zombiess
10 MW
As many know Xie Chang has dropped the EB224 controller and started using a new EB324 controller design. This controller design ends up splitting phase be pretty far apart which is not ideal. I ordered 20 of these controller and sold a few after doing some personal testing on them up to 160A battery and around 200A phase without issues. Before going through and selling more I wanted to make sure these were OK before I sold 20 of these and found out all of them were going to go bad. I'm glad that I'm waited (largely due to my laziness on building controllers).
So I had a customer return 2 controllers that I he had tried out, both failed on him very quickly and at low power, low speed. I took the damaged controllers apart and diagnosed them. To my surprise they all had the exact same failure, phase B (green) had exactly one FET die. Physically the FET looked OK, it was just internally shorted. I repaired one with a matched set of FETs and installed it in my own Greyborg Warp setup for further testing. I rode it for 20 miles and today when I was at my furthest point from home... it died. Let me tell you it is not fun to try and pedal a 100lb bike home 6.5 miles with 52T/16T gearing on 25" diameter tires. Thankfully most of the ride was downhill and I was able to get back home without killing myself too bad.
I just ripped apart the controller for diagnosis and took and educated guess that the exact same FET failed again. Removed it and bingo, rest of them were good. Removed a 2nd one and matched it against my known good spares on the FET matcher and then soldered them both back in. While looking at the controller I thought this is being caused by a design flaw. The FET that has now shorted 4 times on 2 different controllers has the shortest path to the phase wire. All 3 failures were identical in nature, low speed, low load (but high phase current due to being a buck inverter).
I decided to unsolder the phase wire and attach it in the middle of the trace between the high and low FETs. This makes the current path a lot longer from the high side FETs which means the resistive/inductive differences should be closer together percentage wise vs it's stock location. I took some pictures to illustrate what I suspect is a solution to this problem. I need to test this controller for 20-50 miles before I go ahead and start selling the rest of the 24 FET controllers I have. I will be carrying a spare 6 FET controller as get home backup in case I suffer another failure. Pedaling this beast sucks.
At first I suspected I could have been suffering some issues of solder flux conducting enough to turn the gate on of an FET, but I started cleaning it off between all traces and especially around the FET legs. Still had the same FET fail. I leave it alone on the big areas as I am not worried about it doing any harm there.
Pics
High side, FET that keeps failing is the 2nd to last from the right.
Here are the current paths in pink, the shortest one is the one that keeps popping at low throttle
View attachment 1
My attempt to mitigate the differences in current path length
So I had a customer return 2 controllers that I he had tried out, both failed on him very quickly and at low power, low speed. I took the damaged controllers apart and diagnosed them. To my surprise they all had the exact same failure, phase B (green) had exactly one FET die. Physically the FET looked OK, it was just internally shorted. I repaired one with a matched set of FETs and installed it in my own Greyborg Warp setup for further testing. I rode it for 20 miles and today when I was at my furthest point from home... it died. Let me tell you it is not fun to try and pedal a 100lb bike home 6.5 miles with 52T/16T gearing on 25" diameter tires. Thankfully most of the ride was downhill and I was able to get back home without killing myself too bad.
I just ripped apart the controller for diagnosis and took and educated guess that the exact same FET failed again. Removed it and bingo, rest of them were good. Removed a 2nd one and matched it against my known good spares on the FET matcher and then soldered them both back in. While looking at the controller I thought this is being caused by a design flaw. The FET that has now shorted 4 times on 2 different controllers has the shortest path to the phase wire. All 3 failures were identical in nature, low speed, low load (but high phase current due to being a buck inverter).
I decided to unsolder the phase wire and attach it in the middle of the trace between the high and low FETs. This makes the current path a lot longer from the high side FETs which means the resistive/inductive differences should be closer together percentage wise vs it's stock location. I took some pictures to illustrate what I suspect is a solution to this problem. I need to test this controller for 20-50 miles before I go ahead and start selling the rest of the 24 FET controllers I have. I will be carrying a spare 6 FET controller as get home backup in case I suffer another failure. Pedaling this beast sucks.
At first I suspected I could have been suffering some issues of solder flux conducting enough to turn the gate on of an FET, but I started cleaning it off between all traces and especially around the FET legs. Still had the same FET fail. I leave it alone on the big areas as I am not worried about it doing any harm there.
Pics
High side, FET that keeps failing is the 2nd to last from the right.
Here are the current paths in pink, the shortest one is the one that keeps popping at low throttle
View attachment 1
My attempt to mitigate the differences in current path length