T.C. measurements on precision resistors

[MisterDiodes]

You can measure resistors all you want OUT OF CIRCUIT, free hanging in air and maybe come up with similar results.  I don't care if 2 or 10 or 1000 people measure resistors inaccurately and come up with matching inaccurate results, and I especially don't care when basically none of the measures is verified by an absolute, calibrated measure anywhere.  I know all the arguments about "relative measures".  I understand that, and I understand what Andreas is trying to do on a low budget.  That's not the point here, don't get me wrong - I just don't see where "free hanging" resistor tests really show how anything how resistors work in a real circuit - especially on a ratio pair that you are trying to magically extract TC measures down to sub ppm.

The entirety of any precision circuit - especially at PPM levels - is taken into consideration when you're at lower PPM.  Always, and especially since we're in the metrology section.  You must think of the entire circuit as a thermal SYSTEM  first as TiN pointed out.  With resistors and circuits like LTZ's it's all about achieving stable POWER OUT vs POWER IN, and ONLY when power flow is stable do you get as a by product a stable Vref output or stable resistance measure. Unless you have a stable thermal / power situation, your "Free Hang" resistor measure is just not very meaningful in a practical sense, no matter how its done.

My point is:  Especially on ratio resistor pairs - you always measure those pairs AT BIAS and MOUNTED exactly like they will be in final product.  That's because you NEVER know how --each-- resistor in the set transfers power back into the board as heat (which sets the apparent operating point temperature of each resistor), at typical OPERATING BIAS and operating THERMAL condition.  Does each individual resistor in the ratio set transfer heat into the board exactly like it does when it's free hanging?  Usually not - far from it especially for an unbalanced pair like 13k over 1k.  I don't care if you're using a Vishay Foils or PWW or Film or Diffused pair.  The final operating point of EACH resistor in the set matters (sometimes a LOT) when you're reporting TC ratio numbers down into .01ppm.

Even if you see some sort of TC mismatch on your free hanging, unbiased resistor pair test, sometimes you find that once the resistors are actually mounted in position they are more thermally matched and stay at a closer matched internal temperature, and now your ratio TC looks better than the free hang test (or TC even changes direction).  Sometimes you see the opposite, for instance when one resistor in the ratio pair is connected to a cooler ground or power trace than its mate, or maybe the position on the board doesn't allow for good thermal coupling, or maybe there is a fan blowing nearby pulling more heat away from one resistor of the pair.  Now the mounted resistor ratio TC doesn't look ANYTHING like the free-hanging test showed.  This stuff happens all the time in real life when you're exploring the low PPM world.

That's why I suggested: The one and only true REALISTIC and PRACTICAL test of an LTZ heater ratio set is to just run them on whatever LTZ board you've designed, soldered to your traces, mounted in your enclosure, running at equilibrium at your typical operating temperature.  That's the only accurate way to really know what you've got for resistor ratio performance.  At least you know what parameters actually matter in the final evaluation of the Vref.

The next question is - when you build an LTZ with resistor set X or resistor set Y - take a really hard look at your own ability of measuring an absolute voltage value, and see if spending a lot more $$$ on a resistor set will make any difference in the end.  That's all up to what you really need and your lab equipment list - but a lot of times: spending more money on resistors won't make any difference.

Usually what happens is you'll wind up building a Vref that's better than what you can ever really measure on your own equipment (unless you have multiple 732's or similar), or sometimes you realize how noisy that eBay DMM really is @ 7.2V

Have fun!

[Andreas]

Hello,

maybe I will have to improve my ratio measurement setup
since it is the first time that I am using it.

From the professional volt-nuts (like you) we are getting not much
help. Neither a manufacturer measuring setup photo nor any
measurements. You are welcome to contribute with your own
setup and some measurements.

It is not very helpful for me in your posts is that you are
shooting at a broad range of things without analysing
what the most likely error is.

But I also now that many professionals are doing things
from routine (we have done that always this way)
without knowing the background.

Or do you really think that a 200 mW resistor loaded with
constant 2mW (3.5 in the LTZ-cirquit) will have a significant
different behaviour than in final cirquit?
(What I have not shown is that the "free-hanging resistor" is
mounted afterwards in a aluminium case
and additionally shielded against air drafts).

I also really doubt that Jason has been asked by the manufacturer
if he wants to use his divider together with a 7.1V or a 7.2 zener.

If this is really relevant then the so called matched resistor pairs
would have to be considered as "marketing gag" to optimize profit
without any benefit for the customer.

with best regards

Andreas

[Andreas]

Hello,

some further ratio measurements on VHD200 12K5/1K resistors datecode 1804

#2 (2 dots) from 26.02.2018
-0.39 ppm/K linear regression coefficient for the ratio

unfortunately I have killed my ADC#26 by a short cirquit during mounting in the aluminium case.
(I really should not do that with a life system since at the connector there are high voltages (14V) and some processor pins too).
So I had to change to ADC#08 for the further measurements.

#3 (3 dots) from 01.03.2018
-0.57 ppm/K linear regression coefficient for the ratio

#3 (3 dots) reverse mounted (with the 1K resistor towards VRef and the 12K5 resistor to ground) from 02.03.2018
-0.71 ppm/K linear regression coefficient for the ratio (normalized to 12.5/1)

with best regards

Andreas

[Andreas]

Hello,

here the results of the VHD200 1K resistor parts in my standard measurement cirquit with buffered quasi differential 4 wire measurement:

As all results are relatively curvy especially below 25 deg C I have additionally calculated
the average TC from 25 to 35 deg C out of the LMS curve to be comparable to Jasons measurement.

VHD200#1 1K from 02.03.2018

Box: 0.60 ppm/K  (including noise)
LMS: -0.47 ppm/K @ 25 deg C
Box LMS 0.56 ppm/K (without noise)
Average 25-35 deg C: -0.73 ppm/K

VHD200#2 1K from 03.03.2018

Box: 0.41 ppm/K  (including noise)
LMS: -0.09 ppm/K @ 25 deg C
Box LMS 0.31 ppm/K (without noise)
Average 25-35 deg C: -0.37 ppm/K

VHD200#3 1K from 04.03.2018

Box: 0.57 ppm/K  (including noise)
LMS: -0.36 ppm/K @ 25 deg C
Box LMS 0.51 ppm/K (without noise)
Average 25-35 deg C: -0.63 ppm/K

Yesterday I have exchanged the 1K Z201#3 resistor against the 12K5 Z201#1 resistor in the heated reference box.
Today the first 12K5 measurement of the VHD200 resistors started.

with best regards

Andreas

[cellularmitosis]

#3 (3 dots) from 01.03.2018
-0.57 ppm/K linear regression coefficient for the ratio

#3 (3 dots) reverse mounted (with the 1K resistor towards VRef and the 12K5 resistor to ground) from 02.03.2018
-0.71 ppm/K linear regression coefficient for the ratio (normalized to 12.5/1)

I am surprised to see such a difference from reversing the leads!  On the other hand, the 1k resistor in that divider is the only resistors which seemed to behave very strangely (the shape of its T.C. response was different than the other resistors, see the first two graphs in https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg1434076/#msg1434076 ).

----

Not strictly T.C. related, but I think this is the appropriate thread:

Does anyone have any information on the ageing drift of typical 1% metal film resistors?

It would be important to know, if for example you were designing an Kelvin-Varley divider -- at which decade can I get away with dropping to 1% metal film?

[Andreas]

Hello Jason,

I am also surprised. But cannot answer this immediately.

Does anyone have any information on the ageing drift of typical 1% metal film resistors?

This is a wide field:
- cheap metal film resistor? (usually thick film)
- epoxy passivated metal film resistor? (hope that this is thin film)
- glass passivated thin film resistor?

- load life / shelf life conditions?
- humidity (oxidation) at your location.

the better resistors (PTF56, RC55Y) usually have some "typical" values in their data sheet.

with best regards

Andreas

[cellularmitosis]

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

Time to start buying some salts!

[Andreas]

Some further measurements on the VHD200 resistors. (12K5 part of the resistors)

The values are in the attached overview table.

I made up another measurement setup for the VHD ratio measurements as I have seen that the voltage ratio mode of the HP34401A is similar stable as my ADCs.
I use a 2*3-Wire configuration.
A stable 10V reference connects directly to the resistors.
The resistors are kelvin sensed between 0V and 10V (Sense inputs) and 0V and divider out (voltage inputs).
HP34401 is fixed in 10 V range.
Unfortunately I have forgotten to switch the 10 Gig input impedance on the first measurements (already in the table) so do not take these values as too serious. (corrected values will follow).

with best regards

Andreas

[Andreas]

Time to start buying some salts!

I do not think that you want to have that nasty stuff (creeping almost everywhere) near of your precision equipment.
with best regards

Andreas

[cellularmitosis]

I do not think that you want to have that nasty stuff (creeping almost everywhere) near of your precision equipment.

Hmm, good point.

However, I see two items on that list which are relatively safe.  Magnesium chloride is an ingredient in baby formula, and Sodium chloride tastes great on french fries.  Those two would give me 33% and 75% relative humidity, which should be enough to detect humidity sensitivity in resistors.

[cellularmitosis]

As I mentioned in the cal club thread, I got lucky on ebay with a GenRad 1434 decade box which was in a bit of poor condition.  I have decided to sacrifice it, turning it into either some traveling transfer standards for the cal club, or standards which will stay at home and detect drift and shipping-induced hysteresis of other cal club transfer standards (possibly some Ultrohms!).

I made a few quick measurements of three of these resistors, just to get a ball-park idea of their tempco.  The first three 2k resistors showed -5.8, -6.3, and -5.6ppm/C from 25C to 35C.

Because the tempco's were negative, I decided to try my hand at a trick mentioned by zlymex: making copper-wire tempco-compensation resistors: https://www.eevblog.com/forum/metrology/spread-sheet-aided-design-of-compensation-for-7v-to-10v-step-up-resistor-set/msg895856/#msg895856

The results are encouraging!  But also highlight some issues:

- There is a visible thermal lag between the main resistor and the copper-wire resistor.

You can see that when the temperature rises, the resistance initially increases, as the copper-wire resistor responds more quickly to the changing temperature.  Then, the GenRad resistor catches up, pulling the resistance back down to essentially where it started.

- My measurement system in its current configuration has about +/-1ppm of short-term drift.  That's no longer good enough for this sort of work.

(I have a few improvements to this setup which I'm working on right now).

Still, this looks promising for applications like this, where the goal is an isothermal environment to begin with.

Edit: a design using this approach will of course have to account for the resistance of the copper wire before the primary resistor is made.  In this case, its value should be reduced by about 0.14% to account for the copper resistance.

Edit 2: the copper-wire resistor is about 31 inches of 40AWG "magnet" wire, wrapped (bifilar) around a 10meg 1/4 watt resistor (which is just used as a cheap bobbin).

Edit 3: including screenshot of formula for required copper R value.

[texaspyro]

Try some tubing around the copper resistor to provide some thermal lag to the outside world.  You might want to try different types of tubing, shrunk/unshrunk, etc.  Maybe in contact with the resistor.

[Echo88]

Nice copper-compensation-experiments, need to try this myself after ive characterised a few of mine. But i cant really see the thermal-lag-effect you mention. 

[cellularmitosis]

But i cant really see the thermal-lag-effect you mention. 

zooming on the Y-axis helps make it more visible.

[Vgkid]

Interesting , but inside of an enclosure you won't (or shouldn't) see rapid temperature shifts. Especially if you thermally lag the box with insulation inside.

[amspire]

Interesting , but inside of an enclosure you won't (or shouldn't) see rapid temperature shifts. Especially if you thermally lag the box with insulation inside.

What you can get is effects from different amounts of self heating when a current is going through a resistor. If you have a 1ppm/C resistor, and you put enough current through it to raise it by one degree, you get a 1ppm shift. When you do the same to a -5ppm/C resistor and a 5ppm/C copper compensating wire, you can get up to 5ppm shift with the same current. Even if the resistor and the wire are in thermal contact, there will still be a heat gradient from the resistor to the wire.

[Andreas]

Hmm,

wouldnt it be better to wind the copper to the other windings on the mica card instead of a extra bobin?

with best regards

Andreas

[cellularmitosis]

Andreas, of course you are correct . That didn’t occur to me until after I had finished

There might be some value in have a set of standalone copper resistors of various values, so that you could combine them for rapid prototyping, perform a verification, tweak as needed, then cut a final permanent length of copper wirre and wrap around the original bobbin.

[Echo88]

Doh, didnt see the second attached graph, but it was quite late when i was looking at your post. 

[TiN]

Here's data of TCR test of Caddock resistors from Inverted18650.

Resistor	Value	Tempco spec	Number	Measured tempco	Datalog chart
Caddock TF020R	1 kΩ	+/-5 ppm/°C	12	-0.96 ppm/K
Caddock TF020R	9 kΩ	+/-5 ppm/°C	9	+2.18 ppm/K
Caddock TF020R	20 kΩ	+/-5 ppm/°C	7	-2.26 ppm/K
Caddock TF020R	5 kΩ	+/-5 ppm/°C	10	+2.99 ppm/K
Caddock TF020R	100 kΩ	+/-5 ppm/°C	1A	+1.71 ppm/K
Caddock TF020R	18 kΩ	+/-5 ppm/°C	8	+3.59 ppm/K
Caddock TF020R	200 kΩ	+/-5 ppm/°C	2	+3.20 ppm/K
Caddock TF020R	24 kΩ	+/-5 ppm/°C	5	-1.28 ppm/K
Caddock TF020R	100 kΩ	+/-5 ppm/°C	1E	+0.74 ppm/K
Caddock TF020R	22 kΩ	+/-5 ppm/°C	6	+3.80 ppm/K
Caddock TF020R	100 kΩ	+/-5 ppm/°C	1D	+2.75 ppm/K
Caddock TF020R	500 kΩ	+/-5 ppm/°C	3	-2.03 ppm/K
Caddock TF020R	25 kΩ	+/-5 ppm/°C	4B	-1.06 ppm/K
Caddock TF020R	100 kΩ	+/-5 ppm/°C	1B	+3.69 ppm/K
Caddock TF020R	25 kΩ	+/-5 ppm/°C	4A	-0.58 ppm/K

[cellularmitosis]

My computer seems to be allergic to those SVG files.  Here are screenshots in case anyone else suffers from the same problem.

[TiN]

Last 4 resistors tested (with one combined to check idea of action in two opposite TCR resistors in series).
Post with photos and numbers

Live SVG and CSV for each run T1,T2,T3,T4,T5

[Andreas]

Some further measurements of Jasons VHP202Z resistors 9K985 datecode 1739

#1 + #2 are a bit drifty during warm phase. #3 is stable.

with best regards

Andreas

[cellularmitosis]

(with one combined to check idea of action in two opposite TCR resistors in series).

Cool!

I whipped up a little combinatorics in python to look for tempco-cancelling pairs of resistors.  I thought it would be useful among the forum members, so here is the script and the results (attached).

Here's an excerpt from the results for pairs of resistors:

Code: [Select]

1A + 1C = 0.02 ppm/C
9 + 4B = 0.03 ppm/C
10 + 4B = -0.12 ppm/C
10 + 4A = 0.17 ppm/C
1E + 7 = 0.24 ppm/C
9 + 4A = 0.28 ppm/C
9 + 5 = -0.34 ppm/C
1E + 5 = 0.35 ppm/C
2 + 3 = -0.39 ppm/C
11A + 4A = -0.42 ppm/C
12 + 4A = -0.42 ppm/C
11B + 4A = -0.43 ppm/C
1E + 4B = 0.44 ppm/C
1E + 1C = -0.47 ppm/C
8 + 7 = 0.51 ppm/C

Here's what the script thought 9 + 7 should be, so I have hope that my math is correct!

Code: [Select]

9 + 7 = -0.88 ppm/C

Here's an excerpt from the results for groups of three resistors:

Code: [Select]

9 + 12 + 4B = 0.00 ppm/C
2 + 1D + 3 = 0.01 ppm/C
9 + 11B + 4B = 0.01 ppm/C
1E + 5 + 7 = -0.01 ppm/C
8 + 4B + 7 = 0.01 ppm/C
1A + 12 + 1C = 0.02 ppm/C
1A + 11B + 1C = 0.02 ppm/C
1A + 11A + 1C = 0.02 ppm/C
9 + 11A + 4B = 0.02 ppm/C
1A + 4A + 1C = -0.03 ppm/C
6 + 1E + 1C = -0.04 ppm/C
1A + 4B + 1C = -0.06 ppm/C
1E + 4B + 7 = 0.07 ppm/C
1A + 10 + 1C = 0.09 ppm/C
9 + 10 + 5 = 0.10 ppm/C
1A + 9 + 1C = 0.11 ppm/C
6 + 5 + 7 = 0.12 ppm/C
1A + 5 + 1C = -0.12 ppm/C
2 + 1A + 3 = -0.12 ppm/C
10 + 12 + 4A = 0.13 ppm/C
1E + 4A + 7 = 0.13 ppm/C
1E + 8 + 1C = -0.13 ppm/C
6 + 8 + 1C = -0.13 ppm/C
10 + 11A + 4B = -0.13 ppm/C
10 + 11B + 4B = -0.14 ppm/C
10 + 11B + 4A = 0.14 ppm/C
10 + 12 + 4B = -0.15 ppm/C
10 + 11A + 4A = 0.15 ppm/C
8 + 4A + 7 = 0.15 ppm/C
9 + 4A + 4B = -0.15 ppm/C
1E + 4B + 5 = 0.17 ppm/C
2 + 1B + 3 = 0.17 ppm/C
8 + 5 + 7 = -0.18 ppm/C
1A + 7 + 1C = -0.19 ppm/C

A script for groups of four is left as an exercise for the reader

Forum member, if you are new to Python, try seeing if you can modify the script to also print out the total resistance of each combo, as an exercise.

[TiN]

Cool, I can test 9+12+4B next.

Perhaps you can modify code to also calculate final "output" resistance, taking measurement data?
All meters are calibrated within 90 day spec, so you can use absolute values.