EEVblog #1347 - SMD Thermal Jumpers could be GAME CHANGING!

[floobydust]

Is it me, or are the Vishay Thermal Jumper specs out to lunch?
The thermal resistance and conductance are almost the same for 0603, 0805, 2512    I'm expecting larger parts to have proportionately lower thermal resistance.
1.5kVAC hi-pot for an 0603, really? Does it come with unicorns and rainbows too?

[nctnico]

Is it me, or are the Vishay Thermal Jumper specs out to lunch?
The thermal resistance and conductance are almost the same for 0603, 0805, 2512    I'm expecting larger parts to have proportionately lower thermal resistance.

They are also longer so that negates the gain from the extra width. Likely they offer various sizes to be compatible with the parts people use in their designs (like no parts smaller than 0805).

[T3sl4co1l]

If you keep contaminants away -- the voltage standoff is quite high, yes.  As a ceramic, the CTI should be essentially perfect.

In practical cases, yes, such ratings will be rather severely limited by creepage and lower CTI materials (PCB).  You can rout the PCB under the component to address one of those, and fill the thing with potting to address the other.

There are also some resistors rated for pulse voltages, I've seen 10kV on a 0603 -- the datasheet offhandedly noting that you need potting to meet that.


As for thermal resistance, yah, they're all 0.03" thick, what were you expecting?  If they were proportionally thicker you'd expect it to go down proportionally, because bulk resistance has units of m.K/W (multiply by length, divide by cross sectional area, to get Rth).  With constant thickness, the area and length are proportional so the thermal resistance depends only on the aspect ratio.  And the 2512 length is twice its width and the 1225 is half instead, so there's a factor of 4 between them, and indeed that's what they show (within rounding error, or give or take the electrodes).  (This is nothing esoteric, like, we do unit conversions in high school--?)

Tim

[EEVblog]

If you keep contaminants away -- the voltage standoff is quite high, yes.  As a ceramic, the CTI should be essentially perfect.
In practical cases, yes, such ratings will be rather severely limited by creepage and lower CTI materials (PCB).  You can rout the PCB under the component to address one of those, and fill the thing with potting to address the other.
There are also some resistors rated for pulse voltages, I've seen 10kV on a 0603 -- the datasheet offhandedly noting that you need potting to meet that.

Yes, a routed slot under the part is an absolute must if you wanted the full voltage rating here.

[Feynman]

Just to get a feeling: These parts have about 10 times LESS thermal resistance than 1oz copper occupying the same board area. Pretty nice, I think.

[floobydust]

But wait... where are you piping the heat to? Take a DPAK or D2PAK and a thermal jumper to... the nearest iceburg... 
For SMT builds, all I've seen are copper pours, thermal vias to a product's enclosure to deal with dissipating heat.
An app note or design backed up with actual measurements would help.
It could be good for SMPS primary switch which has HV pulses and EMI, where copper-pours radiate and the isolation is important.

[T3sl4co1l]

Correct, you can get it closer to an iceberg without worries of shorting things out.

The final barrier would seem to be, how close do you think your assemblers can get?  Hand assembly, you could butt all these components (transistor, jumper and heatsink) together and let them all get real friendly, but some clearance will be needed on automated assembly.  Or for strain relief, since these are rather brittle ceramic parts and they won't be so happy to have big metal parts and thick solder fillets hanging off them.

Upside, even if you're doing a thermal pad-enclosure heatsinking approach, you can spread the heat out with a thick-ish copper (or other metal) tab.  Far, far more heat spreading than the foil alone can do, so you can get your 5 or 10W out of the thing, even with fairly obnoxious materials like an ABS enclosure or something.  Well, if it's big enough, heh, ABS sucks...


Also neat for fast switching: they make HIP* pure-alumina heatsinks now!  Sounds crazy or ridiculous.  Well, however they manage to make 'em, yeah they're a lot less conductive than metal, but when you're only dissipating a dozen or so watts over a few inches of sink, that doesn't matter very much.  Meanwhile, you have no worry about isolation, just grease and go!  (They usually use spring clamps, no screws needed.)  Very attractive for high dV/dt.

*Hot Isostatic Pressed.  They use an extrusion die (apparently also made out of alumina, but it manages not to weld and erode for some reason), feed in powdered alumina under pressure, heat the middle to just shy of the melting point, and pull out a perfect extrusion from the far side.  Very high density and strength is obtained, as well as apparently almost any shape!

Tim