Sanding DS18B20 to improve its thermal response time?

Crazy question -- has anyone tried to sand/file/whatever the TO92 package in order to improve the ds18b20 thermal response time? Should reduce the thermal mass itself and the epoxy layer between the silicon and the measured surface..

Yes I know that small Pt100 or thermocouple will be much more suited for the task, but 18b20 is so standard and requires no frontend. It is really tempting to hack it a bit.

The minimum time between readings with the DS18B20 is directly related to resolution of the readings. It takes 750ms to get a 12 bit reading, less time with less resolution (see the data sheet). Shaving the sensor will not affect that time.

Whether or not sanding the sensor will improve response would probably best be determined by experimentation.

yes, surely the sensor logic will not be affected.. unless I shave too much :slight_smile: Just hoped someone did that already and can share the experience.

btw no way it can settle in 750 ms to 12 bits. Just to compare, SMT172 has the temperature response time mentioned in its datasheet, and for TO-18(metal can) "in an aluminium block" it's 0.6 sec to the 63% level.

Its epoxy, its a very poor conductor of heat whatever you do. If you need rapid readings go
for a flexi thermistor like one of these: Semitec JT Ultra-Thin Film NTC ThermistorI ATC Semitec

MarkT:
for a flexi thermistor

6 sec response time in still air is not bad, indeed.

It's a shame ds18b20 has nothing on that in its datasheet.

rapucha:
It's a shame ds18b20 has nothing on that in its datasheet.

That's because its available in several packages. The thermal properties of glass-filled epoxy as used
for TO92 encapsulation can probably be gleaned from the internet and stuck in a finite-element package
to simulate its performance.

Its probably easier to measure the time constant though, just let one equilibrate at room temperature and
then poke it through a fridge door seal and graph the response. That will include the free-air thermal resistance,
for liquids its easy to do with warm and cold oil baths.

MarkT:
That's because its available in several packages. The thermal properties of glass-filled epoxy as used
for TO92 encapsulation can probably be gleaned from the internet and stuck in a finite-element package
to simulate its performance.

Its probably easier to measure the time constant though, just let one equilibrate at room temperature and
then poke it through a fridge door seal and graph the response. That will include the free-air thermal resistance,
for liquids its easy to do with warm and cold oil baths.

Yeah.. unlike Dallas or Maxim, AD were not lazy and did that measurements, at least for their TMP35 analogue sensors. The datasheet contains all that data for var. cases, and a simple thermal model.

3 sec in stirred oil bath.

Interestingly, they believe the case thermal capacity is the limiting factor in the thermal response time "in all practical cases". Which makes me think that shaving it is worth trying.

rapucha:
btw no way it can settle in 750 ms to 12 bits.

Really?

Nick_Pyner:
Really?

Check this AD's doc: https://www.analog.com/media/en/technical-documentation/data-sheets/TMP35_36_37.pdf I bet the same is valid for just any TO-92 case.

So it's 3 sec in an oil bath at the 63% level. 63%, I repeat. What good 12 bits makes you if the real, physical accuracy is that low?

Bear in mind this 63% is not anyhow accurate too, so it is impossible to calculate the real temperature just multiplying by some factor. I mean, technically it's easy, and will provide some good estimate -- but nothing close to 12 bit accuracy.

Any temperature sensor measures only one thing -- its own temperature. It's often underestimated how problematic could it be to achieve the equilibrium with the object we want to monitor.

rapucha:
Check this AD's doc:

l'm rather more inclined to read data from the manufacturer of the DS18B20.

Nick_Pyner:
l'm rather more inclined to read data from the manufacturer of the DS18B20.

I'd like to, but they decided not to cover the topic of thermal properties. At least I did not find anything. If you have something, please share.

Explanation and good advice is in reply#3. There has never been any suggestion that DS18B20 isn't up to what is supposed to do and, if it doesn't do what you want, it simply means that you have bought the wrong sensor.

Nick_Pyner:
if it doesn't do what you want, it simply means that you have bought the wrong sensor.

Can't beat this argument. Just note, that thermal response for ds18b20 is not specified in the documentation. No thermal parameters, actually.

Which brings the question -- what is this sensor good for?

I think that they are made to use in HVAC systems where many can be spread far and wide on a single wire. Their thermal response characteristics are not as relevant to that industry. The temperatures that they are made to measure do not change very fast. Choose the sensor for the job if you need specific characteristics.

groundFungus:
Choose the sensor for the job if you need specific characteristics.

Thanks, Captain.

Is that intended to be an insult?

Now I am scared.

The OP has asked for this thread to be deleted as

the topic failed to produce the answer to the asked question,

but as it has several replies containing advice I am going to leave it, but it will be locked to prevent further replies