EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: tarun172 on April 21, 2022, 11:09:37 pm
-
Hi Everyone,
I am trying to instrument up for temperature readings on design that I am currently working on. How much of difference can you expect in temperature reading in case A:Using thermal epoxy attaches thermocouple to PCBA and case B: Using a super glue like Loctite to attach thermocouple to PCBA? All the instruments I have calibration up to date.
Thanks,
tarun172
-
Its going to depend on airflow, temperature being measured, etc. Are you measuring a specific component on the board?
Few other options:
- use thermal compound, and good tape over top to hold the thermocouple down.
- thermally conductive silicone, which is easy to remove after use
Pure cyanoacrylate (super glue) is around 0.1 W/mk: https://www.sciencedirect.com/science/article/pii/S187661961600156X/pdf (https://www.sciencedirect.com/science/article/pii/S187661961600156X/pdf)
Kafuter K-5203 silicone is claimed 1.2W/mk
Good thermal compound may be 5+W/mk
https://www.eevblog.com/forum/projects/epoxy-compound-for-holding-thermocouples/ (https://www.eevblog.com/forum/projects/epoxy-compound-for-holding-thermocouples/)
https://www.eevblog.com/forum/manufacture/glue-to-temporarily-attach-thermocouple-wire-to-fixtures/ (https://www.eevblog.com/forum/manufacture/glue-to-temporarily-attach-thermocouple-wire-to-fixtures/)
-
I am trying to measure board temperature. Yes, I did go through those posts. I was wondering if there is hard data available? How much of error in measurement in degree celsius can we expect from using super glue? and using thermal conductive epoxy let's say for K-Type Thermocouple.
Thanks,
tarun172
-
I suspect it's a balancing act. If there were no conduction down the leads and minimal surface area, almost anything would work. Since it has leads, better thermal contact is necessary. I'd be using silicon thermal grease or a filled epoxy. You can make your own thermal epoxy if you have some alumina powder, or just buy one of the many. Obviously you'll destroy the thermocouple when you remove it, and maybe the surface of the board. Or just cut it off when done.
-
We used special type K TC 30 ga to minimize conducted losses
Oméga makes silver filled paste and epoxy
jon
-
When it's just for temporary measurement/experimentation I just stick the thermocouples with kapton tape. From my non-scientific observations the measurement seems to be within a degree or so unless there's a whole ton of airflow over the board: good enough for my purposes, which is evaluating that temperatures on certain components, heatsinks, etc are within safe limits for reliability, and you don't need crazy accuracy for that.
-
Yes just the kapton tape will be within 1C, well within the accuracy spec for a type K thermocouple. If you need stronger adhesion just use a small amount of metal loaded epoxy, like Araldite steel, Pratley quickset steel or whatever is in the local hardware store. Just a small bit, half the head of a match, will hold the tiny tip in place, and should last long enough to get your temperature readings, the tiny bit will be easy to remove from the sensor and the board, so you can reuse them, or stick it down with the same blob on it to the next board.
Will be fine, just place the sensor out of airflow, probably down under a heatsink near the power device, or even right on the power tab if it is grounded. You will have to be careful if you do not want to create ground loops, so there accept lower accuracy, and place a bit of kapton tape on the board, then tape the sensor tip to this tape with more. Multiple sensors decide the most critical, ie power devices, and measure there, and the rest you just are fine within 2C anyway, just orientation will give more difference due to air flow and radiation to ambient.
-
I’ve had OK results with Kapton tape, squishy gap pat thermal materials (and a little block to push it down) but NOT cheap vinyl tape; that stuff tended to get soft and sticky and let the sensor move around.
I found that it was important to fasten the cable leading up to the probe too, so it didn’t wiggle loose. That might help with thermal conduction along the leads too.
A while ago I was in a jam and tried some thermally conductive silicone adhesives (pH-neutral cure RTV so it didn’t eat the electronics). I noted that as the thermal conductivity went up from basic RTV to thermal RTV to more thermal RTV (sorry can’t remember values :-( ) the silicone got stiffer and stiffer and more prone to cracking.
So don’t just reach for the highest thermal conductivity - there will be compromises in action!
-
Pure cyanoacrylate (super glue) is around 0.1 W/mk
Kafuter K-5203 silicone is claimed 1.2W/mk
Good thermal compound may be 5+W/mk
Except that metric assumes the thicknesses of the adhesive joint are the same. Realistically, the fast setting super glue will end up with a trivial layer of adhesive between the bead/junction and the substrate. Then you start worrying about the heat the thermal probe is transferring to/from the DUT.....
99% of people, tape it on and get close enough.
-
I take it, adding a chip thermistor or RTD to the board is out of the question, either in general, or at this point?
You want a wide enough blob that the junction sees a representative temperature, and that the leads up to it see the same temperature as well, so the cable isn't heatsinking the junction and introducing error.
For a one-off test I just glom some heatsink paste on the TC and hold it flat against the part in question with a little wiggle to settle the paste; this gets within say 10% or so for most things, and obviously it'll be worse for smaller and plastic components, and okay for larger and metallic things (like, nestled in the fin of a heatsink). I guess since you're considering epoxy, you're after something a bit more permanent than that. Same principle, just epoxy based, and with something to fix it in place while it cures.
If you don't have good thermal connection to the junction/cable, then some extra length of it should be coiled up (to sink more heat into the cable, away from the junction), and a tape covering to help insulate it from convection, isn't a bad idea. For which, any not-very-conductive tape would be fine, like, vinyl tape was mentioned above -- its low temperature handling is a weakness by itself, but it's also not very conductive, so if it's not under any stress (because it's covering an epoxy blob), it might be fine for an application like this. Polyimide (Kapton(R)) isn't very conductive either, but isn't usually very thick so YMMV.
Next best insulation, Idunno, got some paper or felt materials? (Paper as in cellulose fiber paper, would probably be fishpaper, but less combustible options would include nylon, polypropylene, etc. Not sure what you'd usually get them as; transformer winding "margin" tape perhaps?)
Mind that, you want enough insulation that the thermocouple reaches equilibrium with the board, but not so much that the board's overall convection rate is impaired, thus reading higher than usual. There's a compromise between accurate reading, and the measurement disturbing the system itself. Which, higher conductivity epoxy and smaller (fine gauge?) TC and blob helps out with, and probably anything more insulating than a layer of paper+tape would worsen.
I've also done it this way before, making up a thermistor by soldering pigtail leads to a chip thermistor, wrapping around with tape, fill the tape with epoxy (any grade; obviously filled/thermally conductive is better), then once that's cured, glob the assembly down onto the e.g. heatsink with more epoxy and secure in place with tape (which will get embedded in place itself, along with some of the lead wires to reduce wire sinking error). The extra tape wrapping ensures isolation (so it's fine for live heatsinks). Another option: do the same, but instead of filling the tape wrapping, insert the wrapped assembly into a ring terminal / lug, then fill. Or just buy such assemblies pre-made, of course!
Have mostly used thermistors, because I don't care about precise measurements, just basic overtemp and monitoring stuff, but yeh, same idea for RTD or thermocouple.
Tim
-
Take a good look at the whole picture of your temperature sensor.
If it has thick leads, then those leads may act as a heat sink, and you will me (partly) measuring the air temperature around the sensor.
If thermal conductivity between the heatsink and the temperature sensor is bad, the difference will be bigger.
Gluing a few cm of the wires to the heatsink too will keep the wires at the same temperature as the heatsink, and even when the coupling is not great, your sensor will still reach (nearly) the same temperature as the heatsink.
Heatsinks also can have a significant temperature gradient over them. There can easily be a difference of 10 degree C (or double that) between different parts of the heat sink, and it does not make sense to try do do it very accurate.
If the function is just to prevent semiconductors from overheating, then a mechanical solid connection is more important than a accurate temperature measurement.
Even if your sensor is a fair bit away from where the semiconductors heat the heatsink, the temperature difference will be somewhat static and can be calibrated out.
If however the mechanical connection is dodgy and it comes loose, then the temperature sensor will not measure the temperature of the heatsink at all anymore.
I don't like cyanocacrylate much. It is a very brittle glue.
Silicone though is both elastic, will stay elastic and has a fairly high thermal conductivity.