Thank you for that lucid explanation! When you put it that way, it all makes perfect sense....
You should use a much larger Gate-Source Voltage than the threshold voltage. The threshold voltage is the voltage at which the MOSFET starts to turn on. To get the best resistance you need more voltage. The NTMFS5C612NL has a threshold voltage of 1.2 - 2 V, but that is at a drain current of 250 μA. To get the typical 1.2 mOhm you need 10 V on the gate.
If it's a negative-earth vehicle, when the dynamo is off, the MOSFET source will need to be a 0 V, so the gate will have to be at 0V as well. When the dynamo is generating, the drain and source will be about 7 V, and the gate will have to be at 17 V to get the best resistance. You don't really need any current at 17 V and you will have the big advantage of little heat being generated, but you certainly won't be able to use the 3.3 V output of the GPU to drive the gate.
That part looks perfect for this application. Unfortunately, they seem to be scarce, like everything these days. I have ordered the single one I could find to experiment with. The price is reasonable: $2.75 in qty. 1 from Newark.
Some great ideas there!For high current traces, you can just have no solder resist on the bits that take high current, and add solder when the components are fitted.
2 oz copper is 70 µm thick, or 0.07 mm. Tin has a lot more resistance than copper, about 5 times, but you can easily get 0.5 mm of solder onto a wide track, which will reduce the resistance by a factor of more than 2 compared to 2 oz copper.
You could even solder down bare copper wire along the track. A bit of 2.5 mm2 solid wire* is about 1.8 mm diameter, so can be soldered along a 3 mm wide track quite easily. A 3mm wide track on 2 oz copper is 0.21 mm2 so the copper wire on top of it reduces the resistance by over 10 times.
I'm planning to bolt the board down to a steel plate with a 0.5mm piece of something like "SIL PAD" in between. (I didn't know it was called that until I watched this video yesterday:A thinner board will transmit heat from front to back better, but it won't make any real difference in conducting heat across the board, as the copper does that. If the heat is conducted from front to back, unless there is something on the back of the board to take heat away, it's not going to make any difference.
The guts of the old regulator are a mine of good solid wire:Tracks inside the PCB don't add to the thermal problems. They can make it a bit better by conducting heat sideways better. Redundant tracks don't help much if the copper is only 1 oz and the heat dissipation from an internal track is worse than from a surface track.
*2.5 T+E cable (Translation to American:- 14/2 Romex) is a good place to get solid wire like that.
No extra charge at all for the shape. I paid extra for some SMT assembly (just LEDs and their resistors), and got hit with "extended" parts charges, which are parts that have to be retrieved for assembly, I guess. You really have to hunt around to find parts that are not "extended". For example, if I'm looking for a 47 nF capacitor, there are 644 choices in stock, and it seems to be a guessing game to figure out which ones are "extended". I have gone through that exercise on the "power" board that I'm working on: https://oshwlab.com/carlk3/volt-reg-1-1_copy_copy_copy_copy_copy_copy. For that one, I plan to use their assembly services as much as possible, since it is all surface mount (and needs to be for the thermal design to work [if it works]).Very nice. I love the way the large diameter, plated through holes have come out.
Did it cost much extra to have the boards cut to that very particular shape?