Temperature test box for component characterization

[janaf]

In the measurements so far, I've been using plain DMMs, no additional circuitry.

In RTD_circuit_3, could have used it's own 25K current resistor but instead I used the U2 buffer which senses the voltage of R_iset. So U2 will sink current from Rref and keep the voltage the same as on the low side of  DUT. In this way, both resistors use the same 25K for current setting, while U1 and U3 have the same reference voltage.

I thought this was better than using two different R_iset which would cause errors if they drift.

So U3 and U2 create a copy of U1 with the same current in both but one through the Rref and one through REF, voltage difference sensed by the InAmp.

But right now, Im not sure. I might use the simpler RTD_circuit_2. Will see...

 

[janaf]

The 1K0 foil resistor I measured the other day looks really good. My DMM say it's 21 ppm above 1K at 23C. So I'm considering keeping it as a reference, instead of putting it in a LTZ1000 circuit.

The TCR of the resistor is negative ,so by ovenizing it, elevating temperature, it could be possible to trim it to "exactly" 1K0. Trimming by 20 ppm does seem to require quite high temperature, but as my DMMs have not been calibrated for years, I don't know how accurate the 21 ppm offset is....

What do you think?

[Vgkid]

^^^The same way Fluke does it in their "A Poor Man's Resistance Bridge"paper
http://support.fluke.com/calibration-sales/download/asset/9010017_a_w.pdf

[janaf]

As a sanity check, I re-did measurements on the same Ultraohm Plus / Pettis resistor. Results essentially the same as before, TCR around +0.6ppm/C and hysteresis, if there its below my horizon of interest.

Generally, as Andreas measured a TCR of -1.1 on the same type of resistor, I think we can simply say the are well within specs and for any better info or we'd need to hand-pick and measure each resistor individually for a circuit.

Edwin, would you say that, without any hand-pick / trimming, if the resistors are made from the same wire at the same time, the measurements would typically be similar? (sorry, I only have one 1K). Or is the kind of spread you see between my measurements and Andreas's measurements something you'd expect?

[janaf]

RS9010 tempco from the calibration certificate compared with a good SR104

It looks like my 1K VHP101T is at least as good as the RS9010 in TCR and could be as good in absolute value as well.

For now I'll keep it, put it in an adjustable oven and then hope to have it tested 

[Edwin G. Pettis]

Janaf,

The resistors I made for you and Andreas all come from the same spools of wire, i.e. 120R all from the same spool, 1K are all from the same spool, ect. for a total of four different spools of wire.  Unlike other PWW resistor designs, all of my resistors exhibit the same characteristics without regard to the wire size, the 120R resistors have the same characteristics as the 1K, 12K or 70K, the majority (typically 60% or more) of the TCR will be within 0±1PPM/°C, therefor I can correctly claim that as a typical TCR with the remaining resistors within 0±3PPM/°C.  I don't play word games with the specs and I use actual measured TCRs as the basis for the TCR spec, no box or butterfly hide and seek.  The <1PPM/°C yield can be higher than 60% of course.  Since PWW manufacturing is significantly different than film/foil, I cannot target the TCR characteristic directly, a given spool of wire may give me a tighter yield on TCR over another spool as such but that is a function of the wire supplier, the only control I have over them is my specification for the wire's TCR which is very specific, it is one of the reasons of many why I can turn out a significant low TCR yield without tweaking.  I do not know of any other PWW resistor supplier that has as good a TCR yield as I do nor are they as linear in TCR, I repeat linear, that is inherent in the design and has been verified by 30 years of production and by customers.  That linearity has been used to discover unknown non-linearity in circuits and measurements.  The curve remains flat until near the outer temperature range when there is a slight change, this is mainly due to the alloy's characteristics.

Vishay does a lot of tweaking to their processes to produce lower TCRs, however, evidence would indicate that they do not have any consistent control over those processes that yield very low TCRs, particularly in the fact that few of their low TCR resistors are within the claimed 'typical' TCR spec.  Whether Vishay actually does any 'cherry-picking' of parts from a given yield for those really low TCR parts, I cannot say for sure, I would definitely not expect Vishay to admit to it publicly or even privately for that matter.  There is nothing wrong with selective picking of parts to meet a customer's specifications so I really don't know why Vishay denies doing it.

Yes, if a customer asks for a TCR yield tighter than my standard yield, then yes, they would have to be 'cherry-picked' from the batch.  I can do things which can improve the yield of low TCR resistors in sets, for instance, if possible depending on the spread of the values, the 1K and 12K resistor set for example.  One of my customers uses this to advantage, he has reported TCR tracking of better than <0.2PPM/°C over a 60°C range (all that he needed), copper adhesive tape was wrapped around the pair.  I was not asked to guarantee a particular tracking TCR for this application, only what could I do to possibly improve tracking without high expense.  As I've mentioned before, the tighter the specs, the lower the TCR specified, the higher the cost.  While I do think that Vishay's oil filled are overpriced, that is the only way they are going to come close and I'd bet a dollar to a donut that those oil filled wonders are indeed cherry-picked.  Are there resistors with TCRs close to zero in my yields, yes, how many?  I really don't know, the distribution of TCR within a given range appears to be somewhat random in nature, there has never been any large scale measurements to mathematically indicate what kind of distribution is actually present.  Given the fact that Vishay's foil resistors are made from essentially the same alloy as mine, they would also have a distribution of both + and - TCRs in their yields, what they are actually selling may be wholly different from that.

[janaf]

Thanks for the info Edwin. It's interesting that they are from the same wire as it gives us a chance to compare!

So back to the question; Can resistors from the same batch / wire have fundamentally different TCR, even if small? His has -1.1, mine +0.7 ppm/C.

Please note that I'm not complaining, both are much better than specs, but I'm curious and interested to know what to expect.

Another possibility is of course that have differences in measurement setup, that we have actually measurement errors / uncertainties. Then also something was learned....

[Edwin G. Pettis]

Yes, the range of TCR, while controlled by the wire's specified TCR (my specification) keeps the resultant TCRs within the specified 0±3PPM/°C, there are other process factors involved which cannot be controlled tightly enough (at least currently) to guarantee an even smaller TCR range than the ±3 PPM/C, the fact that I can control those parameters well enough to ensure a majority of the TCR is within ±1PPM/°C without tweaking or cherry-picking gives credit to the resistor's design.

You are correct in that measuring very low TCRs is difficult particularly when a high quality resistance bridge is not being used, there are a lot of tiny gremlins that can sneak into the measurements that create error and at times, they are not presented to the operator.  Using a bridge minimizes the number of parameters that must be tightly controlled for accuracy, the two most important is measurement accuracy and temperature, these two parameters generally have excellent repeatability and once any other sources of error are identified and/or eliminated (such as thermal EMF), the TCR becomes relatively easy to measure with very small error.  Using indirect methods to measure TCR unfortunately increases the likely hood of error sources as there are more of them to control than with a bridge.  The biggest advantage of the resistance bridge is that it presents an direct reading of the resistance change very accurately, the temperature then only requires repeatability as absolute accuracy is not needed.  Only repeatability of the temperature points and most thermometers are very good at that.  Personally, I still use mercury thermometers, they are accurate and their repeatability is quite high to the point of almost eliminating any error.

I am not surprised at your TCRs, in fact I would have been a bit surprised if they had been between 2 or 3 PPM/°C.

For a bit more information, check my post over in the T.C. Measurements today.

[janaf]

More measurement results attached.

- Ultrohm P (Pettis) 1K, same one for the third and last? time.

- S102K foil 1K. Measured 0.5ppm/C (specified 1ppm/C) and seems very linear at this range, but has some hysteresis? Actually, I made a first measurement, not shown, with higher hysteresis, then cycled this one 25-75C a couple of times and it seems to have helped. I have a batch of the S102K so I can measure several to see what the spread looks like. It's specified 0.05% but the first one measured at +0.1%. However, its a NOS batch, could be decades old....

[Andreas]

Actually, I made a first measurement, not shown, with higher hysteresis, then cycled this one 25-75C a couple of times and it seems to have helped.

Hello Jan,

could be some humidity effect which is baked out at higher temperatures.
On voltage references they give also the hysteresis from the 2nd or 3rd temperature cycle.
But this is not honest since under room temperature conditions the epoxy will soak humidity from the environment.

With best regards

Andreas

[lars]

As Andreas I guess it is some humidity effect on the S102K. I have two S102JT 1K that I have followed about 15 years. They show a variation of about 20ppm for a seasonal change of humidity between 25 and 60%RH. I remember it was difficult to find the temperature coeficient due to "hysteresis". I also have two S102JT 100ohm that has maybe 5ppm seasonal variations and two S102JT 10K that has about 20ppm seasonal variation for the same variation of humidity.

Lars

[lars]

The 1K0 foil resistor I measured the other day looks really good. My DMM say it's 21 ppm above 1K at 23C. So I'm considering keeping it as a reference, instead of putting it in a LTZ1000 circuit.

The TCR of the resistor is negative ,so by ovenizing it, elevating temperature, it could be possible to trim it to "exactly" 1K0. Trimming by 20 ppm does seem to require quite high temperature, but as my DMMs have not been calibrated for years, I don't know how accurate the 21 ppm offset is....

Hello Jan,

I have a 1kohm VHP203 resistor that is -13ppm+-10ppm with +0.2+-0.2ppm/C that I have checked against a calibrated Fluke 5700A and Fluke 8508A. If we happen to live close to each other in Sweden we can perhaps meet. Please contact me off list.

For info the 1kohm VHP203 drifted about -5ppm the first year and the next four years about -2ppm (0.5ppm/year) against a GR1440 resistor I have. As my resistor comparator is a HP3456A, my resolution is limited to 1ppm but as I check them several times a year the statistics helps a little to get a better estimate of the drift.
One guess is that the first year drift was because I soldered wires to the resistors without any heat clamp. I also have four other VHP203 that drifted in the same way except a 100ohm that has about double the drift (with temperature coefficient between -0.1 and -0.9ppm/C). The year after I soldered the VHP203 I soldered two Alpha HK 10kohm hermetically sealed resistors that seemed very similar to the VHP. Now I used a heat clamp. They have drifted less than -2ppm against the GR1440 the last 4 years!

If your resistor is 20ppm high it should be possible to parallell it with 50Mohm. As the sensitivity is only 1:50000 for that parallell resistor(s) it should be ok with standard metal film types.

Lars

[babysitter]

If your resistor is 20ppm high it should be possible to parallell it with 50Mohm. As the sensitivity is only 1:50000 for that parallell resistor(s) it should be ok with standard metal film types.

Lars

Right this week at work I did a similar thing, 120R || (2K+390R) using a Vishay S102 as 120R and a UPW50 as 2K, and a found-in-a-junkbox 390R to make a little no-brain lab standard in a box simulating a Pt100 close to body temperature.

[babysitter]

If your resistor is 20ppm high it should be possible to parallell it with 50Mohm. As the sensitivity is only 1:50000 for that parallell resistor(s) it should be ok with standard metal film types.

Lars

Right this week at work I did a similar thing, 120R || (2K+390R) using a Vishay S102 as 120R and a UPW50 as 2K, and a found-in-a-junkbox 390R to make a little no-brain lab standard in a box simulating a Pt100 close to body temperature.

Also, I abused eevblog knowledge (pointer to the BPR resistors) to build a small oven...

[janaf]

Hello Jan,

I have a 1kohm VHP203 resistor that is -13ppm+-10ppm with +0.2+-0.2ppm/C that I have checked against a calibrated Fluke 5700A and Fluke 8508A. If we happen to live close to each other in Sweden we can perhaps meet. Please contact me off list.

For info the 1kohm VHP203 drifted about -5ppm the first year and the next four years about -2ppm (0.5ppm/year) against a GR1440 resistor I have. As my resistor comparator is a HP3456A, my resolution is limited to 1ppm but as I check them several times a year the statistics helps a little to get a better estimate of the drift.
One guess is that the first year drift was because I soldered wires to the resistors without any heat clamp. I also have four other VHP203 that drifted in the same way except a 100ohm that has about double the drift (with temperature coefficient between -0.1 and -0.9ppm/C). The year after I soldered the VHP203 I soldered two Alpha HK 10kohm hermetically sealed resistors that seemed very similar to the VHP. Now I used a heat clamp. They have drifted less than -2ppm against the GR1440 the last 4 years!

If your resistor is 20ppm high it should be possible to parallell it with 50Mohm. As the sensitivity is only 1:50000 for that parallell resistor(s) it should be ok with standard metal film types.

Lars

Hello Lars.
I do not really trust the 20ppm high as it's based on my DMMs which have not been calibrated for years. So I should have it (mod the resistor) calibrated before any mod, to check my DMMs. As the hermetic seems to have very low hysteresis, I'd could keep it at room temperature, only fire it up if/when really needed. They are specified to drift typical? 2ppm/6 years.
Jan

[janaf]

Actually, I made a first measurement, not shown, with higher hysteresis, then cycled this one 25-75C a couple of times and it seems to have helped.

Hello Jan,

could be some humidity effect which is baked out at higher temperatures.
On voltage references they give also the hysteresis from the 2nd or 3rd temperature cycle.
But this is not honest since under room temperature conditions the epoxy will soak humidity from the environment.

With best regards

Andreas

As I have a batch of these I can treat / measure different ways to see if there is a difference. Time allowing?