I want to monitor the temperature of a large peltier element's heatsink to control a fan. I think I'm going to use a DS18B20.

The 3 pin TO-92 looks like it's plastic, so I'm already worried about low thermal conductivity.* However, how should I secure this to ensure that the chip has the best thermal conduction? Thermal paste crossed my mind, but it'd be pointless to add that with plastic unless there is a big gap there.

I haven't picked out a heatsink yet, but there are plenty of them with a fairly large, flat surface that I could attach the sensor to if I needed. However, I can't glue it onto the heatsink because most glue doesn't conduct heat.

*They make thermally conductive plastics that can insulate (electronically), so the point about the plastic might be moot.

  • \$\begingroup\$ " pointless to add that with plastic unless there is a big gap " air is such a bad heat conductor that even the tiniest air gap will conduct heat much much worse than with thermal paste. \$\endgroup\$
    – PlasmaHH
    Nov 28 '14 at 20:43
  • \$\begingroup\$ Thermal paste. Stronger than glue. And the entire point of a temperature sensor is that it can conduct heat and measure it. Your over thinking this. \$\endgroup\$
    – Passerby
    Nov 28 '14 at 23:05
  • \$\begingroup\$ @Passerby but most thermal pastes aren't adhesive, right? \$\endgroup\$ Nov 28 '14 at 23:31
  • \$\begingroup\$ Tell that to every cpu heatsink ive ever had to remove. \$\endgroup\$
    – Passerby
    Nov 29 '14 at 0:10

That is one poorly defined data-sheet, there is no mention of thermal issue on what is a temperature probe! However, there is this link to the thermal aspects of these one wire devices which give you thermal conduction coefficients. This takes you through self heating effects. Your device is actually listed. enter image description here

Now all they need to do is give you the thermal mass so you can estimate the response time ...

I suspect that you'll just have to measure that. But if you are attaching to a large thermal mass as it is, that mass will dominate.

You can get thermally conductive epoxies made specifically for this task, and bond the flat side down to your plate.

  • \$\begingroup\$ Hmm... the data isn't there for that part (or its case), just an X indicating what package it is (which I already know). Nevertheless, I have something to go off now. I'll have to mess around with the response time... EDIT: Google retrieved this review: a Fast response time. You place your finger on it and immediately the reading starts to rise. \$\endgroup\$ Nov 28 '14 at 20:30
  • \$\begingroup\$ @AnnonomusPenguin Temperature change to a step input is an exponential decay type function. I have no experience with this particular sensor, but some other TO-92 sensors I've used (purely analog) have had time constants on the order of tens of seconds. This means it will take 10s of seconds to change ~63% the difference the temperature change, another 10s of seconds to get to ~86%, and another 10s of seconds to get ~95% (this was with air contact). \$\endgroup\$ Nov 29 '14 at 1:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.