SOA (FET)

Why not use a bigger fet? 3 c/w is terrible. Fets are cheap.

Try to keep the junction temp closer to 100C or 125C. Your IRLI insulated MOSFET package doesn't help.

Right it's got an insulated tab. I'll go search for something else.

I ran it at 150 C for an hour (I'm off the calibration curve for my diode temp sensor).. and then turning it up more, I melted the solder on the fet (air wired) at ~155 C (+36C) ~=190 C...

Poor man's thermal fuse :^)

George H.

Well the FET is attached to the heater. So if I want it to go to 130 C it will be hard to keep it cooler than that. :^)

If not the insulated pac, I'd have to use some other insulation.

Thanks, George H.

You can't PWM? Linear heater control has the square-law transfer function problem, too.

Using fets as the heater, or as part of the heater, is nice because heat becomes linear on current, and no power is wasted.

Right that's what I'm doing. I convinced the "powers that be" to put the FET with the heater resistors... saves me all sorts of headaches.

PWM had been rejected previously.

Just finalizing the design. I'm living in the past, all analog...

George H.

That's OK. What's wrong with a TO-220 resistor in series with the FET doing the job? Then the MOSFET could be closer to your 130C.

Yeah that's what I've got... well actually there are three 15 ohm resistors in series with the FET. I may have to try two 22 ohm, 'cause at the moment there are only three mounting holes on the hot plate. Re-reading the spec sheet it says I can run the case temperature at 150 C and still put 12 Amps through it. Ron (@150 C) = 0.06 ohm, I^2 R = ~8 Watts. I think it will be fine at 130 C. (I'll cook it for a few days at 150 C and if that works.....)

George H.

Fet in series with the resistor, both making heat, is neat because, above half power, the dissipation shifts from the fet to the resistors. At full power, it's all in the resistors. If you control fet current, it's still linear.

Semiconductors age much faster at elevated junction temperatures. Running it at 167 C for a few days isn't going to bring that failure mechanism to light, nor will it make any issues with differential expansion & contraction come to light (if the latter is a problem...)

I'd try to find a better part, if there's something out there.

Back in the old days, the metal-case power transistors often were rated to 200C. There's one FET with that high a rating at DigiKey, SCT10N120 ( a SiC NMOS power device). Kinda pricey, since it replaces with an old 2N3055...

Well I went looking at insulated FET's and unless I go with the bigger to-247 pac I seem to be stuck with ~3C/W junc to case. (there are some space issues that have been designed into the hot plate... I'll have to see if things can get moved a bit.) For the to-247 thermal resistance is like

Another heater story. (As long as no one reports me to the SPCIC*) I use a TIP120 (Tc max = 150 C) as part of a heater for a Rb lamp. Typically the lamp runs at ~120C @~ 2W heater power. But I've run it at 150 C for hours/ days at a time. With out any visible signs of problems.

George H.

Hmm thanks.. not a whole lot of stock out there either. Is Silicon carbide intrinsically better at high temperature?

I think the issue is the glass-transition point of the epoxy. As has been pointed out, the transistor itself is okay at 200C. Now where's Syd Rumpo when we need him?

Cheers

Phil Hobbs

It should...its melting point is about twice that of pure silicon.

Even with metal-case power transistors the derating means you can't actually dissipate much power at those high temperatures. Reliability suffers, too, at high temperatures.

George should really move over to a PWM circuit, then all the heat will be in the power resistors. There are plenty of easy-to-use PWM chips. One by Linear Technology, part of their TimerBlox series, the LTC6992, is really easy to use. See AoE III, pages 432 to 434.

I think he may have noise concerns. I've done PWM with deliberately slow edges, basically a trapezoid, which is good unless you have magnetic field issues.

Slowing the edges is as simple as adding a series gate resistor.

Generating the PWM is easy.

PWM is great in low-to-moderate-performance applications. I use it that way sometimes too. All the advantages you cite are real. However, there are a fair number of applications in EO systems where higher order effects are su per-important, so switching junk is death and destruction.

An approximate analogue square-rooter is a wonderful thing inside heater FB loops.

Cheers

Phil Hobbs

Right. My first choice was a PWM with rolled off edges. (I was thinking of an RC roll-off, being easy) But that caused worry, heater current next to signal lines. I hadn't thought about magnetic fields from the heater. (I think I've told this heater noise story before, but in our noise measurement apparatus, I found that the noise on the heater leads capacitivly coupled to the signal lines... (looking at noise from 10 k resistor in a LN2 probe) and screwed up the measurement. I had to make my own low noise voltage source... Cap. multiplier my noise "hammer".)

Oh thanks! (I was trying to slow down a comparator output, and the National Semi, support engineer told me to hang a cap on the output... good advice.)

George H.