over-temperature shutdown

It's simple.

Vbe of a bjt has a typical temperature coefficient of - 2 ~ 2.2 mV / degree C.

I guess a diode is similar but I always use transistors.

You pass a current through the device - measure Vbe and use a comparator to detect your cutoff temperature. May require calibration but I found devices to be adequately repeatable not to need this.

I gather this is how your PC motherboards measures CPU temp in some designs too ( I think there's a diode junction buried in the processor chip specifically to measure temp ).

Graham

The LM87 is a complex chip with many functions and seemingly thousands of programming registers, and no doubt includes a kitchen sink in there some place. The elegant kT/q bias-current-ratio trick is attractive and worth employing for temperature measurement problems, but a more simple kT/q chip is called for. NSC's LM82, LM83 or LM84 are candidates.

They measure the difference in Vbe at two different current densities (typically with a 10:1 ratio). This causes many of the paramaters in the I vs Vbe relation to cancel, eliminating the need for calibration.

Of course, there are commonly available chips that do this. See, e.g. LM87, etc.

Regards, Allan

In article , Winfield Hill wrote: [....]

A couple of silly thoughts:

(1) You can get thermostats in TO-220 packages.

(2) If you use a PTC thermistor in a RC clock circuit, you can have the clock slow down as the temperature goes up.