New Keithley DMM6500 and now DAQ6510

[trilerian]

Here is a pic reading from both meters at the same time. Sorry, the meters don’t live on the same side of my bench, so there is more in the pic than needed.
Siglent is on the left, Keithley on the right.

[KungFuJosh]

Here is a pic reading from both meters at the same time. Sorry, the meters don’t live on the same side of my bench, so there is more in the pic than needed.
Siglent is on the left, Keithley on the right.

What does your parallel measurement look like with REL off (on both meters)? What does it look like with Filter off?

My meters are busy right now, but I could do a quick uV AC test later on with my DAQ6510.

Thanks,
Josh

[trilerian]

Here is a pic reading from both meters at the same time. Sorry, the meters don’t live on the same side of my bench, so there is more in the pic than needed.
Siglent is on the left, Keithley on the right.

What does your parallel measurement look like with REL off (on both meters)? What does it look like with Filter off?

My meters are busy right now, but I could do a quick uV AC test later on with my DAQ6510.

Thanks,
Josh

No filters, no REL.
3055 - 310uV
6500 - a noisy 530uV

Thanks, I look forward to your test later

[KungFuJosh]

[No filters, no REL.
3055 - 310uV
6500 - a noisy 530uV

Thanks, I look forward to your test later

Eh, I kinda sorta closed TestController by mistake after over an hour of a test running. Oops. 🤦 Anyway, I ran the test with my AWG. It doesn't go below 0.4mV though. That's close enough, right?

There weren't any noise issues for me. The DAQ and the DMM have the same analog board, so you should see similar performance assuming no issues with the meter, and no outside influence causing the noise.. My guess is you might have something interfering with the DMM on that side of your bench. Or the test leads, or out of cal, etc.

I have ferrite cores on both ends of all the power plugs on all my bench gear, and EMI/EFI filtering on the surge protectors. I assume it helps a little, but I dunno if it's relevant to your noise issue or not. 🤷

Have you tried swapping leads between meters? Have you tried twisting the leads to reduce the noise?

Thanks,
Josh

[trilerian]

Maybe my AC cal is off.  My DMMCheck PLUS sheet says it should output 4.999V, my 6500 measures it at: a noisy 4.981.  With a pretty neat uncertainty app from this thread https://www.eevblog.com/forum/metrology/measurement-uncertainty-app-for-keithly-6500-dmm/msg5749551/#msg5749551, I get a high of 4.98733V and a low of 4.975353. 

I never really had the need to measure AC before this.

Is this something Keithley should take care of?  Meter is less than a year old, so still within warranty. 

Anyway, my 3055 measure the check plus at 4.9983.

[trilerian]

[No filters, no REL.
3055 - 310uV
6500 - a noisy 530uV

Thanks, I look forward to your test later

Eh, I kinda sorta closed TestController by mistake after over an hour of a test running. Oops. 🤦 Anyway, I ran the test with my AWG. It doesn't go below 0.4mV though. That's close enough, right?

There weren't any noise issues for me. The DAQ and the DMM have the same analog board, so you should see similar performance assuming no issues with the meter, and no outside influence causing the noise.. My guess is you might have something interfering with the DMM on that side of your bench. Or the test leads, or out of cal, etc.

I have ferrite cores on both ends of all the power plugs on all my bench gear, and EMI/EFI filtering on the surge protectors. I assume it helps a little, but I dunno if it's relevant to your noise issue or not. 🤷

Have you tried swapping leads between meters? Have you tried twisting the leads to reduce the noise?

Thanks,
Josh

I swapped leads and twisted them, no difference.  I'll try switching the power strips they are plugged into.  As far as that side of the bench, I figure it would have less noise, the left side has a switching supply plugged in and on right under the siglent. 

[tautech]

Anyway, my 3055 measure the check plus at 4.9983.

Do you know you can do a user adjustment if you have suitably accurate measurement standards ?

Defpom shows how:

[KungFuJosh]

Maybe my AC cal is off.  My DMMCheck PLUS sheet says it should output 4.999V, my 6500 measures it at: a noisy 4.981.  With a pretty neat uncertainty app from this thread https://www.eevblog.com/forum/metrology/measurement-uncertainty-app-for-keithly-6500-dmm/msg5749551/#msg5749551, I get a high of 4.98733V and a low of 4.975353. 

I never really had the need to measure AC before this.

Is this something Keithley should take care of?  Meter is less than a year old, so still within warranty. 

Anyway, my 3055 measure the check plus at 4.9983.

What temperature does the cal for the DMMCheck+ state? Does your lab temp match close enough?

I think the Keithley warranty might be 3 years. You could certainly tell them what's going on and ask them what they think you should do. IME, their support is excellent.

It doesn't seem like the reading is out of spec. The question is what's causing the instability- assuming it's not also within a normal range. 😉

AC interference can be a huge pain in the ass.

[Kleinstein]

The AC part of the DMM6500 does not look very sophisticated or low noise: no visible capacitors in parallel to the 1 M resistor string and no trimmers at the active divider. The Bandwidth with the related digitizer mode is also not that well defined and gain dependent. After all in automated test systems AC is more of a rare case and low importance. With digital RMS it should still work reasonably well at low voltages - if the softeware handles things like the background noise well. At least digital RMS has a chance to subtract the background noise correctly - with analog RMS chips this does not work well.

It may be worth looking at the AC performance, also at somewhat larger voltages, close to the 5% of FS where the accuracy specs start to apply.
The REL function with AC is tricky - wide band noise and mains hum and a correlated signal add up differently.

[trilerian]

Maybe my AC cal is off.  My DMMCheck PLUS sheet says it should output 4.999V, my 6500 measures it at: a noisy 4.981.  With a pretty neat uncertainty app from this thread https://www.eevblog.com/forum/metrology/measurement-uncertainty-app-for-keithly-6500-dmm/msg5749551/#msg5749551, I get a high of 4.98733V and a low of 4.975353. 

I never really had the need to measure AC before this.

Is this something Keithley should take care of?  Meter is less than a year old, so still within warranty. 

Anyway, my 3055 measure the check plus at 4.9983.

What temperature does the cal for the DMMCheck+ state? Does your lab temp match close enough?

I think the Keithley warranty might be 3 years. You could certainly tell them what's going on and ask them what they think you should do. IME, their support is excellent.

It doesn't seem like the reading is out of spec. The question is what's causing the instability- assuming it's not also within a normal range. 😉

AC interference can be a huge pain in the ass.

I created a ticket with Tek, we'll see what happens. 

[KungFuJosh]

I did a simple test for 30 minutes at 1V and 500mV to give you something to compare to. The DAQ (as usual) was the most stable of my meters.

I'd suggest trying something similar going directly from your AWG to rule other things out.

You might need to be a little patient hearing back from Keithley support. They were not always super fast to respond, but they were very helpful.

Thanks,
Josh

[trilerian]

I'll run these later tonight. Thanks

[trilerian]

I did a simple test for 30 minutes at 1V and 500mV to give you something to compare to. The DAQ (as usual) was the most stable of my meters.

I'd suggest trying something similar going directly from your AWG to rule other things out.

You might need to be a little patient hearing back from Keithley support. They were not always super fast to respond, but they were very helpful.

Thanks,
Josh

It took me a bit to script out what I wanted to do with the readings.  Still forgot to make a max - min, but that is ok.  I'll look into the app you used at some point, seems pretty nice.

Anyway, using the function generator I think I am just left with how unstable the output is, lol.  Both graphs of voltages using the function generator show pretty close to the same thing, all over the place.  So I also did a run using my DMMCheck Plus, which is actually a 10kHz square wave with RMS of 5V, according to my scope.  Interesting, the DMM6500 was much closer to the spec sheet on the check plus today, measuring 4.999113 instead of 4.9892, I got nothing on that...  But from the 5V reference you can see the DMM6500 doesn't fluctuate as much as the SDM 3055.  They both had the same amount of digits output though, I was surprised at that. 

Anyway, the issue only seems to present itself in the lower range, basically under 3mV AC.   

As another test, I did try a different function generator, FY 6900 instead of my SDG1022X Plus.  The readings got closer for the DMM6500, but not quite there. Siglent still outputs around 250uV, where the DMM6500 got to 170uV.  Well, here are the graphs, sorry they don't look as good as that polished app.  I did mess with my Instek milliohm meter for awhile trying to get the temp output added, but it wasn't cooperating with me.  I could open a serial connection to it, but I couldn't get it to return anything...  The temp however maintained a steady 22.2 C during all the testing. 

[2N3055]

I did a simple test for 30 minutes at 1V and 500mV to give you something to compare to. The DAQ (as usual) was the most stable of my meters.

I'd suggest trying something similar going directly from your AWG to rule other things out.

You might need to be a little patient hearing back from Keithley support. They were not always super fast to respond, but they were very helpful.

Thanks,
Josh

It took me a bit to script out what I wanted to do with the readings.  Still forgot to make a max - min, but that is ok.  I'll look into the app you used at some point, seems pretty nice.

Anyway, using the function generator I think I am just left with how unstable the output is, lol.  Both graphs of voltages using the function generator show pretty close to the same thing, all over the place.  So I also did a run using my DMMCheck Plus, which is actually a 10kHz square wave with RMS of 5V, according to my scope.  Interesting, the DMM6500 was much closer to the spec sheet on the check plus today, measuring 4.999113 instead of 4.9892, I got nothing on that...  But from the 5V reference you can see the DMM6500 doesn't fluctuate as much as the SDM 3055.  They both had the same amount of digits output though, I was surprised at that. 

Anyway, the issue only seems to present itself in the lower range, basically under 3mV AC.   

As another test, I did try a different function generator, FY 6900 instead of my SDG1022X Plus.  The readings got closer for the DMM6500, but not quite there. Siglent still outputs around 250uV, where the DMM6500 got to 170uV.  Well, here are the graphs, sorry they don't look as good as that polished app.  I did mess with my Instek milliohm meter for awhile trying to get the temp output added, but it wasn't cooperating with me.  I could open a serial connection to it, but I couldn't get it to return anything...  The temp however maintained a steady 22.2 C during all the testing.

I am confused that nobody pointed out fact that both meters you are using are not going to correctly measure very low AC voltages. Maybe I missed it.

Both of these meters have a specifications that AC measurements accuracy specifications are for AC signals that are more than 5% of full scale.   That is 5mV for DMM6500 on 100mV range, and 10mV for SDM3055 on 200mV range.

Anything smaller than that is going to introduce errors. Measuring  hundreds of µV is both very much into what these instruments are going to have large errors. DMM6500 uses sampling and calculates RMS, SDM3055 uses analog RMS chip.
That means DMM6500 should be more linear with very low voltages, but manufacturer itself does not guarantee it fully belov 5% of FS. Datasheet: "Specifications are for sine wave inputs > 5% of range"...

You  will also have problems having a clean signal in a first place. You are going to have large amount of noise injected into circuit also. You need shielding, grounding etc... Also this is floating measurement, where you rely on instrument CMRR...

You need to make shunt larger, so voltages are at least 10s of millivolts, FS ideally. Also an preamplifier could be used.

As for DMM6500 being noisy:

https://www.eevblog.com/forum/testgear/dmm6500-dmm7510-vs-34465a-34470a-ac-readings/msg3931616/#msg3931616

[Kleinstein]

The accuracy specs for the SDM3055 have a note on values between 1 and 5%. This range still works, but with slightly higher uncertainty.
The analog RMS chips are know to have issue with very small amplitudes well below 1%, but the performance can vary between units.

Normally digital RMS like used in the DMM6500 should also work for less than 5% - at least with correct software. There are just no accuracy specs for this range. I would still expect reasonable performance with lower amplitude with not much variation between units (could still be a wrong constant for the background noise).

The tests in question would be looking at amplitudes from some 0.1 to 5% in the 10 V range. One could use a simple divider after the generator to get smaller voltages. This is not so much about ultimate accuracy, but more about the question how far down it could work with some 1% or 10% accuracy.

[KungFuJosh]

I am confused that nobody pointed out fact that both meters you are using are not going to correctly measure very low AC voltages. Maybe I missed it.

Both of these meters have a specifications that AC measurements accuracy specifications are for AC signals that are more than 5% of full scale.   That is 5mV for DMM6500 on 100mV range, and 10mV for SDM3055 on 200mV range.

I did. 😉 In reply 1557 I pointed out that his DMM results looked within spec, and attached the AC spec sheet for his meter in question.

They both had the same amount of digits output though, I was surprised at that. 

You need to change your settings in the DMM6500 to get the full 6.5 digits. Either under Measure Settings or Quickset. Sometimes the meter is defaulted for higher speed which is usually lower accuracy/digits. You might also want to adjust filter settings to improve stability.

Thanks,
Josh

[trilerian]

I am confused that nobody pointed out fact that both meters you are using are not going to correctly measure very low AC voltages. Maybe I missed it.

Both of these meters have a specifications that AC measurements accuracy specifications are for AC signals that are more than 5% of full scale.   That is 5mV for DMM6500 on 100mV range, and 10mV for SDM3055 on 200mV range.

I did. 😉 In reply 1557 I pointed out that his DMM results looked within spec, and attached the AC spec sheet for his meter in question.

They both had the same amount of digits output though, I was surprised at that. 

You need to change your settings in the DMM6500 to get the full 6.5 digits. Either under Measure Settings or Quickset. Sometimes the meter is defaulted for higher speed which is usually lower accuracy/digits. You might also want to adjust filter settings to improve stability.

Thanks,
Josh

But it is this footnote down the page that seems to address the issue:

19 Specifications are for sine wave inputs > 5% of range.

I looked at the 7510 data sheet and don't see the same restriction.  Any other meters to check?  I am curious which meter I can use for this test, or if I should just go with the reading of the SDM 3055.  It seems the most stable and predictable, and gives different readings for different P2P from the function generator. 

And on a side note, I also tried to use my EEVblog 235 on the mV range to measure and couldn't get a measurement. 

Also maybe as suggested above, amplify the signal from my AWG, larger voltage going in means a larger voltage to measure.  But more voltage going in means more current as well.  Not really sure how that would all this would work.  I haven't researched much into AC circuits as of yet, still learning DC circuits.  I am actually going to take my first class on circuits next semester, getting pretty excited about that.  So far everything I know has been read online. 

[Kleinstein]

One could amplify the AC voltage coming from the DUT. So not more current but AC gain.

A handheld meter may not work that well with a relatively high frequency. Some of the handheld only have a maximum frequency of 1 or 2 kHz.
Some meters also have no AC coupling for the AC mV range - it could work with an added coupling capacitors.
The resolution is usually also not that great, rarely better than 100 µV.

Ideally one would measure with some kind of lock-in amplifier and thus phase sensitive to the driving current.
A DSO in a kind of Boxcar averaging mode can get a similar effect - average over many periods and only than look at the result.
However not all DSOs offer this and there can be limitations.
The DMM6500 with extra AC coupling capacitor in digitizer mode and with a trigger signal from the generator could also be used. The amplitude would than be calculated at the PC side, or maybe as an app in the DMM.
In principle a sound card of some kind with a little protection circuit could be used.

I would first test the limits for the low voltage AC performance at the DMM6500. There may still be a wrong setting, maybe a wrong zero / rel value.
250 µV is well below the range range where accuracy is specified, but in principle the hardware should still work there.
The meters own noise with some 200 kHz BW and maybe 100 nV/sqrt(Hz) as a rather high value would be some 50 µV of noise background - ideally the meter could subtract this part correctly and could read even to below this level, though a bit more noisy. An issue would be excess hum to cause a significant background.

[trilerian]

One could amplify the AC voltage coming from the DUT. So not more current but AC gain.

A handheld meter may not work that well with a relatively high frequency. Some of the handheld only have a maximum frequency of 1 or 2 kHz.
Some meters also have no AC coupling for the AC mV range - it could work with an added coupling capacitors.
The resolution is usually also not that great, rarely better than 100 µV.

Ideally one would measure with some kind of lock-in amplifier and thus phase sensitive to the driving current.
A DSO in a kind of Boxcar averaging mode can get a similar effect - average over many periods and only than look at the result.
However not all DSOs offer this and there can be limitations.
The DMM6500 with extra AC coupling capacitor in digitizer mode and with a trigger signal from the generator could also be used. The amplitude would than be calculated at the PC side, or maybe as an app in the DMM.
In principle a sound card of some kind with a little protection circuit could be used.

I would first test the limits for the low voltage AC performance at the DMM6500. There may still be a wrong setting, maybe a wrong zero / rel value.
250 µV is well below the range range where accuracy is specified, but in principle the hardware should still work there.
The meters own noise with some 200 kHz BW and maybe 100 nV/sqrt(Hz) as a rather high value would be some 50 µV of noise background - ideally the meter could subtract this part correctly and could read even to below this level, though a bit more noisy. An issue would be excess hum to cause a significant background.

Low limit with my AWG is 2mV p2p.  So 0.7mV, my 6500 reads that as 0.8xxmV.  Honestly kind of close, and who knows if it is the AWG that is off or the 6500.  My 3055 reads it as 1.7mV.  I will have to make some dividers to bring the voltage down lower and test.

This whole thing is now about learning, lol.  The actual use of the measurement lost its value a long time ago...  Now I'm just being stubborn.

[KungFuJosh]

What Kleinstein said.

But also, your measurements were all within +/- 1% (IIRC) using your ref.

The test value being so far off was related to the use of REL, and probably your settings and possibly outside interference. Take a look at the measurement settings on your 6500. Make sure it's set to 6.5 digits, and go for slower speed/more accuracy. My settings are attached.

I also attached a test at 5mVAC comparing 3 meters. I assume your 6500 should be capable of this as well. I used two 50Ω attenuators inline to get in that voltage range from my AWG.

Thanks,
Josh

ETA: There is a fast smoothing filter active on the Keysight. No filter active on the Keithley, just the settings shown.

ETA2: There was no warmup time for this test.

[trilerian]

What Kleinstein said.

But also, your measurements were all within +/- 1% (IIRC) using your ref.

The test value being so far off was related to the use of REL, and probably your settings and possibly outside interference. Take a look at the measurement settings on your 6500. Make sure it's set to 6.5 digits, and go for slower speed/more accuracy. My settings are attached.

I also attached a test at 5mVAC comparing 3 meters. I assume your 6500 should be capable of this as well. I used two 50Ω attenuators inline to get in that voltage range from my AWG.

Thanks,
Josh

ETA: There is a fast smoothing filter active on the Keysight. No filter active on the Keithley, just the settings shown.

ETA2: There was no warmup time for this test.

My 6500 is setup for 6.5 digits, there were 6.5 digits in my graph, but that is where I capped it in the script, I can add filters but for the script I was taking a reading every second, filters would slow down the readings and mess the script time up.  Also 5mV reading is fine, but if you would like to test in the 100uV range that would be cool.  I'll do tests for that later this evening, working my real job now and got some classwork to take care of after that. 
Thanks!

[KungFuJosh]

My 6500 is setup for 6.5 digits, there were 6.5 digits in my graph, but that is where I capped it in the script, I can add filters but for the script I was taking a reading every second, filters would slow down the readings and mess the script time up.  Also 5mV reading is fine, but if you would like to test in the 100uV range that would be cool.  I'll do tests for that later this evening, working my real job now and got some classwork to take care of after that. 
Thanks!

Right, I forgot you were pushing into the microvolt range. Scrolling up takes a lot of effort sometimes. 😉

I'm running a test now with measuring freq at 0.01 seconds. I changing the sampling to 1 on the DAQ to keep it fast. The lowest I could go with both terminators in place is around 200uV range, but my Keysight is reading lower for some reason (filter off now anyway).

If you need more accuracy / stability in the microvolt range, you either need to go up in digits or use averaging filters. If you really need the speed, then you need a 7.5 or 8.5 digit meter.

You should try out TestController. There's a little bit of a learning curve to it, but it's free and it works great.

Thanks,
Josh

[KungFuJosh]

Here's the results of that test. I'm a little surprised; the Siglent was clearly the most stable, though I dunno which one is most accurate. 🤷🤣

If somebody wants to do the math: SDG2122X output set at 0.4mVrms, 50Ω output, to two inline 50Ω attenuators. The meters previously agreed (when using a single attenuator) that the output from the AWG was somewhere around 350uV when set at 0.4mVrms.

I'm going to set something up to drop the voltage lower.

Thanks,
Josh

[KungFuJosh]

I used a 50W guitar amplifier attenuator to get the voltage lower. I set it so the DAQ was at 100uV, which knocked the Keysight out of the game. The Siglent is higher by ~50%. I dunno which is accurate, but I would assume the DAQ for no good reason. 🤷

Test setup has the AWG going to a 50Ω attenuator into the speaker attenuator then out to the DMMs. BNC to 1/4" and BNC to banana plug adapters used.

Thanks,
Josh

[Kleinstein]

From the meter I have most trust in the KS34465. This meter also has AC specs down to 1 mV or 0.3% of FS (1 V and 10 V range). It is also know for good frequency response.
The keithley meter is a bit questionable with the frequency response in the digitizer mode. This does not give extra confidence for AC performance.
Normally one should be able to measure the attenuator at a higher voltage and than assume a fixed attenuation. The nominal attenuation is valid with proper termination at the output. So AWG -> attenuator -> terminator -> DMM.
The interesting curve is not so much AC reading versus time, but more AC reading versus AWG seeting or reading with a different meter from before the attenuator. So testing the linearity in AC mode.