Heatsink FAIL delays release of my new power supply for the GFA-565. :^/

So I’ve been sitting on this new design for a GFA-565 power supply for two years, because I haven’t found the time to test and validate the design before I start offering them for sale. I needed to install it in an actual amplifier and test its performance. Recently I found the time to build myself a GFA-565!

One of the key features of this new power supply, is the relocation of the bridge rectifier to the tops of the capacitors, which eliminates about 20 inches of 14ga hookup wire that normally runs from the bridge rectifier mounted on the floor on the left side of the amp, to the capacitors on the right side. In my design, the bridge rectifier is made up of four discrete diodes in TO-220 packages, mounted to a circuit board on top of the capacitors.

My design goal is that the heat dissipation for the bridge rectifier needs to be as good as—or better than—the OEM arrangement, with the bridge rectifier mounted to a heatsink on the floor of the chassis.

As I’m about to find out, this heatsink arrangement isn’t enough.

But otherwise it works great! The amp is performing flawlessly. Dead silent and 957W into 2 ohms!

But I need to compare the heatsinks against the OEM arrangement, and here’s how I did it.

Test theory: Push a constant power through the rectifier diodes, and log their temperature rise over time.

Setup: A thermocouple probe is dipped in thermal paste, and fixed to the heatsink. A power supply is connected across the positive and negative terminals of the bridge, such that a constant power can be applied across two parallel pairs of series diodes at once. (All four diodes will have power through them. Their negative thermal coefficient makes them share evenly.) The lid is placed on the amplifier so convection cooling functions as normal.

Power is applied and logging started at 25C.

I don’t have a constant-power power supply; Mine only does constant voltage and/or current, so as the voltage across the diodes drops with heat, I will tweak the current upwards to keep the power constant at 8W.

Results from the new BFA-565 power supply board:

Approximately 16min to go from 25C to 50C. Temperature maxes out and stabilizes at about 53C after 30 minutes. Sorry for the picture quality, it’s very hard to take pictures of these polarized LCD screens.

Now the OEM arrangement: Leads are hooked to the bridge and the thermocouple fixed in place.

Power is applied and logging begins at 25C.

Ohhhhh…. DANG! OEM wins.

Turns out, the OEM arrangement works really, really good! It’s leveling out at 10C lower than my design. :^( Even though the OEM heatsink is only half the mass, and less surface area than the four heatsinks on my new power supply board, the fact that its mounted to the steel floor of the chassis, really makes a big difference.

Back to the drawing board… well back to Eagle CAD actually. I need bigger heatsinks.

I was using these heatsinks, the Assmann V5220X. (Yes there is really a heatsink company called Assmann.)

Their thermal resistance is 8.6C/W, and the datasheet is a bit lacking, in that it only shows a Power vs Temperature rise graph for the 1″ tall version, and I’m using the 1.5″ version.

And here’s the new heatsink I’m going to use. The Ohmite FA-T220-38E. I think these will be more than plenty, but they too will need to be tested.

Quite a lot chonkier! And spikey. At 3.8C/W, they are less than half the thermal resistance. Here’s the curve…

That should do it.

I’ve already re-designed the board to accommodate the new, larger heatsinks. I made it a little wider on both sides, and I had to re-arrange some components, but it turned out great! Now I need to order some boards and try again…

Leave a Reply