T.C. measurements on precision resistors

[Edwin G. Pettis]

According to Vishay, as stated, every mW of power increases drift up to an additional 0.5PPM (from zero power condition)/1K hours and each mW will raise the internal temperature by 0.1°C by self-heating.  It has nothing to do with electro-migration or anything else.  Vishay's drift specification is with zero power applied, i.e., sitting on the shelf, applying power changes the spec.  If you have any argument with the above statement, take it up with Vishay, it is their statement but you have to talk to Israel because nobody else seems to know anything about it at the rep or distributer level.

Yes, it is the temperature rise of the resistance element itself but it can be affected by its environment as stated by MisterDiodes.

Remember, this is all lumped together when you're measuring resistance on a DVM so what you're seeing on the meter is just a sum total of what's going on in the element, you cannot separate out any of those parameters by simply measuring the resistor.  Each parameter must be measured in isolation or it is invalid.

Above all, don't forget that Vishay data sheets do not tell the whole story nor do they cover every use scenario.  The only real indication is when you put that resistor in the actual circuit under actual operating conditions; measuring it on a DVM in air is not the same thing.  For that matter, measuring it on a resistance bridge does not necessarily produce the same result either unless the conditions are the same as the operating circuit.

You are really wasting time measuring when you should be testing in the actual circuit where it counts, not on a DVM.

[MisterDiodes]

Shock! Surely you're not suggesting that Vishay's datasheet (< 0.1ppm/V) is anything but 'Legal, decent, honest and truthful' (the UK Advertising Standards Authority's mission statement)? I expect the 'values over 1K' exception is clearly detailed in the microdot at the end of the disclaimer section. 

Not only that, but it that VC spec can go to 5ppm/V if you're spec'ing for a defense application - following full MIL test procedures.  In other words, the more critical the final application, the wider the tolerances become on the "As Delivered" resistor datasheet SPECIFIC to that particular resistor design and application.  Even though the resistor itself might be basically the same as a non-MIL model.

When you work directly with Vishay on PMO, you will get a worksheet and define all aspects of your final use, target limits, max temperature range, etc.  At that point you are assigned a unique part number for your resistor model.  That is your Vishay's reference to your company resistor part number & project application and references your company-specific datasheet.  It's not like you're using the general datasheet from the sales department - that's only the starting out point.

Again - it depends on what model resistor, value, how it's mounted, trace width, copper weight, duty cycle etc.  It also depends on resistor size and how much you're under-rating...for instance at any given low power you can go to a larger resistor and generally you see less of an load-dependent effect - at the expense of more board real estate and somewhat higher cost.

The value of 0.1°C temp rise / mW depending on the resistor element size and thermal coupling to board and to ambient.  That should be considered only a "typical" value as described by Vishay when it's mounted on a PCB with 1Oz traces for the VHP101x model we are designing for.  Other resistor designs will be different.   THAT NUMBER IS NOT guaranteed by Vishay because they have no control over final board design thermal flow.  That parameter is marked "As Tested in Application Only" on the delivered datasheet.

Remember that when Vishay describes the resistor value changes you'll always see the words "Up To" - as Mr. Pettis points out every resistor will have a somewhat unique random response.  For our client we have to calculate a very specific documented expected operating uncertainty and maximum drift for the first 1000hr, 2000hr and 1yr over several hundred circuit modules, each containing about 5 to 8 precision resistors.  Because this is a critical application, all calculations and manufacturer expected drift rates, expected MTBF, etc. must be included for review as part of "deliverables acceptance inspection" by customer.

[texaspyro]

Because this is a critical application, all calculations and manufacturer expected drift rates, expected MTBF, etc. must be included for review as part of "deliverables acceptance inspection" by customer.

Soooo... you mean you can't just bung in a random Chinese carbon comp and call it a day?       

[MisterDiodes]

Yeahhhh, well, not really    These customers have a lot at stake and that means they make sure everyone they buy from has to show it's done right...and that means a paper trail for every part, every procedure.

[texaspyro]

Yeahhhh, well, not really    These customers have a lot at stake and that means they make sure everyone they buy from has to show it's done right...and that means a paper trail for every part, every procedure.

Meh... paper grows on trees and ink is cheap.   Tis' easy to dummy up some paperwork and specs... just look at Vishay's (and pretty much every other) datasheet.    And, if caught,  you would probably look fetching in that stripey / checkered / orange prison garb.   

[e61_phil]

Oh - and that VC they say is essentially zero on the standard datasheets?  In reality you find out that's more like a 3ppm/V on some of their VHP and Z-Foil ratio resistors on values over 1k.  Head's up if you're using these on an application where the resistor sees variable bias - say for the input of an ADC front end or similar.  That's why went back to a sealed PWW resistor for that location - for virtually no VC effect!.

Do you have more information about the VC? I used three (I feared the VC) 70k VHP101 in a application in which the resistors see -21..21V. I compared the linearity of the whole circuit against a 3458A and the linearity was much better than 1ppm over the full range. This already included the (non-)linearity of an AD5791 DAC. Therefore, I would expect a much better VC than 3ppm/V from the VHP101.
I also saw around 3ppm/V in the Z201 datasheet. I asked the german VPG guy, but he can't tell me much about it.

[MisterDiodes]

As Mr. Pettis noted, most of the Vishay sales and field engineers won't have much to say (even if they know what VC is) but when you talk to Vishay manufacturing Israel you'll get a more complete answer.  At least that's how it worked in our case.

That "Up To VC" has a lot to do with resistance value,  temp range of your application, the actual resistor model used, bias point,  PMO, mounting method, etc.  That does NOT mean that EVERY resistor will show 3ppm/V over a 42 volt bias change over ALL temperatures and power levels!  However if you DID see that, what Vishay is saying you would still be within normal operating parameters according to their datasheet...and there would be no reason for (their) concern.

If your resistor is under 1k you will probably see smaller effect, at 5k to10k or over you'll see larger effect when working OVER the required operating temp range.   For instance on this industrial test case the required temp operating range is -10C to +60C.  If you're going MIL spec to 125C or 155C or down to -55C (for instance) your datasheet will look different.

Also note that the VC will appear larger on some of the Z-Foil ratio resistor design we were looking at (dual resistors in one package) vs. using 2 discrete resistors.  Yes that ratio resistor construction gets you two resistors in one thermal package, but there are considerations to look at when the resistive elements get smaller - they tend to heat up quickly, but cool down is dependent on thermal flow out of the package.  Sometimes that's an important consideration if you're looking at back-to-back signals at variable bias levels:  If you had a higher bias power signal flowing through the resistors for a few seconds then switch to a lower level - how long does it take for the resistor ratio to cool down internally?  That all depends on how heat flows on your assembly.

Again, you'll have to talk with Vishay Manufacturing Israel on your exact application use for your exact resistor model and at what value.  There are lots of variables in play, and every situation is different.

[e61_phil]

Thank you very much for these informations!

I used the resistors well within their specifications. Temperature is stabilized to 38°C +/-0,01°C and so on. My experience was that the resistors are much better than 0.1ppm/V. But perhaps I had some luck.

My last off topic (VC and not TC) question : Is there an "easy" way to measure the voltage coefficient of resistors? I know some techniques for high values resistors with high voltages, but nothing for resistors with 100k or less.

[MisterDiodes]

That's always the big question - because changing the bias point can give you a VC change, but that also causes a temp change...which could make the end result better or worse, depending on actual TCR at that operating point and at your thermal flow.

But there is one defined measurement method way which actually has an interesting hidden use on Vishay datasheets.

Vishay measures VC using the old MIL-STD-202-309.

In a very condensed nutshell:

The VC is computed as VC = (R-r)100 / 0.9Er.

R= Resistance at continuous working voltage, r=resistance at 0.1 * Rated Working Voltage and E = Rated Continuous working voltage.

You apply 0.1 Rated working voltage and let resistor come to equilibrium in your circuit and grab a measure on "r"...then apply your continuous rated working voltage for no more than 1/2 second in any 5 second interval...Grab your "R" measure as quick as possible during that max 1/2 second pulse...as close to the beginning of the pulse as possible, before the majority of heating comes in.

Run the numbers and that's how Vishay measures VC.

The interesting part is that Vishay claims that there is no VC below 1k resistance value on their foils (which is not exactly true).  The reason they can get away with that is because MIL-STD-202-309 was only defined for resistors of value 1k and over...

Take that for what it's worth.  Vishay BMF parts are not necessarily bad, in fact they can be very good.  The problem is you always have to lift the veil on their datasheets.

[Edwin G. Pettis]

You have to remember, when you are measuring a resistor as a two (or four) terminal device, you are measuring the sum total of what is going on inside the resistor, all of the operating variables are in play and what you are seeing is the sum total.  If you want to measure any one given parameter, it must be done in isolation, which in some instances can be difficult, otherwise you are still just measuring interactions between parameters.  For instance, you cannot accurately measure humidity effects without holding all the other parameters constant, otherwise you will get an erroneous measurement.  Since few if any have done controlled humidity measurements correctly on resistors, I cannot consider any of the statements about humidity effects on any resistors as valid here.

For instance, the interaction of VC and TCR; the TCR could be positive and the VC negative, in which case you'll see the net sum difference which by chance could be quite small, they could also be of the same polarity in which case you would see a much larger difference, which on the outside would appear as a larger TCR.  In the case of multiple resistors operating in a circuit, such as you noted, you will see the net sum total of all of the parameters interacting with each other including environmental.  Again, depending on the individual components, you may see a very small change in output or a larger change in output, it may not be linear either, it could be 'bumpy' and rest assured it will not be the same for each circuit with different components.

This is one of the reasons that we have stated that testing the components in the actual circuit under actual operating conditions will give you a much better idea of actual performance vs. taking a bunch of  measurements on a DVM (or resistor bridge for that matter) individually.  That will tell you very little about how they will perform as a group in a circuit.

Just to set the record straight again, PWW resistors do not have VC nor do they have 1/f noise.  BMF appear to have lower 1/f noise than the other film/foil types, this is based on published papers which I have not verified myself for the BMF resistors.  Like most resistor parameters, VC has a ± limit and any one individual resistor can be anywhere within that range, just like TCR or tolerance, it is unlikely that the VC parameter has an actual zero value though due to the physical effect that generate it. PWW resistors consistently have the lowest inherent noise level possible with the BMF appearing to be very close behind.

[Kleinstein]

That's always the big question - because changing the bias point can give you a VC change, but that also causes a temp change...which could make the end result better or worse, depending on actual TCR at that operating point and at your thermal flow.

But there is one defined measurement method way which actually has an interesting hidden use on Vishay datasheets.

Vishay measures VC using the old MIL-STD-202-309.

In a very condensed nutshell:

The VC is computed as VC = (R-r)100 / 0.9Er.

R= Resistance at continuous working voltage, r=resistance at 0.1 * Rated Working Voltage and E = Rated Continuous working voltage.

You apply 0.1 Rated working voltage and let resistor come to equilibrium in your circuit and grab a measure on "r"...then apply your continuous rated working voltage for no more than 1/2 second in any 5 second interval...Grab your "R" measure as quick as possible during that max 1/2 second pulse...as close to the beginning of the pulse as possible, before the majority of heating comes in.

Run the numbers and that's how Vishay measures VC.

The interesting part is that Vishay claims that there is no VC below 1k resistance value on their foils (which is not exactly true).  The reason they can get away with that is because MIL-STD-202-309 was only defined for resistors of value 1k and over...

Take that for what it's worth.  Vishay BMF parts are not necessarily bad, in fact they can be very good.  The problem is you always have to lift the veil on their datasheets.

That definition of the VC is odd: in the formula it kind of assumes the resistance would linearly change with voltage. However then they use the delay for getting a self heating effect and this would cause an effect proportional to V² instead. No wonder they get odd numbers.

If it is just self heating, there should be a simple connection to thermal resistance and TCR. 

[Edwin G. Pettis]

The VC parameter is not associated with heat, it is applied voltage, the effect is the electrostatic field of the applied voltage on the extremely thin resistive element, it 'tweaks' the element physically causing a change in resistance.  This is in addition to all of the other parameters which have an effect on the resistive element, what you see at the terminals is a sum of parameters on the element.  As shown, each parameter must be isolated in order to measure it accurately otherwise you're measuring multiple parameters at once.

[Andreas]

Hello,

some further measurements
this time PTF56 resistors 25K2 (5ppm/K in data sheet) =     
PTF5625K200BZEB / PTF25.2KDCT-ND  from DigiKey.

with the appropriate 10K counterparts and a little heater this could give a nice 7 -> 10V transfer.

some examples covering the whole range + the overview sheet

With best regards

Andreas

Edit: .csv added

[cellularmitosis]

Very interesting to see not only great TC performance, but also such a variance in the shape of the TC curve.  With a bit of binning, it looks like you could find resistors which had a "zero" TC point near room temp, or combine pairs of resistors which had opposing slopes.

[MisterDiodes]

It's not always all about TC - On those PTF56 / 65's watch out for that little part on the datasheet where load life drift is spec'd in percent, not ppm.  Even at a few mW power these tend to have the wobblies (and a bit more noise) over time.  It just depends on how stable a divider you need, and what noise level you're shooting for - You'll get what you pay for.

[MisterDiodes]

From a production standpoint, by the time you start paralleling wobbly resistors and that added cost of assembly and labor time in sorting resistors etc, it makes more and more sense to just use a much more stable PWW and get rid of "excess" 1/f noise in the first place - and wind up with a much more robust system. 

IF that's important in your application and cost target. Everyone's needs are different.   The way I look at it:  By the time you've bought 3 or 4 PTF56's you might just as well get a decent PWW anyway, if that works for whatever you're building.  That's usually the whole reason why you pay for better resistors for certain applications.

There are other places a PTF56/65 could be used where you need a bit lower TCR but long term drift and more noise could be tolerated, or maybe you need something sort of accurate but economical for AC, etc.  OR maybe you want a Vref where it can be calibrated often, and maybe it only has to be "low drift" for shorter periods of time, etc.

[cellularmitosis]

CAme across this: http://www.conservationphysics.org/satslt/satsalt.php

Time to start buying some salts!

A few days ago I hacked up two small mason jars to act as controlled humidity environments.  I put salt in one and magnesium chloride in the other, then poured in some hot water.  I rigged each lid with a Si7021 sensor (temp + humidity).

This evening I took some measurements:

Code: [Select]

Sensor #1 (NaCl): 82.9% humidity
Sensor #2 (MgCl2): 48.4% humidity

I then swapped the lids, waited about 45 minutes, and measured again.  Here are the results:

Code: [Select]

Sensor #2 (NaCl): 80.0% humidity
Sensor #1 (MgCl2): 51.1% humidity

It looks like this level of humidity regulation is "good enough" to start taking some resistor humidity coefficient measurements   

Edit: Note that this differs significantly from the values listed in literature (should be about 75% and 34%).  Perhaps my solutions are not "saturated".

[ManateeMafia]

From the different articles I have read, the salt mixtures should be a slurry. Perhaps too much water was added?

https://www.allaboutcircuits.com/projects/how-to-check-and-calibrate-a-humidity-sensor/
The article also includes code on how to correct for the sensor's error.

[Andreas]

Hello,

depending on where you are in the absolute humidity a temperature difference of 1 deg C can make up to 6%rH change.
(so perhaps self heating of the sensor ?).

I fear that rH coefficient measurements are a similar wide field as T.C. measurements.
So it would be worth a own thread.
Perhaps with links to some statements that Lars (I think he has most experience with humidity here) already did in the forum. E.g. this one:
https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg1379236/#msg1379236
And note that the rH time constants of epoxy are usually in the >3-7 day range so you should wait some days after each rH change.

with best regards

Andreas

[cellularmitosis]

Thanks Andreas, I’ll start a new thread.

[Andreas]

Hello,

again some PTF56 resistors 10K0 (5ppm/K in data sheet) =     
PTF5610K000BZEB / PTF10KDCT-ND  from DigiKey.
All values again like the 25K2 measured below 2 ppm/K as box T.C.
and there is virtually no hysteresis visible.

Pictures of some examples covering the whole range of T.C. + the overview sheet

With best regards

Andreas

Edit: .csv added

[zhtoor]

Hello,

again some PTF56 resistors 10K0 (5ppm/K in data sheet) =     
PTF5610K000BZEB / PTF10KDCT-ND  from DigiKey.
All values again like the 25K2 measured below 2 ppm/K as box T.C.
and there is virtually no hysteresis visible.

Pictures of some examples covering the whole range of T.C. + the overview sheet

With best regards

Andreas

great work. 

could you test one of these?
https://www.digikey.com/product-detail/en/yageo/MFP-25BRD52-10K/10KADCT-ND/2059114

best regards.

-zia

[Andreas]

Hello,

not in the next time
I have still some PWWs and Z203 on the stack.

On the other side: they will be most probably similar to the RC55Y (15 ppm/K) that I have already tested at the beginning of the thread.
Summary:
https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg462301/#msg462301

They turned to be around 8ppm/K and having a large drift over the cycles.
So the 25ppm/K will have around 12 ppm/K.

If you compare the datasheets of "normal" metal film and the PTF56 then you will find
- stability class 1.5% against 0.04% (load life)
- Special moisture protection for the PTF56

with best regards

Andreas

[rhb]

Andreas,

Would you please post a CSV file of the PTF56 data?  I'd like to do a statistical study of a voltage divider made from a random set of parts to see how predictable the divider is.

Thanks,
Reg

[Andreas]

Would you please post a CSV file of the PTF56 data? 

see above below the overview pictures