FET as linear mode heater

If you're going to cycle it on and off then I suppose a low Ron would be best.

If you want to use the FET as the heater, just do proportional control with a small sense resistor, Watts = Ids * Vds ...Jim Thompson

For coupling, a metal-case transistor would be best; a crufty old 2N3055 will handle junction temperature to 200C and dissipates over 100W. Hey, if you WANT power inefficiency, it's a great part!

It looks good at high temp, but only rated to -65C operation...

Why not use a nichrome wire heater, any kind of accurate power resistor will be pricey and less robust. A soldering iron cartridge would be easy to work into this application.

If you're using the FET in linear mode the Ron isn't going to matter.

It's a Really Good idea to not drive the gate directly from an op-amp. Drive the gate via a 1k-ohm or so resistor (less if the op-amp is known to deal well with capacitive loads). Run a capacitor back from the op-amp's output to the -V input, and a resistor back from the source to the -V input. You'll end up with a device that holds the source at your desired voltage after some settling (it'll look 1st-order unless you really push the cap value down).

Leave out the cap and isolation resistor and you'll have a nice oscillator...

George, maybe I'm particularly dense today but I'm not sure what you want to do. Do you want to control the temperature or the heating power? Do you intend to use the FET as the controlling device or as the heating element? By "linear mode" do you mean no form of switching control?

Dump the series R and let the fet make the heat. Just use a small current sense resistor somewhere.

Shouldn't matter.

Jim, and whit3rd, pimpon, Tim. I clearly didn't state my question clearly.

I'm thinking of a linear mode heater for thermal control. I could do PWM, but the heater is going to be running near other stuff down a probe.. a few feet of cable. I know I'm going to hate PWM at some noise level.

So just a linear heater. To not waste any power I'll attach control(FET) and heater (resistor) to the thing to be heated. I'd really like to-220 packs. I want about ~50W max, and with 2(3) to-220's ... that should work.

(3)Hmm.. I was thinking of splitting max power between control and heater, but maybe more resistance.

So in regards to the FET. I want it's resistance to change between

To stress the FET least it seems I'd like to spread the loss over as much of the channel region as possible. Maybe I should look at even higher voltage?

George H.

Yeah sure I could do that too. Having a resistor down there reduces the load on the FET. It also gives a bigger feedback signal. (I'm thinking 3 to-220's with 2 R's.. I may want more power.) Having more resistance is not always bad. (I think "George's law" should be, "Why not put some resistance between those two nodes?".. of course less resistance is often a good thing.)

OK but one FET will break first, when pushed.... "More power Scotty"* Which one?

George H.

That's my question.. and obviously not my law. :^)

One approach is to put a bunch of FETs in series, with a voltage divider string driving the gates, the way you do to get to high voltage.

Cheers

Phil Hobbs

Yeah, but students will be involved so HV is out. I figure 50V is OK.. I'll touch it... expecting some tingle.

(my DMM blinks HV at ~30V :^(

George H.

I wasn't suggesting HV, just using the same topology.

Cheers

Phil Hobbs

There are monster 300 and even 600 watt mosfets. But if you want to spread out the heat sources, by all means use a bunch of smaller fets. It depends on the thing that you're trying to heat.

Each fet can have a *big* sense resistor in its source. You could size it so the resistor burns half the power at max output. That's still linear, which control loops like.

With multiple fets and big source resistors, all the gates can be driven from essentially the same signal, and sharing will be pretty good. 4 wires into the entire array.

Depletion fets would be fun. Run them at Idss, lots in parallel, just two wires into the whole array. Vary the voltage, with a switcher, to control the dissipation. Pretty linear. Might get goofy at 77K.

m power resistor in series, with feedback from the resistor to control the current.

Bad advice. Power dissipation in a resistor is a lot cheaper than power dis sipation in a FET. With a series resistor, peak power dissipation in the FE T happens at half full-scale current, and it's a factor of four smaller tha n the peak power dissipation in a FET-only heater delivering the same peak power.

And you can run the resistor hotter, which makes it easier to get the power where you want it to go.

ng

ier > >to control.

Smaller on-resistance is usually more expensive ...

No need for resistors, especially if you're dumping all the heat into one place anyway. (And if you weren't, the resistor would make the power quadratic, which sucks for control.)

Example, like the thing in here,

If you need to spread the heat, use more transistor + source/emitter resistor pairs. Plenty of degeneration means you don't care about Vgs/Vbe. Said resistors could stand to be heatsinked too, but they're probably a minor, depending on control voltage versus supply of course.

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

The stated temperature range is from liquid nitrogen to 100C. At low temperatures, you're probably operating in a vacuum (and plastic isn't a great choice for vacuum). Cold, a MOSFET has 'way more gain than you'd like, RF oscillation will be an issue.

At high temperatures, the plastic packages aren't great, but I see a 175 C rating... and 'absolute max' low temperature of -55C.

I'd rather use a Variac for control, and a stepdown transformer for isolation, and run low voltage AC to the heater. Power efficiency, no high frequencies.

MOSFETs with lower RDSon might also have more gate capacitance. All other things being equal, I would suggest a MOSFET with low Cgs and Cgd, because it will not cause as much extra phase shift when driven from an op-amp with non-zero open-loop output impedance. (Of course you could drive it from an OTA which can be made MORE stable by capacitance on the output, but the choices there are not so great, being mostly designs from the '70s.)

It might be worth considering a power darlingon rather than a MOSFET. They might be easier to drive with an op-amp (less input capacitance).

Fortunately in your application you probably don't really care about bandwidth so you can do the stuff that Tim Wescott suggested to make the op-amp feedback be un-lagged by the gate capacitace at high frequencies.

Chris

Not if you are putting the resistor in series with a power MOSFET.

The dissipation in the resistor does rise as the square of the current thro ugh it, but the extra dissipation in the transistor in series with it doesn 't rise as fast (and actually decreases as the voltage drop over the resist or rises above half the supply voltage) so the power delivered to the load is a perfectly linear function of the current through the combination (and that is pretty much what you are controlling - or can be, if you are carefu l about it).

if you arrange for the resistor to match FET dissipation at just over half power, at full power the resistors will be dissipating almost all the power and at half power that's the peak demand on the fet. so a single 13W transistor could run do the whole 50W heater.

but why a fet?