DPAK resistor

We use Caddock resistors. The fine print says it will do 2.5 watts with 1" squared of 2 oz copper on pcb.

George H.

We use Caddock resistors.

What's the capacitance end-to-end, with the tab floating?

It's a 50 ohm resistor, so it would be hard for me to measure that.

Since each end is about 2.3, the two in series must be at least half that. I'll probably always ground the tab, since I need to heat sink it to my ground plane.

Just got some samples of different style DPAK resistors.

The Riedon part is rated for 25 watts, and the Ohmite for 45!!!

I'll test them for capacitances and TDR.

The real issue is, how much power can I allow one of these to dissipate on a PC board? 8 or even 10 watts would be useful right now, but that might get scary.

Maybe I could measure the temperature of the tab with a thermistor and shut down if it gets extreme. What would that be for a PC board? 120C or so?

mandag den 29. juli 2019 kl. 02.33.55 UTC+2 skrev John Larkin:

some info here,

I struggled with 2.5 watts, saw 150C surface temps. After adding a heatsink with dozens of nice prongs, saw much lower temps on the prongs, but still 150C at heatsink base = SMT component surfaces. Only way to get rid of the heat is to add airflow with a fan.

Was that a dpak?

I'm thinking that I might have a lot of vias to a big bottom-side pour, than a gap-pad to the bottom of the aluminum box.

My resistor will get really hot if the user shorts my output (instead of terminating in 50 ohms) so the thermistor shutdown idea looks neat.

That sound like a lot... but you know more about those things than most of us here. (An interesting question would be; How much does each square inch of added PCB copper add to the cooling at the device. Assuming copper is added at the periphery.)

How about 2 or 3 in series or parallel?

George H.

PCB copper doesn't conduct heat very well. 1 oz is about 70 K/W per square. If there is a hot part soldered to a square of layer 1 copper, there is dimishing benefit from making the copper bigger; an infinite sheet will have some asymptotic theta. What helps is to use a bigger part to get a bigger footprint, or equivalently via to additional layers to spread the heat quicker. Lots of vias.

That would help thermally, provided they are not clustered together. But spreading them out has its own problems.

I have a rough idea about thetas.

I could maybe get to 10 k/w, maybe a little less, with some bottom-side cooling help. But how hot might I allow the resistor to get? We reflow solder boards at over 200C, and they survive. Maybe allow 150C under fault (user shorts my output) conditions?

Even clustered you'll still get the benefit of less R-thermal from the heater element to the outside of the package. (well same R_thermal, but shared heat load.)

Hmm, testing some to-220 pac resistors, I melted the solder junction a few times (built in thermal fuse :^). The resistors seemed to survive without a change in R... at least a few times. Eventually I noticed that the resistance had dropped and I replaced it. My assumption is that R's are more durable than IC's (at least temperature-wise)

George H.

mandag den 29. juli 2019 kl. 17.22.39 UTC+2 skrev John Larkin:

you can get pcbs made with copper inserts

We could also hog a hole in the board, under the part, and bring up an aluminum pillar from the bottom of the box. Both would be more expensive than a bunch of vias. Vias are basically free.

I did that on my Pockels Cell driver, cut a big hole for the Cree fets to access the water-cooled baseplate.

Machining is expensive.

tirsdag den 30. juli 2019 kl. 18.56.57 UTC+2 skrev John Larkin:

that's the point of the insert, like a vias it comes with the pcb

we recently used in an RF PA under the main transistor(s) I don't think it cost that much more than a regular PCB

About PCBs, just some weeks ago we got quote for 2 layer thick aluminium PCB

To my amazement they came out cheaper than copper layered PCBs

Cheers

Klaus

From whom?

Cheers

Phil Hobbs