ElectricGod
10 MW
I've had this trapezoidal Kelly controller for several years. It's also pretty strong...something like 10kw. It's seen about 6000 miles of use and has never given me trouble. It has one oddity with regen where if you have it set too high, the controller will disable at random when coasting down hill. It's like regen creates too high a voltage and the controller shuts down at random. Otherwise, it has worked reliably for a long time.
After I see how well this works, I also have a KLS7230S controller I may do this mod to it as well.
I want to run a higher eRPM motor on this controller so I contacted Kelly and asked about the 70,000 eRPM option for my 40,000 eRPM 72221 controller.
I already knew that Kelly products are modular in nature. IE: The lowest wattage controller in a product line and the highest wattage all use the same base controller module. Kelly adds on a daughter board with more mosfets to increase the wattage. In my controller there's 2 sets of phase and battery wires. One set is for the main board and the other for the daughter board. You connect them in parallel outside the controller.
It took some doing to convince Fanny Chen at Kelly that swapping the board was something I could do without trouble. After showing her a few modding threads and other things, she finally relented and sold me the 70,000 eRPM board. It cost me $50 including shipping and arrived in less than a week. Fanny said the faster board has a 32 bit MCU and the slower version an 8 bit MCU. I was curious what was different between the 2 board options. Well now I know...just the MCU and they are both pin for pin compatible.
The boards inside the controller are mated together via pins for the mosfet gate driving only. The left board is the mosfet daughter board. The right board is the actual controller module and it's a complete 12 fet controller that is the exact board in the entire KBS controller product line.
Close-up of the interconnect pins.
This is the 70,000 eRPM board as I got it from Kelly.
This is the 40,000 eRPM MCU and board before I started pulling parts for the 70,000 board. Normally the MCU is somewhat buried under the DC-DC converter which I already pulled off for these images. If you compare parts placements and values they are identical between both boards except the MCU.
I started with the DC-DC and then pulled the large caps and soldered them on the faster board.
I wanted to preserve the 40,000 eRPM board. Who knows...might be useful later. Getting the mosfets pulled was a giant PITA without damaging the board or mangling the mosfets. I have to say that this design of soldering down the mosfets to a single back plane sounds good in theory, but if you ever blow mosfets, repair is way difficult. Among a few other details of this controller, this is not a great design IMHO. Separately installed mosfets that use a ceramic insulator behind each one is much easier to work on...if needed.
The populated new board.
Mated with new pins to the mosfet daughter board. I'll probably get around to testing it tonight.
There's a few design details that are not very good about this design.
Looking at the mosfet legs, you can see they are not pushed all the way down into the board. The thinner section of the legs are not rated for 75 amps, just the thick section. This is going to generate lots of heat in the legs. On the faster board, I installed the mosfets with the legs pushed in all the way. The interconnect pins are a pain. They are difficult to solder in place and cutting them to get the boards apart is the only option. It would be better if some kind of reliable connector was used instead. 2 pin headers would do the job and then the boards could come apart. They are screwed to the shell top and bottom. They can't come apart inside the shell.
The main traces on the board have solder braid on the batt- buss and it's not consistent. There really needs to be reinforced traces everywhere phase and battery power goes and it's not there. This is a 10kw controller! This is 100% of the reinforcing. I'd be putting copper on top of the traces to every mosfet leg! Mosfets are in sets of 4 across the width of the board, but there's no reinforcing between them and the phase connections are on one side only. This is problematic and unbalanced. Both sets of mosfet boards need the same strong current paths.
After I see how well this works, I also have a KLS7230S controller I may do this mod to it as well.
I want to run a higher eRPM motor on this controller so I contacted Kelly and asked about the 70,000 eRPM option for my 40,000 eRPM 72221 controller.
I already knew that Kelly products are modular in nature. IE: The lowest wattage controller in a product line and the highest wattage all use the same base controller module. Kelly adds on a daughter board with more mosfets to increase the wattage. In my controller there's 2 sets of phase and battery wires. One set is for the main board and the other for the daughter board. You connect them in parallel outside the controller.
It took some doing to convince Fanny Chen at Kelly that swapping the board was something I could do without trouble. After showing her a few modding threads and other things, she finally relented and sold me the 70,000 eRPM board. It cost me $50 including shipping and arrived in less than a week. Fanny said the faster board has a 32 bit MCU and the slower version an 8 bit MCU. I was curious what was different between the 2 board options. Well now I know...just the MCU and they are both pin for pin compatible.
The boards inside the controller are mated together via pins for the mosfet gate driving only. The left board is the mosfet daughter board. The right board is the actual controller module and it's a complete 12 fet controller that is the exact board in the entire KBS controller product line.
Close-up of the interconnect pins.
This is the 70,000 eRPM board as I got it from Kelly.
This is the 40,000 eRPM MCU and board before I started pulling parts for the 70,000 board. Normally the MCU is somewhat buried under the DC-DC converter which I already pulled off for these images. If you compare parts placements and values they are identical between both boards except the MCU.
I started with the DC-DC and then pulled the large caps and soldered them on the faster board.
I wanted to preserve the 40,000 eRPM board. Who knows...might be useful later. Getting the mosfets pulled was a giant PITA without damaging the board or mangling the mosfets. I have to say that this design of soldering down the mosfets to a single back plane sounds good in theory, but if you ever blow mosfets, repair is way difficult. Among a few other details of this controller, this is not a great design IMHO. Separately installed mosfets that use a ceramic insulator behind each one is much easier to work on...if needed.
The populated new board.
Mated with new pins to the mosfet daughter board. I'll probably get around to testing it tonight.
There's a few design details that are not very good about this design.
Looking at the mosfet legs, you can see they are not pushed all the way down into the board. The thinner section of the legs are not rated for 75 amps, just the thick section. This is going to generate lots of heat in the legs. On the faster board, I installed the mosfets with the legs pushed in all the way. The interconnect pins are a pain. They are difficult to solder in place and cutting them to get the boards apart is the only option. It would be better if some kind of reliable connector was used instead. 2 pin headers would do the job and then the boards could come apart. They are screwed to the shell top and bottom. They can't come apart inside the shell.
The main traces on the board have solder braid on the batt- buss and it's not consistent. There really needs to be reinforced traces everywhere phase and battery power goes and it's not there. This is a 10kw controller! This is 100% of the reinforcing. I'd be putting copper on top of the traces to every mosfet leg! Mosfets are in sets of 4 across the width of the board, but there's no reinforcing between them and the phase connections are on one side only. This is problematic and unbalanced. Both sets of mosfet boards need the same strong current paths.
