DC-DC layout efficiency

[evandude]

I just have a few questions. the designer of this circuit said that he went from about 60% efficiency to over 90% efficiency just by moving some wires around, and tweaking some component values. now, the values I am using are his "tweaked" values, but I'm trying to improve on the layout in any way I can. This is my current design. Anyone see any glaring issues that I should change in the layout to possibly get better efficiency?
(ignore the bottom capacitor overlapping things, the one I'm actually using is the same size as the rest of them, i just used that model in the layout because of the wider lead spacing to clear the trace going underneath)

Other than that, I just have a few questions. How much power is it OK to dissipate in a TO-220 packaged device without a heatsink? The DC-DC has a schottky diode that has a maximum forward drop of 0.5v at 20 amps I believe, and I am running both diodes in one package in parallel, usually at a few amps... so the diode drop should be even lower than that. In other words, the maximum power dissipation will be a few watts. So far the diode hasn't warmed up enough to be noticeable to the touch, but I also haven't really been pushing the current output very hard. Can anyone point me to some sort of quick estimate of what size/shape of heatsink you should use for a given wattage?

I am currently going on the "when it gets so hot it almost blows up, then i'll upgrade the heatsink" plan... but knowing a particular size I should aim for so i can take care of it once and for all would be nice. thanks...

 

[stevez]

I'm wondering what there is about the layout that has such a profound effect on efficiency.

Regarding dissapation - it all depends on the surroundings. Some of my amateur radio handbooks have fairly simplified methods on heat sinks that might be useful to you. An engineer from Cornell-Dubilier gave me a number in watts per square inch of surface for small components that could be applied. I looked and can't find his email but it might be a fairly well known rule of thumb that someone else might post.

You might determine this experimentally - if you knew how warm/hot the tab is allowed to get. They make sticks (tempilstick is one name) that melt at specific temps. Set up something to simulate the expected situation and test the temp of the tab at various power levels. Don't attach a temp probe because that will act as a heat sink and introduce a lot of error. I think they make resistors in that case size so you could simplify an experiment in that way - just a variable supply.