Author Topic: DMM Noise comparison testing project  (Read 215705 times)

0 Members and 1 Guest are viewing this topic.

Offline TiNTopic starter

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: DMM Noise comparison testing project
« Reply #25 on: April 12, 2015, 04:40:15 pm »
Filtering should be disabled on any case.
Also I noticed HighVoltage's data, it's captured with 10meg input impedance. Should be Hi-Z mode, not 10meg, if available. Updated top post for this detail.

ManateeMafia, you wish list is great, would love to see all them.
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 

Offline macboy

  • Super Contributor
  • ***
  • Posts: 2252
  • Country: ca
Re: DMM Noise comparison testing project
« Reply #26 on: April 12, 2015, 08:14:17 pm »

If you used fresh clean copper wire with good contact, it should be no visible difference.
After you get shorting PCBs you could try that as well, if there will be any difference.

I will add new data to first post and page soon, was busy fixing another 2001  ;) Now I have three which can collect data for us.
I am also using a copper wire loop formed to fit across the two input terminals. The sense terminals are not used for voltage readings and I left them open.

I don't have solid data yet, but I can tell you that after an adequate warm up, I set one 2001 into 8.5 digit mode, 10 plc, no filter, and after 48 hours of readings it showed a max +0.00000085 to -0.00000065 min/max span (using the 2nd display functionality). In the shorter term, the noise looked to have under 100 nV rms amplitude. Enabling a filter of just 10 readings for effectively 100 plc per reading seemed to reduce noise to the last (9th) digit, with the 8th digit changing infrequently.
 

Offline TiNTopic starter

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: DMM Noise comparison testing project
« Reply #27 on: April 13, 2015, 11:11:53 am »
macboy

That sounds about right. Also little offset indicates that calibration likely be good :)

Here some tests with histogram view.
As for me, looks much better and also now can see deviation.



On graph there are  three meters:
Blue - 34461A data from HighVoltage.
Green - K2002
Red - K2001

Offsets nulled from all samples, -66nV for 34461A, +2100nV for K2002, -670nV for K2001.

Also last histogram shows that K2001 and 34461A lack resolution (gaps between bins) ?
« Last Edit: April 13, 2015, 11:19:38 am by TiN »
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 

Offline EEVblog

  • Administrator
  • *****
  • Posts: 37661
  • Country: au
    • EEVblog
Re: DMM Noise comparison testing project
« Reply #28 on: April 13, 2015, 12:23:43 pm »
I'll have to get the DMM7510 and 34670A
 

Offline TiNTopic starter

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: DMM Noise comparison testing project
« Reply #29 on: April 13, 2015, 01:33:02 pm »
Thanks, would be awesome to see some fresh instruments as well.
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 

Offline robrenz

  • Super Contributor
  • ***
  • Posts: 3035
  • Country: us
  • Real Machinist, Wannabe EE
Re: DMM Noise comparison testing project
« Reply #30 on: April 13, 2015, 01:40:11 pm »
8846A  100mV scale.

22 mins before any of the random spikes


2 days including the random spikes. Notice the small sine effect from the daily ambient temp swings



Online tszaboo

  • Super Contributor
  • ***
  • Posts: 7306
  • Country: nl
  • Current job: ATEX product design
Re: DMM Noise comparison testing project
« Reply #31 on: April 13, 2015, 03:15:37 pm »
3458A noise about 20cm plain stupid not plated banana cable
1000 NPLC Autozero off, measurement about every minute.
http://pastebin.com/QrzNP2xK
251nV peak to peak
-131nV offset
RMS 44nV.
 

Offline lukaq

  • Regular Contributor
  • *
  • Posts: 78
  • Country: si
Re: DMM Noise comparison testing project
« Reply #32 on: April 13, 2015, 03:23:26 pm »
34401a

100mVdc, 100 NPLC, auto zero ON, "1s" interval, but it takes far more then 1s to so 100 NPLC anyway 9 and 1/4 hours
I hope it is ok.

Also can someone tell me, why there was that drop at 4h-ish?
« Last Edit: April 13, 2015, 03:30:08 pm by lukaq »
 

Offline macboy

  • Super Contributor
  • ***
  • Posts: 2252
  • Country: ca
Re: DMM Noise comparison testing project
« Reply #33 on: April 13, 2015, 06:24:01 pm »
User Galaxyrise mentioned suspecting a noisy 2001 but didn't have a GPIB interface to log the data.

In this case, try capturing multiple samples into the buffer, then look at the statistical results. With the default 8k memory, you should get around 850 samples (compact format); with the 128k MEM2 option you should get up to 30000 samples. After capturing, you can get the statistics by pressing RECALL, then NEXT (2nd display) multiple times to view Min, Max, Average, and Standard Deviation. Since a standard deviation is defined as the root of the sum of the squares of the difference between each sample and the mean, it is mathematically the same thing as RMS Noise (with a lower bandwidth cutoff extending to inverse of capture time, i.e. super sub-Hz). Ensure that the instrument is well warmed up beforehand to help ensure minimal drift. I recommend 2 hours.

I also recommend using the 2 V range since the 200 mV range adds an amplifier into the signal chain. I also recommend disabling the analog (10 kHz low-pass) filter under Config->DCV to remove that extra circuitry. To emulate a 100 PLC integration time, configure the Advanced filter to 10 samples, 10% window, and repeat mode (not moving). This will return one reading per ten samples.
 

Offline TiNTopic starter

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: DMM Noise comparison testing project
« Reply #34 on: April 14, 2015, 04:14:56 am »
Here's new representation. I think much cleaner now.



Need to add legends and range/settings data.
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 

Offline lukaq

  • Regular Contributor
  • *
  • Posts: 78
  • Country: si
Re: DMM Noise comparison testing project
« Reply #35 on: April 14, 2015, 05:32:45 am »
Also note, that some graphs will probably in # of samples, while others (my) are in % of samples for the Y axis

And we should also have the same number of bins (10 or 20 or something), which I will assume, there already are, to really show the same data in graphs.
« Last Edit: April 14, 2015, 05:37:07 am by lukaq »
 

Offline KedasProbe

  • Frequent Contributor
  • **
  • Posts: 646
  • Country: be
Re: DMM Noise comparison testing project
« Reply #36 on: April 14, 2015, 07:25:16 am »
I thought the 34461A would be better than what this is showing...
Not everything that counts can be measured. Not everything that can be measured counts.
[W. Bruce Cameron]
 

Offline HighVoltage

  • Super Contributor
  • ***
  • Posts: 5453
  • Country: de
Re: DMM Noise comparison testing project
« Reply #37 on: April 14, 2015, 09:10:25 am »
the 34461A test has to be repeated. I did this only with some loose wires as shortage and not a proper shortage block.
I will run a new test soon, also on the 34470A

There are 3 kinds of people in this world, those who can count and those who can not.
 

Online tszaboo

  • Super Contributor
  • ***
  • Posts: 7306
  • Country: nl
  • Current job: ATEX product design
Re: DMM Noise comparison testing project
« Reply #38 on: April 14, 2015, 11:48:36 am »
http://pastebin.com/QgQ55cKc
34410A
NPLC,NullState,Function,Range,ApertureEnable,Impedance,AutoZero,NullValue,AutoRange,
100,False,DC Voltage,0.1,False,10 M,On,0,On,

I've deleted the first ~1000 samples as there was significant turn-on heating effects.
Now, I see a long term change happening on this data, in my experience it has to do something with the air conditioning turning on and off in the building.
Peak to peak :755 nV
Average -370nV
RMS 112nV

 

Offline macboy

  • Super Contributor
  • ***
  • Posts: 2252
  • Country: ca
Re: DMM Noise comparison testing project
« Reply #39 on: April 14, 2015, 01:23:24 pm »
For "apples to apples" comparison, all meters should be running the same NPLC. The K2001 unfortunately only goes up to 10 PLC, but 34401A/34410A data here is using 100 PLC; 3458A data here is using 1000 PLC! No wonder the data is less noisy! If we take K2001 data and average each 10 or 100 samples into 1 (for effectively 100 or 1000 PLC per sample) then it would very much change the shape of its histogram. The 3458A is taking about 1 minute per reading, but the K2001 is doing better than 1 reading per second. This is not level ground.

Also for K2001, ensure autozero is set to "synchronous" which does the autozero on each sample. When set to "normal", autozero is only done periodically which is great for taking high speed measurements, but definitely not great for this purpose.  Setting line-sync to On is also helpful to ensure the greatest amount of line noise rejection. (Both are in the General menu).
 

Offline lukaq

  • Regular Contributor
  • *
  • Posts: 78
  • Country: si
Re: DMM Noise comparison testing project
« Reply #40 on: April 14, 2015, 03:06:16 pm »
Then lets set the rules, that all can use and redo the tests.

Offline ManateeMafia

  • Frequent Contributor
  • **
  • Posts: 730
  • Country: us
Re: DMM Noise comparison testing project
« Reply #41 on: April 14, 2015, 03:29:52 pm »
I agree on the rules. We need data that at least can be compared to the same make/model.

A set of rules that are 'Best Practice' methods that make sense based on the fact some of us are doing this from a home lab.  Temperature and humidity regulation will be difficult to control.
Perhaps listing any HW/FW revs with the readings. Either attach it in a separate file or on the top couple lines of the CSV file. Serial numbers probably should be optional in case people are concerned about privacy.

Also,
I have been working with code for the 3458A and also reading the programming commands. I have run across a minor setting that NANDBlog should try if you haven't already.

Set your NRDGS 1,8

The 8 is set for zero line crossing. My readings dropped significantly making this one change. I am surprised it is not a default setting.

I will take a 1hr reading tonight and I hope the AC does not interfere with the readings too much.

My settings will be...

AZERO ON
NPLC 10
DCV .1, 1E-6
NRDGS 1,8

I will hold off uploading the data until we can agree on a format. I just need to get an idea if my data can be collected reliably.

If there are any suggestions to try, I can try different runs on the same meter. I also have a second that I will run the following night. I have a third that, much to my surprise(not!), has a defective GPIB port. I will be t/s it the next couple of days and will probably be replacing at least one line driver IC.

I will welcome anyone with experience repairing the GPIB port on the 3458A.  The most likely components are U904-U906. There is a HP service note to check the port, but it passes their test. You can bet the replacement parts will be in sockets like the NVRAM.
 

Offline TiNTopic starter

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: DMM Noise comparison testing project
« Reply #42 on: April 14, 2015, 03:35:38 pm »
Ultimately I had a goal to have all ranges and main NPLC ranges measured, and thats what I do on my side. So I dont think measured data was useless.

But yes, its a good point, not all of us have meter without use to collect hours of data, so need set minimum data config to run first:

1. NPLC 10 (or maximum before it if meter dont have NPLC 10, like 2182 have only 5)

2. Autozero enabled, line sync enabled.

3. High impedance mode if  selectable.

4. Base range (one with most accuracy per spec, e.g. 20V or 10V)

5. Use copper short or clean wire to minimize thermal EMF.

If run others ranges, should be as extra.
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 

Offline ManateeMafia

  • Frequent Contributor
  • **
  • Posts: 730
  • Country: us
Re: DMM Noise comparison testing project
« Reply #43 on: April 14, 2015, 04:08:13 pm »
I will be making a slight change to what I will do with the 3458A. Instead of DCV .1, 1E-6 , I will just do DCV .1;NDIG 8

The 1E-6 forces the meter into NPLC=100.
 

Offline ManateeMafia

  • Frequent Contributor
  • **
  • Posts: 730
  • Country: us
Re: DMM Noise comparison testing project
« Reply #44 on: April 14, 2015, 04:18:35 pm »
Question:

Is it normal practice to clean meter binding posts with Deoxit or any other similar product? I have not seen it mentioned in any owner's manual, except for the 34420A. I just thought it might be worth considering.
 

Offline Dr. Frank

  • Super Contributor
  • ***
  • Posts: 2377
  • Country: de
Re: DMM Noise comparison testing project
« Reply #45 on: April 14, 2015, 04:33:50 pm »
OK, finally, some rules?  >:D

Anyhow, one single point does not tell so much, I have to point out again.

See the noise diagram of the 3458A below, how the results may be presented in a systematic and clear manner.

you will get similar graphs for every DMM, I suppose, and also, the relationship to the different DC ranges might also be expressed similarily.
Therefore, several NPLC points should be taken, from 0.01 to 100 ..1000 NPLC, if possible, and in a more intelligent way, maybe.

To gather these data, I think, it's not necessary to blindly log hours of  data.. The statistics will converge sufficiently quickly for higher NPLC numbers, I assume.

It will be sufficient, to gather 10.000 samples for NPLC 0.01,  down to 100 samples for NPLC 100, to get a reasonably stable statistics, and to reduce the measurement time.

That would be a maximum of 6 measurements per range, less than 20 for all three ranges of interest (10V, 1V, 100mV), and about 20 minutes per range.

Then, all the measurements of the participants must be properly and completely specified, which range, NPLC, etc., so that a proper normalization can be done!
In the case of the 3458A, HP normalized to ppm relative to full scale.

Btw.: As the input is shorted brutally, there will absolutely be no difference between 10 MOhm or Hi-Z mode of the front-end.
The 10MOhm divider will be in turn shorted, as will be its noise.

Also, if the measurement time is kept short, the actual offset voltage will not play a role, neither its drift.
Therefore, on the short NPLC numbers, AZERO may be turned off, to get a faster result.

You won't draw any conclusion either from hours of offset / Null stability monitoring, especially if you only want to measure the input noise .

Frank
 

Online tszaboo

  • Super Contributor
  • ***
  • Posts: 7306
  • Country: nl
  • Current job: ATEX product design
Re: DMM Noise comparison testing project
« Reply #46 on: April 14, 2015, 05:13:34 pm »
I have been working with code for the 3458A and also reading the programming commands. I have run across a minor setting that NANDBlog should try if you haven't already.
Set your NRDGS 1,8
This log was made over GPIB, and the actual length of the data is more than 8 digit. The logging script is "combat tested", this is what I had on hand to run quickly, I'll try to modify it to 10 NPLC.On the other hand I dont think the setting would be unfair. I see two meaning of this tests to find out
1. You connect the multimeter to your DUT, you take a measurement, and you are curious how much the multimeter affects the measurement. I agree, 1000 NPLC is 20 seconds, but this is the best measurement the 3458A can do.
2. You are logging some DC voltage, it has some noise, and you are interested, how much of the noise is made by the multimeter. Now you can average the Keithley, and end up with the same noise as the Agilent but all you did was reducing the number of samples to gain some accuracy (oversampling). The 3458A can make 1000NPLC measurement, and this will be your result, so the conclusion: The DMM with best settings will make this amount of noise is true.
It is unlikely for anyone to go to the math menu start averaging for taking a single measurement.
 

Offline TiNTopic starter

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: DMM Noise comparison testing project
« Reply #47 on: April 14, 2015, 05:14:08 pm »
Higher sample count allows to remove initial / spurious drift/noise (e.g. if user turned on AC during measurement process).
Having just hundred samples is enough, but it require more care to run test setup.

As I mention before, having matrix of typical NPLC + ranges will allow to have full chart for each particular meter, but this may need some automation.
I'm not seeing much people switching NPLC's , ranges for dozen times and sitting with timer to capture just enough data, but not too much.
Guess I will be first to write test program to automate testing then.

        0.01 NPLC0.1 NPLC1 NPLC10 NPLC50 NPLC100 NPLC1000 NPLC
200mV       +11    +10   +7      +3     +4      +5       +9   
2V          +11    +10   +7      +2     +4      +5       +9   
20V         +11    +10   +1    Primary  +4      +5       +5   
200V        +11    +10   +8      +6     +8      +9       +9   
1000V       +11    +10   +8      +6     +8      +9       +9   

So to simplify:

1. If you just run one test - let it be 10NPLC, 20V(10V) main range.
2. To do full test - add data setups as per number priority, +1 meaning run first after primary test, +2 run second, etc.
3. Add meter information, mains parameters (60Hz or 50Hz, etc).
4. Add special notes, if meter modified/non-standard (e.g. after repair with different parts, or changed parts, etc)

Autozero should be on, it will reduce time to get stready repeatable. With autozero off I saw little drift on 2001, which can be confused with more noise.
Attached screenshot. Greenline - K2001 calibrated, redline K2002, both max NPLC.  Stready data before "spike" - autozero enabled, syncronous (meter takes autozero measurement on every sample trigger), waivy lines after spike - autozero disabled.

I don't think we need chase board revisions or firmware version, unless there is huge difference in data, repeated at least by two different members.

Averaging by math (either by meter functions or computer-aided) should be avoided, as it's just a pandora box of post-processing and not exactly what this study here for. One can take 0.01 NPLC set of data, and give a median from it with zero noise, but that would not show anything about noise pk-pk.
« Last Edit: April 14, 2015, 05:17:57 pm by TiN »
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 

Offline lukaq

  • Regular Contributor
  • *
  • Posts: 78
  • Country: si
Re: DMM Noise comparison testing project
« Reply #48 on: April 14, 2015, 05:56:31 pm »
It will take time, but i'm for it, will go over all ranges, NPLCs

Will be interesting to see, that's for sure

Offline Marco

  • Super Contributor
  • ***
  • Posts: 6693
  • Country: nl
Re: DMM Noise comparison testing project
« Reply #49 on: April 14, 2015, 06:05:54 pm »
How about just throwing a 0.01 Hz high pass filter over the results?
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf