Mounting an 11W SCR with a Flat Profile

Question

I have two SCRs in my design that will be switching up to 15A at 120-240VAC. The two devices will split the power, with each one conducting a half cycle. I'm considering using the S6025 by Littelfuse.

The power through each device works out to...

(15A * 1.5V) / 2 = 11.25W per device

I'd like a flatter profile for my PCB so it can fit in a slim enclosure. I'm considering a few options...

1. Using a D2Pak. I'm skeptical of a D2Pak carrying so much power because the metal tab is connected to the PCB, not a heat sink. Can this be done?

2. Mounting a TO220 like a D2Pak but with with the metal tab upwards and attached to a heat sink (see diagram). The leads of the TO220 don't seem to have enough clearance to get past an NRTL. They're only ~1.44mm apart in the worst case.

Edit: Just to clarify, I believe both the clearances from leads to heatsink and leads to other leads aren't enough.

3. Using an alternative, wider-pitched package like TOP-3 or TO-247. These cost ~3-10x as much as the TO220 alternatives.

Is there a cost-effective way to dissipate 11.25W from an SCR without a TO220 standing 2cm tall?

Answer 1

Yes, this is fine. Note that you should leave a fairly large hole in the PCB to provide access for the screw and screwdriver, which may make routing more challenging. Alternately, position the device near the board edge, so the screw can be accessed freely (almost as shown).

If insulation is required, the usual route is a thermal pad, with shoulder washer for the screw. With isolated tab an option, the shoulder washer could be omitted, and insulation used just for clearance around the leads if needed (or if thermal pad is more convenient for manufacturing than grease).

Another option is to use spring retention, like these,
MAX07NG Aavid, Thermal Division of Boyd Corporation | Digikey
combined with a thermal pad (plain, no hole). You might prefer the non-isolated version SCR for slightly lower RthJC, which helps make up for the higher resistance of the thermal pad.

The PCB itself can also be used as mounting clamp, though it's a bit sketchy because of straining the board (makes putting ceramic capacitors nearby problematic, for example), plus board material isn't great at retaining stress (FR-4 cold-flows a bit). A typical build might flank the component with two screws into the heatsink. On the upside, this scales well when a row of devices needs to be clamped down, since a screw can sit inbetween pairs.

If the problem is clearance around the leads, partial potting might be considered. It could be gooped around just the base of the part; probably an approved application and inspection process will be needed.

As for SMTs, I wouldn't recommend it for this power level. You need thermal vias and enough pour area to spread out the heat, until it can get into a thermal pad and the bulk heatsink. It's good up to 5W or so, but 10 would be pushing it. Potential improvements include: thinner PCB (shorter vias, lower Rth from board material itself), heavier plating (e.g. 1oz foil plated 2oz+, thus leaving heavy walls in the vias), heavier copper in general (spread heat out further in the first place), heat spreader components around the tab (whether bits of SMT metal, or those AlN chips -- the latter are probably too expensive, but interesting especially where isolation is required), etc. Enough such improvements should get you there, but maybe that starts to drive the fab cost up as well, and you'll want to compare quotes versus other assembly methods.

There's also metal-core PCB. Limitations are few layers (preferably just one!), and possibly poor insulation between conductor and core (I haven't checked offhand what typical ratings are; enough for nominal voltage, but safety I'm not sure?). If nothing else, it could be useful as a heat spreader -- treat the core as live, and use a thermal pad between it and the heatsink.

Also, to avoid clamping forces and hardware, you might consider a bonding TIM like this,
BERGQUIST BOND PLY TBP 1400LMS-HD Technical Data Sheet
parts can be clamped in place (perhaps with a jig for accurate positioning) and is self-adhesive to start, then can be cured for a permanent bond.

Answer 2

Here is what managed to get through UL/CSA in the past, this is your typical 120VAC 600W '$5 dimmer' used in the past for dimming incandescent lamps:

The total maximum dissipation would be around 5W, so about half of your per-device dissipation. It's also designed for use in the relatively benign residential environment, and with modest reliability requirements.

As can see, they've added a robust insulator sheet to deal with the problem you've correctly identified.

I don't like the sharp bends near the package on the thyristor from a reliability point of view, this is pretty crude and nasty work. If the sheet is partially under the triac-diac thyristor, which it appears to be, they're also stressing the package.

Yours is a relatively large amount of dissipation, both per device and total, and you'll need to do careful thermal design and verification. It may be difficult or impossible to get reliable junction temperatures without having at least fins to increase the effective area, especially if you need to deal with an extended temperature range. Of course you could go to a forced air design with a thin fan as you'd find in a notebook computer, but fans wear out, get dusty, cause noise, require holes in the enclosure (holes that don't allow little fingers to get too far in) etc.