At Home Electrical Calibration Lab

[Rhythmtech]

It would be difficult to adequately establish traceability to a national or international standards laboratory without spending a lot of money. It does make sense though that if we are considering purchasing meters that cost half of what having them calibrated would cost we might start thinking of ways to verify their accuracy independently.  For most electronics projects DC measurements are the only ones that concern us, which makes not worrying about trying to build AC voltage and current reference standards acceptable.

Using .1% or .01% precision resistors along with current regulators and voltage references it would be possible to build a simple DC reference that would allow cursory checks of our cheap meters that would be acceptable.

[armandas]

You mean like this?

[alm]

There are various IC voltage references with fairly good stability, but the initial accuracy is usually much worse than the stability, so you really need to have them calibrated for optimal accuracy, and verify the accuracy after some time. It would probably hard to get a commercial cal lab to calibrate a DIY piece of kit. They need to be kept powered on and at constant temperature for optimal accuracy, which is why transfer standards often have batteries and are shipped overnight. A simple 10V reference will only allow you to calibrate the 10V (or 20V/30V) range, and with reduced accuracy the 100V range. For lower ranges, you need a precision divider (resistors with better specs than a simple 4.5 digit DMM are quite expensive). For resistance you need resistors with a lower tolerance than your meter (eg. 0.025% for a simple 4.5 digit DMM). Currents (up to 1A or so) are even harder, although the tolerance of your meter is usually higher.

My strategy is just have multiple precision DMM's (need multiple meters anyway), and use any stable source to compare them. Just connect them in parallel to a variable power supply or function generator, or mains for high-voltage AC. If multiple meters from different manufacturers and places all agree within spec, they either drifted by exactly the same amount, or they're still within specs. Not NIST traceable by any means, but good enough for hobby use in my opinion.

A lot depends on the tolerance, a cheapy 3.5 digit meter with 0.3% accuracy or so is a lot easier than a meter with lots of digits and a very high accuracy. This is usually also reflected in commercial calibration prices.

[Rhythmtech]

You mean like this?

Yes, and like this http://www.syscompdesign.com/REF-101.jpg  oops EDIT - http://www.syscompdesign.com/AppNotes/voltage-reference.pdf

I think what would deserve mention as a blog post would be how to approach calibration, what's the big idea, and why?

For example:
The importance of taking all incoming readings before making adjustments to have recorded everything about the instrument prior to any change and to be able to point at any unexpected changes as a result of adjustment. 
Uncertainty calculations and what do accuracy statements really mean. Sometimes datasheets are vague at best.

Calibration is a lot more than "Does my meter read right?"

[Rhythmtech]

There are various IC voltage references with fairly good stability, but the initial accuracy is usually much worse than the stability, so you really need to have them calibrated for optimal accuracy, and verify the accuracy after some time. It would probably hard to get a commercial cal lab to calibrate a DIY piece of kit. They need to be kept powered on and at constant temperature for optimal accuracy, which is why transfer standards often have batteries and are shipped overnight. A simple 10V reference will only allow you to calibrate the 10V (or 20V/30V) range, and with reduced accuracy the 100V range. For lower ranges, you need a precision divider (resistors with better specs than a simple 4.5 digit DMM are quite expensive). For resistance you need resistors with a lower tolerance than your meter (eg. 0.025% for a simple 4.5 digit DMM). Currents (up to 1A or so) are even harder, although the tolerance of your meter is usually higher.

My strategy is just have multiple precision DMM's (need multiple meters anyway), and use any stable source to compare them. Just connect them in parallel to a variable power supply or function generator, or mains for high-voltage AC. If multiple meters from different manufacturers and places all agree within spec, they either drifted by exactly the same amount, or they're still within specs. Not NIST traceable by any means, but good enough for hobby use in my opinion.

A lot depends on the tolerance, a cheapy 3.5 digit meter with 0.3% accuracy or so is a lot easier than a meter with lots of digits and a very high accuracy. This is usually also reflected in commercial calibration prices.

You are correct, but calibration is all relative to expectation.  Standardizing around something like a NIST standard is for the sole purpose of imposing a uniform definition for everyone. If you are off by the same amount from the standard every time you take a reading your readings are not necessarily bad, just no good for anyone else and no good if you need to refer to design information on a datasheet.  The idea of an at home cal scenario would be to give one's self the ability to spot check your equipment.  You could use the precision resistors as a voltage divider (including error as the root of the sum of the squares) with the precision voltage supply to cover a TTL level cal down to mV level if you wanted.  The whole meter need not be calibrated as a cal lab would, just the channels and ranges for electronics. 

I bet if you built the DIY standard, wrote a cal procedure, and forked over enough money a cal lab would make time for you. I don't know how things are in your area, but the cal labs are hurting here where many businesses have either laid off the person using the equipment - so no cal being done, shut down, or moved.

The multiple multimeter method you describe is good enough, but for a blog post about calibration nobody would want to hear about the history of NIST, computing uncertainty, and traceability if you don't link it to something you've built or neat gadget it just isn't as much fun.   

[allanw]

This site also sells a decent calibration source:

http://www.voltagestandard.com/

[redek]

Here is a link to a Silicon Chip article to built a 10V DC reference.
http://www.siliconchip.com.au/cms/A_111365/article.html

[Rhythmtech]

If you're at a University where they calibrate and have high accuracy bench meters. Technically you could do a non accredited NIST traceable calibration as long as you include the calibration ID info and dates from the cal sticker in your calibration data, which is how it becomes traceable. You just would not have the ISO 17025/9001 certification or A2LA/NVLAP (In the USA) accreditation on your calibration. The certification and accreditation only means that the lab uses procedures and guidelines that ensure traceability and proper handling of that data.

[Joe Geller]

Hi, 

  I noticed a few visitors from the forum and came over for a visit.  Thank you for the link to our SVR reference board!  We sell blank PCBs for those who prefer to collect their own parts as a DIY build.  Typically blank SVR PCBs sell internationally through our outlet store on eBay: http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=320538907673 .  Shipping is just the cost of a stamp, about a $1 US.

  The four wire reference looks good, however, really four wires (two additional sense leads) are only needed if the circuit being calibrated draws current.  For most HI impedance input DMMs (1 to 10 Meg Ohm) a four wire reference is not needed.  The lead errors are easily calculated.  The "lost" reference voltage is the load current times the total connection resistance.

  These days I am mostly focused on our upcoming FDM proton magnetometer experiment for looking at small changes in the Earth's magnetic field related to solar activity.  See the link from our home page.

Regards,
 
Joe Geller
GELLER (Geller Labs)
http://www.gellerlabs.com

[saturation]

Its good to read you here, Mr. Geller. 

Hi, 

  I noticed a few visitors from the forum and came over for a visit.  Thank you for the link to our SVR reference board!  We sell blank PCBs for those who prefer to collect their own parts as a DIY build.  Typically blank SVR PCBs sell internationally through our outlet store on eBay: http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=320538907673 .  Shipping is just the cost of a stamp, about a $1 US.

  The four wire reference looks good, however, really four wires (two additional sense leads) are only needed if the circuit being calibrated draws current.  For most HI impedance input DMMs (1 to 10 Meg Ohm) a four wire reference is not needed.  The lead errors are easily calculated.  The "lost" reference voltage is the load current times the total connection resistance.

  These days I am mostly focused on our upcoming FDM proton magnetometer experiment for looking at small changes in the Earth's magnetic field related to solar activity.  See the link from our home page.

Regards,
 
Joe Geller
GELLER (Geller Labs)
http://www.gellerlabs.com

[McConst]

You may try to use Russian high-precision resistors and dividers.
Manufacturing techniques of precision resistors without glue allow to reach their unprecedented stability while in service and storages. On this site
www.atkp.narod.ru/pageEng2.html
 it is possible to find info about highly stable current shunts and the precision resistors made in the best traditions of military technologies of Soviet Union.