DMM Noise comparison testing project

[plesa]

I also uploaded dataset from K2002, but it is not complete and for second time I received error.
I used old version of scripts, so in few hours will be new run.
Does anybody ported scripts to 34410A, 34411A or 34401A?

[Macbeth]

I ran all the EZGPIB tests overnight on my Keithley 2015 using the K2000 script. I got a few error -213s reported on the meter on some NPLC tests but can't say which as I was asleep through most of it. It didn't error in the script and appeared to log ok. FTP'd to xdevs.

I am running ancient firmware B12 and hope to upgrade to B17 this weekend. Will do another run after the upgrade, maybe the error -213's will be fixed?

I've posted my old ROMs and cal stuff as I know TiN likes to archive them on his excellent xdevs site, and the 2015 stuff is a bit sparse.

[plesa]

The error is due to string conversion float to integer ( NPLC <1 is only affected).

[Macbeth]

The error is due to string conversion float to integer ( NPLC <1 is only affected).

Hmm... I could swear when I spotted it it was on NPLC 5 and 200V range. Also the error was only displayed on the meters VFD, not the app. It was consistently coming up on some NPLC's and not others.

Ok, I just ran the python "parse_noise_sd.py" and it failed parsing the 0.2V range on NPLC 5 ... (I was in bed when it was doing that range  )

Code: [Select]

c:\MyMeasurements\Keithley2015>parse_noise_sd.py
| Multimeter    | Range         | NPLC  |                       Counts  |    Sq.sum     |  Average   |    SD
|    SD,uV      |
| Keithley2015          | .2            | .01   |       5751    | 0.07879803    | 1.370162E-05  | 7.952572E-05  |  79.526       |
| Keithley2015          | .2            | .1    |       5771    | -0.02028955   | -3.515777E-06 | 7.034879E-05  |  70.349       |
| Keithley2015          | .2            | 1     |       5771    | -0.01809268   | -3.135102E-06 | 5.318539E-07  |   0.532       |
| Keithley2015          | .2            | 10    |       1977    | -0.00604425   | -3.057284E-06 | 2.250274E-07  |   0.225       |
| Keithley2015          | .2            | 2     |       5691    | -0.01771140   | -3.112177E-06 | 3.840305E-07  |   0.384       |
| Keithley2015          | .2            | 3     |       5651    | -0.01761382   | -3.116938E-06 | 3.198932E-07  |   0.320       |
Traceback (most recent call last):
  File "C:\MyMeasurements\Keithley2015\parse_noise_sd.py", line 28, in <module>
    sum += float(row[setname[0]])
ValueError: could not convert string to float:

c:\MyMeasurements\Keithley2015>

Looking at the csv files there is definitely data missing on NPLC 5. I will go and rename them all and try again...

[Macbeth]

Ok, removed all the NPLC5's

Code: [Select]

| Multimeter    | Range         | NPLC  |                       Counts  |    Sq.sum     |  Average   |    SD            |    SD,uV      |
| Keithley2015          | .2            | .01   |       5751    | 0.07879803    | 1.370162E-05  | 7.952572E-05  |  79.526       |
| Keithley2015          | .2            | .1    |       5771    | -0.02028955   | -3.515777E-06 | 7.034879E-05  |  70.349       |
| Keithley2015          | .2            | 1     |       5771    | -0.01809268   | -3.135102E-06 | 5.318539E-07  |   0.532       |
| Keithley2015          | .2            | 10    |       1977    | -0.00604425   | -3.057284E-06 | 2.250274E-07  |   0.225       |
| Keithley2015          | .2            | 2     |       5691    | -0.01771140   | -3.112177E-06 | 3.840305E-07  |   0.384       |
| Keithley2015          | .2            | 3     |       5651    | -0.01761382   | -3.116938E-06 | 3.198932E-07  |   0.320       |
| Keithley2015          | 1000          | .01   |       5771    | 94.36460087   | 1.635152E-02  | 8.457514E-03  |8457.514       |
| Keithley2015          | 1000          | .1    |       5761    | 0.92711376    | 1.609293E-04  | 3.231371E-03  |3231.371       |
| Keithley2015          | 1000          | 1     |       5791    | -1.60507862   | -2.771678E-04 | 1.998223E-04  | 199.822       |
| Keithley2015          | 1000          | 10    |       1979    | -0.56855857   | -2.872959E-04 | 1.093811E-04  | 109.381       |
| Keithley2015          | 1000          | 2     |       5781    | -1.44770361   | -2.504244E-04 | 1.380668E-04  | 138.067       |
| Keithley2015          | 1000          | 3     |       5621    | -1.44242376   | -2.566134E-04 | 1.150069E-04  | 115.007       |
| Keithley2015          | 200           | .01   |       5771    | 77.94191925   | 1.350579E-02  | 8.357463E-03  |8357.463       |
| Keithley2015          | 200           | .1    |       5741    | 1.32347618    | 2.305306E-04  | 3.198915E-03  |3198.915       |
| Keithley2015          | 200           | 1     |       5761    | -1.33002973   | -2.308679E-04 | 2.014754E-04  | 201.475       |
| Keithley2015          | 200           | 10    |       1983    | -0.50748620   | -2.559184E-04 | 1.025495E-04  | 102.549       |
| Keithley2015          | 200           | 2     |       5771    | -1.23013294   | -2.131577E-04 | 1.483873E-04  | 148.387       |
| Keithley2015          | 200           | 3     |       5271    | -1.16701184   | -2.214024E-04 | 1.216449E-04  | 121.645       |
| Keithley2015          | 20            | .01   |       5781    | 9.22793018    | 1.596252E-03  | 2.811790E-03  |2811.790       |
| Keithley2015          | 20            | .1    |       5781    | -0.93055499   | -1.609678E-04 | 2.621365E-03  |2621.365       |
| Keithley2015          | 20            | 1     |       5771    | -1.65190293   | -2.862421E-04 | 6.020111E-05  |  60.201       |
| Keithley2015          | 20            | 10    |       1981    | -0.46382246   | -2.341355E-04 | 2.073555E-05  |  20.736       |
| Keithley2015          | 20            | 2     |       5731    | -1.52624034   | -2.663131E-04 | 4.305638E-05  |  43.056       |
| Keithley2015          | 20            | 3     |       5091    | -1.28699979   | -2.527990E-04 | 3.755184E-05  |  37.552       |
| Keithley2015          | 2             | .01   |       5771    | 0.86413454    | 1.497374E-04  | 1.060543E-04  | 106.054       |
| Keithley2015          | 2             | .1    |       5771    | 0.00289789    | 5.021472E-07  | 8.178690E-05  |  81.787       |
| Keithley2015          | 2             | 1     |       5771    | -0.02076525   | -3.598207E-06 | 1.954389E-06  |   1.954       |
| Keithley2015          | 2             | 10    |       1977    | -0.00668119   | -3.379459E-06 | 9.923083E-07  |   0.992       |
| Keithley2015          | 2             | 2     |       5741    | -0.01858068   | -3.236488E-06 | 1.398587E-06  |   1.399       |
| Keithley2015          | 2             | 3     |       5521    | -0.01898148   | -3.438051E-06 | 1.270776E-06  |   1.271       |

c:\MyMeasurements\Keithley2015>

I will take note not to trust the 1000V range at 0.01 NPLC, it looks pretty bad to me 

[plesa]

If someone else has the same problem I reported  in attachment is repaired script.
The 34410A/34411A and 34401A will follow.

[ManateeMafia]

Thanks for posting the fix. The other scripts will have to be fixed as needed.

[plesa]

Initial noise measurement with Agilent 34411A.
Currently I'm dealing with the reading rate, because for valid noise testing we needs all data from e.g 30 min period and my current Python script is not capable to read everything on NPLC 0.001.
So I used my Labview program which can take 50k reading per second.
My modification will take on all ranges and all NPLC data for statistic. I hope that it will be ported to Python soon.

[plesa]

Labview program for noise measurement almost finished. Automatically calculate statistic (St. Dev, Variance, Spread and Mean) save raw data and statistic.
It takes preselected ammount of samples on each NPLC settings and plot histogram, nothing too fancy
Let me know what next should be implemented.

[plesa]

Measured noise on L4411A,34411A and 34410A. The noisiest are ranges 1V and 100V, about 3 times less noise is on 10V range and 1000V range.
Noise calculation has been calculated as st. Dev on NPLC 1,2,10 and 100.
Noise on 100mV was quite high 4 ppm/Range, compared to 0.1 at 10V range.
Source data are in attachment, including .xlsx spreadsheet and Labview program (works only with 34411,34410 and L4411.
Instruments like 34465,34461 and 34401 will follow (measurement data and program support as well).

Program was too big to upload, so I used TiN FTP /Plesa/Agilent

[Dr. Frank]

Measured noise on L4411A,34411A and 34410A. The noisiest are ranges 1V and 1000V, about 3 times less noise is on 10V range and 1000V range.
Noise calculation has been calculated as st. Dev on NPLC 1,2,10 and 100.
Noise on 100mV was quite high 4 ppm/Range, compared to 0.1 at 10V range.
Source data are in attachment, including .xlsx spreadsheet and Labview program (works only with 34411,34410 and L4411.
Instruments like 34465,34461 and 34401 will follow (measurement data and program support as well).

Program was too big to upload, so I used TiN FTP /Plesa/Agilent

You mean 1V and 100V are noisier, than 10V and 1kV?

No wonder.

100V and 1kV is first divided by 100 each. (in mostly each HP DMM)

So 100V range - divider  output is amplified and measured on the 1V range, 1kV on the 10V range.

Therefore, 1V and 100V are practically identically, as 10V and 1kV are.

No use to measure these HV ranges.

Frank

[plesa]

Hi Frank,

my mistake, it was calculated ppm/range and not absolute value.
Everything indicated that 34411 is better than L4411, small difference compared to 34410 ( slightly worse).

[plesa]

Added other Agilent DMM, looks like some marketing material. Because more recent 34465A is better than predecessor 34411A on 10V/1V range at NPLC 100/10/2. At NPLC1 is better on all  ranges 3441XA/L4411A series.
It would be quite interesting to make this measurement on 34470A and compare it to Keithley DMM7510.
Is there any vollunteer with 34470A or DMM7510?

[6thimage]

I've got a 34470A now (cal'd end of July) and I've been running a noise test script on it when I can to try and get the best results (i.e. hoping the temperature doesn't change too much). All have these have been done using the front terminals with a Fluke short, which was left in overnight. The 34470A was warmed up for around an hour and a half before the first data was collected (to collect all takes 4.5 hours).

Data summary:

nplc	stddev	min	1st quartile	median	mean	3rd quartile	max
0.001	2.30602238797611e-04	-1.14229665000e-03	-1.51375433000e-04	1.69011405000e-05	2.09932992126e-05	1.55784426000e-04	1.21614729000e-03
0.002	1.15148534465426e-04	-5.42673578000e-04	-8.19337900000e-05	2.20449659000e-06	2.05098039046e-06	7.16461393000e-05	6.01827570000e-04
0.006	4.66714749030745e-05	-2.44258223000e-04	-2.63069927000e-05	1.46966440000e-06	2.16718026217e-06	3.51249791000e-05	2.27430565000e-04
0.02	1.68439537307131e-05	-8.31095216000e-05	-9.74877382000e-06	3.42921692000e-06	2.01999992659e-06	1.26881026000e-05	8.53385126000e-05
0.06	6.82210110147248e-06	-2.88505975000e-05	-2.93346187000e-06	1.23205398000e-06	1.28789662138e-06	5.96959490000e-06	3.24734229000e-05
0.2	3.86900389895176e-06	-1.38171905000e-05	-1.81753708000e-06	6.84019333000e-07	9.25507110739e-07	3.55690053000e-06	1.72177438000e-05
1	1.27393700250821e-06	-4.12866614000e-06	4.45708276000e-07	1.29020817000e-06	1.29184723605e-06	2.15816639000e-06	6.59179082000e-06
10	3.79754719542073e-07	-1.05562487000e-07	9.89941540000e-07	1.25971234000e-06	1.25422712073e-06	1.50602481000e-06	2.79388714000e-06
100	1.3879052042004e-07	9.35518214000e-07	1.17784276750e-06	1.26674983500e-06	1.26971653494e-06	1.36644774000e-06	1.67961645000e-06

The full data is available on dropbox https://www.dropbox.com/sh/ff4m3vacz5lg1ll/AADX41mw3C1xsYGs4EuEmQQya?dl=0.

The two images show the collected data (plot.png is histograms with fitted gaussians, density-plot.png is histograms with density) - I don't like the look of the density for the 100 PLCs, so I will try and collect it again.

If anyone wants the script, shout and I'll clean it up a little - it should work on all of the 3446xA series, but I haven't tested it on my 34461A yet.

[HighVoltage]

It would be quite interesting to make this measurement on 34470A and compare it to Keithley DMM7510.
Is there any vollunteer with 34470A or DMM7510?

I also have a 34470A and I am willing to do the test as well.
What procedure do I have to do in detail?

[6thimage]

You can use the data logging mode of the 34470A to measure 10 V DC with the input terminals shorted, with auto zero and high impedance on, for different NPLCs. Ideally all, but the main interest is 1, 10 and 100. The difficulty comes in trying to keep the multimeter at the same temperature, so you don't affect the results (I've noticed skewing of the histogram for around 0.2 deg C change).

You can alternatively use the same script that I'm using (which is attached on dropbox here - https://www.dropbox.com/sh/roz4d2nwjoa0sla/AABnY3lGTschvG2dqIN9Wi-3a?dl=0 as the forum doesn't allow python extensions) - it can either use USB or ethernet to communicate with the multimeter (depending on which of lines 5-9 are commented out).

Hopefully the script should work, but I have only tested it on a 34470A with the dig and mem options so far, with the script running on a linux machine using the latest pyusb from github (https://github.com/walac/pyusb). Using ethernet only requires a standard python installation.

[Alex Nikitin]

I've run the tests overnight on one of my Keithley 2015. Below is the result:

Cheers

Alex

[Alex Nikitin]

I will take note not to trust the 1000V range at 0.01 NPLC, it looks pretty bad to me 

Not really. The 100V range is worse, actually, if you look at the ppm of range figure (the worst is 100mV, but that is expected  ).

Cheers

Alex

[Alex Nikitin]

I've tried also to look at the noise of the Keithley 617 electrometer in the Volts mode. However even over GPIB (which adds one more digit and makes it into a 5.5 digits meter) the noise is at most 4LSB p-p at 200mV range (1LSB = 1uV) and below 1LSB on all other ranges  .

Cheers

Alex

[Macbeth]

I will take note not to trust the 1000V range at 0.01 NPLC, it looks pretty bad to me 

Not really. The 100V range is worse, actually, if you look at the ppm of range figure (the worst is 100mV, but that is expected  ).

Cheers

Alex

Actually I think my figures are nonsense - the ranges are all in 2's, so I expect my 100V measurements are in the 1000V range. I used the K2000 EZGPIB script. I had to throw out the NPLC5 stuff.

I have to admit I just ran the script as is, and it was stated as beta.

[Alex Nikitin]

I have to admit I just ran the script as is, and it was stated as beta.

I've run the script (a somewhat modified version for K2001) overnight, so it is possible that some results are wrong. I will try to verify 100V and 1000V ranges separately. 1V and 0.1V results look correct to me thought.

Cheers

Alex

[TiN]

I'll get back on this project tomorrow, so if you have raw CSV data from scripts, that could help to check as well.
Upload as usual, ftp://xdevs.com/ with login and password datashort

[plesa]

Uploaded data from long term measurement on 34411A,34410A and L4411A on ftp. Data has been acquired simultaneously.
What I found on all ranges is strange behavioral between 10<NPLC=<100, in this range the ppm/range error decreased by significant factor.
On NPLC>100 the error is increasing, but these NPLC setting are not supported I suppose.

For direct comparison I propose to add all data into Excel pivot table to add user configurable comparison between available multimeters.

[TiN]

Simple explanation would be if actual NPLC used by ADC was set to 1, instead of unsupported 200. 

[3roomlab]

hi folks, i was trying out a "new" way of "binning" a bell curve, and i think i have hit the internal resolution limit of the keithley 2015.

originally i was "binning" noise count with 1000 bins from -5uV to +5uV, this gave me about 10nV per bin. while i was trying to expand this to widen the scale (to cater to wider noise of 100v and 1000v). i discovered that the internal limit of the DMM is 5nV (the sorters are only filled in approx "5x" intervals). the FETChing are done @ 1NPLC 100mV, but i have found similar limits at 100v and 1000v

has anyone found similar "limit effects" in your DMM?
it is an interesting find for me, hitting the smallest unit of a DMM
for those who own 7.5/8.5 digits, i wonder what are their limits