You can do it from the bottom but you have to be careful to not burn the pcb with the air temp too high, it takes a lot longer this way. I generally only do this for small components that could be blown off the pcb.
For the FET you just set the temp correctly and then aim it directly over the FET, as long as you are using a hot air gun with the proper temperature the FET will be just fine. You can use a piece of normal solder under the FET but solder paste is much easier and makes it a very simple process.
I didn’t do a lot of technical calculations for the FET cooling but it has a fairly large copper pad for cooling and the wires also act as heat sinkes as well. I did check the FET temperature under load and it stayed well within spec even under extreme conditions.
IIRC the highest temps I saw on the FET itself was around 100c or a bit more under max load and it is rated for 150c. So the temps should not be an issue with the real SIR800DP FET like I used.
Interesting. I bet TA just uses trial and error. His hand built units held up to torture testing so it might be safe assume that is not the cause of the failure.
It’s possible the FET that they used has too much internal resistance and that is what caused it to overheat. Or maybe the factory drivers used a thinner copper layer on the pcb which reduced it’s heat transfer ability.
I did some basic calculations but without a major change in design that would be basically impossible it was not possible to significantly increase the amount of copper cooling for the FET so I did indeed focus more on testing then calculations.
A thinner copper layer is possible but I used a 1oz for my test drivers and I doubt they went thinner then this.
A cheaper FET like the one lexel posted is the most likely cause. It was only rated for like 30A so it would make sense it would blow under bypassed springs conditions and that should be around 30A of current.
There are some nicer once that Lexel found recently that I did not know about when I first built these drivers as well for those that order from mouser / Arrow.
I have 10 of those. If I use two fet it will spread the load and iz is more safe think.
Removed the fet with regular soldering iron heated the top and the package turned in to dust. With a lottle help of pliers removed the top and insoldered the big pad.the pcb survived without any damage.
My good fets and hot air station are at work so I will put on the good tomorrow.
The only problem with using 2 FET’s is that the resistors are setup for a single FET and using 2 will cause them to switch more slowly and increase the losses (aka, they will get hotter) and cause the ramp to change some in the low modes.
You can try it and see what happens though but a single SIR800DP should work fine. That is how the lights I built are setup.