kinds of temp ranges not only requires really good sensors but often it
needs to be thermally bonded to what's being measured and often with some
block of material that's makes swings in temperature much more subtle.
Even a breeze in the lab or changes in the AC or heat from area causes big
swings.  Everything needs to be contained.  Measuring temp that way when
doing sub ppm voltage measurements isn't trivial.

I'd be interested too if someone has built a better mouse trap for this?  I
keep my HP3458A right on top of my older 3456A.  The 3456A is still an
amazing piece of gear and GPIB is the way to go with it.  I bought a
Conductus LTC-21 temperature controller for doing temperature
tests... temperature resolution of the LTC-21 is 1 part in 51200.  I'm
planning on using PT1000 RTD with it.  It's one of those projects on the
shelf for now but I've been slowly collecting the special connectors for
the back if it although you can just stick some wire in there.

Bill

On Fri, Sep 25, 2020, 8:29 PM Bob Albert via time-nuts <
time-nuts@lists.febo.com> wrote:

This is the document and the correct P/N that Bert meant:

Application Note 3
Applications for a Switched-Capacitor Instrumentation Building Block
Jim Williams, July 1985
page AN3-6, Precision, Linearized Platinum RTD Signal Conditioner

https://www.analog.com/media/en/technical-documentation/application-notes/an03f.pdf

and

Dual Precision Op Amp (LT1013)

https://www.analog.com/media/en/technical-documentation/data-sheets/LT1013-LT1014.pdf

/tvb

On 9/26/2020 3:02 AM, ew via time-nuts wrote:

A thermistor should do the job.  You can buy them in SMD packages and
down to 0.1% accuracy.
How much resolution you get depends on the measurement range and the
ADC you are using.

A platinum RTD would be another candidate but requires more signal
conditioning.
In 3-wire or 4-wire probe configuration you can compensate for long probe wires.

Any sensor you choose should have a thermal mass less that the item
you want to measure.
Generally, a smaller sensor means smaller thermal mass.

If you really need to resolve 0.01ºC or 0.001ºC then you also need to
pay attention to:

• sensor self-heating
• consider turning off excitation between measurements
• with a thermistor, go with a high R25,  i.e. 100kΩ which will help
  with keeping self-heating low.
• Temperature coefficient and environment of the thermistor's series resistor.
• Stability of the supply/reference voltage.

BTW, 0.1% resistors are sensitive to static discharge. A zap can
easily produce a 0.5% change!

Having really good and stable thermal contact is essential.
The item you are measuring and the sensor should be in an isolated
micro environment.
Airflow or proximity to anything of a different temperature will
potentially cause a temperature gradient between sensor and the
measured item.

All passive temperature sensors require some sort of linearization,
but that could be done away from the sensor or during post-processing.
For thermistors search for the "Steinhart Hart equation".

I am not aware of active (smart) sensors with have better than 0.1ºC
resolution, but I also have not done any significant search lately.

Cheers

Manfred VK3AES

On Sat, Sep 26, 2020 at 12:46 PM Hal Murray hmurray@megapathdsl.net wrote:

--
Manfred  VK3AES

Hi

Going back a bit to get closer to the original request …..

Indeed a thermistor is the “high resolution” king of the hill when it comes to temperature
measurement. Resistance change of 3% (30,000 ppm) per degree is not uncommon, you can do better …
They come in all sorts of packages for not a lot of money. eBay will happily supply you with
a ton of parts. Go with high resistance parts to reduce the self heating issues mentioned below ….

Since the original request was not concerned about accuracy ( only resolution) the cheap
glass body eBay parts may do just fine. You would need to do a bit of characterization /
curve fitting to work out what’s what.

If you are after 0.001C resolution you would need to measure resistance to 30 ppm. That
can be done. Indeed a bridge setup and looking at voltage is a more normal way to do this.
For narrow ranges an op-amp on the bridge will give you lots of voltage. The main issue is
indeed living with the range limitation.

Yes this is a bit of “cobble it together” nonsense. You can indeed get a pretty good setup that
will give you mili C resolution and tens of mili C accuracy. Best guess is that a single sensor
system you can trust will be over $5K. The sensors will be giant RTD probes that are a bit
useless for looking at small devices …. sorry about that ….

======

Since this is time nuts ……

One can build an R/C oscillator with a thermistor. You can measure the period of the output
and get temperature that way. Resolution (obviously) is not going to be hard to get with a
Time Nut grade counter. Don’t bother … the jitter is going to get in the way long before you get
to 0.01, let alone 0.001C ….

Bob

Answered my own question: Ametherm ACC-003 from Tim Hughes’ 2-02-2019 post
to the HPAK Equipment group (thanks, Tim!).

On Sat, Sep 26, 2020 at 10:25 AM Jeremy Nichols jn6wfo@gmail.com wrote:

Jeremy Nichols
Sent from my iPad 6.

• Stability of the supply/reference voltage.

If the reference voltage is also the reference of the ADC, its stability
and precision are much less of a concern, noise will pretty much be the
only issue.

Didier KO4BB