Poor man calibration facility ...

[muvideo]

... and I mean really poor man

First of all I wish to thank king.oslo and all posters of this recent thread bout 100V DC reference:
https://www.eevblog.com/forum/projects-designs-and-technical-stuff/100vdc-reference-circuit/
Also Conrad Hoffman for his series of articles about building calibration devices.

I'm considering to build the instruments to transfer the accuracy of a 10V reference to
the usual DC ranges of a multimeter: .1V 1V 100V and 1000V.
Also I'd like to make the same with a reference resistor of some known value (10kohm? dont know yet)
As for AC no idea yet.

Why?
Learning excercise, learn to trust the meter.

Now my situation is to have a 7 and half multimeter that seems to function properly,
for some reason it lost the cal constants, I changed the eeprom memory and battery
but don't know if I can trust it for maintaining the calibration.
I have no cheap way to calibrate it.
It is the solartron 7061 and is relatively easy to calibrate, having the proper equipment,
and knowing how to do (I think it's not simple to measure things under ppm levels
correctly):
the points of interest are Zero, Low and High of each range,
where Low is 0.075 to 0.21 of range and High is anything from 0.75 to 2.1 range.
DC is calibrated Zero first and after High for each range,
for example 10V is calibrated first shorting the leads and making zero, after connecting a
voltage between 7.5 and 21V and inputting the proper value with keyboard.
Ohms asks only High values, only the highest range asks for open input leads, so no
problem.
For AC the only difference is that instead of zero the instrument asks for a Low range,
so for example 10Vac is calibrated using first a voltage of 0,75 to 2,1V and after a voltage
of 7,5 to 21V.

I could try to ask some friend at university or at some electronics industry to see if we
can organize a test to calibrate the instrument, comparing to some calibrated intrument,
but probably this will be a one off tentative, no possibility to repeat easily, and I dont't
know if they will have any stable voltage source to be used to transfer the accuracy between
the instruments.

Also my lab isn't thermally controlled, so probably there is no meaning to have the instrument calibrated
to last digits.

My main goal is to try to do the best I can with simple low cost and eventually scrap material.

[muvideo]

What I have:
- two Geller SVR, one calibrated and one not, both mounted in ovens the first at 20°C
and the second at 45°C and zeroed to the first, I used the solartron to zero the references.
- the solartron 7061, not calibrated but the ranges are set around the values with random
resistors and voltage sources I had around, the AC and 1000V dc ranges are totally unchecked,
only the 10V range is checked against a Geller SVR.
- another two meters: solartron 7060 totally analog calibration, and unknown state, values approx.
correct, the fluke 8520 repaired by me, but not very trustable

As I said before the first step is to try to have these points:
.1V 1V 100V and 1000V referenced to Geller SVR
And the first step of first step should be to build a 2 step Hamon divider using metal film resistors.

First question: I checked some resistor I had around and seems that the TCR can be positive or
negative, in a metal film resistors run what happens? Is the TCR distributed casually + and -?
Is it about linear? If I test one resistor for +800ppm in 80°C difference (20 to 100°C for example)
will it be about +10ppm/°C  around 20°C ?
I'm asking because I would classify a 100 run resistor pack for value and approx TCR and maybe
use series of them to try to compensate for TCR a little.

Later more questions

[amspire]

Look at the recent thread about the 100v DC reference.

Now 100 deg is too hot for a sensible test. 20, 25 and 30 would make more sense.

A simple test is put two identical resistors in series and put them across the stabilized reference.

See much the divider voltage drifts as the resistors warm up. If it is about 20ppm or less, there is a good chance the Harmon divider will have a reasonable short term stability. The really important thing is the divider need to be made from 10 resistors from the same batch. Three pairs of three, and one on its own.

Richard.

[muvideo]

Richard did you had the time to test these resistors ?
https://www.eevblog.com/forum/projects-designs-and-technical-stuff/kelvin-varley-divider-%28and-precision-voltage-source%29/msg74773/#msg74773

What about metal oxide type? If I understand correctly they are worse than metal film.

Fabio.

[T4P]

Richard did you had the time to test these resistors ?
https://www.eevblog.com/forum/projects-designs-and-technical-stuff/kelvin-varley-divider-%28and-precision-voltage-source%29/msg74773/#msg74773

What about metal oxide type? If I understand correctly they are worse than metal film.

Fabio.

Yes .
Metal film (according to my reference) +/-50ppm and metal oxide , +/-200ppm
The only thing about metal oxide is that higher wattages can be gotten , but it starts at 1/2W 

[alm]

Extra power capability is useful, since it will reduce the heating due to the ~1 mA that the DMM uses for measurement. This can actually be quite noticeable with 1/4 W metal film resistors and 6.5 digit multimeters. I have some ancient tube-era precision wire wound resistors. Just 1% accuracy and TC ~50 ppm, but short term stability appears to be excellent. No idea about long-term stability. I also have some standard resistors (which are just more exotic, larger wire wound resistors) which I picked up for cheap.

[T4P]

Extra power capability is useful, since it will reduce the heating due to the ~1 mA that the DMM uses for measurement. This can actually be quite noticeable with 1/4 W metal film resistors and 6.5 digit multimeters. I have some ancient tube-era precision wire wound resistors. Just 1% accuracy and TC ~50 ppm, but short term stability appears to be excellent. No idea about long-term stability. I also have some standard resistors (which are just more exotic, larger wire wound resistors) which I picked up for cheap.

Normal wire wounds has a TC of +/- 200ppm  , unless it's a aluminium case which is +/-100ppm
Metal oxide can reach 36W yes , but it's civillian rating is around 3W and they are RARE .
And mind you , Metal film you can get many values up to 2W and 50C ppm and you get a TE Connectivity at 200W 440C/ppm at 50$ ... A HELL LOT OF CASH !

[BravoV]

Though I'm not the expert in precision stuff, suggesting to have few high tolerance like 1% or better wire wound resistor from well known brands as a quick & dirty reference. Those no brand even wire wound usually pretty poor in accuracy.

Bought bunch of branded resistors while ago from ebay like Dale/Vishay wirewound high wattage resistor or Caddock high power precision film that are in 1% or better tolerance, found that they're quite precision and I think good enough for short term stability.

For example below pic, the result one of my Dave/Vishay 0.01 Ohm 1% 3 W wirewound, lucky me I have a friend which has a 7.5 digits Japanese benchtop DMM that is calibrated annually for verification, few months ago brought these bunch of resistor and a thermometer for taking a note at the ambient temp there, and then had them verified against my old HP 3468A with the similar temp at home. So far they're proven good enough for "poor man" resistor reference.

Apology for the crappy 4 wires test jig, was lazy to reach the 4 wires probes clip for proper measurement.   

[T4P]

Well you know something ? I wouldn't buy precision resistors from ebay .
I'd rather buy from E14/Premier Farnell or RS .

[amspire]

Richard did you had the time to test these resistors ?
https://www.eevblog.com/forum/projects-designs-and-technical-stuff/kelvin-varley-divider-%28and-precision-voltage-source%29/msg74773/#msg74773

What about metal oxide type? If I understand correctly they are worse than metal film.

Fabio.

I did test them actually. They have a temp coefficient of close to -50 ppm. This is pretty linear from 17 to 30 deg C. Any pair of resistors in the batch match to about 5 ppm / C.

I also tried stabilizing the temp Co. to under 5ppm for a single resistor by winding an extra series resistor of copper a wire around the outside of the resistor, and it sort of worked, but I only had 0.125mm enameled wire, and I need 0.05mm wire or thinner. Not sure where to get it yet.

The copper wire has a +4000pp. Coefficient.

Richard.

[muvideo]

Richard did you had the time to test these resistors ?
https://www.eevblog.com/forum/projects-designs-and-technical-stuff/kelvin-varley-divider-%28and-precision-voltage-source%29/msg74773/#msg74773

What about metal oxide type? If I understand correctly they are worse than metal film.

Fabio.

I did test them actually. They have a temp coefficient of close to -50 ppm. This is pretty linear from 17 to 30 deg C. Any pair of resistors in the batch match to about 5 ppm / C.

I also tried stabilizing the temp Co. to under 5ppm for a single resistor by winding an extra series resistor of copper a wire around the outside of the resistor, and it sort of worked, but I only had 0.125mm enameled wire, and I need 0.05mm wire or thinner. Not sure where to get it yet.

The copper wire has a +4000pp. Coefficient.

Richard.

So the TCR tend to be of the same sign and not distributed around zero.
This is good if it's negative, but if it's positive the copper wire cannot be used.
How about absolute values, did they were uniformly spread?
As for thin wire many years ago I dismantled some ignition coils of cars,
I dont know the english name, back then the coils were insulated
in oil and paper, so easy to open and use, I should try to measure the
diameter, but it's pretty thin.

Fabio

Edit: measured slightly less than 0,8mm diameter.
ReEdit: Naturally I meant 0,08mm

[robrenz]

What about using .1% .25W 1206 25ppm/degree C SMD resistors. relatively cheap and you can use as many as you want in series or parallel to increase your wattage and fine tune your match.  The many I have checked have been extremely consistent. 

[muvideo]

What about using .1% .25W 1206 25ppm/degree C SMD resistors. relatively cheap and you can use as many as you want in series or parallel to increase your wattage and fine tune your match.  The many I have checked have been extremely consistent. 

That should simplify a lot but as I understand (not much to be honest   ) a big "do not" is the mechanical stress on resistors,
I think that is difficult with smd resistors, since have no leads, also the soldering process so close to resistive layer should
bring problems.

Fabio.

[saturation]

Setting up a mini-cal lab as per Conrad Hoffman's article has been something I've been slowly building for several years.   You can find older threads in the eevblog archives and particularly discussions at another forum, volt-nuts.  I think I see you there too, given the last major discussion was about the Solatron 7061.

A problem with building a reference is once you've designed something that you believe is stable, its drift has to be characterized over time in order to define its uncertainty.  You'll have to track data for at least 1 year so you can compare it to the standard specs given by commercial references or meters.  This holds whether you make a voltage or ohm standard.

Voltagestandard.com gives a sample of its reference data over a year, such as this.  Note in this example, the references are rated at +/- 5mV.  Based on these graphs, 5mV is a very conservative rating.  But the key is more the drift pattern, as you can see the references reach better stability past 6000 hours of usage, the standard 200+ hour aging just takes care of the highest rate of change but it continued to climb after that.  Ideally, you need to continuously track your standard or at least sample it to insure it indeed is stabilize and doesn't swing widely some years down the line. 

So if you build a new voltage reference, it may take over 6 months for it to stabilize to use as a reference; it may be shorter or longer depending on what reference and the circuit you choose; you won't really know until you characterize it.   Really good references are aged in years, like wine.

http://www.voltagestandard.com/Tech_Data.html



Geller presumes his reference follows the data provided in the specification sheet.

I've been tracking a Geller SVR since May 2010, and it does not have such a large average rise in voltage, but it swings wildly up to 200uV with changes in barometric pressure and humidity.

Resistance standard have different constrains, but they too need to be tracked over time, and old standards tend to be more stable than young ones.

http://www.ietlabs.com/decaderes-1/resistance-standard/srx.html

[amspire]

What about using .1% .25W 1206 25ppm/degree C SMD resistors. relatively cheap and you can use as many as you want in series or parallel to increase your wattage and fine tune your match.  The many I have checked have been extremely consistent.

The problem is that although the temp co. is good, the thermal mass is lousy.  If you measure say a 10K 25ppm smd on a 6 1/2 digit meter with about 1V test voltage applied, you just see the resistance continue to change while you watch the meter. Physically large resistors change much slower, so you have more time to do useful tests before the resistor heats much.

Also for a Harmon divider, you need to select resistors that match to about 0.01%. What I find with SMD's is after you solder them to the PCB, they no longer match to 0.01%. With leaded resistors, you can carefully bend the leads so they fit into the board with no stress on the resistor, you can leave the resistor leads long so the resistor is off the PCB (more even cooling) and you can heatsink the wires to the resistors while soldering so they do not suffer any thermal stress.

I have an old Leeds and Northrup resistive divider and it is great just because the manganin resistors are so large that they just do not heat that much when you apply a voltage. You attach a 7 1/2 digit meter to them and the reading hardly budges.

If you want SMD type solutions, you are probable better at looking for thick film divider networks.

Richard.

[amspire]

So the TCR tend to be of the same sign and not distributed around zero.
This is good if it's negative, but if it's positive the copper wire cannot be used.
How about absolute values, did they were uniformly spread?
As for thin wire many years ago I dismantled some ignition coils of cars,
I dont know the english name, back then the coils were insulated
in oil and paper, so easy to open and use, I should try to measure the
diameter, but it's pretty thin.

Fabio

Edit: measured slightly less than 0,8mm diameter.
ReEdit: Naturally I meant 0,08mm

I have two different 100 resistor packs of the uxcell 3W metal film resistors. The 3k3 1% all had a temp co of -53ppm +/- 2ppm, and another batch of 8K2 had a temp co of -45ppm (didn't measure a big batch). I did like the fact that the coefficient is so linear.

The testing is a work in progress. When I did get my hands on some fine copper wire, I was going to do some long term stability tests on compensated resistors.

Also I am not actually expecting great long term stability out of any cheap metal film resistors. So I am looking at circuits like the Harmon divider and the Kelvin-Varley divider that can be set up to be calibrated reasonably easily before doing any precision calibration. If I can get good stability over a few hours, that is all that is needed.

Richard.

[muvideo]

Voltagestandard.com gives a sample of its reference data over a year, such as this.  Note in this example, the references are rated at +/- 5mV.  Based on these graphs, 5mV is a very conservative rating.  But the key is more the drift pattern, as you can see the references reach better stability past 6000 hours of usage, the standard 200+ hour aging just takes care of the highest rate of change but it continued to climb after that.  Ideally, you need to continuously track your standard or at least sample it to insure it indeed is stabilize and doesn't swing widely some years down the line. 

So if you build a new voltage reference, it may take over 6 months for it to stabilize to use as a reference; it may be shorter or longer depending on what reference and the circuit you choose; you won't really know until you characterize it.   Really good references are aged in years, like wine.

http://www.voltagestandard.com/Tech_Data.html



Geller presumes his reference follows the data provided in the specification sheet.

I've been tracking a Geller SVR since May 2010, and it does not have such a large average rise in voltage, but it swings wildly up to 200uV with changes in barometric pressure and humidity.

Resistance standard have different constrains, but they too need to be tracked over time, and old standards tend to be more stable than young ones.

http://www.ietlabs.com/decaderes-1/resistance-standard/srx.html

My goal would be to try do be limited mainly by 1 value Voltage Reference and 1 value Resistor Reference.
As I understand to do so it's needed a resistance divider with few hours of stability, and for the calibration
of the hamon divider should be enough a high sensivity voltmeter or an amplifier as per Hoffman's article.
I think that in the process I will encounter many disturbances, maybe I'll not be able at all to reach a reasonable
result, but at least I'll learn many things
So I think that for the initial setup and check the references (voltage and resistance) have to be short-term stable only.
Your warning about Geller reference is worrying for me, but if the variations are spread over some hours,
maybe they can be no problem, if I will have satisfying results, in future I can search for a better reference
(the Geller's cost is almost a miracle here, in south Italy would be difficult just to have it checked by a
calibrated multimeter, let alone measured by a cal lab, at least for me).

Another question:
something like this could be a satisfying resistance reference, at least for short term stability?
ebay # 360440004479
Thinking the low meter's ranges of 100ohm to 1Mohm what value could be best?
I'm thinking using a resistor bridge, one leg is an hamon divider, the other leg made by a reference resistance and a
stable trimmable resistance, using 10kohm I could reach all the ranges with two steps: 1/10->1kohm 1/100->100ohm
and *10->100kohm *100->1Mohm, using the same divider,  is it too ingenuous ?

[saturation]

In the end a similar DMM or better 6.5 digit is the most stable and accurate reference for volts, ohms and more to reference the 7061; it is well characterized, as its older its well aged, the only thing you need to do is align it with the standards, that is to calibrate it.   Buying an uncalibrated but aged and presumed stable reference [ like the eBay resistor] alone won't suffice because you will need as least one verified reading of the reference against a calibrated meter; thereafter a job is to confirm its stability by tracking periodic variations over time and how the environment changes its relative values

With enough data you can calculate a correction factor for changes in the environment, e.g. for several volt references against relative humidity, courtesy of volt-nuts.  The LTZ1000 references track fairly close, while the 2 LM399 ?? have similar patterns but the absolute value varies  between chips in the same model line.


 

e.g. if this 100 ohm will be measured by the Solartron to 30 ppm, then it has a resolution of 100s of micro ohms, e.g., 100.000 3 ohms.

The only way you can cheaply calibrate a meter with the specs of 6.5 digits is to have access to a similar or preferably better calibrated meter; you will then just need to obtain or create transfer values stable for the period while you measure and adjust the DUT, so you can compare each range between the calibrated DMM and the DUT.  This procedure is not a standard calibration procedure, but it can give good results if you are careful to adjust the  DUT to read identical to the calibrated DMM, as possible.   

So, the discussion then comes down to creating DIY transfer references stable enough to make a transfer measurement for each range: VDC, VAC, ohms, and amps.  So far, its practical to obtain low cost VDC, ohms, and low amp ranges, but VAC and 10 amps range are challenging. 

My goal would be to try do be limited mainly by 1 value Voltage Reference and 1 value Resistor Reference.
As I understand to do so it's needed a resistance divider with few hours of stability, and for the calibration
of the hamon divider should be enough a high sensivity voltmeter or an amplifier as per Hoffman's article......
Your warning about Geller reference is worrying for me,
maybe they can be no problem, if I will have satisfying results, in future I can search for a better reference
(the Geller's cost is almost a miracle here, in south Italy would be difficult just to have it checked by a
calibrated multimeter, let alone measured by a cal lab, at least for me).

Another question:
something like this could be a satisfying resistance reference, at least for short term stability?
ebay # 360440004479
Thinking the low meter's ranges of 100ohm to 1Mohm what value could be best?
I'm thinking using a resistor bridge, one leg is an hamon divider, the other leg made by a reference resistance and a
stable trimmable resistance, using 10kohm I could reach all the ranges with two steps: 1/10->1kohm 1/100->100ohm
and *10->100kohm *100->1Mohm, using the same divider,  is it too ingenuous ?

[muvideo]

Saturation if I'm understanding correctly, you are saying that maintaining a single
value "absolute" (not only stable) voltage or resistance reference is not cheaper
or simpler than having (beg borrow or steal   ) a calibrated meter as reference.
The voltage and resistance references have to be charachterized anyway for
enviromental correction and aging etc, and calibrated in a cal lab for absolute value,
periodically.
So the work should be better done on developing the tools necessary to "transfer the
calibration from the calibrated meter to the uncalibrated one" minimizing the accuracy
degradation from the process, sorry if I use improperly the technical terms.
I.E. the development of a short term stable source at the meter's
needed points like for example the voltage and resistance decades.

Fabio.

[saturation]

Yes, that it.   

For short term references to compare against a calibrated meter, I've experimented and found batteries work well, you can easily try it yourself.  With 1-10Gohms input impedance, there is ~ no current flow, and the AA output voltage is stable to uV levels.  I've used resistance dividers with high ohm values to bring it down to 0.1VDC while minimizing current flow that upsets uV levels.

I've also stacked batteries for higher voltage.  The only issue is that as the voltage equalizes between cells, particularly if they are not new, its unstable at 1-10uV level until some time passes.

1 month ago I saw into this video again:





So it gave me an idea to try 11x9V to make 100+ V, and after 5 minutes its was stable to 10ppm or 100uV for 10V, and these are a mix of used & new, cheapo Chinese no-name and good 9V batteries.  In the second video, the user gets nearly 1000V, and the brief reading he showed appeared stable to 1V.



The top 3456a was formally calibrated in 8/11.  The bottom meter was calibrated DIY by me on 5/2010.  Based on these values the DIY meter is in spec but on the outer end of the 1 year accuracy limits, ~ 22ppm. 



I've subsequently acquired an EDC Krohn Hite MV106 voltage reference [ reviewed by Dave elsewhere on eevblog]

https://www.eevblog.com/forum/blog-specific/eevblog-210-krohn-hite-voltage-standard/msg77505/#msg77505

... to replace my resistance divider for generating 10.000 00 V and less.   The middle DMM is the calibrated meter, the top and bottom meters are DIY, and can be seen to track very tightly together after using a DIY KV divider to adjust them over 1 year ago.   However,  I leave for another day, I need to clarify the 21ppm difference from the calibrated meter in the middle; that meter reads 20ppm higher in VDC consistently even against other still in cal meters.

Saturation if I'm understanding correctly, you are saying that maintaining a single
value "absolute" (not only stable) voltage or resistance reference is not cheaper
or simpler than having (beg borrow or steal   ) a calibrated meter as reference.
The voltage and resistance references have to be charachterized anyway for
enviromental correction and aging etc, and calibrated in a cal lab for absolute value,
periodically.
So the work should be better done on developing the tools necessary to "transfer the
calibration from the calibrated meter to the uncalibrated one" minimizing the accuracy
degradation from the process, sorry if I use improperly the technical terms.
I.E. the development of a short term stable source at the meter's
needed points like for example the voltage and resistance decades.

Fabio.

[muvideo]

Than you for explanation, for now only one rapid question:
your 3456 are stacked, I've noted that the solartrons
are highly influenced by the temperature, so stacking them changes
the reading, also with a pair of cm of air in between,
 do you have the meters usually in a fixed position in the lab?

Fabio.

[saturation]

You're welcome.  Sorry for delay.  These devices were made to be stacked but its always good to check they do not overheat. 

The 3456a have vents below and to the side.  Without a fan, the interior temps reach about 90-95^oF if the lab is kept at 75^oF .   When they were being tested individually they would reach 90-95F unstacked, that they stay this way after stacking is amazing.

 Once they are warmed up, all 3 read very close to teach other [ even the calibrated one as its off by a constant fixed 21ppm].  You'd expect the top most one to cool faster, and the bottom most to heat up more, but they are all at the same operating temp and equalized.

I don't get obsessive about the temperature monitoring unless readings are unstable or drifty, as in practice when you work with them continuously you can 'feel' when its hotter than usual, and check with a thermometer. 

You can easily test when meters are warmed properly by using the statistics function on a very stable output: the variance is smallest when the machine is in optimal condition, unless the ambient conditions are unstable [an approaching rain or snow creates unstable uV values due to changing ambient pressure and humidity.]  In practice, I just never turn them off when I'm using them for critical measurements. 

Yes, the meters are in a fixed location because I anticipated using them a lot.  They are also quite wide and heavy, so when I bought them I already decided where I needed to put them!

Than you for explanation, for now only one rapid question:
your 3456 are stacked, I've noted that the solartrons
are highly influenced by the temperature, so stacking them changes
the reading, also with a pair of cm of air in between,
 do you have the meters usually in a fixed position in the lab?

Fabio.