Suggestion: Semiconductors - how hot is too hot?

[SReifel]

Hi Dave.  Thanks for all the great videos, I've learned a lot!

I would be interested in seeing your thoughts on semiconductor temperatures: How to measure, how much to de-rate from "absolute maximum", how to use PCB planes for cooling SMT parts.

I recently designed a buck converter.  Of course I studied the data sheets for info on current ratings and thermal considerations, but how much can I push my design given its layout and cooling?  I've frequently find measuring a component's temperature hard.  I mostly use a thermocouple, but never sure if I'm getting good conductivity to it.  Should I be using an IR thermometer or camera? What is the best method?  Is there a finger test?  If a parts rated to 100°C, I figure if I can touch if for 3 seconds, its probable OK.  Does that make any sense?

Your insights would be appreciated!

[james_s]

I generally try to keep parts under about 70C, it depends on many factors though. Keep in mind the specs are often for the die temperature which can be considerably hotter than the package temperature. Measuring depends on the package, sometimes I use an IR thermometer, sometimes a thermocouple, sometimes I'll stick something like a DS18B20 to the package with a dab of thermal paste.

[thm_w]

https://www.eevblog.com/forum/projects/using-thermal-epoxy-to-fix-thermocouples-to-pcb/
https://www.eevblog.com/forum/projects/attaching-thermocuples-to-a-working-pcb/
https://www.eevblog.com/forum/manufacture/glue-to-temporarily-attach-thermocouple-wire-to-fixtures/
https://www.eevblog.com/forum/projects/epoxy-compound-for-holding-thermocouples/

Thermal camera is best for a live view of everything on the board, but also most expensive.
Thermocouples are good if you have a large number of points you want to track over time, or want to test temperatures inside a sealed enclosure.

[AndyC_772]

I could ramble about this for days, but the short version is: you should be most concerned about the die temperature, which you cannot measure directly, but can calculate.

What you can and should do, is:

- measure the outer case temperature
- measure the power dissipation in the die (apply P=VI to each current loop that passes through the device)
- look up the junction-to-case thermal resistance of the part in its data sheet
- from these figures, work out how much hotter the die must be than the case, given that you know the dissipation and thermal resistance
- add the two numbers together to get the die temperature

Provided the worst-case figure is within the data sheet rating, you should be OK.