While investigating the possibilities of running my amp driver transistors at a constant temperature as an alternative way to keep the temperature sensitive Vbe and therefore the bias current stable I had been thinking about various ways of making a mini oven sort of like are used with crystal oscillators sometimes. I figured that a large fet with a variable current put through it and bolted onto the same block as the drivers would be the go, just pick a suitable temp above ambient e.g 70 deg. Of course you have to measure the temp to regulate it and that is when I had this great idea! In the diagram the fet drain is fed from a constant current source of at least 60 volts. You set the gate voltage according to the temperature you want the fet to run at. Lower temp = higher voltage. The zener and diode are for startup from cold. When the drain-source had say 20 volts across it and the current is always 200mA then it's dissipation is 4 watts and it runs at a certain steady temp. If the ambient cools for instance then the fet starts to cool a little and this raises the gate threshold causing the fet to conduct less (same as lowering the gate voltage). This causes the drain voltage to rise and seeing it has always the same drain current but now a higher drain voltage, the dissipation rises and this pulls the die up to very nearly the temp it was before. The loop gain is very high because of the current source.
There are two very good things about this setup. 1/ the temperature is sensed by the gate junction which is only microns away from the drain-source so it knows what the die temp is *now*, not in several seconds time, so it is very stable with sudden variations in thermal load. 2/ As the set temp is approached the rate of change of drain voltage slows down the closer it gets. At one stage I was measuring 25 deg to 75 deg in about 10 seconds with way under 0.1 deg overshoot. Try that with a conventional feedback loop!
Maybe someone has done all this before but it's original as far as I am concerned. I hope someone finds this to be of use.
GP.
There are two very good things about this setup. 1/ the temperature is sensed by the gate junction which is only microns away from the drain-source so it knows what the die temp is *now*, not in several seconds time, so it is very stable with sudden variations in thermal load. 2/ As the set temp is approached the rate of change of drain voltage slows down the closer it gets. At one stage I was measuring 25 deg to 75 deg in about 10 seconds with way under 0.1 deg overshoot. Try that with a conventional feedback loop!
Maybe someone has done all this before but it's original as far as I am concerned. I hope someone finds this to be of use.
GP.