Equipment for precision resistance thermometry

[bgugi]

Hello, all. I'm looking into building a rig for experiments in resistance thermometry.

Goals:
Measure resistance ratios of pt100, pt1000, and thermistors
Measure resistance ratios near unity between UUT's and a standard resistor with uncertainties better than 40 ppm (e.g. .004  for a pt100), as well as wider ranges (3:1/1:3 and further)
Multiplex standard and UUT resistances
programmable current for zero-power compensation
Electrical test bench cost <700$, used equipment preferred
PC interface, for fully-automatic control.

Architecture selection:
As much as I'd like to play with one, there is a very small used market for automatic resistance thermometry bridges, and their prices reflect that (isotech f- series, fluke "super-themrometers" etc). Bridges requiring manual dialing of decade boxes to balance a circuit are right out, especially if they require manual lead/impedence/thermovoltage compensation. This leads me to DMM's with some supporting hardware. I'm not going to lie, I'm also a bit biased towards purchasing a metrology DMM, as this will also be my first DMM purchase with more counts than I can do with my hands, and will of course lead into other experiments.

I'm planning to build a current-reversing DC bridge, similar to the fluke 1570a/1595a utilizing a programmable current source , a switching/control unit (like an HP 3488a), and some pi/microcontroller action to tie it together.

Current Source Selection:
I think this will likely be scratch-built possibly tucked away in the 3488a. Absolute value, offset voltages, current leakage, etc will have a very small effect. The driving factors will be noise and very short-term drift (on the order of 1-20s). I will likely leave current reversing to the switching unit.

DMM selection:
The DMM should have excellent INL, as this will be the dominating uncertainty component. I think it would be preferable to have a DMM with built-in 100mV ranging, as I doubt I could build a signal amplifier that outperforms the DMM's internal range. Range should have similar INL when measuring negative voltages, as this will allow compensating for thermal voltages without allowing the UUT to cool off from its self-heating. Unfortunately, most meters in my price range will only specify overall 24h specs, each ratio measurement will actually only occur over 1-20s.

HP 3458a:
If I were a rich man... Sadly I don't think my fiance would be too happy with me taking a loan out on my car for my hobby...

HP 3457a:
pros:
- Native 300mV range allows nominal currents of 1mA over most reasonable temperature measurements
- Has excellent 24h specs, ~30 ppm on 100 mV range
- secret 7th digit

HP 34401a:
pros:
- AD linearity spec (2 ppm rdg + 1 ppm range), with data from others on this forum showing sub-ppm INL in the the native (10 V) range, with a fairly smooth curve
- Transfer spec (15 ppm of 100 mV range for readings within 10% of nominal)

What are your thoughts? are there other units I'm forgetting? data I haven't found yet?

[Andreas]

perhaps a specialized unit like this:

https://www.picotech.com/data-logger/pt-104/pt-104-specifications

with best regards

Andreas

[Kleinstein]

With used instruments one has to see what is available in a reasonable time frame.

The HP3457 may also include a relay scanner / multiplexer, which could be handy. Similar there are some Prema DMMs with scanner available relatively cheap.  The linearity may not be as good, but probably sufficient.

[Dr. Frank]

I recommend to first specify the resolution and accuracy of the temperature you want to achieve, like 1mK resolution and +/- 10mK accuracy, or so, and then calculate, how accurate your resistance measurement has to be.
This initial T specification is directly related to the PT100 sensors (NTCs will never be as precise) you have to buy. and to their price, of course.
Common ones usually have about 0.3 K accuracy, maybe a bit better, so for higher accuracy, you'd need calibrated ones, with an individual R(T) table, which you would then have to process by a special program.
Lake Shore offers calibrated PT100, maybe even to 1mK uncertainty, also Fluke bought a company which deals with temperature metrology and also offers all that stuff.

Using standard PT100, the requirements for precise resistance measurement is not that high.
You only need 4 wire measurement and offset compensation features, as especially in thermometers, big thermal voltages occur, and PT100 has very low resistance change vs. temperature change, i.e. ~0.39Ohm / K, so 1mK resolution @ 1mA equals 0.4µV, which ultimately requires compensation for thermal voltages.

If you'd  use a 34401A, you then will need an external precision 1mA DCI source, which either can be switched on/off or reversed, under program control.
Or you directly buy a DMM with Offset Compensation and PT100 table built in, like the 34465A. Again, more precise PT100 need specific R(T) table.
The 3458A also has all these features, surely, but its internal PT100 tabes are outdated, as in between these 30 years, the standardized Pt100 tables have been changed. I don't know how much, and if it's really relevant.

If you build your own current source, it needs to be precise to about 0.01%, as a rule of thumb, what companies like Lake Shore also offer for their own T-sensors.

I've done that for all kind of temperature sensors, so you'd perhaps also need 10µA and 100µA, especially to reduce the self heating effect of your sensor.

I've published my circuit somewhere here on eevblog.. that source initially could be easily digitally interfaced because the switching is done by reed relays.

Frank

[branadic]

I wonder if temperature measurement down to 1mK resolution does really make sence. With respect to this video:



ITS-90 is off, so each resistive temperature measurement is also off. Hopefully temperature definition will be updated in recent years.

-branadic-

[bgugi]

perhaps a specialized unit like this:

https://www.picotech.com/data-logger/pt-104/pt-104-specifications

Accuracy is close, and only rated at 100  .

Dr. Frank: You've brought up a number of things i've considered. My 40 ppm / .004  figure is equivalent to the 10 mK accuracy i'd like, and resolution being higher than accuracy is necessary to realistically hit that uncertainty. The goal of the project is to experiment with rtd's and fixed point cells. Since I'd be doing my own characterization and interpolation calculations, I'm not worried about pre-calibrated probes, or instruments having interpolation calculations/tables on the instrument. is this https://www.eevblog.com/forum/metrology/recycling-of-precision-current-source-noise-reduction-for-low-burden-shunts/ the supply you were referencing? I also looked at the lakeshore 110, as i might be able to add in a substitution circuit for the programming resistor. I wanted one that I can program so that I can make measurements of self-heating effect. NTC's can be surprisingly precise and stable... but they rely on interpolation from an RTD to actually perform their calibrations (due to their small range)

I wonder if temperature measurement down to 1mK resolution does really make sence. With respect to this video:

{Video}

ITS-90 is off, so each resistive temperature measurement is also off. Hopefully temperature definition will be updated in recent years.

-branadic-

I've actually seen that video. Errors between the ITS and thermodynamic temperature are very interesting... especially with the recent redefinition of the K related to the boltzmann constant. Even without that, we're pretty overdue for a new ITS scale (previous scales were standardized in '48 and '68). Even with the inherent issues, if/when we get a new scale, or solid corrections between its-90 and the thermodynamic K... I should be prepared for convert my data to whatever new scale becomes the international standard... That's still years away, though.

ONE MORE THING... I forgot to mention this in my original post... does anyone know where I can get wirewound (not thin-film) RTD elements with 4-wire leads and a .00392 alpha? I don't have the equipment/lab space to do high-purity autogenous welds...

[Vgkid]

 Look into omega, heraeus, sensing devices, burns engineering

[branadic]

Companies like Delta-R and A+S are making them on request. But you can get wirewound PT100 also at RS Components, though with 2 wires only, but no problem attaching 4 wires to them: 2938480, 6117867. I used them in my thermal chamber.

-branadic-

[awallin]

ITS-90 is off, so each resistive temperature measurement is also off. Hopefully temperature definition will be updated in recent years.
-branadic-

huh? The SI system revision from May 2019 does redefine temperature!? It's directly related to the velocity/energy of atoms, so you just need to measure the velocity distribution and you are done (given that the velocity distribution relies on the meter and second and can be realized precisely, and you might need the mass of the atoms also..)

(in practice, before there are primary thermometers across the range you want to work, one needs practical scales like ITS-90 and similar...)

[mzzj]

Companies like Delta-R and A+S are making them on request. But you can get wirewound PT100 also at RS Components, though with 2 wires only, but no problem attaching 4 wires to them: 2938480, 6117867. I used them in my thermal chamber.

-branadic-

Those are alpha 0.00385 sensors conforming to IEC-751 industrial Pt100 standard.
"lab grade" pure platinum 0.00392 alpha are  harder to find.

"secondardy standard" from Fluke is around 2000eur/usd 
https://eu.flukecal.com/products/temperature-calibration/probes-sensors/platinum-resistance-thermometers-prts/5626-5628-seco?quicktabs_product_details=1

 

[mzzj]

Some industrial IEC-751 standard Pt100 sensors are good to maybe 20mK if calibrated with limited range like 0...300cel. (and taken first trough some preconditioning steps)
Error is quite curvy and hard to predict so these would need calibration every 30...50cel or so.

[Dr. Frank]

Dr. Frank: You've brought up a number of things i've considered. My 40 ppm / .004  figure is equivalent to the 10 mK accuracy i'd like, and resolution being higher than accuracy is necessary to realistically hit that uncertainty. The goal of the project is to experiment with rtd's and fixed point cells. Since I'd be doing my own characterization and interpolation calculations, I'm not worried about pre-calibrated probes, or instruments having interpolation calculations/tables on the instrument. is this https://www.eevblog.com/forum/metrology/recycling-of-precision-current-source-noise-reduction-for-low-burden-shunts/ the supply you were referencing? I also looked at the lakeshore 110, as i might be able to add in a substitution circuit for the programming resistor. I wanted one that I can program so that I can make measurements of self-heating effect. NTC's can be surprisingly precise and stable... but they rely on interpolation from an RTD to actually perform their calibrations (due to their small range)



I've actually seen that video. Errors between the ITS and thermodynamic temperature are very interesting... especially with the recent redefinition of the K related to the boltzmann constant. Even without that, we're pretty overdue for a new ITS scale (previous scales were standardized in '48 and '68). Even with the inherent issues, if/when we get a new scale, or solid corrections between its-90 and the thermodynamic K... I should be prepared for convert my data to whatever new scale becomes the international standard... That's still years away, though.

ONE MORE THING... I forgot to mention this in my original post... does anyone know where I can get wirewound (not thin-film) RTD elements with 4-wire leads and a .00392 alpha? I don't have the equipment/lab space to do high-purity autogenous welds...

Yes, that's about the circuit which I have used for my thermometry 30 years ago. You only have to add a digital interface to steer it via PC.
It's relatively cheap, and by aid of a 3458A, can be very accurately calibrated to several ten ppm @ 1mA.

Lake Shores current sources are quite mediocre (~ 0.05%) and very expensive.

In our physics institute, we used Time Electronics current sources, like the modern model 1024:
https://www.timeelectronics.com/portable-voltage-current-instruments/1024-dc-current-calibrator/
which gave 0.02% accuracy, probably at a similar price as Lake Shoe, but w/o interface. (Really? 2600$  )

The 3458A that I bought in 1989, probably still had the old IPTS-68 scale built inside the RTD85 math function, so it also lost its temperature calibration on 1.1.1990, besides the Volt and Ohm, when SI-1990 and ITS-90 were both released on that date.
As its firmware probably was never really updated concerning this RTD table, these differences of several 10 mK between IPTS-68 and ITS-90 exist up to date.

I never used the 3458A for direct PT100 temperature measurements, but a 34401A, with my precision current source box, and calibrated sensors from L.S., like PT100 and DT470 diode, to measure temperature down to 2 Kelvin.

The ITS-90 scale is still valid for mise en pratique, obviously, when you go to the bipm page.
The NPL video is very interesting, as it demonstrates where the deviations and kinks in the standard PT100 curve come from.

Frank

[notfaded1]

ITS-90 is off, so each resistive temperature measurement is also off. Hopefully temperature definition will be updated in recent years.
-branadic-

huh? The SI system revision from May 2019 does redefine temperature!? It's directly related to the velocity/energy of atoms, so you just need to measure the velocity distribution and you are done (given that the velocity distribution relies on the meter and second and can be realized precisely, and you might need the mass of the atoms also..)

(in practice, before there are primary thermometers across the range you want to work, one needs practical scales like ITS-90 and similar...)

This was all written about in the paper I posted about the other day:  https://www.eevblog.com/forum/metrology/the-ampere-and-the-electrical-units-in-the-quantum-era/  Kevin was also redefined.

[arcnet]

ONE MORE THING... I forgot to mention this in my original post... does anyone know where I can get wirewound (not thin-film) RTD elements with 4-wire leads and a .00392 alpha? I don't have the equipment/lab space to do high-purity autogenous welds...

Some manufacturers I know are
TDI http://www.t-d-i.co.uk/
Netsushin https://netsushin.co.jp/en/ (very small wirewound PT10 0.4 mm diameter, 0.5 mm length, 4-wire construction, but also normal PT100, high temperature and/or high stability ones, glass and ceramic)
Hayashidenko https://hayashidenko.co.jp/en/rtdelement.html (glass and ceramic)

[Henrik_V]

..
Those are alpha 0.00385 sensors conforming to IEC-751 industrial Pt100 standard.
"lab grade" pure platinum 0.00392 alpha are  harder to find.

"secondardy standard" from Fluke is around 2000eur/usd 
https://eu.flukecal.com/products/temperature-calibration/probes-sensors/platinum-resistance-thermometers-prts/5626-5628-seco?quicktabs_product_details=1

Some weeks ago I visited a temperature calibration lab... they told me that Fluke (and others on the market for SPRT) isn't able to deliver the high end range ... it seems hard to get the platinum in the quality needed. 
 

[e61_phil]

A couple of month ago I had to measure temperature gradients in a vacuum application with lots of Pt100s. I used an Agilent 34972A and a Agilent B2902A SMU (because I had it). I had all the Pt100 and a 100R Vishay Z201 in series (to keep the number of vacuum feedthroughs low, but have 4W connection). The SMU was switched between +800µA and -800µA to compensate offsets. I used less than 1mA to stay below 120mV (100mV DCV Range of the 34972A).
This way everything was relying on the stability and accuracy of the Vishay resistor (measured with our Fluke 8508A before). The current source and also the 34972A needs to be short term stable only.

The 34970A and 34972A are 34401As with relays in priniciple. They are often not expensive on ebay. There are also simple relay modules available for these instruments which could be used to reverse the polarity of a DIY current source.

[nfmax]

The 34970A and 34972A are 34401As with relays in priniciple. They are often not expensive on ebay. There are also simple relay modules available for these instruments which could be used to reverse the polarity of a DIY current source.

IIRC these instruments also have an offset-compensated ohms function, which the 34401A does not - presumably because of the relays (and the measurement bus backplane connectors)

[e61_phil]

The 34970A and 34972A are 34401As with relays in priniciple. They are often not expensive on ebay. There are also simple relay modules available for these instruments which could be used to reverse the polarity of a DIY current source.

IIRC these instruments also have an offset-compensated ohms function, which the 34401A does not - presumably because of the relays (and the measurement bus backplane connectors)

Yes, they have. But I wasn't able to use it due to the number of connections.
The other problem is to stay in one ohms range at higher temperatures with a Pt100.

[nfmax]

Well yes, OCO is not applicable in the setup you describe, but it is there for setups where it can be used. It's a pity these are only really 6 1/4 digit meters, not proper 6 1/2 digits. Otherwise they are great - I've been using them in different jobs for the last 20 years!

[mzzj]

..
Those are alpha 0.00385 sensors conforming to IEC-751 industrial Pt100 standard.
"lab grade" pure platinum 0.00392 alpha are  harder to find.

"secondardy standard" from Fluke is around 2000eur/usd 
https://eu.flukecal.com/products/temperature-calibration/probes-sensors/platinum-resistance-thermometers-prts/5626-5628-seco?quicktabs_product_details=1

Some weeks ago I visited a temperature calibration lab... they told me that Fluke (and others on the market for SPRT) isn't able to deliver the high end range ... it seems hard to get the platinum in the quality needed. 
 

Could be. But better SPRT delivery times have been rather long always.

We use only the "inexpensive" secondary SPRT's that are in 2-3 keur price range, primary SPRT's are nowadays probably close to 10 keur including calibration. 

[nfmax]

Some industrial IEC-751 standard Pt100 sensors are good to maybe 20mK if calibrated with limited range like 0...300cel. (and taken first trough some preconditioning steps)
Error is quite curvy and hard to predict so these would need calibration every 30...50cel or so.

The best thermometer I have here at home is one of these - an Isotech 935-14-13 https://www.isotech.co.uk/assets/uploads/Data%20Sheets/NEW%20Angled%20Semi%20Standards.pdf. However mine does not have individual calibration, so I am dependent on its adherence to the IEC 751 curve.

[HighVoltage]

I have used the Keithley DMM7510 in combination with a calibrated Fluke RTD 5616-12 for precision resistance measurements.

The RTD was for several hours in the middle of 1 liter of water in a clear glass container to reach equilibrium.
The DMM7510 showed a flat line.

Then I used a small LED flashlight, to shine a light at the tip of the RTD from about 1m away.
Probably not very accurate but definitely precision for comparison.