1. Home
  2. Electronics Design
  3. FET and thermal stuff

FET and thermal stuff

Hi all, this is related to my FET w/ thermal shust down question.

I mounted the FET to a metal tab on the case that holds the pcb. Here's a picture.

formatting link

I screwed up the first time and had to hack away the pcb. The case is grounded and there is a sil pad between the FET and tab. This did not work as well as I hoped. The case is some sort of steel. (not the best thermal conductor.)

This is ugly and I was thinking of adding a copper pour to the pcb, attach the FET to that, and then attach the pour to the case tab.

formatting link

I'll still need a dielectric break (sil pad) somewhere. Under the FET? or between the case tab and pcb?

Oh as a final 'silly' question. I assume it's a bad idea to think about using the solder mask as a dielectric break.

George H.

Reply to
George Herold
Loading thread data ...

What's the net theta?

How many watts will that transistor dissipate?

Copper pours have a lot of thermal resistance, about 70K/W per square for 1 oz copper.

Sil-pads are terrible. Read the data sheet and multiply theta by 2 or so; they always lie.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics
Reply to
John Larkin

Of the metal tab? I was pumping in ~10 Watts on Friday. The case of the Fet got to ~165C (!) But with a ~2C/W sil pad. So ~14 C/W. total.

Typical (max) is 2-3W, but worse case (someone shorts the output) is 10 W. (If I can't do that much I can reduce the max current or voltage.)

Hmm that does not sound good. I was going to stitch the top and bottom together, so two planes. I've got ~1.5" by 3" that I could use. It's hard to know how much thermal resistance that will add.... I was thinking that it would radiate/convect some.

Maybe best to just leave it bolted to the tab? (I've got these thermal shutdown FETs on order, if those work it won't matter as much.)

George H.

Reply to
George Herold

On a sunny day (Mon, 5 Jun 2017 08:02:35 -0700 (PDT)) it happened George Herold wrote in :

I wires is an option then mount the TO220 first on the side of the case with isolation washers etc, and solder wires to to PCB later? Less thermal expansion problems, better against vibration perhaps. Solder joints can get lose with PCB mounted hot stuff.

Reply to
Jan Panteltje

yuck. I wonder why you didn't mount the fet on the other side of the pcb, & put the bolt through fet then pcb then case lug.

way way too hot. Putting a pcb in the way would make it a lot worse. This is no go. Either mount fet to case - but I doubt that would be enough - or give the FET its own ali heatsink, or a sheet of ali between fet & case to spread the heatflow out.

NT

Reply to
tabbypurr

A TO220 mounted on an infinite sheet of 0.062" thick aluminum, no insulator, will have a theta around 2.5 K/w, measured at the hot spot on the plate. So no aluminum chassis can do better. That tab, on a steel case, will be much, much worse.

A bigger transistor footprint would help if the transistor were bolted to a flat surface. TO247 or something. A heat spreader, preferably copper, would help too.

Or a real heat sink.

formatting link 

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com
Reply to
John Larkin

Right. there's some room on the Aluminum back panel. But the pcb gets crowded there and I don't want air wires. If I attached it to the pcb I could solder it in place. I'm going to wait for the thermal shut down FET.

George H.

Reply to
George Herold

Well it will look a lot nicer when the pcb house makes the cut-out and not my dremel. There's a PEM nut boss on the other side of the case tab.

Grin, It ran at 165 for several hours til I shut it off. I was thinking of other heat sinks.. but with no air flow in the case I'm not sure how much a heat sink will help.

George H.

Reply to
George Herold

I did wonder about distributing Pdiss over several devices fwiw.

NT

Reply to
tabbypurr

Mount the FET on the PCB with 10 mm copper band around it on the same layer. The stitch the hell out of that band with bias and use an internal layer with big size to distribute the heat

Thermal resistance of internal layer is double that of an outer layer, but that doesn't matter much when the layer is large

See if it is possible to fill up the vias to reduce thermal resistance

Cheers

Klaus

Reply to
Klaus Kragelund

How many C/W will thet get you?

NT

Reply to
tabbypurr

Hi Klaus, I can't do any fancy pcb's.. I'm making ~10, maybe 100 pieces. I was thinking I could put a sheet of Al (say 3/32" thick ~ 2.5mm The max our little shear can take.) 2"x 4" stick that to the tabs, (there's another tab 4" (10cm) away.) and stick the fet to that, with a sil pad, right at one tab. I guess I'd have to take it off and paint the sheet black too. How much does coupling to the air, (without a fan.) buy you? (I'll try and find out, for this case. NPI)

George H.

Reply to
George Herold

What's the thermal convection resistance of a slab of air, say 1"x4"x6"? (heat across the thin dimension.) (of course it depends on delta T, in some non-linear way.)

George H.

Reply to
George Herold

Klaus Kragelund wrote on 6/5/2017 5:51 PM:

Isn't it just the opposite, as the size of the plane approaches infinity the thermal resistance is defined by the thickness, no? Or are you saying with a large plane the thermal limitation is elsewhere?

--

Rick C
Reply to
rickman

I don't have a number, but you won't find a *lot* of conductivity across 1 inch of air, especially in such a small area. Make it 1/4 inch and I think you will find the thermal conductivity will be at a minimum and *very* low. At that thickness there is virtually no convection. This may differ considerably depending on the orientation of the slab. With the 1 inch in the vertical direction convection will be reduced compared to having the vertical direction being 6".

The non-linearity is very abrupt. Read a little about chaos theory where the constructed exactly this sort of thermal cell. They find the air is very still until a threshold is reached where convection starts as a smooth motion. As the temperature delta increases the air movement becomes very chaotic. It's been a long time since I read about this, but I think I have it right.

--

Rick C
Reply to
rickman

Inner layer have ~1/2 the thickness of outer layers (of 1 oz pcb.)

I wish you'd ask questions, rather than comment without 'knowing'.

George H.

Reply to
George Herold

I didn't say anything about inner vs. outer layers. You didn't actually understand what I wrote. Try reading it again slowly and make sure you understand what I am writing about.

--

Rick C
Reply to
rickman

Sorry, perhaps I mis-understood. (I have only a small interest in what you are talking about.) I understood Klaus as talking about the copper thickness of inner layers. If I'm wrong I figure he'll correct me. George H.

Reply to
George Herold

On a sunny day (Mon, 5 Jun 2017 14:51:39 -0700 (PDT)) it happened Klaus Kragelund wrote in :

I have something against those trannies with internal limiting. Sure I have used linear regulators that have that internal limit even foldback. But once you put more and more in one design / piece of equipment, the more total heat generation can happen.

It is not that hard anymore to add a current sense and trip using a micro (PIC of course ;-) ) and a neat error report via network or display and do a 100% shutdown of the related device.

Reply to
Jan Panteltje

The idea is to use a small surface for the component on the top layer, just the size of the component and a 10mm extra ring around it. You thermally couple that ring to the component.

Then you place a lot of vias in that ring to transfer the heat down to layer 2. In layer 2 you place a very large copper plane, say 100x100mm.

The heat will transfer from the junction of the component to the band ring, to the vias, and finally to the layer 2 copper. Then you get a hot inner layer, which will couple heat through the FR4 to the surroundings

Try to draw up the thermal resistances, it will show you if it can work

Regards

Klaus

Reply to
Klaus Kragelund

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.