liveforphysics wrote:This is a controller rated for 350amps for 2minutes when bolted to a huge 5lbs heatsink.
It's got 42 mosfets in it. It's got 24 big ass low ESR caps positioned between each row of FETs that have an excellent thermal path to the heat sink.
It's got massive fat high current busses running all over the place.
Thats a nice piece.
liveforphysics wrote:RC controllers are for toys.
Most of them probably are. But I want to argue, that if done properly, you can make something very powerful in a small package like the ESC in question.
Lets make a rough comparison.
The Sevcon has 7 FETs in parallel. I don't know which FETs they are, so I'm going to assume IRFB4115
. They have an RDS of 9 mohm, which is doubled at high temperatures - 18 mohm. 7 in parallel makes 2,5mohm. The switching losses on these through hole FETs are quite high. On 120V they are probably roughly the same as the conduction losses, so I'm going to count that as a doubled RDS - 5mohm. The high side and low side are in series - 10mohm. Now it's easy to estimate the heat losses:
At nominal 140A: 140A^2 * 10mohm = 200W
At 350A: 350A^2 * 10mohm = 1200W
I don't know which heatsink Sevcon is using, so I'm assuming something similar to this 300x300x40mm
, which is 0,28K/W (natural convection). This puts the heatsink at 56K above ambient at nominal current, and would put it at 340K above ambient if used continously at 350A. The thermal mass of the heatsink makes it possible to use it for 2 minutes. If the mass is 2500g aluminum, then the sink would rise about (1200 - 200)W * 120s / (2500g * 0,9J/(g*K)) = 53K above the nominal temperature, but in practice less, since the heatsink moves more heat as the temp diff to ambient increases. I'm too lazy to solve the equation. 14 FETs with a thermal resistance junction to case of 0,4 K/W is 0,03K/W. At 350A, the FETs will get 1200W * 0,03k/W = 36K hotter than the heatsink.
It looks like the FETs junctions are approaching their maximum temp (175°C) after 2 min of 350A. However, at 50A per FET, the max allowed temp is about 125°C. My numbers are probably a bit worse than reality, but I think they are in the right ballpark. With some airflow around the heatsink, it gets better.
The SMD ESC could have 10 FETs in parallel, or more. I'm going to assume IRF7769
. They have an RDS of 3 mohm, which is doubled at high temperatures - 6 mohm. 10 in parallel makes 0,6mohm. The switching losses on these smd FETs are not very high. On 80V they are probably roughly 30% of the conduction losses, so I'm going to count that as a 1,3 times the RDS - 0,8mohm. The high side and low side are in series - 1,6mohm. Heat losses:
At 140A: 140A^2 * 1,6mohm = 32W
At 350A: 350A^2 * 1,6mohm = 200W
Again, I don't know what heatsinks they are, so I'm assuming something similar to this 40x40x16mm
, which is 1.93K/W (forced convection). With two of them, it's about 1K/W. This puts the heatsinks at 32K above ambient at 140A, and would put it at 200K above ambient if used continously at 350A. The thermal mass isn't much on these heatsinks, they are 21g each, so about 40g aluminum total. If we allow the sink to rise 60K above above the nominal temperature, we will be able to use 350A for (40 * 0,9J/(g*K)) * 60K / (200W - 32W) = 12s , and in practice more, since the heatsink moves more heat as the temp diff to ambient increases. 20 FETs with a thermal resistance junction to drain of 0,5 K/W is 0,025K/W. At 350A, the FETs will get 200W * 0,025K/W = 5K hotter than the heatsink. This is assuming they are soldered to an aluminum PCB, and the heatsink is mounted on the other side of the PCB.
12 seconds of 350A is not much, but if you increase heatsink mass from 40g to 400g, then you have 120 seconds, which is on par with the Sevcon.
To me it looks like the SMD stuff could compete with the through hole stuff, if done properly. I have yet not seen it done properly.