neat SMD heatsink

We've used them, but my enthusiasm is lukwarm. Theta of the sink is high, and the pcb copper from the dpak to the sink adds more. Still, they are sometimes the thing you want.

Here's another version:

and

imes

They're not perfect, but I've got a 6-layer board, packed, 85c worst- case ambient, still air, and about a watt that needs dumping. A little air-borne copper really helps.

Looks like yours has more surface area. I considered the (larger) D2PAK heatsinks, but was skeptical of leaving a gap--that's potentially hard to solder. Adds theta too. The little fellow I picked was good enough, so I stopped there.

Use the layers! Lots of copper around the dpak, tons of vias to pours on inner layers and a big pour on the bottom. The idea is to conduct heat vertically through thin layers of FR4 into the ground plane and any copper power pours, to spread the heat out laterally.

Looks like the major isssue here was the spreading thermal resistance of the copper on the bottomside of the board, which encourages using multiple copper layers for heat spreading. Note the hot spot on the bottom side.

The big diagonal band in T750_t5 is a strip of kapton tape on the bottomside of the board, opposite the resistors. The kapton has a high emissivity in the thermal IR, which allows reasonably accurate imaging of the copper temperature. You can't use an IR thermometer or imager on copper; it's almost a perfect mirror at those wavelengths.

e:

.JPG

gh, and

metimes

inner

ly

rs, to

I'm using the layers, but that's my margin of safety. A great deal of the difficulty has been that it's unknown, and pretty well the worst case scenario (very small box, low clearances, sealed, plastic, small air space, possibly horizontally mounted). It has to work, first time, and I don't get to test it.

Not much room for copper pours on the bottom--that's paved with parts. The top side too, mostly. The inner layers are fully engaged as heat-spreaders--that helps. I calculate the thermal resistance of vias is surprisingly high--you need a lot of 'em.

heat

Nice pictures!

the

pper

The copper patch looks like 0.5x.3"? That's not much area at all--the semi mfr references I've been reading would've put that patch at around 60C/W Your 30C rise is surprising.

I can stand 40C rise, exactly what AAVID says the heatsink provides alone. The modest top-copper + middle layer spreading adds the margin that should make it solid. The refs I've been reading say inner-layer spreading gains about 30%.

ide of

ature.

That experiment of yours is a pretty good sales pitch for a FLIR.

and

sometimes

inner

to

of

temperature.

Anyone know a good current FLIR for this sort of thing? The ones I have seen are all for long-range stuff like looking at buildings. Or do you have to buy a special lens for it?

There is a new cheap type available, but it is no good for closeup work.

The "Technical Data Sheet" is useless, too:

impressive ascii art job there.

The photo I linked showed the part inverted, so I figured a diagram would make it all clear.

(The more I try to explain technical things in words, the more I appreciate the value of pictures.)

and

sometimes

inner

to

heat

copper

of

temperature.

The Fluke people that I talked to didn't seem to know anything about their IR imagers.

One of the guys here recently bought a lower-price thermal imager. I don't know how well it works on PC boards. Extech has some low cost imagers, too.

The Flir will resolve the hot-spot temp on an 0603 resistor, but it cost us $12K or some such.

high, and

sometimes

inner

pours, to

heat

the

copper

bottomside of

temperature.

$12K

I use quarter inch male spade lugs. They are plated copper so they take solder really well, and hence couple the thermals really well. They are about 1mm thick.

Instead of solder, I use silver filled conductive epoxy. The same thing they use to attach the chip die to the back side of the thermal interface surface! Stay away from the pins!

and

sometimes

inner

to

heat

copper

of

temperature.

How many pixels??!!

bottomside of

temperature.

Oh they tell you that. For the visual camera!

Like I said, useless.

I got one in earlier in the week, out of curiousity (everything is "sale or return" from Farnell in the UK). It resolved my steaming coffee cup quite impressively. But no chance at seeing small parts.

bottomside of

temperature.

I wonder if it uses this:

bottomside of

the

temperature.

perfect

Dunno, could be.

Actually just a pinpoint measurement together with a visual picture would be pretty useful for occasional use. So you could point it at each suspect part in turn and get a reading.

bottomside of

y in the

temperature.

perfect

ve

rk.

sorta like this?

-Lasse

bottomside of

the

temperature.

perfect

Nice, and a slow scan would be perfectly feasible for looking at circuit boards. But again it looks like it is not good for small parts.

Here's a bare-bones DIY imaging approach that uses the same Melexis

It provides the thermal data, but it does not offer an archivable image as a result ( though you can photograph the results ).

Enjoy...

Frank McKenney