Where to put the temperature sensor.

If you want accuracy, put the sensor on the end-element whose temperature you care about. Of course, that has worse dynamics (diffusive lag) than closing the loop on the heater temp.

Thermal loops are nasty because of the heater mass and the lag between heater and device/sensor.

One trick is to close the loop near the heater and apply a feed-forward correction based on ambient.

Wouldn't you want a PID scheme to use the end-element sensor for the I part, and P from the heater-element sensor? That limits the forced bandwidth pretty well. D from the heater-element if the heater could overtemp and damage something...

The real win is to close an inner loop around that sensor, and then run the outer loop sensing the temperature you actually care about. (For extra points, spill some sweat optimizing the thermomechanical design to make the two more nearly the same.)

Cheers

Phil Hobbs

Yes. Sensors on both is best.

Inner loop around a sensor at the heater, outer loop around a sensor on the thing you care about. The inner loop only needs to be a proportional, or a proportional-derivative. The outer loop probably just wants to be proportional-integral.

Not only does it give you much better loop dynamics, but it lets you hold the heater temperature down to whatever you feel is reasonable -- only controlling on the temperature of the thingie that you're heating means that your heater might overheat, particularly if it's transistors and not just resistors.

As great as the TC difference is between the remote sensor and the heater sensor, you may have overshoot if you drive the heater hard and then back off just when your remote sensor reads the correct temperature. So -- fun fun with making that all work well.

The wide differences in time constants indicates that the remote thingie is rather removed, thermally, from the heater. That tells me that the difference in temperature between the thingie and the heater will probably not just be due to time lag, but will be affected by ambient temperature -- that's why you don't just want the sensor on the heater. If you want to test that, control your heater temperature while looking at your thingie temperature (what is the thingie, BTW?). Do it with the whole assembly in an oven, then do the same thing with the whole assembly in an ice box. You _will_ see a difference in the thingie's temperature.

Yeah, that's what I did on our diode laser. (for a bit I was hacking apart commercial controllers and sticking in my own bigger integrating cap.) I'm not sure it was the right decision, the diode is a heat source.. that varies...

Anyway in this case it's a thermal can... hunk of copper with Al shield attached. My piece of copper is bigger than the real thing, but in the 'scrap' heap. I figure we can pick one place, and set the temp there, why not right at the heater?

Huh... OK that means a second sensor. (I dream more about nested control loops)

George H.

Yeah that's my dream... I think I understand, but until I do it I can't be certain. This piece does not warrant such attention.

George H.

Right, you only need one integrating cap on the outer loop.

For sure! I think that's my number one reason for the sensor on the heater. Then (unless something breaks) it can never get hot enough to melt the solder on the heater resistors. (It's my job to make sure that can't happen on the Fet pass element. melting the solder... solder the ultimate thermal fuse. :^)

OK so one simple loop on the heater is fine.

Right sorry, in this case the thingie, is a metal thermal can.. my hunk of copper is only my model.

I hope some day to try a double loop thing.

George H.