Over temperature protection

[nono]

These days I'm designing a high power audio amplifier. I'm thinking of using a thermistor for identifying excessive temperature in power transistors. There are two types of thermistors available in the market as NTC and PTC thermistors . Im not sure which type is most suitable for the application. Can someone help me?

[Thecomedian]

PTCs increase in resistance with heat. NTCs decrease in resistance with heat. If you want thermal protection for something getting hot, PTC. If you want thermal protection against "surge" as a high voltage or current is switched on, use NTC.

http://en.wikipedia.org/wiki/Resistance_thermometer#Values_for_various_popular_resistance_thermometers

[nono]

Thanks for your help.

[earthtonesaudio]

Whatever type you use, take a look at the rise/fall times.  Depending on how you intend to protect, you may want a faster or slower response.

Also note that for best accuracy you have to put the thermistor in direct thermal contact with the hot thing, which may require special hardware.

I recall taking apart a power supply that had the thermistor bolted to the heatsink right next to one of the power transistors, so that's one possible way.

[R.G.]

All very good advice.

NTCs have pretty smooth changes in resistance per change in temperature, so they are very good for set-to-any-temp trip points. PTCs tend to have a more sudden transition, where they're low(ish) resistance up to some temperature, then the resistance increases dramatically with each increase in temperature. That makes them better suited to applications where you know ahead of time exactly the temperature you want them to "trip" at.

I would probably go for the NTC for the easier adjustability as long as the supervisory circuits that read the NTC were easy to adapt to it and also easy to adapt to doing whatever shutdown/cool-off method you're using.

I've thought about this exact question a number of times. Here are few of those thoughts in random order.
- The old standard way to overtemp protect things was with a bimetal snap action temperature switch. These were used to open the incoming AC power. They worked well as long as you could stand the slow-ish time lag between the output devices heating and the switch opening, and were OK with running high voltage AC power (= hummmmmm) around back in your audio signal circuits.
- Silicon diode junctions have a predictable temperature coefficient too. This is why they're often used in bias compensation networks. An extra silicon transistor on the heat sink, perhaps used as a Vbe multiplier to put out a higher voltage, also gives a good temperature signal. It's negative-coefficient too, as its voltage decreases with increasing temperatures.
- There are some three-pin temperature reading ICs designed to work as programmable thermostats. They give you a clean logic high (or low; can't remember) when their temperature crosses a boundary.
All of the temp reading devices except the bimetal switch need some circuitry to feed them the right voltage and sense when an overtemp happens, and then to do the right things about the overtemp happening. Going analog works fine for the NTC, PTC, and semiconductor junctions. Analog also works for the thermostat ICs if you have some way of telling it ahead of time "trip at 73.4C". The analog will consist of some kind of comparator circuit to sense the trip point and then some circuitry to turn power on/off or limit/mute signal, whatever you want to do to cool down.

These days, I'd probably program a PIC to read the temperature device, then do whatever I wanted, but that's only feasible because I've gotten over the hump of being able to simply and easily program a single PIC (or other uC).

[amptramp]

There used to be an item called a Globar resistor that started at high resistance and went lower as it heated up.  It was used to limit the inrush surge in tube televisions.  Since the last tube television (in fact, the last mass produced tube anything) was the 1982 GE Portacolor, these are generally found as new old stock.  There was a version that was plugged into a wall socket and had a socket itself so the globar was in series with whatever you plugged in.  But it served mainly as inrush protection, not as a temperature sensor.

[R.G.]

These are still around as inrush current limiters. They're especially needed for the inputs of switching power supplies where inrushes can be ...huge...

They're pretty common.

But you're right, they're not all that useful as temp sensors, being mostly used in the self-heat mode and fairly massive. For sensing temps, you want low mass to heat quickly.