3
\$\begingroup\$

Sanity check: I need to measure the junction temperature (or as close as possible) of an SMD LED. Can I solder the tip of a thermocouple to a leg of the part? Assuming there is no ground loop (the meter is battery operated, standalone), is there anything I should be aware of?

I am assuming that the thermocouple being in connection with the circuit will not influence the reading, is that a correct assumption? The LED is driven by a 25kHz PWM signal. I'll also be connecting another thermocouple to the metal enclosure by drilling a hole and pinching the tip in the hole.

I have tried thermal glue previously and that was a unreliable so I would want to avoid glue.

\$\endgroup\$
2
  • 2
    \$\begingroup\$ Do you need to measure the temperature permanently in production, or just during testing? \$\endgroup\$ – posipiet Sep 2 '11 at 7:15
  • \$\begingroup\$ This is just for measuring the temperature during testing. \$\endgroup\$ – morten Sep 2 '11 at 19:56
3
\$\begingroup\$

Solder a type T thermocouple to whichever pin of the LED has a better thermal conductive path to the LED die.

Type T thermocouples are copper-constantan, and are solderable, whereas the more common type J and type K are of metals that do not form a solder joint.

\$\endgroup\$
5
\$\begingroup\$

Two things. First, a thermocouple is not the best way to measure LED temperature. Thermocouple signals are very small. The main advantage to thermocouples is that they are simply two metals, and as such can stand high temperatures. Thermistors give you much bigger signals and can still cover the temperature range of a LED.

Second, LEDs are diodes and therefore have a significant temperature dependence themselves. If the LED is driven with a known current, its forward voltage drop will tell you its temperature after some calibration. You said the LED is driven with PWM. What does the driving circuit look like? Is the current reasonably fixed when on? If so, you only need to measure the voltage during the on phase to determine temperature. You will have to test this at a few known temperatures to calibrate, but once that's done it should be pretty repeatable.

\$\endgroup\$
2
  • \$\begingroup\$ Never thought of doing this indirectly, by measuring the voltage drop. It is an interesting alternative to the on-board thermistor. One challenge would be that the Vf varies a bit from LED to LED so some sort of auto-calibration might be necessary. \$\endgroup\$ – morten Sep 2 '11 at 20:16
  • \$\begingroup\$ Sorry for not being clear, but the question is about attachment of a probe for testing only. \$\endgroup\$ – morten Sep 2 '11 at 20:18
1
\$\begingroup\$

The best way to go is to measure the cathode pin temperature: http://www.photonics.com/Article.aspx?AID=29860

\$\endgroup\$

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.