Keithley DMM7510 7.5-Digit Multimeter Review, Teardown & Experiments

[HighVoltage]

Ok, I did this measurement with 5 NPLC in the video. Did you also see the plot I posted in this thread about the DMM noise as a function of the NPLC?

Yes, I saw that and would put this in to consideration during the analysis.
Would be great, if you could repeat this measurement at 5, 10 and 100 NPLC and store the data as .CSV and may make a visual screen shot of the statistical analysis.

Thanks

[dr.diesel]

^ I could do the same for the K2450/K2010, but IIRC the K2010 has a max of 10PLC.

[HighVoltage]

^ I could do the same for the K2450/K2010, but IIRC the K2010 has a max of 10PLC.

Yes, please this way we get some good data comparison on 7 1/2 digit meters.

[Vgkid]

Thanks for the review, the experiments were the most interesting part.

[KedasProbe]

By the way, from the DMM user manual:

The Model DMM7510 has a nonlinear shape for its speed versus noise characteristics. The Model
DMM7510 is optimized for the 1 PLC to 5 PLC reading rate. At these rates (lowest noise region in
graph), the Model DMM7510 will make corrections for its own internal drift and will still be fast enough
to settle a step response of less than 100 ms.

I don't really understand what this graph means.
Does this mean 100PLC is worse than 1PLC then what is the point of having 100PLC?
I assume not, but what does it mean, that a change from 1PLC to 10PLC will be a lower improvement than what you mathematically expect?

Are 10 x 1PLC better than 1 x 10PLC?

[Dr. Frank]

Dear Shahriar,

at that point, I just wanted  to emphasize, that I really enjoy your other videos about scopes and other HF instruments.. that's always a great pleasure!

In this case anyhow, it would really be very interesting, if you please would measure the noise performance for different NPLCs, from 0.01 to 100, or so, as proposed by xdev already.

I can really not understand, why the noise should go up for NPLC >5, and why Keithley did not specify that range .. there's something fishy. Compare that to every other DMM, especially the 3458A, which noise is on the same order of magnitude , but I suppose to be still much better (decreasing numbers) towards NPLC 100.

The old Keithley 8 1/2 DMM is nearly a magnitude worse in terms of stability and noise , I still hope that Keithley improved in that category.

regards Frank

[dr.diesel]

^ I could do the same for the K2450/K2010, but IIRC the K2010 has a max of 10PLC.

Yes, please this way we get some good data comparison on 7 1/2 digit meters.

I didn't see much use in letting it run for very long!  If this was not your intent, please let me know.

(forum doesn't allow for an .ods file, converted to .xls, looks good on my end)

[nfmax]

By the way, from the DMM user manual:

The Model DMM7510 has a nonlinear shape for its speed versus noise characteristics. The Model
DMM7510 is optimized for the 1 PLC to 5 PLC reading rate. At these rates (lowest noise region in
graph), the Model DMM7510 will make corrections for its own internal drift and will still be fast enough
to settle a step response of less than 100 ms.

That plot looks a bit like an Allan variance plot, showing the transition from white-noise behaviour at short averaging times, to drift at longer times. I wonder how they measured it? If that is the case, it looks like the Keithley ADC cannot (fully) compensate for drift that occurs during the integration period. Is there any chance of collecting data for an Allan variance analysis covering the entire range of integration times, ideally comparing results with the Keysight/HP's?

[HighVoltage]

^ I could do the same for the K2450/K2010, but IIRC the K2010 has a max of 10PLC.

Yes, please this way we get some good data comparison on 7 1/2 digit meters.

I didn't see much use in letting it run for very long!  If this was not your intent, please let me know.

(forum doesn't allow for an .ods file, converted to .xls, looks good on my end)

Thank you for the data, but something is badly wrong
May be a wrong connection or the output not enabled and this is just noise?
Hmmm...

Max value in your data is +3.03 E-9
Min value  in your data is -6.93 E-9

It should be very stable around 1.00 E-9

[dr.diesel]

The K2010 lowest current range is 10mA.  Same connection, if I set 1uA it's pretty solid, varies a few counts. 

At 1nA it's just noise.

[HighVoltage]

The K2010 lowest current range is 10mA.  Same connection, if I set 1uA it's pretty solid, varies a few counts. 
At 1nA it's just noise.

Ahh, I see. Sorry I did not study the Keithley 2010 datasheet.
Then it is not possible to make this measurement.
Thanks anyways.

[TiN]

I tried at my site, 2400+2002, with rather long cable BNC-banana, bounce around 150-200pA.

Update:

2400+2002, StdDev = 50.31pA, Avg = 951pA, Min = 818pA, max 1119pA, 500 samples, 50 NPLC
2400+2001, StdDev = 22.11pA, Avg = 1025pA, Min = 969pA, max 1076pA, 500 samples, 10 NPLC

Measurement steps:
2400 set to source zero current, Vcmpl = 12V. Multimeter set to REL to null offset current. SMU set to 0.001uA, measure 500 samples on DMM.

[dr.diesel]

The K2010 lowest current range is 10mA.  Same connection, if I set 1uA it's pretty solid, varies a few counts. 
At 1nA it's just noise.

Ahh, I see. Sorry I did not study the Keithley 2010 datasheet.
Then it is not possible to make this measurement.
Thanks anyways.

Yeah, i'd forgotten myself, or I would have knew better to even try! 

[HighVoltage]

Here are two pictures of a very short term measurement of the 1 nA
Current source: Keithley 2450
DMM: Keysight 34470A

I am traveling right now but when I am back I will post some more long term measurements of the 1 nA source.
The Keysight 34470A seems to be extremely stable and repeatable in this low current range.
 

[Hugoneus]

I spoke with Keithley engineers. It is in fact true that the optimum operating point is 5 NPLC. This provides a balance between 1/f noise and white noise of the analog front-end.

There was nothing wrong with my measurements. It was at the optimal point. I will make capture some data and post it here.

[KedasProbe]

I spoke with Keithley engineers. It is in fact true that the optimum operating point is 5 NPLC. This provides a balance between 1/f noise and white noise of the analog front-end.

There was nothing wrong with my measurements. It was at the optimal point. I will make capture some data and post it here.

Maybe you can provide enough data of your unit to be added in this set of data:
https://www.eevblog.com/forum/testgear/dmm-adc-noise-comparison-testing-project/

[VintageNut]

By the way, from the DMM user manual:

The Model DMM7510 has a nonlinear shape for its speed versus noise characteristics. The Model
DMM7510 is optimized for the 1 PLC to 5 PLC reading rate. At these rates (lowest noise region in
graph), the Model DMM7510 will make corrections for its own internal drift and will still be fast enough
to settle a step response of less than 100 ms.

I don't really understand what this graph means.
Does this mean 100PLC is worse than 1PLC then what is the point of having 100PLC?
I assume not, but what does it mean, that a change from 1PLC to 10PLC will be a lower improvement than what you mathematically expect?

Are 10 x 1PLC better than 1 x 10PLC?

The graph shows the optimum NPLC setting for lowest noise. If you look at the 2182A manual, the same shape graph appears. Chances are the same front end topology is in both instruments.

This NPLC setting is not the panacea for every measurement. It is just information that may be of benefit if you require the lowest noise possible.

I borrowed a 7510 for short while. I installed a factory 4-wire short and let it percolate on DCV auto. The average reading was -3.5 uV and the standard deviation was 200nV. This was many thousands of readings at 10 NPLC.

[ElectroIrradiator]

Have only had time to watch the first half of the video, will complete it ASAP.

Noticed a regular pattern of 'dots' / impressions(?) on the first few shots of the LCD. Wondering what this could be and whether it is noticeable to the naked eye. See for instance clip starting at 09:52. I don't recall seeing anything like that in the reviews of the Keithley 2450 and 2460 SMUs.

If you like, Shahriar, then I would like to donate[1] a long rod of solid 6mm^ø Teflon/PTFE to the cause. You can sharpen it and use it as an alternative pointing tool (at least until people start worrying about buildup of static electricity) instead of that Tungsten tipped marker.

[1] Serious offer, I just need to dig out the bundle I have. Somewhere...

[TiN]

Array of dots on screen are touchscreen spacer dots. They are on all touchscreens, just different size and visible on specific angles.

[splin]

I borrowed a 7510 for short while. I installed a factory 4-wire short and let it percolate on DCV auto. The average reading was -3.5 uV and the standard deviation was 200nV. This was many thousands of readings at 10 NPLC.

200nVrms looks to be way out of spec which is .5ppm Vrms (typical) noise for 1 and 5 NPLC, 100mV range = 50nV rms. I doubt that 10 NPLC would be much different to 5 NPLC.

Its interesting to compare the noise specs with the 34470A which are guaranteed RMS values as far as I can see, so I've taken the 7510 'additional peak noise uncertainty' values and divided by 3.3 to get RMS (apart from the 1 NPLC which is the typical RMS figure).

Assuming I got my numbers right, the Keysight is generally between 10 and 30x noisier than the Keithley. The 7510 datasheet shows that it has a noise 'problem' in the 100V range (in comparison to the other ranges) as pointed out in the review. I'm surprised they didn't resolve it before release, though it still beats the 35570A - at 1 NPLC and below at least. It would be interesting to see actual noise measurements at slower but equal update rates - eg. 1 and 5 NPLC plus averaging for the Keithley versus longer apertures for the Keysight.

All values in Vrms, ppm of range

          DMM7510      34470A            Ratio
100mV
1                .5               6                  12
.2             3.3              35                 10.6
.06           5.2              50                   9.6   
.006        12.7             90                   7.1

1V
1                .09            1.5                16.7   
.2               .48            8.0                16.7
.06             .64          23.0                35.9
.006         5.45          54.0                  9.9

10V
1                .07            1.05              15
.2               .33            5.3                16.1
.06             .64          20.3                31.7
.006         3.33          50.4                15.1

100V
1                 .4             1.005              2.5
.2              2.85           5.03                1.76
.06            5.15         20.03                3.9
.006        30.3           50.04              16.5

1000V
1                 .07           1.0005          14.3
.2                .33           5.003            15.2
.06              .73         20.003            27.4
.006          5.45         50.004              9.2

[ElectroIrradiator]

Array of dots on screen are touchscreen spacer dots. They are on all touchscreens, just different size and visible on specific angles.

Interesting, thanks. I wonder why I have never noticed those before.

[KedasProbe]

This NPLC setting is not the panacea for every measurement. It is just information that may be of benefit if you require the lowest noise possible.

The question is more is it better to use the average of 100 x 1PLC values than use 1 x 100PLC value.
It seems the answer is yes, makes the 100PLC setting more or less pointless.

[Hugoneus]

A few points:

The dots on the screen are not normally visible. They are part of the touch-screen but due to my lighting and camera angle you could see them. I never notice them.

I have not made the NPLC measurements yet, but I will. However, according to the people who designed this instrument my measurements were optimal and earlier criticisms were incorrect.

[blackdog]

Hi,

I have the same problems as Dr. Frank about reading the display of the DMM7510...
I do not care, that the best PLC is about 5...

These jumping around digits have no function anymore, and witch human can read it?
Ofcourse, is see that its usable for industrial test setups, but on my bench no way!

I hope that there is average possible, otherwise this instrument is useless in my eyes...

Keep up the nice video's, thanks for that :-)


Kind regarts,
Blackdog

[splin]

I have not made the NPLC measurements yet, but I will. However, according to the people who designed this instrument my measurements were optimal and earlier criticisms were incorrect.

I have to agree that earlier criticisms were correct, I felt exactly the same when watching it. Most measurements shown were only stable in the first 4 1/2 digits and some, including the 1G resistor were only stable to 2 1/2 digits - hardly a good demonstration of the resolution and stability of a 7 1/2 digit meter. It was almost impossible to follow the changing display on the 1 NPLC measurments. Ideally the averaging function would have been used to increase the resolution and stability of the measurements.

No doubt most of the noise was coming from the sources as the Li-poly battery measurement was stable to 7 1/2 digits. That doesn't explain the very noisy 1G ohm resistor measurement and neither does the datasheet. Looking for the noise specifications on the 10G range, surprisingly it turns out there aren't any specs for the ranges above 10K (that I could find at least). The question then is that an error or is it deliberate? Its not in the reference manual either.

Having said that there was plenty of interesting stuff in the review so thanks for doing it. Incidentally I expect the dropping battery voltage shown was very unlikely to be due to self discharge and much more probably drift due to small temperature changes - battery voltage is very temperature dependant and the changes were only of the order of a few ppm.