Basically, what I need is a way to turn a cheap potentiometer into a large wattage rheostat. The difficulty here is that there is no separate power supply. It must take power from the circuit. (e.g. as if it were a rheostat with 2 pins)

The "rheostat" needs to be able to handle 20 watts, and vary from about 0-20 ohms.

One requirement is that there can't be much drift. If it is set at something, it should stay there +/- 15% at most. Even with changes in temperature in any active elements.

Suggestions on how to proceed?

 

I would not want to pretend to know the best solution, but this sounds like you could use a power BJT in the linear region.

If you connect the gate to the power trough a resistor in series with your pot, then it should act nearly as a larger pot with a bit of a dead zone at the bottom. Well, and also only work for on current direction.



Please wait for someone more knowledgeable than me to confirm though, as I am still slightly unconfortable with BJTs.
As a side remark you will never get a perfect 0 with any transistor.

 

SuperFrog posted his while I was working on mine.

All plotted vs resistance of R1 (the pot)

Blue=Dissipation in the Pot (i.e 1W Pot)
Lt Blue=Effective resistance of the network
Purple=Dissipation in R3
Green=Voltage across the network
Red=Dissipation in the 3055

Attached Thumbnails

 

__________________
Mike ML.

 

Thanks!

I'd like to get the potentiometer into the 250mw range or less though. Can I use a darlington to decrease the base current?

And how stable are BJT's with respect to temperature? If I set this circuit to "10 ohms" when it's room temp (25C) and everything heats up to 80C, how will that effect the circuit?

 

Sure, hire a consultant.

__________________
Mike ML.

 

The gain of a BJT varies significantly with temperature (typically >30% from 25C to 125C), thus you would need a large heat-sink to minimize the transistor temperature change, for MikeMl's circuit to work within your limits.

__________________
Carl
Curmudgeon Elektroniker

 

I would have thought a power MOSFET, voltage controlled at the gate by the bias of the potentiometer and a series resistor, current limited at the D-S, would be a good solution.

Do you guys care to expand on why the BJT is better? Pros, cons.

 

I thought so. Same problem with mosfets I believe...

Other than an opamp with a separate supply, can this problem even be solved?

 

^^^

Even if it's just to point to some reading, I'm okay with that. Trying to learn here......

 

The main problem with a MOSFET is that they typical take several volts to start turning on. That means your resistor wouldn't start conducting until the voltage reached at least the MOSFET gate threshold voltage.

A bipolar transistor only takes about 0.65V to start turning on so it would start conducting at a much lower voltage than a MOSFET.

__________________
Carl
Curmudgeon Elektroniker

 

If you are just dimming a 12V lamp, then a 2 wire PWM dimmer can be built like this:
EDN Access--10.23.97 Two-wire dc lamp dimmer replaces rheostats
Not a very good image but you can find a better one if that is what you are after. What exactly are you trying to "control" with the rheostat?

__________________
Inside every little problem, is a big problem trying to get out.