18FET to 12FET TO247 controller conversion

[oldswamm]

I kept seeing people question the feasibility of putting TO247 FETs on an Xie-Chang type board. I was thinking of making a 'here's how I would do it' post, then decided to just do it.
Anyway, here's how I'd do it. :wink:


This is a prototype for an EB818 converted to TO247. Drilled every other hole larger, removed most of the traces near the FETs, soldered bus bars (flattened copper tubing), to the board, drilled holes near each source and drain lug and soldered 10ga solid so it contacts the lead up to the FET case (will change that part of the design so they're easier to solder) and to the bus bars. The gates are driven through ferrite beads (removed from a prototype for a FET audio amp I built 15 or 20 years ago).

I didn't have 'ultimate' FETs so used 75A rated parts. I didn't even have 12 identical parts, so I used Fairchild FDH3632s on the top (faster and easier to drive), and Fairchild HUF75652s on the low side (slower, but much better heat dissipation).

Capacitors are mounted upright on top of the board (I decided right off that it wouldn't fit in a standard case), and enough are added to the bottom, connected to the bus through flat copper to bring the total up to about 4000mfd. There are also some 14 MLCC added at various places as near the high side drains as possible. I used 20 .01 1W smt resistors for the shunt, because I had them. Among other changes on farther controller builds of this kind, I would use .090" flat copper instead of flattened tubing. The transistors are soldered to copper heat spreaders (I haven't attached it to the heat sink yet, just laying there....) Some of the things that are strange, such as extra resistors in the PS are because I intend to bed (or at least heavily coat) the board, and wanted to reduce peak temperatures.

Oh, I should mention that the prototype doesn't work (yet). With power on, not much happens, with the hall connector hooked up, it's like 'lock' is on. The wheel can be turned by hand, but stops at the next pole. The throttle has no effect. I need to put some test points on it so I can 'look' at it without shorting it (the only o-scope probe I can find has a tip like a multimeter, and no 'tips/adapters'). I've been spending most of my time working on the rest of the bikes electronics (and keeping the one I'm ridding operational) (and trying to stay warm, it never got above 0F in Nov, then started raining in DEC ), and only seem to pick this board up every week or two. It was two weeks between 'finishing' it and soldering on the power and phase wires, then more than a week before I soldered on the throttle and hall wires and a programming header, then I had to work up the courage to connect it up.....

Laying on one of the heat spreaders is one of the driver boards for the 6FET I'm planing to build next. These are the first PCboards I've built in over a decade, and I had problems at each stage, including having to buy an HP printer (Brother apparently uses a higher temp toner ). Will use Fairchild FDH055N15A FETs. Interesting device, 2 FETs in one case! 5.5mOhm (max) 150V. The driver board uses FAN7392, and can connect to the gate through ferrite beads or the gate lead can be bent out, cut, and soldered flat to the board, in which case there is probably less than 10mm from the driver case to where the gate enters the transistor case, including the gate resistor.

I have a driver board designed for 18 to 12 FET conversions, but as long as I'm using 9C motors, I don't really need it......

If someone has a 9 or 18 FET, programmable board that has excess damage in the FET area, or bad drivers, but a good CPU, that you could let me have cheap, please PM me. Money (lack thereof), is the main thing limiting both this and the control/interface electronics development.

I also have some FGA180N33A I would love to build one of these up around. A 180A 330V IGBT with only about 1Vec, and easy to drive. The driver boards are designed for 300V plus.

Here's the url for the electronics mentioned earlier: ebike control electronics
Thanks for reading,
Bob

 

[John in CR]

How about 15fet to 12fet conversions. I have 4 15fet controllers that I'll send for free, and even spring for some of the better fets, and if you get more than one going, I get one. I looked once and the 15 looked like it went to 12 of the larger FETs easily.

 

[oldswamm]

I don't have any 15 FET, but unless it's virtually as long as the 18, it won't work.
Notice in the pic how close together the FETs are. Actually barely enough clearance for 100V!

 

[John in CR]

I've got some 18 fetters too, though not as many, but I'm sure I have 2 not totally destroyed. What do you think about the one for you and one for me deal if I spring for 24 TO247's? Personally I'd like to see some good snubbers, since I have low inductance motors that are hard to drive.

As an added bonus a 2 or 3 way switch on the rear of the controller that bypasses the shunt for half, 1/3rd, or none of the current feeding the current sense. ie Bypassing half with 2 equal resistors bridging the shunt would give double the pre-mod current. That would give me the ability to change current limits of the controller without plugging into the computer. I don't understand why this isn't already a common thing. This isn't a show stopper if you don't feel like doing it, but it sure seems like a useful, needed, and easy controller option to me that I plan to start adding on every controller I open.

John

 

[drunkencat129]

I kept seeing people question the feasibility of putting TO247 FETs on an Xie-Chang type board. I was thinking of making a 'here's how I would do it' post, then decided to just do it.
Anyway, here's how I'd do it. :wink:
View attachment 1

This is a prototype for an EB818 converted to TO247. Drilled every other hole larger, removed most of the traces near the FETs, soldered bus bars (flattened copper tubing), to the board, drilled holes near each source and drain lug and soldered 10ga solid so it contacts the lead up to the FET case (will change that part of the design so they're easier to solder) and to the bus bars. The gates are driven through ferrite beads (removed from a prototype for a FET audio amp I built 15 or 20 years ago).

I didn't have 'ultimate' FETs so used 75A rated parts. I didn't even have 12 identical parts, so I used Fairchild FDH3632s on the top (faster and easier to drive), and Fairchild HUF75652s on the low side (slower, but much better heat dissipation).

Capacitors are mounted upright on top of the board (I decided right off that it wouldn't fit in a standard case), and enough are added to the bottom, connected to the bus through flat copper to bring the total up to about 4000mfd. There are also some 14 MLCC added at various places as near the high side drains as possible. I used 20 .01 1W smt resistors for the shunt, because I had them. Among other changes on farther controller builds of this kind, I would use .090" flat copper instead of flattened tubing. The transistors are soldered to copper heat spreaders (I haven't attached it to the heat sink yet, just laying there....) Some of the things that are strange, such as extra resistors in the PS are because I intend to bed (or at least heavily coat) the board, and wanted to reduce peak temperatures.

Oh, I should mention that the prototype doesn't work (yet). With power on, not much happens, with the hall connector hooked up, it's like 'lock' is on. The wheel can be turned by hand, but stops at the next pole. The throttle has no effect. I need to put some test points on it so I can 'look' at it without shorting it (the only o-scope probe I can find has a tip like a multimeter, and no 'tips/adapters'). I've been spending most of my time working on the rest of the bikes electronics (and keeping the one I'm ridding operational) (and trying to stay warm, it never got above 0F in Nov, then started raining in DEC ), and only seem to pick this board up every week or two. It was two weeks between 'finishing' it and soldering on the power and phase wires, then more than a week before I soldered on the throttle and hall wires and a programming header, then I had to work up the courage to connect it up.....

Laying on one of the heat spreaders is one of the driver boards for the 6FET I'm planing to build next. These are the first PCboards I've built in over a decade, and I had problems at each stage, including having to buy an HP printer (Brother apparently uses a higher temp toner ). Will use Fairchild FDH055N15A FETs. Interesting device, 2 FETs in one case! 5.5mOhm (max) 150V. The driver board uses FAN7392, and can connect to the gate through ferrite beads or the gate lead can be bent out, cut, and soldered flat to the board, in which case there is probably less than 10mm from the driver case to where the gate enters the transistor case, including the gate resistor.

I have a driver board designed for 18 to 12 FET conversions, but as long as I'm using 9C motors, I don't really need it......

If someone has a 9 or 18 FET, programmable board that has excess damage in the FET area, or bad drivers, but a good CPU, that you could let me have cheap, please PM me. Money (lack thereof), is the main thing limiting both this and the control/interface electronics development.

I also have some FGA180N33A I would love to build one of these up around. A 180A 330V IGBT with only about 1Vec, and easy to drive. The driver boards are designed for 300V plus.

Thanks for reading,
Bob

did it with a 12fet with stevo lol kmx flys almost killed him self to as we where doing a speed run lol wiped the kmx its on youtube lol did 107km so far keep amps low as posible u will see 100 amps easy even tho ur shunt is a 50a shunt still messing with the controller tho

 

[oldswamm]

John,
PM sent.
I think a fetter is a 'snubber' for the ankles of cows, people, and other animals. :wink: :lol:
I guess an 18fetter would be for a chain gang.....

With faster switching with the drivers I intend to use on future versions of this, a snubber will become more important. I have no experience, but am studying. Would be different for different FETs, and from what I read, probably should be tweeked using an O-scope, and connected to the motor it's going to be used with.

I think the shunt idea would be easier and more accurate to use 2 (or more) shunts in series, and switch which tap connects to the processor ADC.
edit: I'll use a voltage divider, as suggested by bigmoose in Modifying a shunt with controlled results, for current 'proof' testing, now that I understand what John was getting at.

Drunkencat,
I've seen at least one photo of that one (I think). Really, my idea is just to eliminate the wires (as well as a bunch of other details). The reason for the low (current) shunt value is I want everything very overrated, because I intend to 'bed' it so it can survive short term immersion, which will both limit heat dissipation, and make repair next to impossible. For current testing prior to bedding, I intended to solder extra shunts in place.

Bob