High resolution thermometer

[Catoss]

I have a need for a high-resolution temperature data acquisition for chemical process analysis.
The temperature spectrum of interest is 273-373K. Absolute accuracy is not of concern beyond 0.1K but the resolution I need is 1mK or let's say 10mK noiseless.
(for the design phase I can always scale back if the lesser resolution is sufficient).

I am a beginner when it comes to analog data but looking at products like the one in the link below resolution should be doable.
https://www.traceable.com/6412-traceable-platinum-ultra-accurate-digital-thermometer.html#product_tabs_additional_tabbed

Is using an RTD system a bad idea?
I have found a few reference designs i understand notably
https://www.ti.com/lit/ug/tidu969/tidu969.pdf?ts=1607286543809&ref_url=https%253A%252F%252Fwww.google.com%252F
https://www.analog.com/media/en/reference-design-documentation/reference-designs/CN0337.pdf
https://www.analog.com/media/en/reference-design-documentation/reference-designs/cn0383.pdf

For the TI design delta of 100K would need a resolution of 17ENOB for 1mk resolution is this achievable with 24bit ADCs and that design.

Thanks in advance.

[TimFox]

Is your resolution requirement to see a reasonably quick change of 1 to 10 mK, or do you need a long-term stability on the order of 1 to 10 mK?  The former may be dominated by noise, and the latter by DC drift.

[HighVoltage]

I have been using a Keithley DMM7510 with a calibrated Fluke RTD probe.

[Benta]

Is your resolution requirement to see a reasonably quick change of 1 to 10 mK, or do you need a long-term stability on the order of 1 to 10 mK?  The former may be dominated by noise, and the latter by DC drift.

Very pertinent question! The OP needs to let more out of the bag.

[Catoss]

Thanks all forgot to talk about that aspect my bad.

The stability required is a maximum of 5min (projected reaction time). Due to noise being the primary concern I assume ratiometric measurement is the way to go.

As a bit of extra info, the system is based on a flow-through reactor that is well isolated from the environment, and the sensor sees a constant flow of liquid whose temperature is changing based on reaction parameters happening "upstream".

The problem with Benchtop DMMs is I do not have a good idea how to log the data from DMM in sync with other sensor data already running on the experiment (other parameters).

Thx again

[jmelson]

I have a need for a high-resolution temperature data acquisition for chemical process analysis.

Will the data logging equipment be far from the sensor?  RTDs require 3-wire or 4-wire connection for long cable runs.
My preference for this would be to use a current-output sensor such as the Analog Devices AD592 sensor, which outputs a current proportional to absolute temperature, so 273 uA at 273K.  I have used these in a remote temperature sensing application, although not anywhere near the resolution you are asking for.  I use a simple one op-amp current to voltage converter circuit to precondition the signal for an ADC.
You will have to calibrate your sensor(s) with something that can be trusted to deliver your level of accurate reference temperature.
That, in itself, is a SERIOUS measurement/calibration issue.

Jon
Jon

[jbb]

The DMM7510 looks like a very nice instrument. If you’re happy with the performance and can keep your lab mates from thieving it, then maybe you should be asking “how can i synchronise the DMM with my other test equipment?”

I expect it will have some communications - USB or LAN? - which you can use for remote control. There may also be a ‘trigger’ input on the back which can be configured to, for example, “measure the temperature every time there’s a voltage pulse.”

You can make your own digitiser, of course, but it could be quite time consuming to get it just right.

[Vojtech]

There are also DAQ modules  for instance AD24USB: http://janascard.cz/aj_Vyrobky.html#AD24USB, its main advantage against other modules and multimeters is its switched current source which eliminates parasitic thermoelectric voltages and input offset, so resolution 0.1 mK is possible with Pt100 and it also have more inputs so more sensors can be measured. I am a manufacturer so I am biased a little bit  .

[nfmax]

You haven't mentioned it, but from the description of your application the thermal response time of the sensor itself may be an issue. For this reason, and to obtain your required temperature resolution, I would suggest using a miniature glass-encapsulated precision NTC thermistor. A resistive sensor, this can be measured using a 2-wire connection to DVMm or preferably a data logger. A multi-channel data logger such as a Keysight DAQ970A or a Keithly DAQ6510A can record time-stamped measurements from the thermistor and may be able to record your other parameters as well.

The NTC thermistor has one of the highest 'sensitivities' - change in output (resistance) for a given change in temperature - of all thermometers.

[Catoss]

Thx for imput all of you. I was thinking about RTDs since i have found an ideal glass coated RTD for the use in our flow reactor.
This rtd can be mounted in our system without modifications since we have ports available for dosing that use 1.5mm PEEK capillary connections and we can mount sensor instead of PEEK tubing.
According to datasheet response time is fast enough.

The DAQ posted by Vojtech looks like a winner atm but since the covid situation is easing up atm i may be able to work with colegues in other parts of the uni for imput if i am approaching this from the right angle.

As far as calibration goes i have acces to certified calibration bath for the range the system would be used in.

[HighVoltage]

The German manufacturer JUMO make really nice and small precision temperature sensors, based on Platinum.
Here is a datasheet of some of their sensors I have used in the past. (Sorry German)
And their response time was not bad at all.

[EmmanuelFaure]

Achieving 1mK resolution with RTDs is doable but tedious. You should use a high stability thermistor instead, they're the way to go when you want to measure very small temperature variations. mK resolution is achievable with a very simple, very low cost circuit, kind of : Arduino + 16 bits ADC.

[Kleinstein]

For directly sensing the liquid the chemical stability of the sensor can be a big factor in the sensor choice.
Another factor to look for is the self heating of the sensor - some (especially older)  DMMs use a relatively high current. This can be a problem with PT1000.

[Doctorandus_P]

Have a look at atmospheric pressure sensors such as the Bosch BMP180 (and variants).
These chips have a pretty decent ADC onboard for the pressure measurement and also a temperature sensor for compensation purposes.

They do not quite get to the spec you want, but close, and they're very easy to interface (Just I2C).

[sam[PS]]

Have a look at atmospheric pressure sensors such as the Bosch BMP180 (and variants).
These chips have a pretty decent ADC onboard for the pressure measurement and also a temperature sensor for compensation purposes.

They do not quite get to the spec you want, but close, and they're very easy to interface (Just I2C).

One of my colleague is currently designing a project around those sensors, and from his roe the electrical interfacing is easy (as you said just I2C) but the software part is far from what i would call easy interface as those chip have so many functionalities with each configurable parameters and not all combination are valid. So all in all i would not call that "easy to interface".

Just my 2cts...

[Conrad Hoffman]

I used to use a thermometer made by the Think corp., no doubt long out of business. They were probably a small garage shop company. It had a resolution of 1 mK and was reasonably stable. What was in it? A fairly ordinary bead thermistor, a network with a bunch of pots to linearize it, and an off-the-shelf digital meter. Somewhere I still have a manual for it. I vote thermistor.

[wizard69]

The problem with Benchtop DMMs is I do not have a good idea how to log the data from DMM in sync with other sensor data already running on the experiment (other parameters).

Thx again

That throws a spanner in the works.   It is one thing to measure a temperature and hope for valid measurements over that range.   Building up an entire data acquisition system morphs this project into a much bigger problem.   Now given that many DVM's do have sync operations or capabilities, both hardware and software.   The problem then becomes just how closely do yo need to sync.   Many DVM's can operate to high resolution but they may not be fast enough for your needs, not to mention a sensor that is fast enough.

You described a process time of 5 minutes over all but did not describe how quickly you will be sampling data.   If you are talking in the Hertz range you  might get decent values out of many sensors.   However if you are sampling data fast and it is important to have temperature data that is accurate at the same time other things are happening (data being collected) then you will need you will likely need a very fast transducer.

I generally have mixed feelings when people suggest that they need temperature readings to these extreme level of precision.   The problem is everything in the acquisition loop is temperature sensitive.   That is you can't look at what is happening at the transducer and hope that no outside factors have influenced it.   In otherwords your need really good lab practices in the hopes that the results will be unmolested.   I once ran into this problem with an R&D engineer that wanted a temperature display to be calibrated "perfectly".   The problem was the equipment was very temperature sensitive to what was happening in the room.   We would get accuracy shifts every time the air handling system kicked in.   So I would expect that whatever equipment you are using is orders magnitude better than that old hardware from the 1970's, even so you want to make these measurements over a short period of time at a resolution that will likely be impacted by the environment.

This is why I like the suggestions above to go with a high resolution thermistor or other device that has high resolution.   I'd also supplement it with another transducer as relying upon just one sensor for critical data does not leave me all warm and fuzzy.   This assumes that your can get fast enough response times for your process requirements.