrated at 0V and +10V comparing a 10V reference to the K3458A (artifact calibration), however -10V doesn't fit and therefor the slope in the deviation plot. To make it more clear, the plot of deviation [V] is meter reading - set voltage. The resulting slope from +10V to -10V is exactly my issue and I have no idea how to get rid of that.

-branadic-

Kleinstein · « **Reply #605 on:** August 27, 2019, 06:08:18 pm »

Is the zero offset OK ? The first curve for the 6581 looks like there may be quite some error at 0 V.

The deviation plot also has a slope for the positive side, so the error is no just for the negative side.
A turn over error suggests that the positive and negative behave different. However the curve does not look that asymmetric. Still it is hard to judge how much is from the calibrator and how much from the meter(s).

A possible problem could be from fast versus slow measurements. Thermal effects could take some time and thus behave different for just a zero reading and the zero reading as part of the AZ cycle. So there could be some settling effects, that might effect the calibration measurement. An error in the zero (e.g. thermal EMF from the relay) could also be a factor if rather large. This could improve with the relay fix, especially if the internal correction for the offset is somewhat handled the wrong way during calibration.

branadic · « **Reply #606 on:** August 28, 2019, 05:53:44 am »

I think I had a temperature issue and temperature changed over time, since the measurement was done with NPLC100, 10 values per voltage step and 20s delay time between each voltage step for the calibrator to stabelize and from +11V to -11V.
I will repeat this measurements with NPLC20, only 5 values per step, 10s of delay between each step and an alternating voltage step strategy, which means +0V, -0V, -100mV, +100mV and so on. This way I should get similar results for both, positive and negative values without temperature variations effecting the measurement to much and since the Datron 4000A switches output polarity only results will become better, hopefully.

-branadic-

e61_phil · « **Reply #607 on:** August 28, 2019, 06:06:20 am »

20s seems a bit short to me. My Fluke 5440B takes much longer to get very stable. For the first experiments I took measurements with the 3458A and I did that until three succeeding 3458A measurements are within a given range.

If you want to test transfer accuracy it might be better (or another experiment) to jump between 0, 1V and 10V with some random numbers on top. Such a meausrement is done in a couple of minutes and don't need ultra stable temperature enviroment. And the result is more meaningful than these nice looking, but not much saying INL curves. They are nice for comparisions, but for absolute numbers to calc with, I prefer the steps.

Kleinstein · « **Reply #608 on:** August 28, 2019, 07:58:56 pm »

To reduce the temperature effect and make is visible one should run the same points several (e.g. 5 x) times. This may require a shorter measurement (may not be that practical as the settling time can be significant) or less points. (e.g. 0.25 V steps and fine steps only for a smaller range, possibly as a separate run).

branadic · « **Reply #609 on:** September 03, 2019, 07:53:34 pm »

Finalized my binding post mod. The adapter is 3D printed and designed to match the LowThermal binding posts. It looks like Advantest initially planed to install this higher quality posts, but decided for the brittle plastic solution instead. It's still no 3458A, but definitely higher grade now and I like it.

-branadic-

EDIT: Attached are the CAD files for the adapter.

branadic · « **Reply #610 on:** September 05, 2019, 08:36:46 pm »

And just for completeness I received two 3D printed adapters with a ceramic filled resin today, so I changed the front adapter and also installed the rear binding posts. There is no drilling or other modification needed, just simply install the binding posts to the adapter and screw it down to the chassis.

-branadic-

MiDi · « **Reply #611 on:** September 06, 2019, 10:56:54 am »

A very nice update, it gives even a higher value look.
Could you reveal where to get this posts and their price?

branadic · « **Reply #612 on:** September 06, 2019, 01:23:40 pm »

You can get the binding posts here: http://www.lowthermal.com/cables-and-connectors.php
Please contact them yourself for a quote.

-branadic-

branadic · « **Reply #613 on:** January 19, 2020, 12:40:30 pm »

While measuring one of my LTZ references @25°C in a thermal chamber over a longer periode with my R6581D I noticed that there is quite some t.c. visible. Therefor the internal temperature was logged together with the voltage readings. It turned out this t.c. is due to the meter itself and about 0.8ppm/K. A measurement of the t.c. of the internal reference itself revealed, that this t.c. is not due to the reference, but something else. Time to upgrade R6581!
Several improvements are documented here:
http://bbs.1ppm.cn/topic/23/advantest-6581磨机-zy的得意杰作
http://bbs.38hot.net/forum.php?mod=viewthread&tid=28698&highlight=6581

It just so happened that pipelie organized a group buy to replace R234 which boosts the zener voltage to 10V and R200 as part of the voltage to current converter.

R234 original: Alpha Electroncis SLD (7k34 X F 10k0)
R234 replacement: Vishay VHD200, hermetically sealed

R200 original: Alpha Electronics SS103F20k00 XF
R200 replacement: Vishay 300193ZT

The resistors arrived on friday so the weekend was used to replace them. Unfortunately there is no drop in replacement for R200, so the leads of the 300193ZT needs to be bend slightly into the correct position, while the VHD200 fit perfectly.
Now that the replacement is done it‘s time to let the meter settle with the new components installed, which is said to be about a month and see how this modification turns out.

Not to hide anything, removing the networks is a real pita and needs careful work and a lot of solder to heat up all pins equally. The vias are also very small and hard to clean. So unless you are well experienced in such work I would not recommend it, as you can damage a lot.
I use drill needles to finally remove last solder pieces inside the vias while heating the pad.
Also note that after that modification the meter needs calibration. To be continued ...

-branadic-

Kleinstein · « **Reply #614 on:** January 19, 2020, 04:37:46 pm »

That change is a brave and somewhat risky change. These two resistor sets (especially R200) are really critical. R200 not only effects the overall gain, but can also have an effect on the INL.

I am not so sure the meter would really need a full calibration. Most of the changes should be compensated for by ACAL.

For longer term drift measurements one should consider a ACAL call from time to time anyway. This should remove much of the TC, at least the part caused by R200 and R234.

branadic · « **Reply #615 on:** January 19, 2020, 05:43:38 pm »

It sucks, if the meters t.c. dominates a 24h measurement of a LTZ1000 reference, so that was at least the reason for me to change this dividers, hoping that overall t.c. improves. However, this modification were already proven by the chineese R6581 group (a Tencent QQ group) to improve the t.c. of the meter. It was the second run of a group buy on such resistor dividers and I had the chance to participate. Unfortunately the progress on the meter made by this group are barely visible to the rest of the world. The meter don't need a full calibration, but an artifact calibration (short, 10V, 10kohm).

This modification still doesn't improve reversal error (+10V, -10V), that is hopefully another story for the future.

-branadic-

guenthert · « **Reply #616 on:** January 19, 2020, 06:47:47 pm »

Quote from: branadic on January 19, 2020, 05:43:38 pm

It sucks, if the meters t.c. dominates a 24h measurement of a LTZ1000 reference,
[..]

It does? Isn't the whole point of a voltage reference to be stable? More stable than the measurement tools available?
You know that you will need now a new, more stable voltage reference, don't you? And then ...

e61_phil · « **Reply #617 on:** January 19, 2020, 07:47:13 pm »

Unfortunately, the TC of my 3458A is also not that great (but better than 0.8ppm/K). Therefore, I use the 3458A as a ratio machine only. I build a low thermal multiplexer and the 3458A just compares voltages. No matter if I compare resistance or voltages. And that is what the 3458A is best for.
During a single comparision point the temperature will not change significant. And it is much easier to put the reference you compare against in a stable enviroment than a big 19" DMM like the R6581 or 3458A.

serg-el · « **Reply #618 on:** January 28, 2020, 08:08:27 am »

Incorrect scheme, from the manufacturer.
Expand the capacitor.

Wrong.

Circuit diagram.

Right.

branadic · « **Reply #619 on:** January 28, 2020, 09:01:39 am »

Have you double checked that it's not only a wrong screen printing, but the component is installed right? I would expect a blown capacitor if installed in the wrong direction.

-branadic-

serg-el · « **Reply #620 on:** January 28, 2020, 09:03:50 am »

Twice, three times ...

Measured voltage.
The result is the same.
Incorrectly installed.

quarks · « **Reply #621 on:** January 30, 2020, 09:06:57 am »

I just checked my R6581 and silkscreen and orientation of cap is matching.
Therefore this problem seems to be corrected.

gamalot · « **Reply #622 on:** January 30, 2020, 07:17:04 pm »

My 6581 is also wrong!

branadic · « **Reply #623 on:** January 30, 2020, 08:27:40 pm »

Had to swap the capacitor in my unit too.

-branadic-

serg-el · « **Reply #624 on:** January 31, 2020, 11:33:49 am »

Quote from: quarks on January 30, 2020, 09:06:57 am

I just checked my R6581 and silkscreen and orientation of cap is matching.
Therefore this problem seems to be corrected.

Well, you're in luck.
What year is your multimeter?

But TiN, Mickle_T, ramon, deadlylover, chihaxinh and I’m out of luck.

https://www.eevblog.com/forum/metrology/advantest-r6581-8-5-digit-dmm-mini-teardownrepair/
https://www.eevblog.com/forum/repair/pls-help-me-repair-advantest-r6581/

They need to fix their multimeters.

I wonder from what year, and from what serial number did the Advantest correct this error?


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: Advantest R6581 8.5 digit DMM mini teardown/repair (Read 203591 times)

Share me