FET and thermal stuff

That's a good geometry. Fat vias are helpful.

More heat spreading layers are better, if available. Copper foil has a lot of in-plane thermal resistance.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Nice app note:

Cheers

Klaus

Thanks Klaus. (I mis-understood a bit.)

I mounted the FET to an Al plate inside the box. Turned the power to full (~10W) and let it come to equilibrium. (covers on box no air flow.) There are some pics here.

Summary* 'bare' FET to case tab 175C area 0.060" thick 0.032"

4"x2" 110C 113C 4"x1" 120C 125C 4"x1" (sil) 139C

Putting the sil pad under the fet rather than between tab and AL plate made it ~15 to 20 C warmer about in line with a 2C/W resistance for the pad.

I also tried some 'protected' FETs.

VNP10N07: No transients on turn on and off.. same as IRF520 Thermal shut off at case = 155C. FET cycled on and off (automatic reset) on for ~1sec, off for 250 ms.

AUIPS1031: Some ringies at turn on, turn off was slower (more gate capacitance perhaps.) Thermal shut down at case temp of

72C! This was much too low for me and not usable. Maybe there is a lot more resistance from die to case. (I didn't read spec sheets in depth.)

BTS133: Again some ringies at turn on and slower turn off. Thermal cycle at 150C case temperature. There was no automatic reset and current had to be shut off for device to work again.

I'm going to redo the pcb similar to what you suggest. (only 2-layer pcb) And I'm thinking of big fat via's that we'll fill with solder. (I have to do some calc's.)

And thanks for the nice app note (in a later reply) The numbers for a double sided pcb 40C/W for 2.5 in^2 is in line with my smaller 4x1 inch Al slab. ~30C/W.

George H.

*numbers are maybe good (consistent) to 5C. I tried to tighten the nylon screw the same each time. I broke several screws. :^)

On a sunny day (Fri, 9 Jun 2017 06:42:30 -0700 (PDT)) it happened George Herold wrote in :

I absolutely do not understand why anyone would use those horrible glow ball worming heater MOSFETs. Even if you do not know how to program, you can use simple logic to detect and signal a fault condition, and shut the system down, for example:

in this case make sure there is some RC or something on the reset switch users do sometimes keep the button down...

Trump being president cannot be an excuse to mindlessly heat up the environment.

Wow, that's as bad as some of my scribbling. :^) I'm not sure what I'm going to detect. I guess if I had had more fore thought I could have put some temp sensor on the FET. But for a few bucks the VNP10N07 lets me move on.

George H.

Build an analog computer to calculate fet power dissipation in real time. Sum in heatsink temperature and add a suitable time constant to simulate fet thermal mass. The result is simulated Tj. Shut down when that gets too high.

A few cheap parts can let you push expensive fets and heatsinks far better than a simple current limit. You can do the same math in a uP if you have the resources.

Now *that* was mindless.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

On a sunny day (Fri, 9 Jun 2017 08:23:24 -0700 (PDT)) it happened George Herold wrote in :

Yea, but is is GREEN.

The voltage drop across the source resistors in case excessive current (load short) opens simple NPNs wired ORed to a D flopflop.

Already we had some real hot days here last week... be responsible :-)

BTW the LED is a depletion type (using the proprietary drawing symbol '2 green arrows', not the 'O', so it is 'on' when no voltage across it [1].

[1] Depletion type LEDs work like this: A tritium light with a liquid crystal panel... Very efficient, no glow ball worming.

On a sunny day (Fri, 09 Jun 2017 09:10:31 -0700) it happened John Larkin wrote in :

Very impressive!

I ordered 10 LM35 sensors last week, I only needed one, but OK, ebay you know. But this is much much better.

I wonder if he lowers the car emission standards if he pays back the fine to Volkswagen.

Not a fancy Fet but the vnp10n07 does it for a few bucks.

I'm not sure I can make things (heat flow) better by sticking the FET on a pcb. With no pcb, Fet to tab I have 15C/W or thermal R. Coupling to the pcb I can gain about 30C/W in parallel. (10C/W total) But I then have to couple the heat across the pcb. And if the resistance is more than 5C/W, I lose. A big via 1.5 mm (diam) filled with solder has a resistance of ~17 C/W. (Which is inline with James A's calc's here.. previous thread.)

I can stick four big vias under the to-220, but that barely gets me to break even. I can put more further away, but with diminishing returns as the heat has to flow over more surface area.

Oh well,

George H.

Bigger fets (TO247 maybe) or multiple fets help spread the heat out. Spreading thermal resistance is the killer when dumping heat into relatively thin sheets. As the contact area approaches zero, theta approaches infinity.

Bigger fets also increase sil-pad area.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Right I'll have to try that.

"My" thermally protected fets don't turn off correctly. They are wrapped in an opamp current source loop. When I ask for zero (current) the opamp output goes a bit negative -2.5V. and there is some reverse protection feature in the thermal fets.. this pulls about 3 mA from the drain. I can't have that. grumble....

OK. I'll try sticking the fet on the Alum. back panel. (Right I said no air wires, but I'm allowed to change my mind.)

George h.

Sockets on the PCB can help.

A heat spreader can help too, to make the panel effectively thicker, sort of make the transistor footprint bigger. A couple of square inches of 1/8" aluminum or copper. What's neat is to insulate the spreader from the chassis, not the transistor from the spreader.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com