Epic 7kw+ 12 fet controller...or there abouts :)

ElectricGod

10 MW
Joined
Nov 1, 2015
Messages
3,404
Location
3 OH 3
This is a test of an interesting idea. Excessive cooling...

That 7kw+ won't be peak...that will be continuous.

Mosfets can handle a ridiculous amount of current, if you can keep them cool. Whatever the die limit is for the mosfet, that's really how much current it can handle...aka 100 amps or 180 amps or whatever. There are a few limiting factors. For example the legs on a TO-220 mosfet can't deal with more than 75 amps. So no matter what the die limit is of the mosfet, you really can't get more than 75 amps out of any TO-220 mosfet. The mosfets legs will burn out or rather get so hot they will cause the mosfet to overheat. AKA leg limits and cooling are the BIG limiters to any mosfet.

The TO-220 75 amp leg limit is still a LOT....gobs...oodles...loads! In a 12 fet controller...that's like 300 phase amps...if all you have to think about is the 75 amp leg limit. Reality is obviously a bit less than this.

It's silly to imagine 300 phase amps coming from a 12 fet controller. BUT you can get somewhat closer to it than the 4-5kw continuous I typically see from 12 fet controllers after I mod them. What's needed is making all the power busses very stout and as close to the mosfet legs as possible. Also the mosfets need to be kept cool so they can be massive current switches. Short legs on the mosfets...as short as they can possibly be helps with this. And of course loads of heat sinking. Typical controller shells are designed for practical size instead of maximum heat sinking.

1. A typical 12 fet controller can't conduct sufficient current to get remotely close to 7kw. I need larger shunts. These are 12 awg. All I care about is keeping the total shunt resistance at 1.25 mOhms so the the MCU can read current reliably. These are about 6 mOhms each so 5 in parallel is 1.2 mOhms...close enough. I think I might get away with 6 in parallel. I did an experiment some time back with 12 awg copper wire in place of the shunts. The MCU had issues at very low current levels and would stop running the motor. Get the motor going a little bit so current draw could get to about 10-20 amps and it ran just fine. I was at about .5-.75 mOhms with straight copper in place of the shunts.

Controller%20upgrades%201.jpg


2. All the current paths need to be beefed up. In a PV 12 fet, they already have .5mm of added copper on all the main traces. I intend to beef them up with some solid copper wire and extend the wire to the actual mosfet legs. Currently the phase and power traces extend to the mosfets with some added solder.

12%20FET%20controller%204_zpsd3kmt6ly.jpg


3. The internal heat spreader is fine for internal, but the shell needs a lot of help. Imagine adding a second heat spreader on the outside of the controller that then extends to LOTS more heat sinking. I'll be using 1/4" thick 6061 for the external heat spreader and for extending to a larger heat sink.

top%20of%20board%20-%2010%20awg%20wires_zps6owg40pm.jpg


4. The secondary heat sinking will be comprised of 2 CPU heat sinks. I doubt I'll need the fans since lots of direct airflow will happen while moving. In an active airflow path, this ought to provide a LOT of cooling. I'll mount them something like this. With the vertical arrangement, a fan between them will pull air in one heat sink and out the other one...if I need added cooling.

CPU%20heat%20sinks%202.jpg

CPU%20heat%20sinks%207.jpg


The rest will be what I typically do to a 12 fet to get it to work at 4-5kw continuous.
10 awg phase and battery wires being the big thing still needing done. The controller already has top tier TI mosfets.

What the controller is actually capable of...well that we will see! I built up a 12 fet controller for a friend in the UK. He is consistently seeing 6kw since his is mounted directly to a thick piece of steel in his battery box. His controller temps are rarely more than air temperature.

More to come soon...
 
Best wishes.

I tried to maximize the power out of a 12FET with 3077's and kinda hit a wall when i realized that the battery cables were the biggest issue. Ran 8 gauge but even those still got hot on 70 amps.
You'd have to drill out the board and run the battery cables out of a separate hole to get bigger.

This is of course after doing the usual beef up the traces routine.

ZombieSS had a lot of threads on delta matching the FETs so that they output more power reliably, if you really want to get extreme and build the most uber 12fet ever.
 
neptronix said:
Best wishes.

I tried to maximize the power out of a 12FET with 3077's and kinda hit a wall when i realized that the battery cables were the biggest issue. Ran 8 gauge but even those still got hot on 70 amps.
You'd have to drill out the board and run the battery cables out of a separate hole to get bigger.

This is of course after doing the usual beef up the traces routine.

ZombieSS had a lot of threads on delta matching the FETs so that they output more power reliably, if you really want to get extreme and build the most uber 12fet ever.

Thanks for the info...I'm sure I'll run into limiting factors.
 
When I was space constrained, I went from 12 to 24FET that has the FET rows parallel. Then I proceeded to cut the case down a bit, so it would be only as big as the circuit board. Not to mention the traces/wires are bigger from the get go, the bus is solid copper, and 7kw is easy to get.
The cooling interface is crap on these. You need a different computer grade thermal paste and possibly rough up the surfaces.

No fit.
gPNSuIP.jpg


Won 2 inches.
QrAHTI0.jpg


Perfect fit.
bgUxIul.jpg
 
I'm not sure what you are presenting here. 7kw on a 24 fet should be no problem at all...even with low grade mosfets. I'm talking about getting 7kw out of a 12 fet controller. Cooling being the main issue and the reason for the CPU heat sinks. How does anything you presented have anything to do with making a 7kw 12 fet controller? I guess you were trying to present that your small controller shell was a problem?

You mentioned thermal paste...I use the cheapest stuff I can get. Heat conduction is negligibly different between the expensive and cheap stuff. It's whole purpose is to fill tiny gaps between surfaces. More than that, get the surfaces flatter. Thermal paste was never designed or intended to fill large gaps. Realistically for best heat conduction, you want direct metal to metal contact.
 
One approach would be to replace the existing heat spreader in the 12F with a piece of thick angle aluminum so you can get more heat to the top side where there are fins (or your CPU coolers). The heat path needs to be as short as possible through the spreader.

I run one of the 12F units on my A2B. Limited by the battery voltage and motor resistance to about 60A on the battery side. I used aluminum pieces to make a bridge to the bike frame which acts as a huge heat sink. The controller never gets more than barely warm by the time the motor is overheating.
 
Note that motor current (sometimes called phase current) also flows through the FETs, so suffers the same lead current limitation (300A motor current exceeds the lead current specs). Since motor current is always greater than or equal to battery current it is the primary limitation on current in the controller.

Another issue is that paralleled FETs don't divide current evenly. So the current through whichever FET is higher determines the overall current limit at something less than twice the single FET current. Here's a paper on FET balance: https://www.infineon.com/dgdl/para.pdf?fileId=5546d462533600a401535744b4583f79
 
ElectricGod said:
I'm not sure what you are presenting here. 7kw on a 24 fet should be no problem at all...even with low grade mosfets. I'm talking about getting 7kw out of a 12 fet controller. Cooling being the main issue and the reason for the CPU heat sinks. How does anything you presented have anything to do with making a 7kw 12 fet controller? I guess you were trying to present that your small controller shell was a problem?

I was guessing the point of this experiment is that you are space constrained? So I wrote down how to make a smaller package of a bigger controller. But apparently this is a novelty project.
 
fechter said:
One approach would be to replace the existing heat spreader in the 12F with a piece of thick angle aluminum so you can get more heat to the top side where there are fins (or your CPU coolers). The heat path needs to be as short as possible through the spreader.

I run one of the 12F units on my A2B. Limited by the battery voltage and motor resistance to about 60A on the battery side. I used aluminum pieces to make a bridge to the bike frame which acts as a huge heat sink. The controller never gets more than barely warm by the time the motor is overheating.

Angle aluminum...I posted having done this in the PV review thread in September 2017. Maybe that's where you saw this?

Heat%20spreader%202_zpsm8emddvg.jpg

Heat%20spreader%203_zpslvemvgwy.jpg


Inside the shell is pretty crowded. I'm looking for some angle aluminum that is 1/8" thick and 4" wide. I'll cut down the aluminum to just fit the insides of the shell. That will help some in spreading around heat. I might even be able to use it directly to mount the CPU heat sinks.

An external heat spreader screwed to the side wall at the mosfets will bring out heat better. I think a chunk of heat sink like this could be used at the mosfets. I'll have to find it, but I have a heat sink more or less just like this to cut a section from.

2018-11-24%2021_32_47-Used%20large%20finned%20aluminum%20heat%20sink--%20measure%2011-1_4X11-13_16X2-1_2%20_%20eBay.png
 
Alan B said:
Note that motor current (sometimes called phase current) also flows through the FETs, so suffers the same lead current limitation (300A motor current exceeds the lead current specs). Since motor current is always greater than or equal to battery current it is the primary limitation on current in the controller.

Another issue is that paralleled FETs don't divide current evenly. So the current through whichever FET is higher determines the overall current limit at something less than twice the single FET current. Here's a paper on FET balance: https://www.infineon.com/dgdl/para.pdf?fileId=5546d462533600a401535744b4583f79


Thanks for the FET balance article. I'll read up on this. It's something I knew about, just never read about in detail.

300 amps...I did say that's ideal. I'm not expecting to get 300 phase amps. Everything you said here, I said too.
 
Tommm said:
ElectricGod said:
I'm not sure what you are presenting here. 7kw on a 24 fet should be no problem at all...even with low grade mosfets. I'm talking about getting 7kw out of a 12 fet controller. Cooling being the main issue and the reason for the CPU heat sinks. How does anything you presented have anything to do with making a 7kw 12 fet controller? I guess you were trying to present that your small controller shell was a problem?

I was guessing the point of this experiment is that you are space constrained? So I wrote down how to make a smaller package of a bigger controller. But apparently this is a novelty project.

Space is not my issue at all. This is a test build...pure and simple. I know it's doable, now to actually do it. I've been pushing controlelrs further than their advertised continuous limits for a couple of years. It's time I up my game a little. Heat is the big issue here...in the mosfet legs and in the mosfet die. Overcome them and 7kw+ in a 12 fet is doable. That's the goal...7kw in a 12 fet.
 
ElectricGod said:
Space is not my issue at all. This is a test build...pure and simple. I know it's doable, now to actually do it. I've been pushing controlelrs further than their advertised continuous limits for a couple of years. It's time I up my game a little. Heat is the big issue here...in the mosfet legs and in the mosfet die. Overcome them and 7kw+ in a 12 fet is doable. That's the goal...7kw in a 12 fet.

If the controller can handle it, I would do it with a 84v nominal battery. Much less amps need to be pushed. Fets and caps are 100v on my infineon clone.
 
Tommm said:
ElectricGod said:
Space is not my issue at all. This is a test build...pure and simple. I know it's doable, now to actually do it. I've been pushing controlelrs further than their advertised continuous limits for a couple of years. It's time I up my game a little. Heat is the big issue here...in the mosfet legs and in the mosfet die. Overcome them and 7kw+ in a 12 fet is doable. That's the goal...7kw in a 12 fet.

If the controller can handle it, I would do it with a 84v nominal battery. Much less amps need to be pushed. Fets and caps are 100v on my infineon clone.

Yes...more voltage is better since it keeps phase amps lower for the same wattage. My first runs will be at 66 volts since I have this scooter already set up with one of these 12 fet controllers. I'll just be swapping in the modded controller in place of the existing one. Later on, once I see reliability, I have an RV-100 outrunner. it's a good bit stronger than the C80100 currently on the scooter. I expect 6-7kw will be doable on this motor while the C80100 can do a little less than 5kw. It will make the Currie scooter insanely strong. It already accelerates like the cars flooring it and tops out at 45mph on level ground.

Currie%2010181117%2024.jpg
 
ElectricGod said:
...

300 amps...I did say that's ideal. I'm not expecting to get 300 phase amps. Everything you said here, I said too.

As long as we are talking TO220 FETs with 75 amp lead current limit, simple math says motor current should be safely limited to less than 150 amps, nothing close to 300 amps. 300 amps would be ideal for a 24 FET (4 parallel FETs times 75 amps). 150 amps would be ideal for a 12 FET. But imbalance will reduce that further. Other FET limits often kick in before the lead current limit is hit as well. Cooling the leads may be helpful, if you want to tackle that.
 
Found this aluminum angle on ebay....4" x 2"x1/8"...a little too large, but I expected to need to cut down whatever I found.

https://www.ebay.com/itm/4-Pieces-3-16-X-1-1-2-ALUMINUM-6061-FLAT-BAR-12-long-T6511-187-Mill-Stock/223225719275?hash=item33f94a29eb:g:sQIAAOSw-YZa0giU:rk:44:pf:0

The vertical space inside the controller shell is fairly tight. I'll have to work out something to get the board a bit lower inside the shell to clear the large caps. I have some .3mm fiber board sheet to put under the controller so it can't short on the shell. There's sufficient width to add the heat spreader behind the existing one. At least the insides of the shell are nice and flat. That will make adding the internal heat spreader easy.

heat%20spreader%20space.jpg
 
I did this to another 12 fet some time back. The through holes in the board are not large enough for 8 awg wire. 10 awg wire is a fairly close fit. This is the solution I came up with to get 8 awg wires in 10 awg holes. Use a crimp barrel that fits 8 awg wire and then crimp in a section of 10 awg into the other end of it. Solder in the 10 awg section of wire into the controller. The netire section of 10 awg is about 1/2" long, but most of that is inside the barrel and soldered in place. That leaves a short 1/4" long section that passes through the board to be soldered in place. I tin the insides of the crimp on barrels to make sure that solder flow is optimal. The wires get tinned before getting crimped and soldered in place as well. I want best possible current flow.

8%20AWG%20wires%202.jpg

8%20AWG%20wires%201.jpg

8%20AWG%20wires%203.jpg


I pulled off the heat spreader and then unflowed all the crappy Chinese solder from the main current paths. While I was at it, I reheated the solder at each mosfet and pushed them as far into the board as possible. Then I added a little bit of flux to all the legs on the mosfets so I could then get solder to flow up onto them. This eliminates any amount of small leg section still being above the board surface. It also creates a bit more path for heat out of the mosfet legs and into the board traces.

Mosfet%20legs%201.jpg

Mosfet%20legs%202.jpg


The controller comes with 2 14 awg shunts and I'm replacing them with at least 4 12 awg shunts...maybe 5.

New%20shunts.jpg


More reinforcing to come soon...
 
I totally removed the other casing and added a COMPUTER GRAPHICS CARD COOLER TO IT. What do you think about that? Then I fastened it on a plastic plate and added wires and a swich for +12v so to activate the cooler. Sadly I don't remember what feels my controller uses if it where 4410 or 4110 so its sad however they are 24fet infusion controller bought from greentime beefed up "resistor" for about 80A at 20 or 22s battery pack at 10c or above. I can't get a picture right now.
 
leffex said:
I totally removed the other casing and added a COMPUTER GRAPHICS CARD COOLER TO IT. What do you think about that? Then I fastened it on a plastic plate and added wires and a swich for +12v so to activate the cooler. Sadly I don't remember what feels my controller uses if it where 4410 or 4110 so its sad however they are 24fet infusion controller bought from greentime beefed up "resistor" for about 80A at 20 or 22s battery pack at 10c or above. I can't get a picture right now.

A GPU cooler would do the trick. Mount it direct to mosfet heat spreader...yup...would work pretty well.

IRF4110 is a mosfet...is that what you meant? It's made by infineon. They are 100 volt parts so 22S is doable.

I have more pics to post. I messed with more "beefing up" on this controller. Now it has 4 shunts in it...each 12 awg...should handle a good bit more than 80 amps. It also now has 8 awg wires for power and phase. All the power traces have much more copper on them.
 
2" x 4" x 1/8" aluminum angle will arrive this week.
Battery and phase wires are now all 8 awg.

The power traces are beefed up with added copper. All the traces to mosfet legs are reinforced. The phase busses have 2 lengths of copper wire on them. The battery busses just one piece if wire. The phase busses had much less copper on them than the battery busses so they needed double the copper. Also phase amps are in those busses...should be beefier and more stout anyway. Not really needed, but I extended the copper down to the filter caps.

Reinforced%20busses%20and%20shunts.jpg


4 12 awg shunts in place of the 2 14 awg shunts. These shunts were a tight fit in the through holes. They were not intended for 12 awg shunts! They look like they touch, but there's a gap between them.

Bigger%20shunts_1.jpg
 
ElectricGod said:
4 12 awg shunts in place of the 2 14 awg shunts. These shunts were a tight fit in the through holes. They were not intended for 12 awg shunts! They look like they touch, but there's a gap between them.
It won't matter if they touch.

Looking good. It would be nice to make some laser cut copper sheet pieces to beef the traces. But wire is much cheaper...
 
fechter said:
ElectricGod said:
4 12 awg shunts in place of the 2 14 awg shunts. These shunts were a tight fit in the through holes. They were not intended for 12 awg shunts! They look like they touch, but there's a gap between them.
It won't matter if they touch.

Looking good. It would be nice to make some laser cut copper sheet pieces to beef the traces. But wire is much cheaper...

I have some .5mm copper sheet...somewhere. LOL...that was my first thought for beefing up those traces and cut it with some tin snips. Alas...couldn't find the stuff so copper wire it is!
 
Back
Top