Measuring MOSFET temperature w. sensor

[Jdo300]

Hello All,

This is a new thread related to my output stage setup. Only this time, rather than determining if something went wrong, I simply want to monitor the temperature of the MOSFET on the output stage. I have been doing a bit of research to see what kinds of sensors are available out there and I can see there are a lot of choices ranging from thermocouples to simple silicon temperature ICs with both analog and digital outputs.

In my situation, I have one (possibly two) MOSFETs that will be mounted to this heatsink here:
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=HS349-ND#HS349-ND

I would like to somehow get a reasonably accurate temperature readout of the heatsink so that I can monitor the temperature of the MOSFET(s) that are mounted to it. So far, I would like to use a temperature sensor that has a digital output so that I won't have to add an extra ADC to the board (the output stage is isolated from the digital side) to measure the temperature value. For that, I have found one good sensor that would work perfectly for the job (TC74A0-5) as it comes in a TO220 package and could simply mount on the back side of the heatsink to measure the temperature.

The problem is that I really want to save that space for a second FET. Sooo... I found another sensor, the LM74, which also offers a digital output that I can read. It comes in a SOIC package, BUT I'm wondering how I would accurately take a temperature reading with it. Would it be sufficient to simply locate the sensor somewhere close to the heatsink or is there a better way to do it? I'm surprised that there are so many SMT sensor ICs out there but I'm dumbfounded to know how one would actually get an accurate temperature reading on a specific 'something' using those package types?

I also noticed that there is a nice selection of thermistors out there which I could also play with, but I think I'll still run into the same problem with those.

Any suggestions appreciated.

Thanks,
Jason O

P.S. as far as accuracy is concerned, I'm only looking for a resolution of 1 degree C.

 

[vk6kro]

I like the 18B20 which can read temperature to 1/16th of a degree and is directly readable with some of the Picaxe chips.
It is digital and already calibrated. It works from -55°C to +125°C
I think they cost about $6.

It is in a small package like a TO92 transistor, so you could probably find a space for it on a heatsink. This would not read the FET temperature, of course, but it would give an indication if the heatsink temperature suddenly rose, that the FET was also getting hotter.

You can get thermometers that can read temperature without touching. These are like you might have seen doctors or nurses use for taking a temperature in your ear. You could use one of these to compare the heatsink temperature to the actual FET temperature.

 

[Mike_In_Plano]

Hello Jason,

I used to build power amplifiers as a hobby. You basically ended up with the same problem - you'd get 60% heat sink 15% on that side of the board and 20% of the other and so on. A lot of times, you'd see the sensor held in place with a clip that fasted to the heat sink on either side. Or, you'd see small TO packages (like a 2sd669) that were screwed to the heat sink. Or, you could purchase NTC thermistors from US Sensor that came built into something like a screw lug. These tracked pretty well.

Of course, if you have a sensor that's mounted, you can always tape a bit of insulating foam over it.

 

[Jdo300]

HI Guys,

Thanks for all the great suggestions. I have decided to go with an approach similar to what Mike_In_Plano suggested. I found a sensor that comes in a TO-92 package (the LM34) which has an analog output voltage. It is small enough that it can fit snugly between the MOSFET and the corner of the heatsink where the fin is. If it is mounted on the PCB, it could be close enough to almost touch the heatsink, but because it is sitting in the corner, the ambient air temperature may get high enough to give a pretty close reading. The other nice thing is that the position allows the sensor to be somewhat shielded from any cooling fan induced airflows which may help make the reading more accurate. It's the next best thing I could come up with short of mounting it to the heatsink directly.

- Jason O