Steve's Things

  After being apprised of an issue with heat transfer through PCB traces, being the stubborn idiot I am, I had to try and solve the problem. The issue is with a particular PCB that has a cutout section with a thermistor placed in a partially isolated section within the enclosure, unfortunately the traces themselves are serving as a heat transfer conduit especially so since copper is a particularly good conductor of heat.
 
  After thinking about it for a while I came up with a couple of ideas like calibrating the calculated output to accommodate for the temperature discrepancy, however that probably wouldn’t work without a second thermistor closer to the heat source, accounting for the difference. Another idea I had was to use an infrared thermopile but they’re kind of large and expensive and would require a heat channel mounted above it to prevent measuring the temperature of the case.
 
  After thinking for a while about it, the simplest answer I could come up with would be a heatsink to mitigate the heat transferred to the thermistor. While it is theoretically impossible to totally remove all heat difference, I believe that a large portion of the heat transfer can be eliminated. So I wanted to see if heatsinks actually help equalize heat much at all so I devised a bit of a rough experiment.
 
The Experiment
 
  I decided the quickest way to test this would be with materials I already had. I have a selection of thermistors so I used a glass axial thermistor.

This is the thermistor board used in the experiment. the thermal leads aren't attached.

The circuit used for the test. Beautifully illustrated

 
The thermistor used is a glass 10kOhm NTC Thermistor in series with a 1k resistor in order to detect current/divide voltage. With the setup above, the voltage detected from GND to the divider is 438mv. the temperature in my basement is pretty steady since the furnace isn’t running. So now I attach the leads.
 

This is the board with the two 22AWG leads attached.

I also performed the test with the leads closer to the thermistor.

 
  I used Teflon coated wire to prevent conduction between the two sides and used my soldering iron as a heat source since it’s temperature controlled. I let the heat soak up through the lines for 10 minutes to ensure that the heat stabilizes at the thermistor. I also performed a test with the soldering iron closer to the sensor.
 

 
Finally, I decided to place a heatsink on the line. I was going to solder aluminum shims to the leads in order to provide for heat sink. That turned out to be difficult at best, so I used coiled up copper wire and soldered it to the leads. I performed the same two tests with the position of the soldering iron. Anyways, here are the results of the test.

Test Type	Voltage Recorded
No leads, ambient	438mV
With Leads, ambient	438mV
Pressing thumb on thermistor ~33.1c	554mV
Heat with long leads, No sink	496mV
Heat with short leads, No sink	625mV
Heat w/ long leads, w/ sink	476mV
Heat w/ short leads, w/ sink	544mV

 
  It should be said that this test is far from perfect and doesn’t prove anything quantitatively with any real degree of accuracy, however I wanted to see how effective even a rudimentary heat sink would be in a situation like this. It should be noted that while thermistors are inherently non-linear, we have voltage drops of 43% and 34% with the short and long leads respectively. I simply thought it was an interesting experiment. I have a few ideas on how to sink the heat or account for it but I think the heatsink is the easiest plan.
 
  Here’s a beautifully rendered image of the idea to reduce thermal linkage between the PCB and the thermistor ;).
 

 
  If there were a heatsink over the exposed traces and the traces made as long as possible, it should be possible to bring the traces fairly close to ambient. Like the experiment showed, even a rough heatsink was able to reduce the heat going to the thermistor by a substantial degree.
 
Feel free to comment.