1

I'm using SMD based analog temperature measuremet ICs (a version of the old TMP36). It will be mounted on a tiny and thin (.5mm) PCB I've cut to hold the IC and a few mfgr recommended parts and a 3 pin header. That assembly will be fixed with epoxy to a circular band of aluminum, which will be clipped around a 2 inch wide PVC pipe carrying water to and from a pool pump. The obvious purpose is to get a reasonable approximation, over time, of the water temperature flowing in the pipe.

So for my adhesive, and also to cover and shield the small PCB from the element, it seemed that thermally conductive and electrically insulating epoxy would be the first choice. But after seeing what I would call astronomical prices for small amounts of epoxies with these specifications, I'm thinking maybe I don't need it?

First of all, it seems to me I've often covered boards with many ordinary epoxies for moisture protection, and I've never found any that proved to be a bad electrical insulator.

Second, considering my aluminum strap will go around a 2" PVC pipe (whose wall thickness is already about 7mm thick) and considering I don't care if it takes 5 minutes or more to get a reasonable temp reading, and most important that I'm only looking for a reasonable approximation, I'm thinking just about any cheap epoxy will do.

Thoughts?

Randy
  • 3,205
  • 3
  • 21
  • 38
  • FYI - you can make your own thermal epoxy by mixing something like Arctic silver paste with regular long cure epoxy - I've done this using JB Weld and it works reasonably well. – Norm Jun 27 '18 at 19:04
  • Any chance though that the addition it could cause it to be electrically conductive? The TMP36 doesn't like to drive more than 50uA, which at an approximate output of 1V, it would only take about 50K of conductivity across closely spaced pins to cause a problem. – Randy Jun 27 '18 at 19:30
  • You could use something like Arctic Silver Ceramique which is non-conductive. – Norm Jun 27 '18 at 19:52

1 Answers1

4

What you need for accuracy is a large difference between the thermal conductivity to the water and to the surrounding air. The hose will be a pretty good insulator so I don't think the epoxy will matter that much. Accuracy will be slightly affected by self-heating but probably not to an important degree (so to speak).

For fast response you need low thermal mass and low absolute thermal conductivity to the water.

Metal-loaded epoxy like JB weld is available in consumer-friendly packaging and prices, but not sure about its insulating properties.

If I might suggest- make the back of the PCB flat and pour an insulated plane, screw it to an aluminum block with a semicircle cut out of it and a flat top, attach the block to the clamp with screws. No epoxy of any kind. Then surround the assembly with foam insulation.
enter image description here

The best solution, of course, is to cut the hose and put a small-diameter sensor (maybe not that one) in direct contact with the water. If the L/R ratio is not high enough (10:1 is good), then bend it so it goes in line with the water.

Spehro Pefhany
  • 376,485
  • 21
  • 320
  • 842
  • I agree. the thermal insulation to air must be 100x greater than to plastic tube for 1% error due to. Ambient differences. Low mass is best. Perhaps with this solution you might get 4x Rth thermal resistance. Delay time will be fast enough several minutes after pump runs. All you want is 0.5’C accuracy ? Then calibrate it and use water intake. – Tony Stewart EE75 Jun 27 '18 at 19:06
  • 1
    I'd definitely surround the sensor (pressed to the pipe) with some kind of closed cell foam wrap to make sure the probe is reads the pipe temp. Oh. And that also reminds me... emissivity and "open night sky" definitely could affect things. So one might go further and make the outside surface of the closed foam wrap have a low emissivity specular surface. – jonk Jun 27 '18 at 19:11
  • @jonk I have some crazy good aerogel stuff that works very nicely. – Spehro Pefhany Jun 27 '18 at 19:24
  • Wow! I've wanted to play with such stuff and spent some time considering setting up to make some easier forms of it. But that never happened and I've only been able to *wish* about it. – jonk Jun 27 '18 at 19:27
  • 1
    Thanks. Well its not a hose, but a 2" PVC pipe. So cutting it would involve a significant extra effort to repair. But I did get a VERY thin and small PCB cut, which easily bends around the aluminum surrounding the large 2" diameter pipe. As far as the metal infused JB weld, I'll have to try embedding a pair of wires and see what the conductivity is like after cure. The TMP36 is a little fussy about loads, not wanting to drive more than 50uA. So at about 1V output, 50K would have a major effect, and I'd want to stay at least 1/10 that (500K) to feel good about it. – Randy Jun 27 '18 at 19:40
  • And @TonyStewartolderthandirt, the purpose here is simply to be able to turn off the pump when a certain temp is reached for a certain time. Since I'm doing this with a microcontroller (an Arduino NANO clone FYI), I've already made provision in my user setup, with LCD and buttons, both to set the time/temperature, AND to calibrate the Temp output. So all things considered, even a few whole degrees off can still be made to work. – Randy Jun 27 '18 at 19:45
  • I think you can tweak it so it doesn’t cycle more then x times per hour and average the temp to reduce errors and be able to choose 1 Or 2degrees if you prefer. The key to the theory is the resistance ratio as a comparator. So solar or ambient temp does not influence errors too. Without great insulation it will just cycle more often and have bias towards ambient changes, unless you wish that, which is also – Tony Stewart EE75 Jun 27 '18 at 22:38