T.C. measurements on precision resistors

[Andreas]

Hello,

and just the next DSMZ (#2) measured with new ADC #15 dividing the reference voltage ratiometrically.

18.5uV change over 33.9 K giving 0.55uV/K or 0.22ppm/K (so better than the previous DSMZ)

This competes with the LTC1043 #1101_003 which has:

3.2 uV (max.) change over 31.8K giving 0.1 uV/K or 0.04 ppm/K
(and would be much better when the switching on of the cooler did not produce that spike at 32 degrees as can be seen in the drift over time diagram).

Offset drift of LTC1043 #1101_003 is:
1.2 uV change over 32.4K giving 37nV/K or 7.4ppb/K referred to the 5V ADC input (= divider output).

ADC #15 alone measured 0.3uV/24K offset drift (17-42 deg C)  and 2.2uV/34.5K gain drift.

With best regards

Andreas

[Andreas]

Hello,

I have repeated the measurements of the LTC1043 #1101_003 and DSMZ#2 divider.
(the jump on LTC1043 was not plausible to me).

The gain drift of the LTC1043 of 07.10.2014 is now without any jumps.
So I guess that I had a problem with one of the connectors on the measurement of 30th of September.

Without regarding measurement noise I get  0.5uV/31.8 K = 16nV/K = 6.3ppb/K for the LTC1043 gain drift.

DSMZ#2 was repeated on 08.10.2014.
T.C. of the divider is similar to the first measurement
17.2uV / 32.3K = 533nV/K = 213ppb/K (2.5V)

The ageing drift/hysteresis ist much smaller than on the first measurement. 0.7uV = 0.28ppm
So either the drift was due to longer time after soldering or due to relaxation after the first thermal cycle.

With best regards

Andreas

[branadic]

Hi Andreas,

still very interesting results but somewhat confusing. Can you give a brief summary what resistor is best in your measurements?

Regards.

[Andreas]

still very interesting results but somewhat confusing. Can you give a brief summary what resistor is best in your measurements?

I think it's still a Work In Progress [WIP], but I too am interested in the "Reader's Digest" version of which [so far] are the best resistors.  But then, we have to define what "best" means, but to me, it is excellent TCR and super-stable resistance over time-- but that is because I will not be running the resistors under high power or pulsed power conditions [at least in what I am doing now].

As with many things in the world of engineering, the definition of "best" depends on how you are using the resistor.

I would be interested in what Andreas found with Edwin Pettis' resistors, which appear to be the best of the Evanohm-wire based PWW resistors available.

Hello,

I can only speak for the resistors that I have measured.
And there is no clear winner up to now.
Each resistor even from the same series behaves more or less individually.
And yes its a WIP I have planned to do further measurements at least on some Z201 and if I get some samples also from 8G16. (But there we have one measurement from Frank with 2.2 ppm/K. So this would be only to see if hysteresis will be better with slower temperature change).

From Edwin I still have no idea where the pricing will be for small quantities. And I fear that the shipment costs from US to Germany will be tremedous high especially when using UPS or similar service who does the handling with customs. And I do also not know if I could pay per PAYPAL.

From trend I would give the following statements today:

Z201 will be my first choice for low T.C. + hysteressis without selection.
UPW50 is second choice. And hand picked devices may exceed average Z201.
UPF50 are the surprising cost/performance winner regarding T.C. + hysteresis.

My UPW25 samples although similar to UPW50 seem to have problems.
S102 and USR2 have also measured far below UPW50.

I have no clue how long term stability will be.
(And up to now I have no reference grade resistor to compare with).
And I have no hermetically devices in my samples.

For dividers I would not use individual resistors but something like DSMZ.
The open point is: how will they behave when being soldered on a PCB?
Perhaps it would be better to use them dead bug mounted.

Another point:
the cylinder style resistors seem to be more critical with thermal coupling. (and perhaps large temperature gradients).
I have the impression that thermal grease is necessary to keep them at same temperature as the isothermal block, whereas a rectangular style resistor showed lesser difference (shown as "hysteresis") between "dry" and "greased" state.

With best regards

Andreas

[MK]

Perhaps you could use medicinal paraffin to thin down some thermal grease to make it easier to keep the round resistors in contact with the aluminium block?

[Dr. Frank]

Andreas, DiligentMinds, branadic

up to now, this thread was a very instructional and successful demonstration, how such delicate measurements like the determination of these low T.C. of resistors could be done with relative simple means, instead of using heavy equipment like an 8.5 digit multimeter, or ultra stable and precise resistance bridges..

But at that point,  it is no way of judging, which resistor technology or brand is better, or best in class..

Reason for that is the mere lack of statistics.. means, too few different samples per technology / brand have been used.

Elsewhere, I already have provided 5 T.C. values for the hermetically sealed, oil filled VHP202Z, and found out, that these values showed a big spread, and all were an order of magnitude above the promoted typical value.
(Andreas, if you like, I may add these measurements here as an additional item)


To evaluate, what the different technologies are really capable of, say in statistical terms of 'typical' values and max. spread, you need much more samples than 2, or even 5 ones.

Therefore, up to now I would not draw such a conclusion, as DiligentMinds did.

Maybe others also can add such measurements...

Andreas, if you don't have an 8E16 resistor yet, we may discuss, how to organize that.. I have a bunch of 50 Ohm, .., 10k, 30k, 100k in stock.
Either I do some much better measurements on them,  (I mean, than the quick *n dirty one that I already did  ) using my equipment, or we may exchange some samples..


The basic problem with these tubular case resistors is that the wires obviously transport most of the heat into the resistor, compared to the mould component.
That effect can be reduced by attaching the leads thermally to the isothermal block also.. there are several possibilities to do so.

Anyhow, this subject is still very interesting, so keep on going!

Frank

[texaspyro]

Speaking of the Fluke 5450...  does anybody know of a source for the DL02614 displays used in these?  One of mine has a bad segment.

[MK]

Andreas, DiligentMinds, branadic

"...it is no way of judging, which resistor technology or brand is better, or best in class..."

<snip>

Therefore, up to now I would not draw such a conclusion, as DiligentMinds did.

<snip>

Hi Dr. Frank,

Sorry, when I posted that I guess I kind of implied that I was basing my conclusions on the work done here in this thread by Andreas.  I apologize for that.  That isn't what I meant to say.

My conclusions were [and still are] based not only on Andreas' work, but long years of experience and hundreds of hours scouring the available literature for relevant material.

In a private email from Edwin Pettis to me, Edwin complained that many engineers over-spec the resistors they are using in their circuits.  I think he is correct.  You do not need a super stable hermetically sealed, oil-filled, 0.001% resistor for an LED ballast resistor.  Part of the art of electronics is knowing when to specify better parts, and in order to do that you have to know *why* you need a better part.  Then, you can use your engineering skills to ask the question: "how good do I need this part to be?"  You also need to know what parts are available, and what the environmental effects on each resistor type are.  In some applications, you also need to know what the long term drift might be for each resistor type.  Without this information, you might get lost in all of the data, and end up specifying the wrong part.  In some applications, you would want to know how to accelerate the aging of the resistors and other components in your design so that your design is very stable coming off of the factory floor.  There are different techniques for doing this, and not all of them apply to all resistor types [and/or other components].

I too think that the work Andreas is doing is important, and interesting.  Andreas, please keep going!

I agree that overspecifying is expensive, the problm I believe is a lack of trust in the datasheets, like some of the current TO220 devices that claim to be able to dissipate 200-300Watts, the same general weasel words have crept into many other datasheets too.

But this thread grew out of the LTZ1000A thread where there are many other error sources that creep in. For an LTZ1000 my current thinking is a mix of econistors/squaristors and some Edwin resistors in the most critical positions, and then jellybean resitors for current limit, psu etc.

[TiN]

Just won a broken Keithley 2510 TEC SMU, will be a nice tool for little controlled temperature chamber for TC testing.

Thinking of building plastic insulated box with BNC inputs and RTD sensors around to make accurate -20....+100°C controlled temperature chamber.

[ManateeMafia]

TiN,

I picked up one of those earlier this year for the same purpose. Mine appears to be fully functional. If you need me to do anything to help with the repair, let me know.

I have a small aluminum enclosure milled for a small peltier assembly. I have also purchased one of these   http://www.ebay.com/itm/Liquid-Thermoelectric-Peltier-Enclosure-Dual-Fan-With-Heat-Sink-/140898947623.

[Andreas]

Hello,

I have done a repeated measurement on S102#1 just to see what happened in the mean time.

The measurement setup is slightly changed just to make it easier to change the resistors.
I do not use thermal grease; the resistor is pressed against the isothermal block with a spring.
The measurement lines are not soldered but I use 4 clips for the 4 wire measurement.

Measurement on 18.01.2015 compared to the measurement of 13.07.2014 shows
a different hysteresis curve and a drift of -12.6 ppm @ 25 deg of either the S102#1 or the reference resistor Z201#1.

Measurement on 19.01.2015 is more similar to the old measurement of 13.07.2014 regarding hysteresis.
Drift is reduced to -11 ppm @ 25 deg.
So both effects together seem to be related to the S102#1 resistor.

With best regards

Andreas

[Andreas]

Hello,

another repeated measurement with Z201#2

Comparison is to measurement of 01.08.2014

On 20.01.2015 the drift is slightly negative: -0.3 ppm @ 25 deg C
after 3 further temperature cycles:
On 23.01.2015 the drift is slightly positive: +0.9 ppm @ 25 deg C
Slope varies also slightly (but may be also different by using sligthly different minimum and maximum temperatures).

All in all the two Z201 (reference #1 and DUT #2) have much lesser drift than Z201#1 against the S102#1.
So either the Z201 are both very stable or at least have the same ageing drift.

other coefficients:
01.08.2014
A 0 = -7.81060638217116E-0001
A 1 = -2.31032022287009E-0001
A 2 = -4.31093591373655E-0003
A 3 =  2.78270315402516E-0005

20.01.2015
A 0 = -1.1077309086686037E+0000 -> drift -0.3 ppm @ 25 Deg
A 1 = -1.9762440120142556E-0001  -> slope -0.03ppm/deg
A 2 = -3.9960281447573684E-0003
A 3 =  1.7247792207858591E-0004

23.01.2015
A 0 =  1.4344628522426358E-0001 -> drift + 0.92 @25 deg (01.08.2014)
A 1 = -2.0906922064354618E-0001 -> slope -0.02ppm/deg
A 2 = -3.0228719576275895E-0003
A 3 =  1.2306553962915563E-0004

With best regards

Andreas

[Andreas]

Hello,

the measurement clips brought me to a idea:
What is the influence of the lead length on the resistor T.C. ?

When doing the test with soldering I used the maximum possible lead length in order to make minimum thermal stress to the devices.
But now with the clips I am more flexible. (And when soldering finally to a pcb I would never take the full lead length).

So from datasheet of a Z201.
lead length outside from case:  about 25.4 mm (one inch) with AWG#22 copper.
(54.7 mOhms/meter TCR: 3880ppm/K)

Within case the lead is additionally up to 9 mm each.
So minimum usable lead length is 20 mm in sum.
Maximum around 70 mm in sum.
Difference is 50mm resulting in 0-2.7mOhms variable resistor with 3880ppm/K
That sounds not much but for the 120 Ohms resistor of a LTZ1000
2.7mOhms are already 22.7ppm variable resistor with 3880ppm/K T.C.
Giving from 0 to +0.088ppm/K additionally T.C. on the 120R resistor.
So slight negative T.Cs could be trimmed by the resistor lead length or by the PCB trace length.
(Ok this is probably not the best idea since on pcbs the trace thickness has large tolerances during production).

For the 1K resistors the lead lenght short/long gives also a 0.01ppm/K change.
If one would use a 10 Ohms Z201 the leads give up to 1ppm/K change in T.C. which is far beyond the "typical" datasheet spec.

So at which lead length the T.C. of a Z201 is specified?

With best regards

Andreas

[MK]

The z series are supposed to be mounted touching the pcb, so if we assume single sided with the copper opposite the component, it will be used about 1 mm from the body.

[babysitter]

The Econistors say in their datasheet that the connection for calibration is "made - 10mm along the lead-out wires from the body", and also give a figure of Resistance of termination leads depending on Type.

[Dr. Frank]

Hello Andreas,

I'm really sorry to disagree to your statements and your setup.
It's not at all from electrical, but from pure physical aspects.

Your (really wild) hysteresis figures show, that all of your measurements were always done in a very big thermal disequilibrium between your measured temperature (sensor?) and the resistor element.
It's always a dynamic thermal measurement (everything is changing rapidly), instead of near static measurements.
It's also caused by the fact, that these type of resistors were mostly heated / cooled by their wires, not over the body. So the thermal mass, as you use it, does not help.

You either have to couple the wires also to your thermal block, or you have to do the measurements much, much  more slowly, in equilibrium, i.e. not always by forced heating / cooling, but by very slow thermal convection.
Or you may put temperature sensor and DUT in one closed metal box and slowly heat/cool the whole box.

As a consequence, your T.C. curves do not represent the actual behaviour R(T) of the DUTs, and also your statement, that T.C. has changed between the two dates, I think, is unsustainable.


For a volt-nuts from U.S., I've just recently calibrated  his NTC against a precision PT100, and then later determined the R(T) of his PWW by using this calibrated NTC.

I append the hysteresis curve for the NTC vs. PT100 measurement, where you can nicely see, how forced cooling / heating cause hysteresis, but during slow convection (taking several hours for one direction), there is no hysteresis at all.

In the last chart (raw data) you can see, how well you can supress hysteresis for the R(T) measurement (it's the PWW in a aluminium box), if you simply make the measurements very slow.
The forced cooling is visible as an offset, but even the slow, forced heating by an IR lamp causes no  hysteresis.

The loop starts at room temperature, cooling down quickly to 15°C, then relaxing to room temperature, slowly heating to 32°C, and relaxing again to r.t.

Frank

[Andreas]

The z series are supposed to be mounted touching the pcb, so if we assume single sided with the copper opposite the component, it will be used about 1 mm from the body.

The datasheet is of opposite opinion ;-)
http://www.vishaypg.com/doc?63187

At least for the resistance value they spec on page 4:
Lead test point: 0.5" (12.7 mm) from resistor body

I doubt that they change their test fixture for measuring T.C.

With best regards

Andreas

[Andreas]

Hello Frank,

one temperature cycle already needs a whole day. 11-12 hours for the cycle + 1-2 hours for overlap until I do not have a "open" hysteresis curve.
So I am happy when getting a hysteresis below 1 ppm and take the interpolation in between as "truth".

Maybe you are right for the Z201.

But the S102 components are (with similar shape and test fixture) so much worse that I consider that the inner construction of the S102 is much worse (some interaction between molding compound and resistive element which gives something as a time constant).

Also Vishay states up to 20 ppm seasonal effects for molded resistors due to humidity.

http://www.vishaypg.com/doc?63171

With best regards

Andreas

[Dr. Frank]

Hello Andreas,
maybe I did not explain precisely:

The basic problem with all of measurements is, that you never get some repeatable curves or parts of curves in ALL of your measurements.

So you can't identify, which effect is really hysteresis (e.g. by the metal foil technology), and which is thermal difference between the resistor and its wires, or the thermometer.
Even if you measure slowly, the wires may always be on a different temperature, than the thermal mass and the thermometer.

I don't dig it from your measurements, but all these curves look quite strange, if I recall temperature measurements, and also hysteresis measurements  I have done.

Maybe the S102K has a very huge hysteresis, but the 201Z definitely has to have much lower hysteresis (and only for bigger temperature excursions as +/-15°C only)


I propose that I repeat your measurements with my Veltins thermal chambers  and see what happens.

Frank

[Andreas]

The basic problem with all of measurements is, that you never get some repeatable curves or parts of curves in ALL of your measurements.

Even if you measure slowly, the wires may always be on a different temperature, than the thermal mass and the thermometer.

I don't dig it from your measurements, but all these curves look quite strange, if I recall temperature measurements, and also hysteresis measurements  I have done.

Maybe the S102K has a very huge hysteresis, but the 201Z definitely has to have much lower hysteresis (and only for bigger temperature excursions as +/-15°C only)

I propose that I repeat your measurements with my Veltins thermal chambers  and see what happens.

Hello Frank,

For the repeatability I look at the UPF50 (metal film resistors) on page 1. Those have virtually no hysteresis.
Ok since they are cylindrical the setup is slightly differen. But they should even be worse because the thermal contact is worse.

All in all you are right: the results look strange.
I have tried many different setups but I did not find a setup with much better hysteresis up to now.

When looking at the Z201 chart at 25 degrees C. The hysteresis in x-direction is around 7 degrees C  (or +/- 3.5 degrees in each direction).
7 degrees at 0.12K / minute slope correspond to nearly 1 hour (or +/- 30 minutes in each direction).
I really doubt that either the temperature difference between temperature sensor and resistor element is 3.5 degress.
It should be more in the +/- 0.5 deg C range when the fan is active.
And I also doubt that the thermal time constant is around half an hour with active fan.

Perhaps my fault is that I rest too long at the minimum and maximum temperature so that the hysteresis can "charge".
Ok I will have to check the temperature sensors. Perhaps they have not survived the treatment with the thermal grease.

If you want to spend your time, I will send you the Z201#2 and S102#1 for measurements.
It would be very interesting how the setups compare.
And if repeatability is related to measurement setup or resistors.

With best regards

Andreas

[Dr. Frank]

Hello Andreas,
I look forward to get your resistors!

What's really strange is the lack of a virgin path if you really observe the hysteresis, which is a companion of the metal bulk foil technology.

Then, the Z201 should nearly be free of hysteresis in the temperature range where you measure.

And I can really not believe, that the S102K should be that bad, because for a claimed T.C.  of 2ppm/K (even with Vishays butterfly method), or so, such a big hysteresis should make them useless.

What might cause such behaviour, is a foregoing history of extreme temperature excursions.. remember, how I received my VHP202Z back from Vishay Precisison/ Germany, he had heated them asymmetrically to 125°C, and they were stuck at +5ppm higher resistance, until I unloaded this memory / hysteresis effect.

Do you remember, what happened to your resistors?

Maybe, that you already have a prepared aluminium box with jacks for the resistor and the sensor?

Otherwise I have to build one.

That will be an exciting experiment, again. 

Frank

PS: I may probably measure these two comrades during that session, the left one is also a Z201.
Has yours the same case?

[babysitter]

Maybe closer inspection  of Andreas' isothermal box including heaters and sensors would gain some insight?

How big are the gaps between the block and the parts? I suppose you already reduced it as much as reasonable and that the parts are touching one part of the wall due to gravity, but how would you describe it? Thermal grease works well usually only in very, very thin layers. Also I would recommend some thermal isolation of the block, even if it is just that kind of sticky tape that is used for mirrors where you can put it without closing holes.

Andreas: More photos, please ! I am just aware of one.

[Andreas]

Do you remember, what happened to your resistors?

Maybe, that you already have a prepared aluminium box with jacks for the resistor and the sensor?

Otherwise I have to build one.

PS: I may probably measure these two comrades during that session, the left one is also a Z201.
Has yours the same case?

Hello Frank,

Treatment of my side for Z201#2 was as follows:

26.07.2014 in the evening
soldering 4 wire with crocodile clips on the leads for protection.
Thermal grease for trying to get better thermal contact to isothermal block
temperature setpoint 45 degrees (first hysteresis test).

27.07.2014
thermal cycle 8.5-46 degrees setpoint beginning from room temp 0.12K/minute

28.07.2014
3 thermal cycles 8.5-46 degrees setpoint beginning from room temp 0.3K/minute
Temperature measurement error due to failure in USB device -> no evaluation possible

29.07.2014
3 thermal cycles 8.5-46 degrees setpoint beginning from room temp 0.3K/minute

01.08.2014
thermal cycle 8.5-46 degrees setpoint beginning from room temp 0.12K/minute
in the evening:
desoldering with crocodile clip to cool the leads

some days later:
degreasing of the device with solvent (Kaltreiniger).

20.01.2015
first repetition measurement without thermal grease and with IC-grabbers instead of soldering.

At the moment I am doing measurements on a never greased device (Z201#3)
Hysteresis seems to be better. (so that I can blame it on my measurement setup).
So the question is if I better wait until these measurements have finished.
You would get a 3rd device.

And no: since I want to use the resistors afterwards for a LTZ1000 I have no test fixture.
At maximum I could send you a spare iso thermal block.

According to PCN you have a device before 2006
Mine are the black version of Y1453

http://www.vishaypg.com/doc?63200

By the way in 2014 they changed again the molding compound:
http://www.vishaypg.com/doc?63254

of course this has never a influence on the electrical specifications 

With best regards

Andreas

[Dr. Frank]

Hello Andreas,

well your Z201 has onviously seen no excursion from R.T. more than +/- 20°C.. therefore, the Z foil resistors should not show hysteresis, from what I learnt from the VHP203Z.

Especially not so huge ones.


Well, leave yourself time, and then send me more samples, idf you like.. I will prepare a box then.. careful soldering at the end of the wires, with a copper pincer is allowed, I hope?

I also would like to get more insight in your heater/cooler assembly.. One sentence drew my attention.. you mentioned, that you run the fan.. all the time?

Does that imply, that either the thermal heat or sink is always on, also the fan?

Then there might lay the reason..

As I indicated, it's better to leave the system alone w/o forced heating or cooling, and especially no air draught.

That's the main purpose of my beer thermal box.. it stores the heat in the box, so that you get a constant, very slowly varying air stack, cold at the bottom, warmer at the top (R.T.), or warm metal box, which slowly cools down by thermal air diffusion.

Well, it's total speculation at the moment, let's simply do the test at my premises.

 regards Frank

[Andreas]

I will prepare a box then.. careful soldering at the end of the wires, with a copper pincer is allowed, I hope?

I also would like to get more insight in your heater/cooler assembly.. One sentence drew my attention.. you mentioned, that you run the fan.. all the time?

Does that imply, that either the thermal heat or sink is always on, also the fan?

Then there might lay the reason..

As I indicated, it's better to leave the system alone w/o forced heating or cooling, and especially no air draught.

Hello Frank,

soldering is allowed (I did it before).
I hope you do not want to use the Pb-free solder wire with higher melting temperature.

Attention: According to PCN the size of black resistors is smaller than the green ones.

Yes the "inner" fan is running all the time (with about 9V voltage for a 12 V low noise fan).

Most of the pictures on page 1 are still valid for the setup.

The fan of the cooling box is only running during cooling phase which goes from +32 degrees C (downwards) to 8.5 degrees and from 8.5 degrees (upwards) to 15 degrees.
The heater works with 1 second on/off time to do the fine adjustment of the actual temperature against the setpoint ramp.

And when looking at my "failed" measurement of January 22nd where I forgot to switch on the inner fan. (I thougth it was useless to look at it). I fear we have found the reason. Above 32 degrees where the fan of the cooler is also off I see no hysteresis at all.
Damned that hurts: I have learned a lection. I will have to fix my setup.

The open question is: how can I cool down the resistors with a defined ramp with minimal air draught within the cooler?

with best regards

Andreas