Bench DMM Noise & Stability Comparisons with TestController

[Salient666]

Here is my test of the Siglent SDG1062X "versus" the DMMCheck Plus.

- All meters have a warm-up time of approx. 1.5 hours
- The SDG1062X has a warm-up time of approx. 1.5 hours
- The DMMCheckPlus has a warm-up time of approx. 1 hour
- Output of both is: 5.000Vrms, 100Hz, Square Wave
- All meters set to: ACV, AC-Filter: 20Hz (Keithley=30Hz), no other filters
- Both devices tested at 5 minutes and 30 minutes

[Salient666]

Quiet around here...

In the meantime I've made another simple test for DCI this time.
I used my DMMCHeck Plus which has a very stable 1mA DC output. I let the meters warm-up for a couple of hours and the DMMCheck Plus for about one hour.
Everything is with AutoZero ON an 10PLC.

Again, the winner for lowest noise figures is the Keysight. This seems like a common theme.
I still prefer the DMM6500 though for it's much quicker start-up time and it's better UI.

[Martin72]

Hi,

I still prefer the DMM6500 though for it's much quicker start-up time and it's better UI.

You can do that with a clear conscience, because I'd say the visual spectacle created by the “Noise” measurement is more of an entertaining feature.

[KungFuJosh]

Here's some early tests of @TylerPeppy's ADR1399 based reference. I'm calling it the "Almost 10V" ref.

Looks good so far. I'm mad at Windows though. I was doing an overnight test to see how it went with larger temperature drifts, and Windows decided to update, restart the computer, and screw up the com port the 2010 was connected to. 

Tyler measured 9.9840125V, and I'm seeing an average 9.984094V on the 34465A, and an average 9.9841397V on the 2010. Neither are recently calibrated, but I'd say that's pretty good anyway.

I stepped the voltage (16.5V to 19.5V) on the SPD3303X supplying the ref, and there was no shift on output voltage.

Thanks,
Josh

ETA: Math channel setup the same as Tyler with 1799 points.

[TylerPeppy]

Looks good so far. I'm mad at Windows though. I was doing an overnight test to see how it went with larger temperature drifts, and Windows decided to update, restart the computer, and screw up the com port the 2010 was connected to. 

I had to do this on an IT deployment I administered last year. You can disable auto and scheduled updates on most windows 10/11 systems. You can modify a key in the registry under "HKEY_LOCAL_MACHINE\SOFTWARE\Policies\Microsoft\Windows\WindowsUpdate\AU" called "NoAutoRebootWithLoggedOnUsers" set to 1 and it should prevent that.

I'm calling it the "Almost 10V" ref.

There is a story behind this... I ended up assembling the units I got purely without external trim to get an idea of the spread on Vz from 3 different ADR1399 parts. Turns out my main one is 9.97, Josh's is 9.98, and another is 10.003. The ADR1399s aren't spec'd for tight tolerances on initial Vz so I expected that anyways. In the project description (https://github.com/T-622/uDR1399-10V-Reference-Standard), you can use R12 + R13 to trim very slightly. More instabilities with the TCs of those resistors can be expected though.

Last I measured though, the 9.9840125V was an average taken at 72 or 96hrs of burn in. My meter was calibrated in September of 2025, so might need to get it calibrated soon again. I aim for 1yr intervals on the DMMs if possible.

It does look like in some tests with higher SD, the temperature drops from 25C to 23.4C? The voltage on the 34465A seems to drop, while the K2010 is the inverse behavior?

Tyler

[KungFuJosh]

My K2010 is still unsealed. The case is on, but not screwed in, and the rear feat that seals the back is not on either. I expect it to respond more to ambient changes. For the most part, they seem to move the same way, with the K2010 being more dramatic (as expected).

I haven't decided what I'm going to do with the K2010 yet. I might wait a year or two before I do anything more to let the new reference age. I dunno. I guess it depends on boredom/inspiration.

[TylerPeppy]

My K2010 is still unsealed. The case is on, but not screwed in, and the rear feat that seals the back is not on either. I expect it to respond more to ambient changes. For the most part, they seem to move the same way, with the K2010 being more dramatic (as expected).

I haven't decided what I'm going to do with the K2010 yet. I might wait a year or two before I do anything more to let the new reference age. I dunno. I guess it depends on boredom/inspiration.

Might not be a bad idea to screw the case in for now when doing tests, it also seems there is a temperature correlation. But burning it in for a while probably isn't a bad idea either.

[KungFuJosh]

Test 12 with the same settings as Test 11. CSV file also added.

Thanks,
Josh

[TylerPeppy]

Test 12 with the same settings as Test 11. CSV file also added.

Thanks,
Josh

This looks clear as day, can someone else more familiar with this part confirm the jumps for the longer intervals are "popcorn" or telegraph noise?

I notice the previous post from Josh indicated there were temperature jumps in ambient, although my unit is thermally isolated, and I wouldn't expect that fast of a temperature rise internal to the case.

Thanks,
Tyler

[KungFuJosh]

Whether it directly effects the ref or not, the ambient temperature will effect the meters to different degrees.

[KungFuJosh]

Here's what seems like a pretty nice example of a K2750 with the VREF10-001.

[KungFuJosh]

Here's the K2750 with @TylerPeppy's "Almost 10V Ref" - and it's powered with a portable battery backup / power bank thing.

I'll eventually do some more with the filter off.

ETA: That power bank is annoying. It shuts itself off after about 2 hours. 

[TUMEMBER]

You're finally presenting a "grown-up" device. I'm a little jealous. My Solartron 7062 has a simpler keyboard layout, but fewer advanced features. It does have a nicer display, though the K2750 has better stability.

[KungFuJosh]

You're finally presenting a "grown-up" device. I'm a little jealous. My Solartron 7062 has a simpler keyboard layout, but fewer advanced features. It does have a nicer display, though the K2750 has better stability.

Grown-up?? You only like full rack width meters? 

I was impressed with the K2750's stability. I'm probably not keeping it though. It's too grown-up for my free space. We'll see if TEAS wins or not.

[TylerPeppy]

You're finally presenting a "grown-up" device. I'm a little jealous. My Solartron 7062 has a simpler keyboard layout, but fewer advanced features. It does have a nicer display, though the K2750 has better stability.

Grown-up?? You only like full rack width meters? 

I was impressed with the K2750's stability. I'm probably not keeping it though. It's too grown-up for my free space. We'll see if TEAS wins or not.

The 2750 is actually one of those form factors I would put in my lab without a doubt, as the multiplexing features are invalueable when doing long-term product testing. For regular measurements, it also looks like a nice form factor, as long as your gear can stack well or you have the rack space.

The mux relay card on my DMM6500 gets consistent use already, a welcome addition. The thing that the 2750 would also be very good for is large numbers of thermocouples. At work, I use 34970As and Fluke Hydras with lots of thermocouples epoxy bonded to cards. While their form factors are nice, the lack of a multimeter interface on the front and lack of full banana plugs in the 34970a is a bit annoying at times.

Tyler

[KungFuJosh]

Well, if/when I sell the 2750, I will probably include two 7700 cards. I probably won't list it here though. I think I can get over $2k on feeBay, and maybe $15 here. 

[TylerPeppy]

Well, if/when I sell the 2750, I will probably include two 7700 cards. I probably won't list it here though. I think I can get over $2k on feeBay, and maybe $15 here. 

Math works out, 2k on feeBay is roughly 10$ profit, right? 

[KungFuJosh]

Basically. 😭

[TylerPeppy]

Some more data from my 3rd unit (3 of 3). This unit seems to actually have the lowest noise of the bunch. it has been burned-in for about 432hrs so far, and the DMM6500 is exhibiting the lowest noise so far with it it would seem. I am seeing less jumping compared to other tests (that I haven't bothered publishing).

The Std-Dev30 is actually using 180 samples, and I noticed that Josh was using 1799 ~ 1800. It's fine, just that you won't see as much variance in the standard deviations as such. This unit also seems to have the higher Vz as well, much closer to the datasheet specified 7.05V nominal.

I don't have the greatest temperature logger, I need a more precise + accurate one. My lab is 23C at best anyways... I am on the lookout for a better one already. I don't feel like putting extra wear on the front relays with the dual measurement though. Maybe a UEI, Fluke, or Omega?

Edit: The meter has been on 24/7 since I got it a couple weeks back if anyone is curious as to the test setup.

Edit2: 10V Range, AZ ON, Line Sync Enabled, Moving Filter 10, 10PLC

Tyler

[KungFuJosh]

I used 1799 to match your previous test post. I'll switch to 180 on the next one. I'm probably going to mutilate the 2010 a little more, and see if I can fix my laser to make the changes I want before more testing.

Thanks,
Josh

[Salient666]

While you guys are testing the expensive really good stuff, I decided to test my Ali Express AD584JH reference @ 10V.

All meters are at NPLC=10, No filters, AZ=On

[KungFuJosh]

Here's some more data for the uDR1399 ref. The K2010 is closed up now, but the overnight included warmup time. Still not too shabby. Looking even better this morning. Math set to 180 samples now.

Thanks,
Josh

[TylerPeppy]

Here's some more data for the uDR1399 ref. The K2010 is closed up now, but the overnight included warmup time. Still not too shabby. Looking even better this morning. Math set to 180 samples now.

Thanks,
Josh

Much better stability than what I can reach. Might need to get myself a more stable meter at some point. Interesting the 638min test has the "bump" in the middle. It jumps and then settles again. I haven't actually seen that a whole lot in my data, but even with the filtering, my values ares still covered in noise.

Best I have been able to do is use two seperate references and subtract their values to get an idea of the noise, attempting to remove the effects of the measuring instrument. When I did that, I am getting relatively similar results to Josh's data.

Tyler

[TUMEMBER]

The time base in this program for "hours" is broken. 6.74 h?

[HKJ]

The time base in this program for "hours" is broken. 6.74 h?

The time base is seconds, but shown in hours. If you want a nice hour scale add a math column that divides second by 3600
or use hh:mm:ss scale