DMM Noise comparison testing project

[TiN]

It's not limit effects I believe, it's resolution effect, as with faster sample rates, A/D resolves less bits, so your minimum LSB become bigger.
Use slower NPLC, and you should see smaller "steps".

I did first test on my 3458A, so I leave these graphs here as well:

Simple AWG26 copper wire short.
Added ManateeMafia's data from July 2015 on his 3458A for reference (red lines on graph).
This is standard deviation data, not RMS values.

100mV range.



1VDC range.



10VDC base range



100VDC



1KV range

[dr.diesel]

Data from a K2450.  No windows machine to run the automated tool, 10 NPLC, 2V etc.  Full setup data included below.

Happy to do more, also have a K2010 if desired.

[TiN]

Yey, desired everything you have

[dr.diesel]

Results from a K2010, in cal, Keithley 4-term shorting block.  Instrument setup is mirrored in the output text.     I initially started at 30 min captures, then realized that was a bit excessive when noise and not drift was the focus.

I partially automated this, I will redo the K2450 run tomorrow in the same manor, except all at 10 min runs.

(my lab is not particularly quiet either  )

EDIT:

Range 2 on a k2010 is 10v, FYI.

[alanambrose]

Hmmm, anyone got ezgpib to work with Win 64 bit / Agilent visa / 82357B?

A.

[dr.diesel]

Results from a K2450.  Slightly different as I forgot the k2450 doesn't have rs232, 

Instrument warmed up ~2 hours, 2V range, 6.5 digit mode, autozero, no filter/math, Keithely 4-term shorting block, 10 min each.

[3roomlab]

in my separate noise "adventure" with long boring posts (related to this DMM noise logging), i was trying to tinker with my K2015 to lower/find noise problems with my pre-loved korean set. i think i have hit a place where i would need some input from anyone with some relevant data.

i have summarised some of my noise "findings" of my slightly modified K2015 (see pic, atm still collecting 1NPLC data). but i am unsure as to where i stand in terms of pk-pk noise of the DMM (it is influenced by my environment AC noise, as i have yet to get any AC isolation done).

the calculations i used in my sheet are using opencalc, the opencalc SYNTAX for this peak-peak is [MAX(data-range) minus MIN(data-range)] to get the value. it does not exclude outliers

i hope this is not too off topic. if anybody can put out some numbers of your peak-peak noise for comparison, it would be greatly appreciated TYVM (esp 1v/10v range 10NPLC, and also if any solartron 7081 n any 7.5digits owners , i am particularly curious at solartron for it is older than 3458a in technology). i recall forumner Robrenz mentioned in a thead he has his tek4050 log 152hrs with only 1uV pk-pk noise, and that was amazing.

then again, should pk-pk data become a part of the main noise data here? (it did help identify popcorn noise in my DMM, thanks to kleinstein)

*edit : there is an error in the PPM range column, 0.1v range PPM should be x10 more

[TiN]

I was reading your adventures, interesting stuff.
Can you upload your data, I'll definately make comparison graphs with datasets people already had sent.

[dr.diesel]

i hope this is not too off topic.

It's a really fun topic set!

Perhaps if we standardized on a file format, TiN's website could automagically parse datasets?

@3roomlab, your thread has inspired me to source a busted DMM, just for modding.   

[dr.diesel]

i recall forumner Robrenz mentioned in a thead he has his tek4050 log 152hrs with only 1uV pk-pk noise, and that was amazing.

I have a vague memory of this, was it done without filtering/math?

Either Way, the 4050/8045A is a very underestimated instrument IMO.

[3roomlab]

i recall forumner Robrenz mentioned in a thead he has his tek4050 log 152hrs with only 1uV pk-pk noise, and that was amazing.

I have a vague memory of this, was it done without filtering/math?

Either Way, the 4050/8045A is a very underestimated instrument IMO.

iirc it was inside either the resistor or LTZ discussion thread . it was the 1 i noted specifically because i was still discovering how low STDEV could K2015 go as all these are new territories to me. i was speed reading thru and saw his STDEV was 76nV, and in the other thread, i think he had another test and it was 73nV. and it was for 152 hrs ! wow !

(and then now, the peak peak value, i think this could be the 1 that will really interfere with alot of serious measurements, and i think mine is crazy. some of my logs, there are 10v range pk-pk as high as 7uV. making a voltage reference will be useless as i dont think i can log anything useful with this noise, am i right?)

(@ TiN : edit and a package for NPLC10 in xls/ods, first column is time elapsed. the wonders of PYTHON 2.7!)
(edit ** : the file extension in XLS, rename to ODS/CSV where applicable)

(@dr.diesel ... i missed the chance to grab a used solartron 7081 for under 1.5k.  ...)

on a sidenote, my measurements are done w/o special thermal considerations in terms of special cables etc. so it could be i have higher thermal noise in shorting plug or 4mm jacks etc. i do notice that my 4mm socket on the DMM is always shedding material when i wipe it, just black stuff on the cotton buds. it is making me itch to not use the sockets at all

[TiN]

Can you save in excel or at least CSV? I don't have this openy-thingy-software, sorry

[dr.diesel]

.csv files:

[dr.diesel]

Wife bought me a DMM7510 for X-Mas, so those tests coming up once it arrives.

[SharpEars]

Wife bought me a DMM7510 for X-Mas, so those tests coming up once it arrives.

Fucking finally, already!

Pardon my French, but I've been waiting for a DMM7510 to be (measured for noise and/or) added to the list for forever. Dave tested it, Shariar (of SignalPathBlog fame) tested it, others tested it and noone but a single random internet engineer has said dick about or measured its noise characteristics.

Nothing but empty promises of "I'll take a look at it and report back." Hopefully you will not become another data point in that trend of promising and not delivering.

[lukier]

Pardon my French, but I've been waiting for a DMM7510 to be (measured for noise and/or) added to the list for forever. Dave tested it, Shariar (of SignalPathBlog fame) tested it, others tested it and noone but a single random internet engineer has said dick about or measured its noise characteristics.

Nothing but empty promises of "I'll take a look at it and report back." Hopefully you will not become another data point in that trend of promising and not delivering.

Well said 

Teardowns or playing around with simple experiments on the camera is nice, but not a replacement for proper testing and pushing the instrument to its limits. Maybe that is not a content that would generate enough views, no entertainment value, thus ignored by video bloggers?

On the other hand this is of great importance when talking about the instrument like DMM7510. Potential buyers would like to know if it is really worth spending quite a bit extra to get DMM7510 instead of 34470A.

[dr.diesel]

Nothing but empty promises of "I'll take a look at it and report back." Hopefully you will not become another data point in that trend of promising and not delivering.

Not to worry, but these come directly from Keithley, so maybe 2-3 weeks, atleast that's why my K2450 took to arrive.

[TiN]

Wife bought me a DMM7510 for X-Mas, so those tests coming up once it arrives.

Fucking finally, already!

Pardon my French, but I've been waiting for a DMM7510 to be (measured for noise and/or) added to the list for forever. Dave tested it, Shariar (of SignalPathBlog fame) tested it, others tested it and noone but a single random internet engineer has said dick about or measured its noise characteristics.

Nothing but empty promises of "I'll take a look at it and report back." Hopefully you will not become another data point in that trend of promising and not delivering.

I did not tested it I asked Keithley though about loaner, no reply at all. Perhaps they did not like my DIY repair endavours  .
Can't wait to see it's data. It's yet to be determined if it can compete with K2002.

[deadlylover]

Shamelessly stolen graph from Dr. Frank, I hope you don't mind, many many thanks in advance. =P

Small teaser from the Advantest R6581 8.5digit meter, statistics done in instrument. (where the hell is my GPIB-USB cable grr)

100NPLC done with 100 total samples, 10 and 1NPLC done with 1000 samples, 10k for the rest. Auto Zero was on, I thiiink it was the same as what Dr Frank did. I will try them again with more samples and on other ranges when my damn cable shows up.

I ran the 10V 100NPLC test a few times, ranged from 0.0084 to 0.0092 ppm/range stdev. I was using the Fluke shorting block, and you know what, I was expecting the R6581 to be much worse than the 3458A on 100NPLC but it's not that bad is it? Maybe I don't need want a 3458A after all...

I'll also do the same with Advantest's R6871 and R6561 7.5/6.5 digit meters, but going through their instrument menus is hell.

[dr.diesel]

So Keithley has outdone themselves, DMM7510 will be here tomorrow according to UPS tracking.

Any interested in standardizing on a .csv format?  Items like what header info, submitter info, sample rate, instrument setup, etc?

[lukier]

So Keithley has outdone themselves, DMM7510 will be here tomorrow according to UPS tracking.

Any interested in standardizing on a .csv format?  Items like what header info, submitter info, sample rate, instrument setup, etc?

Check TiN's website for more info and useful scripts: https://xdevs.com/article/dmm_noise/

[dr.diesel]

Ok, unless otherwise requested I'll use the expected format of TiN's parse_noise.py script.   

[TiN]

I think there is need to revamp my python knowledge to make it more useful though.
Would be nice if you just follow existing format for now.
What you will run to get GPIB data? I can try to make an script for 7510, but haven't read it's docs yet.

[Dr. Frank]

Shamelessly stolen graph from Dr. Frank, I hope you don't mind, many many thanks in advance. =P
..
Small teaser from the Advantest R6581 8.5digit meter, statistics done in instrument. (where the hell is my GPIB-USB cable grr)


I ran the 10V 100NPLC test a few times, ranged from 0.0084 to 0.0092 ppm/range stdev. I was using the Fluke shorting block, and you know what, I was expecting the R6581 to be much worse than the 3458A on 100NPLC but it's not that bad is it? Maybe I don't need want a 3458A after all...

Hello, you are very welcome using this diagram!

Do you have any idea, on which basis this R6581 does the Auto Cal , in comparison to the 3458A?

Frank

[deadlylover]

Do you have any idea, on which basis this R6581 does the Auto Cal , in comparison to the 3458A?

Unfortunately I have no idea, even in the Japanese manual there is barely any info. Maybe I can run the Auto Cal and write down the processes it displays on the screen if it will help at all? The whole process takes about 11 minutes.

I think it is done more simply than the 3458A, some clues are in the 24hr range error specification. From 10V>1V>0.1V the range error is 0.1>1>10 ppm. I don't know if the difference are from a worse ADC, or different Auto Cal scheme, or both.

If you have any free time with the 3458A, can you check the error from applying 100mVDC to the 10V range and then reading it from the 100mV range? I get a difference of about 2-5ppm of reading even after an Auto Cal.

Another deviation is that the 3458A has a 40k reference resistor inside right? The R6581 has a 10k reference inside instead, the value can be pulled over GPIB same with the LTZ1000 reference data.
I didn't know the R6581 had a clock inside that stores the last external calibration time, I should have read it before I sent it for calibration!

When I find another cheap R6581/D I will get it and see if I can do any investigation. There is a D model (no AC functions) available for ~USD800 but it's too expensive for me, the AUD is too weak right now.