T.C. measurements on precision resistors

[Overspeed]

I guess what seems to be visible as drift is the thermal time lag between copper and resistor. Attached is the final temperature profile, followed by an 8 h stability measurement at lab temperature. As can be seen it takes about 5.5 ... 6 h for the whole thing to equalize.
This time could be reduced by improving the thermal transfer inside the case. I was thinking about filling the case with ceramic balls (3-4 mm in diameter) or ceramic pouder, thermally conductive but electrically insulating.

Edit: Added the initial thermal resistor behaviour for comparison.

-branadic-

Hello
Filling the enclosure with any product will increase the time of temperature equilibrium between the temperature chamber and the part temperature .
In mechanical metrology , parts are let 24H in the lab with controlled temp before measurement time to get stable temp , that the same for chemical testing on energetic material , propellant are let 24H before testing .

Can be interesting to run differential temp measurement with one sensor on the resistor itself ( drill the outside tube ) and the chamber temp

Regards
OS

[branadic]

While verowire (0.2 mm in diameter) perfectly worked for the 100 Ω and 1 kΩ, it didn't for the 10 kΩ. The wire length needed was over 1 m and what I think happened was that this long wire, while wound bifilar around the resistor, introduced a second order term due to thermal expansion. I've now replaced it by 60 µm enamelled copper wire which resulted in way lower wire length and the residual t.c. is now slightly positive and almost linear. I will give it some additional trimming and call it a day afterwards.

-branadic-

[Overspeed]

While verowire (0.2 mm in diameter) perfectly worked for the 100 Ω and 1 kΩ, it didn't for the 10 kΩ. The wire length needed was over 1 m and what I think happened was that this long wire, while wound bifilar around the resistor, introduced a second order term due to thermal expansion. I've now replaced it by 60 µm enamelled copper wire which resulted in way lower wire length and the residual t.c. is now slightly positive and almost linear. I will give it some additional trimming and call it a day afterwards.

-branadic-

Hello
if you wind ( coil winding ) at the temperature reference , your wire dilation shall be negligible IF if the winding core is stable ( no expansion ) and IF you don't strain the wire by too much tension .

Does someone have tried loose winding in epoxy cast ? so a no stress / strain on the wire .

Second parameter is the self heating generate by the electricity power that an interesting as the 10 mW power calibration ( I suppose limited in the aim to limit self heating )  is not a universal value as that directly linked to the size/dims of the tested resistor a S102 is not at all a SPR221 regarding the power which can be handle. I agree that not wire resistor but that an example ratio power vs size .

Some studies demonstrate that resistance value change ( in the last digit ) with current value , a 1 Ohm resistor is a good candidate to check values with 1 mA / 10 mA .

Regards
OS

[branadic]

After trimming, here is what I've ended up with. I'll leave it over here.

-branadic-

[The_Dall]

Hi everyone,

I wanted to share my TCR (Temperature Coefficient of Resistance) measurements of 34 different resistors, ranging from precision types to general purpose ones.

Test Setup & Equipment:
I used a pair of temperature calibrators (RTC-157B and ATC-156B) as a very stable thermostat. The resistance was measured using two Russian precision temperature meters — MIT 8.15 and MIT 8.30 (which are comparable to a Fluke 1595A in terms of measurement capability) and an HP 34401A.

To ensure good thermal contact and avoid lead resistance errors, I had a custom fixture machined from duralumin. The DUT is placed inside this block, which then goes into the calibrator chamber. Connection to the MIT instrument is made via a 4-wire Lemo 4-pin connector.

The calibrator was programmed to step through the following temperature points: 18°C → 23°C → 28°C → 23°C(Sometimes 23°C → 18°C ​​→ 28°C → 23°C). At each point, the temperature was held for 30 minutes to allow the resistor to fully stabilize.

I cannot test standard resistors like the R3030 or Fluke 742A with this setup simply because they are too large to fit inside the calibrator chamber — this method is only suitable for individual through-hole or SMD resistors.

In the Excel file, you may come across a label like "Vishay Solartron". This is a resistor that I salvaged from a Solartron instrument, and since I don't know its exact part number, I named it that way for identification.

https://docs.google.com/spreadsheets/d/1s8FnWospWLVpeyA4uamXXwtao5fr2WR0/edit?usp=sharing&ouid=115112181287416241546&rtpof=true&sd=true

[TUMEMBER]

Hi everyone,

I wanted to share my TCR (Temperature Coefficient of Resistance) measurements of 34 different resistors, ranging from precision types to general purpose ones.

Test Setup & Equipment:
I used a pair of temperature calibrators (RTC-157B and ATC-156B) as a very stable thermostat. The resistance was measured using two Russian precision temperature meters — MIT 8.15 and MIT 8.30 (which are comparable to a Fluke 1595A in terms of measurement capability) and an HP 34401A.

A lot of work.
Have you checked the temperature stability of your oven? Just because the display shows XX,XXX and it's stable doesn't mean the temperature is actually stable. I doubt the "oven" maintains it within +/- 0.01 degrees, let alone 0.001 degrees, as shown on the display. Temperature fluctuations certainly affect the stability of the resistance readings for resistors around 100 ohms, as I assume the resolution of the resistance meter reveals this (it's sensitive enough in the 100 ohm range). I didn't notice you using a temperature sensor in the resistor channel, so there's no reference to the actual instantaneous temperature.
I'm attaching screenshots of temperature stability measurements of technologically similar Fluke ovens. They tend to go crazy around 23 degrees Celsius – they don't know whether to heat or cool – the Peltier elements go into a pseudo-stable state.
https://www.fluke.com/pl-pl/produkt/przyrzady-do-kalibracji/kalibratory-temperatury/fluke-calibration-9171

[The_Dall]

I actually have stability data specifically for the ATC-156B, measured using an MIT 8.30 precision temperature meter and a reference platinum resistance thermometer (PTSV-1-2, rugged design, with a nominal resistance of 100 Ω at the triple point of water). Over a 30‑minute period, I observed a peak‑to‑peak temperature variation of only 0.007 °C.

I also tested the RTC-157B earlier, but unfortunately I didn't save the logs. However, I can say it performs even better — peak‑to‑peak stability was around 0.002 °C or less. The RTC‑157B is a newer model, and its ADC is capable of resolving such tiny temperature fluctuations down to one thousandth of a degree Celsius. Of course, the manufacturer's specifications are conservative; the numbers I've given are what I actually measured.

I'm attaching a screenshot showing the temperature stability of the ATC-156B.

I highly recommend downloading the Excel file, because the online preview does not display the graphs correctly, and the resistor photos are missing.

[TUMEMBER]

Thanks for the reply. This clarifies some of the "uncertainties." It's good to know that someone knows that what's shown on the screen doesn't always fully match what's actually being measured.

[The_Dall]

My detailed measurement process:

I carefully soldered each DUT to the wires going to the Lemo connector, then cleaned the solder joints to remove any flux residue. After soldering, I let the resistors sit for 3 days to allow them to recover from thermal and mechanical stress. I prepared them in batches of five.

Before starting any measurements, all instruments were powered on for 2 hours to stabilize. The MIT meters were logging data via their dedicated software throughout the entire process. The temperature calibrator was programmed to automatically step through the setpoints, so my only tasks were to process the data afterward and install the next resistor for testing.

For each temperature point, I took the average resistance value during the stable period (after the 30-minute soak) and entered it into the Excel table, where the TCR is calculated automatically.

I followed the calculation method described in the Fluke 742A Series Instruction Manual, which also uses the 18–28°C temperature range. This seemed like a well-established reference, so I adopted the same approach.

About the MIT meters:
The MIT precision temperature meters have two important features:
Thermoelectric EMF compensation and built-in constants that allow them to calculate how their internal references change with ambient temperature and apply real-time corrections automatically.

Why I didn't measure the actual temperature continuously:
I decided to rely on the calibrator's setpoint rather than measuring the temperature inside the block for every resistor. My reasoning is simple: I doubt any meter could resolve a resistance change caused by a 0.007°C temperature variation when measuring a precision resistor. This is clearly visible when measuring something like a VHP101T — the graph shows mostly instrument noise, not actual instability of the DUT. So I use a single temperature value for calculations, trusting the calibrator's stability data I provided earlier.

You may notice two sheets in the Excel file: "100 Ом EE (HP)" and "100 Ом EE". The "(HP)" suffix means I used the HP34401A for those measurements. The average result was 2.34 ppm with the HP versus 1.99 ppm with the MIT 8.30 — a difference of about 0.35 ppm.

The reason I did this comparison is that the MIT 8.30 has a 10 kΩ measurement limit, so for any resistor above 10 kΩ I'll have to use the HP34401A. I wanted to understand the offset between the two instruments.

I plan to continue testing various resistors and will update the table as new data comes in.

[Kleinstein]

For resistor TC measurements (except temperature sensors) the critical part is the resistance measurement at the DUT and not the temperature measurement. The temperature measurement system looks really good, but the limiting case is more the 34401 to measure the DUT.  It is only with high TC parts like temperature sensors that the temperature part really matters.
With temperature that can be related effects (humidity internal stress) that can have an effect with delay and hysteresis. For this it makes sense to look at more than just the resistance a few stable temperatures.

[The_Dall]

Thanks for the valuable input — you've raised some important points.
I understand that besides TCR, there are other effects like hysteresis, humidity influence, and long-term drift. In my tests, I specifically checked the return to 23°C and saw no hysteresis within the 2.5-hour cycle, so I didn't include those intermediate 23°C readings in the graphs — they would have just overlapped with the initial ones, adding clutter without new information. For deeper analysis, longer tests and humidity control would be needed — that's the next level. Thanks for the insight!

[The_Dall]

I've updated the table with 12 more resistors — the red cells on the side mark the newly tested ones.

There are two new leaders: a 10kΩ AE XT and a 20kΩ S102K. This was quite unexpected — very interesting results!

I also remeasured the 10kΩ VHP 101T to compare the DMM7510 against the MIT 8.30. There was no difference between them. All new measurements were done with the DMM7510 in Offset Compensation mode.

https://docs.google.com/spreadsheets/d/1s8FnWospWLVpeyA4uamXXwtao5fr2WR0/edit?usp=drive_link&ouid=115112181287416241546&rtpof=true&sd=true

[Andreas]

Hello,

on my measurements the S102 types had a large hysteresis. (+/-5 ppm or 10 ppm total)
But hysteresis is also dependant (some say quadratically) by the temperature excursion range.
https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg462303/#msg462303


Since in my "lab" I have typically from 18-33 deg C I usually try to test from 10-40 deg C to have some margin.

with best regards

Andreas

[The_Dall]

I see we're approaching this from slightly different angles — you're testing over a wider temperature range (10–40°C) and looking at hysteresis, while my setup is limited to 18–28°C, more from the perspective of treating these resistors as "resistance standards" within a narrow window.

I haven't observed significant hysteresis in my tests, but that's likely because my temperature excursion is much smaller.

[Andreas]

Hello,

it also has to do with relaxation time for the whole cycle.

You may see lesser hysteresis with shorter cycles.

With best regards

Andreas

[TimFox]

Bulk foil resistors (e.g. S102) achieve very low tempco by cancellation of resistivity tempco with mechanical tempco of the substrate.  The former is instantaneous, but the latter is slower, producing hysteresis of the actual resistance.

[dietert1]

The closer you look the more you will discover. One example of what you may see when looking at the 0.01 ppm level was reported here: https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg2751598/#msg2751598.

I remember a discussion on the glue used to fix the foil to the substrate. Those are not metal film resistors!

Regards, Dieter