DMM Volts/Ohm Calibration Club - anyone interested?

[enut11]

Based on the posted responses so far, the original idea to fund the calibration of a member's high end DMM and use this to calibrate roving standards did not get enough support.

Change of Scope
1) Most like the idea of a roving calibration system and maintaining a log to compare various members gear.
2) An accurate voltage and resistance standard to be built using best practices of eevblog forum members.
3) Members are divided on whether the roving standard(s) should be limited to a country or region to simplify customs and shipping.
4) System supply voltage be based on a local 18-24v supply.
5) Shipping container to be ruggard and protect delicate stadards, eg blow moulded case with foam cut outs.
6) Characterised resistors of 100r, 1k, 10k, 100k and 1m
7) Characterised voltage standard based on LM399 to provide unamplified output (7v?) as well as 1v and 10v.

So, we need a source of parts and someone with metrology experience to build the cal standard.

Also useful would be a headcount of those interested in participating as well as country of residence.
enut11 (Australia)

[TiN]

Hold on sourcing parts 

Few important parts are missed:

* Integrated captive cabling to external DUTs. E.g. Box should have shielded teflon insulated cable (L=1-2m) to have same connection to terminals. Good luck finding copper banana plugs for decent price (perhaps somebody with access to CNC/lathe could roll own ones?)
* Integrated environment condition sensing. Raspberry Pi Zero + BME280? That will need separate 5Vin jack for supply. Pi has to log data to SD everytime it's powered on. Let's say once a minute.
* Software support. Many of nuts may not be happy to poke buttons on DUT all day long to measure the box. So has to be automated. Best case: plonk USB GPIB to Pi, enter "/run_test" command and job done hour later. Further improvement: have switching function for functions in the box, so once wiring connected to DMM 4W port - it takes care of all DCV/DCI/OHMs automatically. Big iceberg hidden here, as DUT libs need to be written for each different DMM Vendor/Model.
* Budget - calculate estimated project cost. I'd expect this will reduce amount of participants 10 fold... Ballpark figures - resistors 20$/each, LM399 ref - 50$ (total), 7V-10V-1V (50$), Pi+BME (40$).

[2N3055]

Hold on sourcing parts 

Few important parts are missed:

* Integrated captive cabling to external DUTs. E.g. Box should have shielded teflon insulated cable (L=1-2m) to have same connection to terminals. Good luck finding copper banana plugs for decent price (perhaps somebody with access to CNC/lathe could roll own ones?)
* Integrated environment condition sensing. Raspberry Pi Zero + BME280? That will need separate 5Vin jack for supply. Pi has to log data to SD everytime it's powered on. Let's say once a minute.
* Software support. Many of nuts may not be happy to poke buttons on DUT all day long to measure the box. So has to be automated. Best case: plonk USB GPIB to Pi, enter "/run_test" command and job done hour later. Further improvement: have switching function for functions in the box, so once wiring connected to DMM 4W port - it takes care of all DCV/DCI/OHMs automatically. Big iceberg hidden here, as DUT libs need to be written for each different DMM Vendor/Model.
* Budget - calculate estimated project cost. I'd expect this will reduce amount of participants 10 fold... Ballpark figures - resistors 20$/each, LM399 ref - 50$ (total), 7V-10V-1V (50$), Pi+BME (40$).

As for head count, Croatia  (EU)

I would suggest serial (rs232, can easily be made optically isolated ) on RefBox  port and Python script on a PC side.. Simple enough to fix and poke around, works on windows and *nix-es.  I connect everything with Ethernet, USB and serial..

So RefBox would have it's own commands to read and write and all  the custom stuff everybody does in their own lab. Of course, we would help each other with the setup..
For example, someone poste DP832 + Keithley DMM calibration script, took me 2 hours to make it work for DP831 +DM3068..

In that case we could even drop an Atmega on board no need for Pi...

Thoughts?


Sinisa

[e61_phil]

I don't like the idea having such digital stuff next to some precision analog stuff.

Do you really need an automatic DIY multifunction calibrator for this purpose? I think having a box of artifacts is more useful. Therefore, everyone is able to compare the single resistors against each other and is also able to build known dividers and so on. Furthermore, connecting five resitors to the DMM isn't that time consuming, I think.

All these digital stuff, relais and switches will just increase the uncertainty of the stuff and a defective artifact will block the whole box.

[2N3055]

I don't like the idea having such digital stuff next to some precision analog stuff.

Do you really need an automatic DIY multifunction calibrator for this purpose? I think having a box of artifacts is more useful. Therefore, everyone is able to compare the single resistors against each other and is also able to build known dividers and so on. Furthermore, connecting five resitors to the DMM isn't that time consuming, I think.

All these digital stuff, relais and switches will just increase the uncertainty of the stuff and a defective artifact will not block the whole box.

ATMega running of the 32kHz crystal, put in a sleep mode is for all the practical purposes off... We would use latching relays, so pulse on change then again quiet.. TiN said so because the more of the signal path is done right and characterised, it more likely it will have repeatable results.. If I don't do interconnections right, I will get huge errors just by using wrong wire...

[e61_phil]

You need a wire from the box to the DMM in any case.

And connecting a small box with 4 terminals (like the 10k reference Dr. Frank built) isn't harder to connect than a box with 4 terminals. I would invest my money rather in better artifacts than in expensive relays with very low thermal EMF. But this is just my opinion.

[2N3055]

You need a wire from the box to the DMM in any case.

And connecting a small box with 4 terminals (like the 10k reference Dr. Frank built) isn't harder to connect than a box with 4 terminals. I would invest my money rather in better artifacts than in expensive relays with very low thermal EMF. But this is just my opinion.

You have a valid point there, and I value your opinion.. Problem is that high quality connectors are going to be most expensive part....

[Nuno_pt]

Regarding the 100V part that enut11 talk about, Dr. Frank has build an precision divider box of 10:1 & 100:1 to get 100V & 1000V with very few ppm of the 10V with the LTZ, I know it's somewhere here in the forum but I can't find it now, I believe he even put the board layout in PDF and the schematic with the parts use, maybe Dr. Frank will stop by later and point to the right direction, this could be a good ideia for one to have at home, to calibrate the cardinal points at moderate cost.

I'm with e61_phil no need to make the things automated, with expansive reed relays and other stuff, but the logging should be automated like TiN suggested, logging the readings and Temp, Pressure and Humidity inside the box.

The resistors could be order from Edwin (member here on the forum) maybe he can do good prices for this.

As for the reference what would it be? LM399, 2DW23x, LM299, since the LTZ like TiN said could be expensive board.

[TiN]

There's no much point to spend an effort for box without environment logging integration.
Here I plot simple example of expected use scenario with few measurements, let's say 10V.
Red dots are ambient temperature when measurement was taken. Blue romb is voltage output reading.



I'd suggest e61_phil could estimate and tell us :

* How much reference output drifted after it was shipped back to User A ?
* What is uncertainty of transfer between User A data and rest of users?
* Which user's measurement is bad? 

Digital section would be fully isolated from analog, so there is no point in optical isolators/protection/etc. That's why I mentioned separate 5V jack, not take power from main 24V (two SLA battery) to analog section.

Reason for pi - easy and fast binding to internet (RJ45/Wifi), no problems storing megabytes of logs and data on internal 32GB SD card, small size, cheap.

Therefore, everyone is able to compare the single resistors against each other and is also able to build known dividers and so on.

Compare to 100ppm? Maybe. To 10ppm? Less so. To 2ppm? Very few members here able to do so.
Building dividers need expensive stable resistors too. If one has them and DMM already, they would unlikely be interested in discussed box anyway.

This brings bigger question. What specs are required for each of artifact. 1ppm? 10ppm? 50ppm? Design (and cost) would change a lot by this spec.

[e61_phil]

It is of course necessary to monitor the temperature and so on. But is a temperature and humdity logger which is shipped together with the other stuff not enough?

If you wan't to log all the data with the raspberry pi you have to support all the meters and the right interfaces (GPIB for example). Isn't it easier to set up a webpage which collects the data? Everyone can have a look at any time.

I'd suggest e61_phil could estimate and tell us :

I don't like it to talk that way. Let us stay objective. I don't want to say the approach with a DIY box is bad. I just wan't to ask critical if this way is really the best way to do it or is it much more effort than one need. Most of the people here in the thread have much more experience than me (especially you, TiN).

[TiN]

Excuse me if that sounded too harsh. I'm just looking on most nuts logs and posts and see often the case when temperature data is neglected. One temperature value for complete dataset is NOT enough, so temperature data has to correlate to measurememt logs.

From my experience, nobody has perfect stable 24.0c lab when doing voltnuttery, so it's important. Of course if we trying to get at least 10ppm. If it's 100ppm project, then things can simplify a lot.

Also when I start to comment here, idea of multiple same boxes was kept in mind. E.g. to send different regions or for sanity cross-checks. So having that BME280 data tied and unseparated from reference unit makes things much easier. If go all this effort to make international transfers, why skip on details?

Gpib/meter support can be added by participants separately as project evolves.

P.s. online data availability already done part for me, but I cannot say it's easy to understand data for everyone. 

[2N3055]

@TiN

I suggested low power uC, because I thought that with low power chip and 2032 battery (or supercap) it could be made to log even in transit.... With Mega32 i also can write to sd card easily, And with 328 we could even load Arduino bootloader if someone fancies that...

That was my rationale .. But that was only suggestion.

[Kleinstein]

Temperature logging is important. Here one might even need more than just one sensor, e.g. one for the room and one close to the resistors.

However I am not convinced that one needs to have the logging unit and much of the support along with the reference part. The logging part really depends on the meters, at least the software part.  For many the could something like a Raspi or an even smaller µC based unit. It does not make much sense to move this around as it really makes sense to have such a logger.

I would consider just moving around the actually critical reference part and keep most of the rest in the lab, even if this means you have to build it a 2nd or 4th time.

So in the extreme case the exchanged part would be only the voltage reference (with direct support circuit and likely 7 V -> 10 V step),  the dividers for 10 V -> 100 V and 10 V -> 1 V -> 0.1 V (if needed), a few resistors (maybe even shared with the dividers) and 1 or 2 temperature sensors.

The rest of the circuit to make this a small calibrator should not be that expensive and should be pretty much exchangeable.  So no need to also ship this part around every time - this could ease on shipping costs and ease on customs.

[guenthert]

From my experience, nobody has perfect stable 24.0c lab when doing voltnuttery, so it's important.

Well, the obvious solution (to me) would be to integrate an oven with controller.  Less than 0.5K drift can be achieved with a simple NTC and TL431.

[Nuno_pt]

That would be good to put all in one box, 7V->10V buffer, the dividers for 10V->100V->1000V and 10V->1->0.1V, and also the 100R, 1k, 10k, 100k, 1M, 10M, 100M resistors, the logging can be done with one BME280 inside the box, and an connection for an room BME280, that way it would log the 2 temp parameters, along with the rest of the measures.

About the resistores maybe the Econ 8G16A 0.01% 3ppm/C, but this only go to 500k, or ask the prices to Edwin maybe we could do some good prices.

For the dividers we could use the schematic from Dr. Frank, with is permission.

For the Reference seems that we're at LM399 (4 free samples from LT) or 2DW23x (cheap on Aliexpress), or maybe LTZ since it doesn't need the expensive Z-foil from Vishay, and we can get very good results with WW resistors, for the boards we could use some of the TiN or Andreas or other designs with LTZ or LM399.

[lukier]

I don't see the usefulness of such project. Correct me if I'm wrong. My thinking goes as such:

- if one needs to briefly check low to mid range DMMs then possibly the DMM Check or similar products are the way to go,
- for calibration, as someone mentioned already, dozens of calibration points are necessary, from DC to AC to current to ohms, for example just resistors required for my meters:
K2015 (K2000): 1k, 10k, 100k, 1M
K2001: 19k/20k, 1M
34401A: 100, 1k, 10k, 100k, 1M, 10M, 100M
3457A: 30, 300, 3k, 30k, 300k, 3M, 30M

that's a lot (15) of rather expensive resistors, that's why there are Fluke/Datron multifunction calibrators.

I see that RS will calibrate a bench meter for 62 GBP:
http://img-europe.electrocomponents.com/uk/img/site/campaigns/servicecentre/EMEA_0009_Calibration1-7-FINAL.pdf

That is cheaper than a DMMCheck Plus or building and shipping various gizmos. If one got a decent 6.5 digit meter for 200-300 GBP on eBay (typical prices for K2015 or 34401A) then 62 GBP is not that bad.

I'm not sure they do 6.5 digits (will be asking soon), but probably they do (they have 6ppm on 10VDC so that's probably a Fluke calibrator:
http://www.ukas.org/calibration/schedules/actual/0310Calibration%20Single.pdf)

So we've covered handhelds and bench meters up to 6.5 digits. For more serious stuff things get even worse, as calibrating 7.5 digit and higher requires very precise standards and often is only done properly by the manufacturer.

The only noble exception where this endeavour might make sense is HP 3458A, as it can calibrate itself from 10V and 10K. But in this case the travelling standard must be top-notch, on par with Fluke 732B and ESI SR-104 (maybe VHP101). I think forum users affluent enough to have 3458A would rather pay for Keysight calibration, maybe even the more expensive Loveland calibration against references tied to the Josephson Junction standard (0.014 ppm uncertainty on 10VDC).

[2N3055]

Well this would be like DMM Check for 6.5 dig DMM.. That is the point... And would be calibrated on a regular basis by people that know what they are doing.
That would give me enough confidence to know whether I need to send some equipment for adjustment or not..
That would be great help.. Also I would make one or two to keep and that way establish my own monitoring process..

I think that is worth a lot, and appreciate someone is willing to help with such effort..

Of course it would be easier to just send equipment to get calibrated.. If you can afford it..

[lukier]

Well this would be like DMM Check for 6.5 dig DMM.. That is the point... And would be calibrated on a regular basis by people that know what they are doing.

I see. Actually, I'll be building TiN's 2xLM399 board soon for a similar purpose (to check and perform user calibration on my K2001 that I've almost finished repairing).

Still 6.5 is can be quite demanding. All my meters have LM399 and being often sourced from eBay recyclers these are very old thus well aged references. I think to have decent TUR the travelling standard would have to be either preselected/aged multiple LM399 or LTZ1000 and it's where it becomes expensive. And then there are precision resistors, amplifier/divider for higher ranges, current sources and I don't even have an idea how to do a precise AC source.

Anyhow, I bet such DMMCheck 6.5 will be more expensive to build than just to pay for calibration

[Nuno_pt]

Well this project if it lift off and we can do this, will serve as a learning point both in Metrology and Electronics, and also to be able to *Cal* some of my gear and make measurements on how mine compare to the others or how much ppm it deviate from the last check, or how much drift relate to the Temp in the little house where I've my working bench.

The set WW precision resistores for an LTZ should be around 50 to 60€ maybe less, some parts can be sample from LT, probably Edwin can help also with WW for the decades at moderate cost.

I'm also be building 2 to 4 boards of TiN with 2xLM399, I've order 2 of each samples from LT, I'll order also samples for the parts that I can from LT or other places.

Next will 2 to 4 LTZ.
 

[BobsURuncle]

If you get enough people, someone in the group will always have a new and/or recently calibrated meter.  For example I have a new Keithley 7510, a 7.5 digiit DMM which is less than a year old - so still in calibration.  I would be willing to check some standards for some people if said persons pay for all the shipping costs.

[VintageNut]

If you get enough people, someone in the group will always have a new and/or recently calibrated meter.  For example I have a new Keithley 7510, a 7.5 digiit DMM which is less than a year old - so still in calibration.  I would be willing to check some standards for some people if said persons pay for all the shipping costs.

On the DMM7510, all of the ohms are 2-year same as 1-year. Fort DCV, the 2-yr is de-rated some from 1-yr.
You can use the first year to dial-in all of your DCV sources and then coast for year 2.

[hammy]

If you get enough people, someone in the group will always have a new and/or recently calibrated meter.

We could use a wiki page. Eyeryone who joins the group could add his dmm, including the calibration dates, to this wikipage. We send two or three standards in a circle always around, and everyone could enter the meassurements into the wiki page. If someone gets the standard, he can check his DMM against the others.

We do this in our country, and the shipping costs are low. Once a year we send a standard over the ocean to another measurement club and they send us vice versa their standard. In this way we could do a cross check.

I would join such a "measurement club" here in germany/europe. And I can offer hosting the wiki. Anyone interested?


 

[e61_phil]

I would like to join your idea hammy.

[Nuno_pt]

Well since I'm in EU, I'll join EU *club*, since the most near guys from me are in Germany.

[Nuno_pt]

Looking at some old threads from LTZ and looking at TiN website, I've gathering the following:

BOM from TiN website price today at Digikey;

311-10ERCT-ND - $0.99 -> 100
311-1.0KERCT-ND - $0.99 -> 100
311-1.0MERCT-ND  - $0.99 -> 100
SAM1061-04-ND - $10.99 -> 10
MAX6610AUT+TCT-ND - $21.17 -> 10
399-3905-1-ND - $19.04 -> 10
LTC2057IS8#PBF-ND - Samples LT -> 2 (here one can ask for 2 free samples at LT website)
P402KBCCT-ND - $9.40 -> 10
399-6417-1-ND - $11.39 -> 10
1N4148UR-1-ND - $19.60 -> 10
PZT3904CT-ND - $3.30 -> 10
1189-1818-1-ND - $15.25 -> 5

Total = $112.75

Resistors from Edwin prices from 2015;

120R - $6.85
1k - $6.34
10k - $7.43
12.5k - $7.43
70k - $12.38

Total = $40.43 for each board or 161.72€ for 4 boards.

LTZ1000ACH - $54.50 from LT
LTZ1000CH - $42.85 from LT

The order from Digikey can be split between all, that way would be cheaper to all, the WW resistors would be bought by each one depending on how many boards one want to build, the same with the LTZ.

I'll have to make the BOM for the 2xLM399 from TiN also.