T.C. measurements on precision resistors

[Andreas]

Hello Alex,

are you shure that you are not below the dew point
at 16 deg C and below when doing the measurement.
Or why is the resistance first rising and then going down?

With best regards

Andreas

[Alex Nikitin]

Hello Alex,

are you shure that you are not below the dew point
at 16 deg C and below when doing the measurement.
Or why is the resistance first rising and then going down?

With best regards

Andreas

Essentially this change is below my ability to measure it accurately. I don't think this drop is due to a leakage. Here is the full graph going back to +23C as of this morning. This rise could be just a transient effect due to a reasonably fast cooling.

Cheers

Alex

[CalMachine]

I'm looking to get a small benchtop incubator/oven to start doing some TC measurements!  Anyone know of any, fairly decent, cheap units available on the market? 

[dr.diesel]

I'm looking to get a small benchtop incubator/oven to start doing some TC measurements!  Anyone know of any, fairly decent, cheap units available on the market? 

Not too hard to custom build.  If you end up buying something make sure and get a controller with some kind of IO, you can you script/program/automate the testing.

[Andreas]

>> For me its not clear: what effect do you mean with "end cycle hysteresis"?

"end cycle hysteresis" is the resistance change measured after one or more thermal cycles.

If you get the following example you can see that at beginning (step 2) resistance was 20011.70 ohm circa.
After the temperature going up and down two times resistance was 20011.39 ohm circa (step 47).
So the resistor has lost 15ppm after thermal cycling.

Ok, thanks for the explanation.

What I see in the diagram is
a) one part with immediate response to temperature
b) a effect in opposite direction with a much longer time constant.

the opposite direction could also be a smaller CTE of
one of the components that build up the resistor.

As you cannot change the readily built resistor
I would ask myself if I have actually such temperature risings in my end application.

In my lab I have measured temperature risings of up to around 2 deg C/hour.

With best regards

Andreas

[Alex Nikitin]

Hmm, it looks like either I have an exceptional example of 10G RX-1M, or these resistors are really that good. I've ramped up the temperature in the chamber up to 49C (measured on the resistor enclosure) and again, the resistance change is in order of 100ppm for over 25 degrees... . I have another sample of 10G resistor at work and also a couple of 100G of the same type - will measure these eventually. Right now I am measuring the same 10G cooling down, but with a different method - using the I/V mode on the Keithley 617. The difference at 20V supplied is very little comparing with the normal (constant current) method but the cable leakages are less of a problem.

Cheers

Alex

[mimmus78]

Andreas,

I measured this "overshoot" also with 5° K difference but it was really marginal.

I think with normal natural temperature cycling this overshoot has almost not effect on our LTZ circuit and because of the more slow temperature changes it almost do not reveal.

What I think is that this will also have some impact on long time stability of the resistance. But this will takes many years to prove it.

This effect, the hysteresis due to thermal cycling and maybe drift due to humidity  suggest me that if you want to reduce drift you must keep resistors always at the same temperature. The hole circuit must be in a hoven.

Now I'm checking what is effect of humidity on the resistance of a UPW50. I'm drying a resistors with some "medical" grade desecant. Actually registered -20ppm in two days.

Tomorrow will be third day and I will fire up my MF calibrator to countercheck the resistance readings, but I'm sure the drift is there.

My plan is to leave the resistor to dry for one week, check the "drift" and than countercheck if it creeps back as it will reaccumulate the humidity and in how much days.

[mimmus78]

@Alex Nikitin

That's very odd to me, never expected this to be the case.
From charts it seems you are measuring stability of the Keithley 617 and not of resistor :-)

[Alex Nikitin]

@Alex Nikitin

That's very odd to me, never expected this to be the case.
From charts it seems you are measuring stability of the Keithley 617 and not of resistor :-)

Yes, the accuracy and stability of the electrometer are limiting factors. The spec on the K617 at 20G range is 1.5%  . Nevertheless, the temperature in my home lab is reasonably stable, so it is mostly the noise and the resolution of the Keithley. Here are two more graphs - a full 22C-10C-50C-25C cycle today and the R(T) curve from that cycle that is reasonably fits a parabola after all. There are obviously some transient effects (the resistor is coupled to the enclosure only by its leads connected to two BNC sockets).

Cheers

Alex

[Alex Nikitin]

And here are the graphs for the 100G resistor of the same Ohmite RX-1M series. A very different behaviour, essentially a linear -230ppm/C tempco. Vertical scale 0.1%/div . I may re-measure this resistor in the V/I mode of the Keithley 617 just to confirm this data and check if there is any obvious leakage influence.

Cheers

Alex

P.S. I am considering now if I should build a 100G reference resistor from ten 10G ones. The price however, would be very high - £600 (£500 + VAT) for the resistors only!

[mimmus78]

Well this last chart from the 100G has much more sense. :-)

[Alex Nikitin]

The results for the second sample of Ohmite RX-1M 10G resistor. A different tempco shape - almost linear - but still very low, about 12ppm/C.

Cheers

Alex

[mimmus78]

I run another experiment to check "overshoot effect" with 40°K temperature difference and a 10K UPW50 resistor.
Experiment than proceed with small steps of 5°K (up) and 2°K step (down).

I still have doubts that this overshoot are coming from my setup so I set a thermistor in contact with one of the two resistor leads near the case of the resistor. I wanted to check that lead temperature is similar to the one of the DUT.
The thermistor is very small NTC that I got from a body thermometer. To keep minimal thermal mass, resistor was soldered with very thin wrap wire. Resistance of this thermistor was measured by the K2001.

I scaled thermistor resistance reading with following equation Y=75-(R/1000) to have it charted on the graph on the same scale of the temperature.

As usual the green line is a digital temperature sensor that is place by side of the DUT resistor.
Yellow dotted line is the scaled NTC resistance.
Blue is the DUT resistance.
The other two dotted lines are 3458a temperature and ambient temperature.
Resistance is measured by the 3458a with 4W, 30 NPLC, offset compensation on and delay of 0.1 s.

RESULTS:

- "overshoot effect" with dt of 40°K is near 6ppm
- "overshoot effect" with dt of 5°K is 1ppm circa and visible in every step even at 2°K
- "overshoot effect" is not visible on the resistance of the NTC
- resistance of the NTC really track very well temperature measured by the digital temperature sensor

I attach three charts from this experiment.

 - first image is the overall experiment
 - second one is the zoomed in part when the DUT reach the fist 40°K temperature increase
 - third is resistance vs temperature

[Andreas]

Hello,

got some goodies ...

will have to do some T.C. measurements ...

with best regards

Andreas

[mimmus78]

Ohhh nice ... Courius to know how they act :-)

Inviato dal mio Nexus 6P utilizzando Tapatalk

[Andreas]

Hello,

T.C. according to data sheet:
+/- 0.2 ppm/K typ +/- 2 ppm/K spread.
hysteresis not specified.
http://www.vishaypg.com/doc?63120

First sample:
VHP202 1K #1  Datecode B1629-  top marking (hand written) "S179"

17.02.2017 Measurement from 11 deg C to 43 deg C (25 deg C -14/+18 deg C)
near linear T.C. of -2.5 ppm/K
hysteresis about +/- 1ppm (or 2 ppm span)
The hysteresis is related to the hot phase of the resistor so the resistor does not like the 43 deg C temperature.
The hysteresis recovers "over night" so a measurement at the following day shows the same behaviour.

18.02.2017 reduced temperature range of around 25 deg C +/-7 deg C
near linear T.C. of -2.5 ppm/K
hysteresis within noise (below 0.5 ppm)

The T.C. is somewhat at the upper limit of the datasheet and far away from the "typical" value.
And this resistor should be only used either with very slow temperature gradients or not be heated above 32 deg C.

The second 1K resistor seems to be better in T.C. (results will follow ...)

with best regards

Andreas

[mimmus78]

Anyway this little histeresys (if always recovered) is almost ininfluent within a ltz1000 circuit.

Inviato dal mio Nexus 6P utilizzando Tapatalk

[Andreas]

Hello,

second sample:
VHP202 1K #2  Datecode B1629-  top marking (hand written) "S164"

23.02.2017 Measurement from 13 deg C to 43 deg C (25 deg C -12/+18 deg C)
near linear T.C. of -0.24 ppm/K
hysteresis about +/- 0.8 ppm (or 1.5 ppm span)
The hysteresis is related to the hot phase of the resistor so the resistor again does not like the 43 deg C temperature.
The hysteresis recovers "over night" so a measurement at the following day shows the same behaviour.

24.02.2017 reduced temperature range of around 25 deg C +/-7 deg C
near linear T.C. of -0.24 ppm/K
hysteresis within noise (below 0.5 ppm)

So this sample is much nearer to the "typical" spec of +/- 0.2ppm/K

with best regards

Andreas

[Andreas]

Anyway this little histeresys (if always recovered) is almost ininfluent within a ltz1000 circuit.

Hello,

I wanted to have a humidity insensitive "standard" resistor.
For that the T.C. of the first sample is much too high.
And I have also to take care that the temperature change is not too fast for the 2nd sample.

with best regards

Andreas

[babysitter]

Hi Andreas,

ebay recently presented a image of three old, scavenged Ultrohm Resistors (pay attention to the missing plus, so Edwin might has been a bit younger in that era, how much is Ultrohm plus linked to this older shop?), and at less than 5€ I gave them a try. The absolute value (just measured with 2 wire mode even!) seems to be okayish after all these years, I wonder If you would like to give those well-seasoned, funny-shaped bobbins a try on your T.C. measurement setup.

Eye candy attached.

[Edwin G. Pettis]

These are old dinosaurs, for awhile Ultronix used the 'Ultrohm' term on some of their resistors, it was not in use during the time I worked at Ultronix and I do not recall ever seeing a resistor form like these made during my tenure  there.  They probably date from the 1960s, there was one other product line (which to my knowledge) that never became commercial which bore the Ultrohm name on it.  Ultrohm never trade marked the name so it was available, hence my line became the Ultrohm Plus, it bears no design or relationship to the old Ultronix 'Ultrohm'.  I don't see a tolerance on the resistors nor any tolerance code so what tolerance these have is anybody's guess.  These are molded in the same Ultronix blue alkyd that was used on all of their molded resistors.  I have quite a few of them still laying around, very few of them are still in tolerance after all these years, not to be unexpected as molded resistors will continue to drift with age never becoming completely stable.

[Mickle T.]

... I don't see a tolerance on the resistors nor any tolerance code so what tolerance these have is anybody's guess. 

Standard tolerance code "B" = 0.1%

[Edwin G. Pettis]

Oh yeah, I haven't used those codes in years for the under 1 % ranges, forgot about them, we didn't use the letter codes at Ultronix all that much even for the military parts we made.  With a ±0.1%, they could still be within tolerance if they weren't too close to the limits to begin with.

[babysitter]

@Mickle: Plausible!

@Edwin: Thank you for the historic info and the guesstimation of their age. So they are just loosely related.

What really catched my eye was the unusual shape, actually. like a hybrid of "ye olde open-air wire-on-erected-rod" and "plastic over all things!".

With a little side of "certainly older than me".

Now that the kids are sleeping I found some time to hook up 4 wires instead of just two.

My 3456A in 4 ohms OC mode, ambient around 20°C measured:
specimen "362R0B" at 362.028 ohm, "390R0B" at 390.138 ohm, "144R0B" at 143.987 ohm.
I'd say they aged in dignity!

[Edwin G. Pettis]

Just how big are these, looks like the diameter is over 0.5" and around an inch long?  That would indicate some rather large diameter wire in there, probably why they haven't drifted too much over time, molding wasn't quite as nasty to larger wire sizes.  Most of the old Ultronix resistors I have are somewhat tighter in tolerance and much smaller in size.