Bench DMM Noise & Stability Comparisons with TestController

[Salient666]

Here's a 25-hour test of the 1399, NPLC100, with temperature measurement.

[Salient666]

I'm gonna test these soon, just for fun.
It's an AD 584KH and a TL431 reference from Ali Express.
I will test the AD584 at it's 10V and the TL431 at 2.5V since those values are what can be considered 'native' and eliminates the rest of the (questionable?) components on the PCB and both at 15V input voltage from a linear bench supply.

Quick check already reveals that input voltage actually determines output voltage a little on these references but that's normal behavior as I understood.

[TylerPeppy]

I'm gonna test these soon, just for fun.
It's an AD 584KH and a TL431 reference from Ali Express.
I will test the AD584 at it's 10V and the TL431 at 2.5V since those values are what can be considered 'native' and eliminates the rest of the (questionable?) components on the PCB and both at 15V input voltage from a linear bench supply.

Quick check already reveals that input voltage actually determines output voltage a little on these references but that's normal behavior as I understood.

I have a similar 584KH board I haven't tested in recent memory, but would be interesting to see against yours.

Tyler

[Salient666]

Here are the test results. Tested after I've installed the refs in a case with foam inside as well.

Tests done for 2.5 hours @ both NPLC=10 and NPLC=100. Everything with AutoZero=ON

[KungFuJosh]

I made some mods to my K2010 to test it out. I had the ADR1399 using a 1uF cap and 5.1Ω resistor (tented together over the cap pads). Instead, I swapped the cap/resistor combo with a single 10uF tantalum cap (with a rated ~5Ω ESR). I also Crapton taped some foam over the ADR1399 ref.

This is @TylerPeppy's reference, AZ on, 10PLC/10Filter. Warmup overnight.

Thanks,
Josh

[TylerPeppy]

I made some mods to my K2010 to test it out. I had the ADR1399 using a 1uF cap and 5.1Ω resistor (tented together over the cap pads). Instead, I swapped the cap/resistor combo with a single 10uF tantalum cap (with a rated ~5Ω ESR). I also Crapton taped some foam over the ADR1399 ref.

This is @TylerPeppy's reference, AZ on, 10PLC/10Filter. Warmup overnight.

Thanks,
Josh

This looks very promising.

I am assuming that you were using a 1uF tantalum cap before? I would be interested to see what a film cap or C0G / NP0 ceramic might be able to do. I was messing with the stabilisation network on my references beforehand, and a higher capacitance semeed to help across the zener a bit, but I didn't try removing the stabilisation resistor in the RC snubber.

Tyler

[KungFuJosh]

I am assuming that you were using a 1uF tantalum cap before? I would be interested to see what a film cap or C0G / NP0 ceramic might be able to do, as tantalum still has a high TC. I was messing with the stabilisation network on my references beforehand, and a higher capacitance semeed to help across the zener a bit, but I didn't try removing the stabilisation resistor in the RC snubber.

No, I was using a high-quality C0G/NP0 1uF cap. 10uF tantalum is only because it has a higher series resistance, and negates the need for the 5.1Ω resistor. It's covered in the datasheet for the ADR1399. The tant cap I chose was rated 5Ω in its datasheet, and measured that at 100Hz on my ST42 tweezers.

A tantalum cap is a much cleaner choice for upgrading an LM399 with only a cap to ground. I didn't want to keep the tent I pitched as a long-term solution. I should have read further in the datasheet before ordering stuff in the first place.

Thanks,
Josh

[KungFuJosh]

Here's @TylerPeppy's ref with longer stability tests.

Thanks,
Josh

[Salient666]

I've got my Ebyte temp logger now and seems to work fine.

[TylerPeppy]

Here's @TylerPeppy's ref with longer stability tests.

Thanks,
Josh

Yeah, your 2010 seems to look exactly like my 34465A in terms of noise now. I guess stripping the regular snubber network and using the reccomended 10uF tantalum was a good move. I might give that a try in my unit to see if we see increased stability too.

[KungFuJosh]

Here's the final 2 from that test. Now to mess with the K2010 some more.

[TylerPeppy]

Here's something interesting people might be able to use (or not). Usually I would have Cursor or something similar chug at this sort of task, but I decided to do it manually...

I spent yesterday building a Python utility to do the stability tests on the DUTs, as I couldn't get over how manually I was doing things before. It ingests an IP address for the DMM, and is currently set to use a TA612C over serial for temperature logging. You can configure the test run duration, sampling interval, and number of repetitions. It will poll all of the data and then spit out a CSV after every repetition with the raw data, and a PNG of the measurement graph and temperature.

At the end of all repetitions, it will spit out a CSV with the statistics (Min, Max, Mean, STD-Dev (selectable sliding window size)) with all the testrun info. While I can't get the exact same std-dev as testController (might be a function of pandas' calculation algorithm), it still seems to work quite well for most purposes. I added some sample outputs here. I don't claim to be a professional (those who know me, know I am still a student), so the software could always be better.

PS: The TA612C_read.py is just an example pulling temperatures over serial from all 4 channels and printing them, sampling every 1 second. Might be a good starting point if anyone needs it. The TE_Read_Script.py is the one you need to run.

Reqs:
- matplotlib
- colorama
- pyvisa
- pandas
- serial

[KungFuJosh]

I spent yesterday building a Python utility to do the stability tests on the DUTs, as I couldn't get over how manually I was doing things before.

I'm confused. What weren't you able to do with TestController?

[TylerPeppy]

I spent yesterday building a Python utility to do the stability tests on the DUTs, as I couldn't get over how manually I was doing things before.

I'm confused. What weren't you able to do with TestController?

The example of the 5 sets of runs I did previously for the stability tests on my reference works well. I had to manually start the logging, export the data, take the screenshots, and clear + reset before further runs. This script is designed to do all of the above and can be programmed to do so automatically.

This will do the 5 sets of runs, with programmable durations, and export the CSV data for each run, along with a visualization of the data. Very much "click-and-run" and wait for the data to be complete.

[KungFuJosh]

I spent yesterday building a Python utility to do the stability tests on the DUTs, as I couldn't get over how manually I was doing things before.

I'm confused. What weren't you able to do with TestController?

The example of the 5 sets of runs I did previously for the stability tests on my reference works well. I had to manually start the logging, export the data, take the screenshots, and clear + reset before further runs. This script is designed to do all of the above and can be programmed to do so automatically.

This will do the 5 sets of runs, with programmable durations, and export the CSV data for each run, along with a visualization of the data. Very much "click-and-run" and wait for the data to be complete.

Pose that as a question in the TestController thread.

[TylerPeppy]

I spent yesterday building a Python utility to do the stability tests on the DUTs, as I couldn't get over how manually I was doing things before.

I'm confused. What weren't you able to do with TestController?

The example of the 5 sets of runs I did previously for the stability tests on my reference works well. I had to manually start the logging, export the data, take the screenshots, and clear + reset before further runs. This script is designed to do all of the above and can be programmed to do so automatically.

This will do the 5 sets of runs, with programmable durations, and export the CSV data for each run, along with a visualization of the data. Very much "click-and-run" and wait for the data to be complete.

Pose that as a question in the TestController thread.

Figured I might as well, but came up with a quick solution to try myself, never hurts to do things that way

[Martin72]

After “a bit of a drama,” I'll finally have Tyler's reference in my hands in a few days.

I would then perform the same measurements as with the Circuit Valley reference.
Maybe just with the two 7.5-digit meters—we’ve seen plenty of evidence that the 4065A performs well.
Last time, the measurements were taken with 10PLC, 100PLC, AZ on, and AZ off.
Anything else?
For example, I’d let the two meters warm up, but start the reference cold and thus record the stabilization phase as well.

[KungFuJosh]

For example, I’d let the two meters warm up, but start the reference cold and thus record the stabilization phase as well.

That's good to measure, but for the regular tests, I'd let the reference warmup a few hours also.

[Martin72]

That's clear.

[TylerPeppy]

“a bit of a drama,”

DHL in CAN nearly had me pulling my hair out telling me that the item was "seized", when all German customs told me is that they needed a phone from you. I only got this information after calling about 5 or so times to get ahold of someone.

Maybe just with the two 7.5-digit meters—we’ve seen plenty of evidence that the 4065A performs well.
Last time, the measurements were taken with 10PLC, 100PLC, AZ on, and AZ off.
Anything else?
For example, I’d let the two meters warm up, but start the reference cold and thus record the stabilization phase as well.

I would be very interested to see a warmup curve, as I have tried to have the fewest power cycles for my DUTs in reason of some unrelated research on the effects of heating and cooling the zeners on their stabilities.

Apart from that, I believe all tests you listed are more than sufficient. A side-by-side of the duts would be quite nice when heating up, and then having had stabilised for a little while.

Martin, I will shoot you a quick PM after about something unrelated to the testing...

Thank you,
Tyler

[Martin72]

Coming soon to theaters....
Probably this weekend.
The outfit looks good so far.

[KungFuJosh]

The outfit looks good so far.

Would have looked better if I remembered to hide a penguin inside the panels. 

...I was too excited my laser was working again, and forgot about the important things.

[TylerPeppy]

Coming soon to theaters....
Probably this weekend.
The outfit looks good so far.

Exciting times! At least it arrived

penguin inside the panels. 

There should have been a surprise to that effect in there 

[Martin72]

Probably this weekend.

Sunday, to be exact—Everything's already set up...
And Josh was right; a 5-port network hub like that quickly reaches its limit.
If I were to connect the SDM4065A now, all the ports would be in use.
As I said, my plan is to first record the power-up torque until the “Tyler reference” stabilizes, and then continue recording as usual.

[TylerPeppy]

Probably this weekend.

Sunday, to be exact—Everything's already set up...
And Josh was right; a 5-port network hub like that quickly reaches its limit.
If I were to connect the SDM4065A now, all the ports would be in use.
As I said, my plan is to first record the power-up torque until the “Tyler reference” stabilizes, and then continue recording as usual.

Josh can comment on this, but how much luck (if any) have you had with using batteries and hotplugging? I am still weary after what I think Josh did which was kill one of my rev1 units with hotplugging.

Tyler