Author Topic: T.C. measurements on precision resistors  (Read 271990 times)

0 Members and 1 Guest are viewing this topic.

Offline babysitter

  • Frequent Contributor
  • **
  • Posts: 808
  • Country: de
  • pushing silicon at work
Re: T.C. measurements on precision resistors
« Reply #200 on: January 27, 2015, 09:03:44 pm »
Thank you for re-pointing to the first posts with the setup images.

My suggestion:

Put some thin foam, say 1 to 5 mm,  around every available spot on the isothermal block and try again.
Lack of foam on non-available spots intentional, the ramp shall be done before eternity is over. You might even consider putting holes in the foam where not necessary if it improves symmetry versus the necessary ones.

This should help a lot in my opinion.

« Last Edit: January 27, 2015, 09:19:07 pm by babysitter »
I'm not a feature, I'm a bug! ARC DG3HDA
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #201 on: February 01, 2015, 08:26:59 pm »
My suggestion:

Put some thin foam, say 1 to 5 mm,  around every available spot on the isothermal block and try again.

Hello,

I did some tests. But with the results up to now I think that it is not sufficient to only put some foam around the isothermal block and let the contacts outside the shielded room.

What I did up to now:
First: put a additional temperature sensor right between the legs of the DUT.  (bound with isolated wire).

27.01.2015
The fan from the cooler box (outer fan) is a horror for stability.
The inner fan cannot compensate fully for this.
So I decided to use a carton box for the heat spreader and the isothermal block.

30.01.2015
Due to (large) temperature differences between the leg sensor and the isothermal block I decided to put a further heat spreader (TEKO A3 enclosure) around the DUT and isolate the resistor. Intention is to keep the resistor and the legs at the same temperature.

31.01.2015
Still not sufficient I put a copper bar (around 5-6mm thick) above the DUT and below the TEKO A3 enclosure to get further improved temperature distribution.

01.02.2015
I got the suspicion that there are Seebeck voltages which generate the hysteresis.

Edit: 1uV thermocouple voltage may change the measured resistance by 0.4ppm.
The copper-beryllium contact has about 0.8uV/K against copper.
The solder junction at the other end 3-4uV/K against copper.
So a 2 ppm hysteresis might be a +/-2.5uV thermocouple voltage.

So I started a measurement (still running) with reverse polarity of the reference voltage for the resistor divider.
Perhaps I will have to use the cirquit on page 7 of LT AN96:
http://cds.linear.com/docs/en/application-note/an96fa.pdf

@Frank: does the HP3458A any measures against thermocouple voltage like current reversal or pulsed measurement

with best regards

Andreas
« Last Edit: February 01, 2015, 08:39:14 pm by Andreas »
 

Offline Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1726
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #202 on: February 01, 2015, 10:18:21 pm »
Hello Andreas,
Yes the 3458A has OFFSET COMPENSATION for Ohm measurements.

The reference current is switched on and off, and thermo voltages were subtracted to give the true resistance.

For voltage measurements, that's of course not available, You would have to reverse the DUT manually

Not totally clear, what You Want to measure in this instance.

Frank
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #203 on: February 03, 2015, 11:02:00 pm »
For voltage measurements, that's of course not available, You would have to reverse the DUT manually

Not totally clear, what You Want to measure in this instance.

Hello,

according to AN96 I could also reverse the "(half) bridge" voltage.
Or switch off the VREF alternately.
But this would increase the overall noise level since I have not so many values to average.

As you noticed : As you increase the thermal mass, you have lower the slope of temperature accordingly, or phase lag will appear. So : The smaller test system, the better.

Hello Emmanuel,

thank you for your comments. This gives me a idea how to improve the thermal coupling of the leads.
I think in my case with the fan from the cooler I will need some thermal mass (at least around the resistor and the junctions) to equalize temperature.
On a thin sheet plate you can easily measure temperature differences of some degrees.
I have ordered some thick walled Hammond enclosures for better heat spreading in case it is necessary.

And I have also some small aluminum bars 19x19 mm where I can put the resistors in between with some silicone rubber when I get some slot milled for the resistor.
So we will see.

With best regards

Andreas
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #204 on: February 03, 2015, 11:22:51 pm »
Hello,

in the mean time the measurement from 01.02.2015 with reversed polarity shows the strange effect that the hyseresis is halved but no change in direction as I would have expected for thermocouples. So there seems to be another fault.

02.02.2015: measuring offset by shorting both ends of the resistor line to AGND.
Shocking: the offset is relative large -15 .. -23 uV and depending on input impedance.
The low impedant pins have the -23 uV and the resistor taps have the -15 uV.
One resistor tap (red) shows a hysteresis of 1uV and nearly no other temperature dependancy.
So only around 1uV effective thermocouple voltage visible.

The ADC itself has usually below 5 uV Offset and usually in positive direction.

The -15..-23 uV "Offset" reminds me of my first measurements with the capacitive voltage divider:
The capacitive switching input of the ADC generated the offset by rectifying the noise at the protection diodes of the buffer OP-Amp output.
So in this case the multiplexer protection diodes are affected by the switching noise of the sigma delta ADC.

First measurements with a filtered buffer amplifier between MUX and ADC show a equal offset of all mux channels at room temperature.

With best regards

Andreas

 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #205 on: February 07, 2015, 02:25:24 pm »
Hello,

I have made a offset measurement with a ADA4638 buffer.
Against unbuffered measurement all 4 channels show similar behaviour over temperature.
Only one channel (the voltage pin on the measured resistor) shows around 0.5uV hysteresis.
(which would give 0.2 ppm resistance change).
This is similar to previous measurement (the red curve of unbuffered measurement).

So I think that thermal EMFs are not the primary problem in my setup. Since all differences nearly cancel out.

With best regards

Andreas

 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #206 on: February 07, 2015, 02:30:15 pm »
Hello,

I also did a comparison between unbuffered TC curve of Z201#3 and with the ADA4638 buffer:

So there are no significant differences in the diagrams.
« Last Edit: February 07, 2015, 02:32:24 pm by Andreas »
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #207 on: February 07, 2015, 02:51:17 pm »
Hello,

I did also do a comparison with reversed supply of the resistor divider:

What I do not understand is that the hysteresis differs (is nearly halved) from previous measurements.
Between the 4 measurements I did no mechanical change to the setup within the cooler box.
Any ideas?  Are resistors with welded connections in reality some hidden diodes?

The only thing that I have is that there are around 0.4 deg C difference between different NTCs at the same temperature depending on rising or falling temperature.
But 0.4 deg C at <0.3ppm/K would not explain a 2 ppm hysteresis opening.
As calculated on the first page the self heating of the 1 K Z201 resistor is around 1 deg C.

I will follow the advice of Emmanuel for the next: using better thermal contact to the wires of the resistors.
In this case I cannot use the test clips and will have to solder again.

But all in all I think searching for the reason of the hysteresis is more a academic theme.
In all diagrams I have more or less 10 ppm difference between the endpoints for 33-34 deg C temperature difference.
So if I interpolate in between the hysteresis I will be not far from the truth regarding T.C.
And around 0.05ppm/K ( .. 0.1ppm/K) seems to be the limit what can be reproducable measured with a 24 Bit ADC.

With best regards

Andreas
 

Offline ManateeMafia

  • Frequent Contributor
  • **
  • Posts: 719
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #208 on: February 07, 2015, 03:53:53 pm »

Any ideas?  Are resistors with welded connections in reality some hidden diodes?


This may not be pertinent to your measurements but Edwin Pettis had written an article where he mentioned that some resistor welds acted like diodes...

http://www.edn.com/design/analog/4427940/1/The-last-half-century--Wirewound-resistors-Part-two
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #209 on: February 07, 2015, 09:42:10 pm »
Hello Ken,

as described on page 1 I am using a LTC2400 and not a 3458A.
Since I am measuring ratiometric I use the VREF of the LTC2400 for the resistors the MUX and the ADC itself as supply.
The input voltages when measuring T.C. are all within 0.5mV to VRef-0.5mV -> within conversion + supply range of the ADC.
So far away from leakage of ESD diodes.

With best regards

Andreas
 

Offline janaf

  • Frequent Contributor
  • **
  • Posts: 339
  • Country: se
Re: T.C. measurements on precision resistors
« Reply #210 on: February 08, 2015, 02:58:34 pm »
Hello Andreas,

Apologies if I'm somewhat lazy, but is it possible to make a brief summary of your results? The amount of data is kind of overwhelming  :)

A couple of related things: The S102 are epoxy molded. Do you have any thoughts on improving the humidity transfer by for example  dipping these in additional layers of epoxy? Epoxy dipping may of course also have other effects, influencing hysteresis. It may be of for DIY and small scale use. If someone could test, I could supply some S102K or S from the same batch, with and without additional dipping into epoxy.

I also started a thread relating to your post #1 on alternative methods of bonding sensitive components like these, onto PCBs.

https://www.eevblog.com/forum/testgear/methoods-for-connectig-sensitive-components/

Jan
my2C
Jan
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #211 on: February 08, 2015, 04:41:07 pm »
Hello Jan,

from time to time I post some collected overviews.
The last was in post:
https://www.eevblog.com/forum/projects/t-c-measurements-on-precision-resistors/msg509712/#msg509712

Just dipping into epoxy will not really help. (May only change the time constant).
The only way to really avoid humidity effects is to use a hermetically sealed package.
Metal or ceramic with glass sealing for the wires.
The wires have to be out of Kovar in this case to adapt the thermal expansion to the sealing.
Unfortunately the Kovar wires have a large thermal EMF (39 uV/K) against copper.
So avoiding the humidity problem will need further effort at the thermal layout side.

To really test humidity you will need a stable (hermetically) reference resistor.
(Which I do not have). And much time.
The reaction time to humidity is usually in the range of several (3-7) days.

With best regards

Andreas
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 415
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #212 on: February 08, 2015, 06:08:17 pm »
I have to disagree about hermetic seals are the only way to avoid humidity effects.  The design of the resistor determines the sensitivity to its enviroment, in the case of film resistors, hermetic is the only way to completely avoid it because the film elements are sensitive to humidity.  In the case of my resistors, there are no significant effects, unless they are put into a high temperature, high humidity enviroment for a prolonged period and even then, the effects are very small compared to all other resistor types.  It all has to do with the design and construction of the resistor.  In a normal environment, humidity can be ruled out as a effect to worry about.

In film resistors, the resistive element is very small and the line widths/thickness are also very tiny, the element has no protective covering, it is bare and unless sealed against the enviroment, reacts to anything coming inside of its packaging.  Although wire wound resistors can use extremely tiny wire sizes, as small as 0.0004" diameter, all wire is enamel coated and pin holes are required to be few and far between.  The main culprit of humidity sensitivity in precision wire wound resistors was the lack of a true welded joint at either end of the resistor, this allowed many problems to affect the resistor's stability in the past.  That problem was solved years ago but only one manufacturer has that technology.
 

Offline Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1726
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #213 on: February 08, 2015, 08:30:31 pm »
metal foil resistors in mold compound package may age from oxygen and humidity.
That's something, Vishay themselves explain, when they describe the advantages of their hermetically sealed AND oil filled VHP packages..

(These truely give < 2ppm/ 5years stability.)

To protect ordinary molded types, by an additional epoxy around , you would first have to bake them, to expel humidity, as you have to do that with each IC, which was exposed to air, in order to reflow solder them, without cracks.
The mold compound is like a sponge, which stores humidity.

I think, PWWs are protected quite naturally, because good quality wires have a very tight and relatively thick lacquer around them.

The equivalent were conformally coated metal foil types, maybe, like AE = Alpha Electronis (now also Vishay) offers them.
This is obviously an additional epoxy coating against humidity, etc.
See attached document .
The stability figures are not stellar, either. 


Mr. Pettis, you claim 5ppm/year stability for your resistors.. would you mind explaining, how you measured that, or how this stability figure can be derived from the mechanical / electrical design of your resistors?

Thanks a lot!
Frank

« Last Edit: February 08, 2015, 08:41:06 pm by Dr. Frank »
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 415
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #214 on: February 08, 2015, 08:57:23 pm »
Hello Dr. Frank,

There are three methods for testing drift, the 'sitting on a shelf' at room temperature with periodic measurements, works great but takes a long time and also has increasing uncertainty in the readings over time if you are splitting hairs.  The next method is putting the resistors in an oven at a specified elevated temperature over a period of time and taking periodic readings, this takes less time but costs more to do, has the advantage of faster results and minimal changes in uncertainies.  The third method, which is what was used with my resistors, is the thermal shock testing, it is more expensive but provides results in a few hours and the uncertainty factor is essentially the same as when the testing began.

As I mentioned elsewhere, a third party subjected 50 resistors to 50 thermal cycles (+125°C to -55°C, 30 minutes, with no more than 2 minutes between temperature, well beyond the normal 5 cycles, the resistors were found to be superior in stability to every other resistor they had previously tested (even at 5 cycles) and with a zero failure rate.  I can state that most resistors had the equivalent drift of <2 PPM/year, but I still specify 5 PPM as some units can be a little higher.  It was also found in studies conducted by institutes of standards (Australians for one), that the Evanohm alloy (of which all of mine are made from or a derivative) has by far the best long term stability of any resistance alloy IF USED PROPERLY.   The actual design of the resistor (i.e. construction) has a very significant bearing on how well the resistor performs, while it is not impossible for other resistor houses to produce 0 ±3 PPM/°C resistors (check their specs, only over a limited range in most cases), their stability is significantly worse because of their construction.  It has been and still is believed by virtually all other resistor houses that they produce a welded, stable PWW resistor, when if fact, their own specifications prove otherwise.
 

Offline janaf

  • Frequent Contributor
  • **
  • Posts: 339
  • Country: se
Re: T.C. measurements on precision resistors
« Reply #215 on: February 08, 2015, 09:06:36 pm »
Andreas, thanks for the summary!

Epoxy; is that a physical limit, i.e. diffusivity of oxygen & vapor through epoxy well defined? There are tons of different epoxies available. I imagine the epoxies used in electronics are subject to dozens of requirements, diffusivity only one of them. Resistors for the LTZ1000 circuit, would not need high temperature life, fast curing, suitable for molding, good heat transfer etc.

Unless already well tested; I'm thinking setting up a bridge of two dipped, two un-dipped resistors, measure the ratio, soak them for a few weeks, leave measure for a few weeks to see if the ratios change. If there is no clear difference, I'd leave it at that.

I don't have a climate chamber so that's not an option for me.

I did ordered hermetical resistors about a month ago, 16 weeks delivery time :-| The dip idea is mostly out of curiosity.

Just dipping into epoxy will not really help. (May only change the time constant).
my2C
Jan
 

Offline Joe Geller

  • Contributor
  • Posts: 24
Re: T.C. measurements on precision resistors
« Reply #216 on: February 08, 2015, 09:18:08 pm »
Andreas, 

Hysteresis should cause the curve to end at different resistance values depending on how that point was approached.  For example cycling the temperature from 30 c to 28 c would give one resistance and cycling the same resistor from 26 c to 28 c would give a different end value.  That is different resistance values at the same Temperature dependent on the path taken to get to that same temperature.

On the other hand, if cycling the temperature causes a circular or oval pattern, that might not be hysteresis, but rather changes in the system such as different parts arriving at the new final temperature profile along their own slightly different paths.

Even with relatively slow cycling, I observed such curves with my small thick walled Hammond aluminum box in foam which I use as a micro-environmental chamber.  Yet, if I allowed the system to stabilize at a particular temperature, the voltage references came to the same voltage, independent of path.  If it was hysteresis, there would have been different stabilized voltages, depending on the temperature path to get to the same temperature.

One way to check your system is to choose different temperature points on the now rounded curves.  Approach the point from above or below, however once at the point, stay there for an hour or more.  It might end up that you see the curve eventually settles to a common value independent of the path to get there.  If so, that is not hysteresis, where the final path dependent values would be two different values.
« Last Edit: February 08, 2015, 09:19:55 pm by Joe Geller »
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 2430
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #217 on: February 08, 2015, 10:15:12 pm »

One way to check your system is to choose different temperature points on the now rounded curves.  Approach the point from above or below, however once at the point, stay there for an hour or more. 

Hello Joe,

nice to hear from you.

That is exacly what I did the last 2 days.

yesterday:
Setpoint 25 deg C for 3-4 hours
then Setpoint 47 deg C for 3-4 hours
then Setpoint 25 deg C for 3-4 hours.

today:
25 deg C for 3-4 hours with cooler on
7 deg C for 3-4 hours
25 deg C for 3-4 hours with cooler on
Just will have to evaluate ... tomorrow.

With best regards

Andreas
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 1496
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #218 on: February 09, 2015, 08:00:27 am »
Quote
Epoxy; is that a physical limit, i.e. diffusivity of oxygen & vapor through epoxy well defined? There are tons of different epoxies available. I imagine the epoxies used in electronics are subject to dozens of requirements, diffusivity only one of them.

Normally expoy is modified to match thermal expansion factor of the substrate, in case of IC packagings the expansion factor of the mold is matched to that of silicon. In addition the mold is filled with glas beads to minimize humidity influence.

Regards, branadic
Fluke 8050A | Prema 5000 | Prema 5017 SC | Advantest R6581D | GenRad 1434-G | Datron 4000A | Tek 2465A | VNWA2.x with TCXO upgrade and access to: Keysight 3458A, Keithley 2002, Prema 5017 SC, 34401A, 34410A, Keithley 2182A, HDO6054, Keysight 53230A and other goodies at work
 

Offline macfly

  • Contributor
  • Posts: 30
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #219 on: February 09, 2015, 06:51:36 pm »
Hi volt-nuts,

the discussion comes very often to the same point: hermetically sealing.
So, why not putting the hole circuit in an oil filled case, perhaps additionaly
filled with dry nitrogen ?
My old standard resistor from L & N ist filled with 'medicale grade white oil',
which is used since today e. g. by WIKA / Germany for their resistance standards
(Castrol WOM 14).
But 'white oil' is not suggested to use with pure copper over a long time period.
I thought about a medical grade silicon oil.

Has anyone practical experience for sealing electronics with oil ?

Regards,

macfly
Genius is one percent inspiration, ninety-nine percent perspiration (Thomas Alva Edison 1903)
 

Offline janaf

  • Frequent Contributor
  • **
  • Posts: 339
  • Country: se
Re: T.C. measurements on precision resistors
« Reply #220 on: February 09, 2015, 08:28:58 pm »
Seems simple  ;D

Can you get the glass "frit" in small quantity? What temperature is needed to melt it??
my2C
Jan
 

Offline macfly

  • Contributor
  • Posts: 30
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #221 on: February 09, 2015, 10:07:25 pm »
Wow, a clear and straight concept  8)
But for an hobbyist like me, it sounds a little bit oversized ... :)

I thougth more on a tin can case and using ceramic feed through cap's
which could be useable to get the signals in / out of the box.
These cap's are also available with copper / tin legs.

The questions at the moment are:

1. How do I handle various air pressure.
2. How do I handle temperature depending elongation.
3. Is there any type of corrosion between the used materials and silicon?
4. Will the feed through cap's be leak-tight enough for the next 20 years (or so) ?

... and possibly a lot more questions.

Regards,

macfly
Genius is one percent inspiration, ninety-nine percent perspiration (Thomas Alva Edison 1903)
 

Online splin

  • Frequent Contributor
  • **
  • Posts: 820
  • Country: gb
Re: T.C. measurements on precision resistors
« Reply #222 on: February 09, 2015, 10:12:42 pm »
... and don't forget to do a helium leak test - it would be unfortunate to go to all that trouble and then spend the next several years characterizing the unit only to discover you have a microscopic pinhole.  |O
 

Offline janaf

  • Frequent Contributor
  • **
  • Posts: 339
  • Country: se
Re: T.C. measurements on precision resistors
« Reply #223 on: February 09, 2015, 11:00:10 pm »
PTFE is very "airtight" to oxygen, probably to steam too. The bad news is that it can't really be formed by anything but mechanical machining. Warming it up does not help, it simply deteriorates.... It might be possible to make seals with PTFE and quality O-rings. Oxygen tanks / systems are made that way, both for liquid and gaseous oxygen.

Leakage test: Maybe better to fill with Argon. Almost as simple to test as Helium, cheaper. He is the second smallest molecule of all, will leak through "anything". But still, you can only detect relatively "large" leakages. One that empties the can in a few months, half a year would be very hard to find.

I thought of having a thin metal membrane showing inner pressure. As long as there is over-pressure the metal membrane would bulge out and you'd be safe from in-leakage and . A product someone?
 
my2C
Jan
 

Offline texaspyro

  • Super Contributor
  • ***
  • Posts: 1312
Re: T.C. measurements on precision resistors
« Reply #224 on: February 09, 2015, 11:24:53 pm »
Helium leak detectors can be exquisitely sensitive.  The best ones are basically mass spectrometers and can detect leaksof just a few molecules.  Helium is a very slippery little molecule and hates to be confined... if there is a way out, it will find it.

A poor man's leak detector can be a halogen (freon) leak detector used to find leaks in refrigerent systems.  Mine can detect leaks smaller than 1 gram/year.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf