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

[ElectricGod]

Last night...I was looking at the shunts on the bottom. Solder flow on them was sparse. NiChrome wire does NOT take solder particularly well. Pulled them off, re-scraped the ends, retinned and resoldered...better, but still NOT great. I have some smaller .005 ohm shunts, the ends are something else...chrome welded to the NiChrome? The ends on them solder just fine. There must be a trick to tinning NiChrome reliably. The factory shunts look like they are solid NiChrome and the ends are tinned pretty well. I have 10 feet each of 12 and 14 awg that is a giant pain to solder/tin. Solder kind of flows and clumps on the Nichrome, but it doesn't just flow over the entire bare end. Put solder on some copper and they flow together really well. This stuff...the affinity for solder is pretty low.

Things I've tried....
Scraping off the ends with a razor
Sanding off the ends
Grinding off the ends with a dremel wheel
2 different kinds of flux
3 kinds of solder:
Cheap Chinese solder sucks for everything...including this.
Real lead/tin (60/40) was slightly better.
Silver solder (3% silver) was about the same as 60/40.
More heat/ less heat...no change

I have some 100% pure sulfuric acid...no water or other dilutes. Maybe that will work? Last time I used it on steel, it turned the surface black. I doubt solder will flow onto that!

 

[ElectricGod]

Sorry for the grainy images...massively blown up to get in close.

I clean off the leg, then dip it in flux, then into liquid solder, then back in the flux and back into liquid solder to get flow like this. You can see the leg is shiny and polished clean. This stuff just does NOT take solder very well. Uncleaned NiChrome is a dull grey color. Scrape it with a razor and it's golden/straw colored. Polish it and it's shiny silver.

This is pretty typical...very poor solder flow. This leg has just one little blob of solder on it.

The leg shows some glossy high spots. That's the slightly better solder flow I got on that one. Still pathetic...the rest of the cleaned off NiChrome is completely bare.

 

[fechter]

Most of those shunt wires are Manganin, which generally solders pretty easily. NiCr is like stainless steel. You can solder it by using special flux and solder that does NOT contain rosin flux. Rosin flux poisons the joint.
The other (better) way is to get some real silver solder and stainless steel flux and tin the end with silver solder. Once that's done, it will soft solder nicely.

Why not use SMD shunt resistors like all the new controllers?

 

[ElectricGod]

Most of those shunt wires are Manganin, which generally solders pretty easily. NiCr is like stainless steel. You can solder it by using special flux and solder that does NOT contain rosin flux. Rosin flux poisons the joint.
The other (better) way is to get some real silver solder and stainless steel flux and tin the end with silver solder. Once that's done, it will soft solder nicely.

Why not use SMD shunt resistors like all the new controllers?

Rosin flux would at least get some amount of solder flow. Without it, 60/40 would not flow at all. I'd dip the leg in the pool of solder and I might as well have been trying to use 60/40 on aluminum!

This stuff was sold as NiChrome...probably not manganin.

I have no real silver to try out...just 3% and that did NOT work. Jewelry solder is probably the stuff I need. Is there a silver solder you recommend?

The board bottom could possibly take SMD shunts and I have quite a lot of them too in 10 watt. That's assuming they will span the gap. If I remember right, they are 3/8" long. I didn't think about using them since I bought the 12 awg NiChrome just for this controller build. I probably could use those 10w SMD shunts I have. Clean off the coating on the ground traces, reinforce with more copper and those shunts might just span this gap.

 

[ElectricGod]

This thing is getting beastly!

Feels like overkill...8 awg wires on a 12 fet controller. I don't use 8 awg on an 18 fet controller, but hey, I'm going ALL out here!

This port out of the controller is PACKED! I had to move all the smaller wires to the very bottom row of this silicon gasket. I then cut out all the rest of the silicon to make room for these monster wires. It's packed in there pretty good. I was concerned that the metal end cap might cut through the silicon gasket so I added some heat shrink...just in case.

5.5mm bullets...that ought to do the job. They hold up nicely at 150 amps...don't care what the specs say...I've used them there before and they are just fine. Make sure they are getting a snug fit when connected together. I spring the 6 prongs a little. It makes a tight fit, but they will do that amperage all day long.

 

[fechter]

I think with NiCr they normally weld a piece of copper or nickel to the end then solder to that. Manganin is easy to solder.

Go for the SMD resistors. On larger controllers I've seen them use a row of up to 10 parallel. With 10W ones, you won't need that many.

8ga does look like overkill. It won't be the weak link for sure.

 

[ElectricGod]

I think with NiCr they normally weld a piece of copper or nickel to the end then solder to that. Manganin is easy to solder.

Go for the SMD resistors. On larger controllers I've seen them use a row of up to 10 parallel. With 10W ones, you won't need that many.

8ga does look like overkill. It won't be the weak link for sure.

At this point, I'm wondering what the "weak link" will be?

I haven't pulled the mosfets out to see if they all match up or not. I'm taking that on faith that they are OK...ish...as is.

I've bought a fair number of mosfets from legit sources and when testing them, found significant variance between identical parts from the same order. You'd be surprised how stating that on ES upsets folks. LOL! I've had this happen with Infineon, IXYS and AOT. I've seen Vgs vary a lot. Specs will say it's 3.6v and none in the batch will be less than 4.2v and few as high as 4.7v. Rds is another...turn on the mosfets with 10v on the gate and one mosfet will be 4.5 mOhms and another 3.2...and spec is 3.6 mOhms. Want to work people up into fits...tell them stuff like this! I had a guy tell me that all mosfets are perfect. Umm...OK, but...they aren't. And how can he claim that when no manufacturer EVER tests 100% of the components they make.

Anyway...ignoring the mosfets...for now...fingers crossed.

I have the aluminum angle now. Hopefully the internal heat spreader will get made over the weekend. It will span the entire top and wall inside the shell. Worst case it will make for loads more metal to get heat soaked and to radiate well before things over heat. I'll be increasing the heat spreader by about 400%.

The current heat spreader is held in by 3 M3 screws, but it has threaded holes for 5. The extra weight here could use all the support and clamping pressure it can get. I'll make this use all 5 threaded holes. I want to mount an external heat sink on the side of the controller on top of the internal heat spreader and the angle piece.

The 3 holes in the side of the shell raise the board up a good bit. I've made this worse by cramming the mosfets down into their through holes tightly. I'll need to stretch those 3 screw holes downward to get the board lowered inside the shell. I need all the vertical space I can get in here. The bottom of the board will sit on top of an insulator and then be pressed down on the bottom of the shell. The angle aluminum will take 1/8" height and the those caps are really close already. The 8 awg wires are a close fit too. Since the unspported edge of the board will no longer be in that channel, I'm thinking some high density foam on top of the caps will act to press the board down so it can't move. This area inside the shell will all be filled with the aluminum angle.

 

[ElectricGod]

A bit more work done on the controller...

Last night was the first time the controller has been spun up since I started upgrading it. Good thing it still all works! I've spent a ton of time on this thing. It would truly suck if I somehow broke it!

First thing I did was remove the 12 awg shunts since solder was not flowing well to the NiChrome. I replaced them with SMT 10 watt shunts. I may see if another shunt can be added. But I'm currently at just slightly higher than 1 mOhm. There's traces under the SMT shunts. I figured a little heat shrink on them wouldn't hurt.

The side wall of the shell used to have fins on here. I wanted to make this space as large as possible for the external heat spreader. I broke them off and then scraped the surface flat.

Heat spreader parts cut to size.

Test fitting...pretty good!

Almost 3/8" thickness at the wall where the mosfet heat spreader mounts. 3/16" across the top of the shell.

I need to make a piece of aluminum for this space so that the CPU heat sinks have a place to mount.

Internal and external heat spreaders. 6 screws hold the internal heat spreader to the top of the shell. All mounted with heat sink grease.

I was using an end mill to make flat areas in the fins and the shell got away from me. The end mill messed up that corner area on the shell top...grrrrr. Everything else worked out OK.

Once I had the internal heat spreader in place, then I could see where the board was going to sit. There used to be a thin channel here to hold the edge of the board. I knocked it out so I could get the board to sit lower in the shell. Later I filed and scraped the area flat.

I was concerned that there might be a possibility of something touching and creating a short on the internal heat spreader. The phase and battery wires touch and the large caps come pretty close. A little Kapton will take care of that.

Before I could get the controller board to fit, I had to take off about 1/16" from the floating edge. A good bit of careful filing did the trick. The edge of the board beyond the ground plane used to be 1/8" wide.

It's a tight fit, but it's all in there! Under the bottom of the board is a piece of fiberglass board to act as an insulator. The large caps have about 1/16" space above them. The large cap was so close to the inside of the shell, I put a bit of foam tape on it so there wouldn't be any chance of abrading and wearing through it's covering. A little foam tape under the board lifts it up a little. The free edge of the board is jammed tight against the wall. It can't move.

 

[ElectricGod]

I'm not sure if I'll do this or not. There will be a piece to go in that gap in the top as a minimum for the CPU heat sinks to mount to. There will also be a heat sink mounted to the side wall over the mosfets. I'll have the heat sink in a few days.

 

[ElectricGod]

All closed up, configured and tested...complete with a telemetry module. There's still a couple more things to do, but I'm waiting on that heat sink before I go any further. The silicon grommet gave way. I think I'll add a second layer of heat shrink to the wires.

 

[fechter]

Looking good. It will take a while for that to overheat.

 

[ElectricGod]

Looking good. It will take a while for that to overheat.

That's the plan! A VERY LOOOOONG time hopefully.

I found a GPU heat sink designed for fanless use. Cost me $7 and it's just a couple mm longer than the controller shell. Trimming it down to fit the controller should be easy. I might not even bother with the dual CPU heat sinks if the GPU heat sink is big enough. The plan was to mount the GPU heat sink to the side of the controller and the CPU heat sinks to the top of the controller. I want compact and high wattage. I have that right now, but then it's not seeing any load yet either.

 

[fechter]

Have you considered using pyrolytic graphite?

https://www.mouser.com/new/panasonic/panasonicthermalgraphite/

 

[ElectricGod]

Have you considered using pyrolytic graphite?

https://www.mouser.com/new/panasonic/panasonicthermalgraphite/

Too bad you can't use this stuff direct on the mosfets where it would matter the most. However, it's conductive...so no bueno. From the heat spreader to the shell matters, but less so.

 

[ElectricGod]

HUGE new work done!!!
OK...no not really...lol.

The silicon grommet issue was bothering me so I looked around for a simple solution...such as a second layer of heat shrink on the wires. It's all sooo crammed in there that slidng another layer of heat shrink onto the wires was not possible. I also worried that the bare metal edge would still wear through after a while and that would be catastrophic! Under the big wires isn't a problem, All the small wires are down there and they a buried in lots of silicon grommet. It's just on top of the big wires that there is no protection and lots of cutting force on a narrow strip of steel. I finally settled on a strip of poster card and wrapped it in several layers of kapton tape. Even then, cramming it in between the metal edge and the wires was not particularly easy. I clipped off the dangling portion of the grommet and shoved in the cardboard "shield" in it's place. A zip tie keeps it all together. It's about 3/4" wide and half of it is inside the end plate. Adding the zip tie actually helped with the outward pressure on the bare metal edge.

My one mistake. FYI...don't ever try to use an end mill while holding the object with your hand. There's an excellent chance that it won't go well. This could have been far worse!!! I'm an idiot...lol.

Probably overkill...all the shell screws are M3x25mm. I figured since the end plates carry the weight of the controller and this controller is about to gain some weight once those heat sinks get added, that I should make sure the screws are secure and have lots of threads. I'll use a few more of these to secure the side plate and heat sink to the side of the controller.

5.5mm bullet and IP68 connectors. I do this to all my controllers. Those two IP68 connectors are on any controller I've ever messed with in the last couple of years and wired identically. Power and phase connectors are exactly like this too. It means any controller can plug into any EV and work interchangeably. The 6 pin IP68 connector has the motor halls and motor temp sensor in it. The 8 pin has enable, throttle and 3 speeds in it. I never use the brake or cruise connections to the controller. They are not needed IMHO. However, since there are 3 ground connections in here, I could eliminate 2 of them for some other purpose. The individual solder points are quite small inside the connector and I have no need for more functionality here, so I've never bothered. These aren't "big current" connectors, but they are rated for 5 amps...so that's pretty decent.

This is what I do inside the 8 pin connector...just in case. Depending on the controller, enable can be BATT+ connected to the internal DC-DC converter. The red and orange wire are those connections. Right next to them, the black wire is BATT-. The rest are hidden, but any ground or voltage wire gets heat shrink on it. A friend in San Francisco used my IP68 wiring for himself and neglected to seal up around the wires going onto the connector. He rode in driving rain and got water in the connector through the cable end...which of course was not water tight. This caused some kind of internal partial short and made his throttle act weird part way home. The EV still worked except for the glitchy throttle. I soon suspected what the the problem was. He found inside, the connector was full of water and of course he never heat shrunk around any of the wires. I don't know, but high voltage wires right next to low voltage wires and no isolation between them, doesn't sound like a great idea to me. A little heat shrink provides a few 100 volts of isolation.

Those IP68 connectors are not overly expensive and they are keyed so there's no connecting things wrong once your wiring is correct. I use them on any exterior connection on my EV's. All my lights, if there is an exposed connector somewhere use an IP68 connector.

Here's the 6 pin and 8 pin versions I use. Pretty cheap to not have any water intrusion in my connections!

6pin...less than $4 each: https://www.ebay.com/itm/IP68-3-4-5-6-Pins-Waterproof-Electrical-Cable-Wire-Connector-Plug-Socket-Outdoor/332568173662?hash=item4d6e9b9c5e:m:m6GTPWYMlkNaNy3drcTtLgQ:rk:3f:0

8 pin...less than $8 each: https://www.ebay.com/itm/M14-IP68-Assembled-Waterproof-8-Pin-Electrical-Cable-Connector-Plug-Socket/391529045046?hash=item5b28f31c36:g:MJMAAOSw65FXqo9w:rk:11f:0

How I wire them for the controller.

Not directly related to this controller build, but IP68 connector related. This is my 4kw Currie scooter. That IP68 connector goes to the tail lights. I have a Nomad bike trailer. It will soon have LED lights on it for brakes, directional and running. This connector will get a Y attached to it and then it can run the scooter tail lights and the trailer lights. since I want to leave the car home as often as possible, I'll also make the exact wiring in this connector standard to all my other EV's so I can connect up the trailer to any of them.

 

[madin88]

At this point, I'm wondering what the "weak link" will be?

Yeah it should definitely not be temperature related with those big heatsinks and wires

Maybe i missed that, but what battery voltage you going to use?
Because if you want to push high phase currents close to max voltage rating you should think about using better caps (like Rubycon ZLJ) and adding additional ceramic caps close to the FET's in order to reduce dangerous voltage spikes.

I haven't pulled the mosfets out to see if they all match up or not. I'm taking that on faith that they are OK...ish...as is.

I've bought a fair number of mosfets from legit sources and when testing them, found significant variance between identical parts from the same order. You'd be surprised how stating that on ES upsets folks. LOL! I've had this happen with Infineon, IXYS and AOT. I've seen Vgs vary a lot. Specs will say it's 3.6v and none in the batch will be less than 4.2v and few as high as 4.7v. Rds is another...turn on the mosfets with 10v on the gate and one mosfet will be 4.5 mOhms and another 3.2...and spec is 3.6 mOhms. Want to work people up into fits...tell them stuff like this! I had a guy tell me that all mosfets are perfect. Umm...OK, but...they aren't. And how can he claim that when no manufacturer EVER tests 100% of the components they make.

hm.. If the measurements are correct (assumend the correct measuring equipment was used !!), those FET's are probably non-genuine which would be no surprise as i had two similar 12F controllers that came with such. I had to replace them all because some of them suddenly got a short without any reason. I hope that you will not have similar troubles..

 

[ElectricGod]

At this point, I'm wondering what the "weak link" will be?

Yeah it should definitely not be temperature related with those big heatsinks and wires

Maybe i missed that, but what battery voltage you going to use?
Because if you want to push high phase currents close to max voltage rating you should think about using better caps (like Rubycon ZLJ) and adding additional ceramic caps close to the FET's in order to reduce dangerous voltage spikes.

I haven't pulled the mosfets out to see if they all match up or not. I'm taking that on faith that they are OK...ish...as is.

I've bought a fair number of mosfets from legit sources and when testing them, found significant variance between identical parts from the same order. You'd be surprised how stating that on ES upsets folks. LOL! I've had this happen with Infineon, IXYS and AOT. I've seen Vgs vary a lot. Specs will say it's 3.6v and none in the batch will be less than 4.2v and few as high as 4.7v. Rds is another...turn on the mosfets with 10v on the gate and one mosfet will be 4.5 mOhms and another 3.2...and spec is 3.6 mOhms. Want to work people up into fits...tell them stuff like this! I had a guy tell me that all mosfets are perfect. Umm...OK, but...they aren't. And how can he claim that when no manufacturer EVER tests 100% of the components they make.

hm.. If the measurements are correct (assumend the correct measuring equipment was used !!), those FET's are probably non-genuine which would be no surprise as i had two similar 12F controllers that came with such. I had to replace them all because some of them suddenly got a short without any reason. I hope that you will not have similar troubles..

I've bench tested at 82 volts so far, but the Currie scooter runs at 66 volts...where I'll put it through real world tests. Once it passes muster at 66 volts, I have a 50cc sized moped that is currently running an 18 fet controller and 6kw at 82 volts. That will be it's next "test bed".

I was seriously thinking about adding an external capacitor bank. The 4 internal buss caps are fine for the 4kw this controller normally ever sees, but for loads more phase amps...I definitely need more power filtering. The insides of the shell are crammed full as is. There is no way to get more stuff in there! I'll have to put caps outside the shell. I could fit some more ceramic caps inside, but more large electrolytic caps...no way. I'll see what I have on hand, but I bet I can scrounge together at least 14,000uF of 100 volt caps.

The mosfets in this controller are sourced from Arrow or Digikey in the USA and then soldered in place in the USA. The chances of them being Chinesium is highly unlikely. This is a Chinese made controller, but they come to the USA with zero mosfets in them.

Regarding legit components varying a lot...those too were purchased from legit sources in the USA by me. I have seen lots of mosfet component variation. I bet with $10,000 test equipment or $30 test equipment that would be borne out either way.

 

[ElectricGod]

Popped a few mosfets last night doing a test...

I ordered 30 CSD19536KCS. I didn't have any on hand and I was down to 5 AOT290's and didn't have any of the mosfets already in the controller. Of course the actual mosfet installed was out of stock everywhere.

 

[madin88]

Popped a few mosfets last night doing a test...

I ordered 30 CSD19536KCS. I didn't have any on hand and I was down to 5 AOT290's.

Oh thats sad to hear. Could you say more about the "test" and did the FET's really pop?

Because mine had just a short and no burn marks etc. Those damn cheap fake part's!! The most annoying thing was that Vadim boasted that he bought the FET's for 0,30$ each (AOT290) while i had to pay 10 (ten!!) times as much for genuine ones from digikey. What a shame..

 

[ElectricGod]

Popped a few mosfets last night doing a test...

I ordered 30 CSD19536KCS. I didn't have any on hand and I was down to 5 AOT290's.

Oh thats sad to hear. Could you say more about the "test" and did the FET's really pop?

Because mine had just a short and no burn marks etc. Those damn cheap fake part's!! The most annoying thing was that Vadim boasted that he bought the FET's for 0,30$ each (AOT290) while i had to pay 10 (ten!!) times as much for genuine ones from digikey. What a shame..

yes...lost 2 of them. I wanted to try a larger motor so I grabbed my AP 12090. I'm still doing unloaded testing while I'm adding cooling to the controller.

The controller looks kind of silly now. Last night I made up the mount for the CPU coolers and got the external mosfet wall and heat spreader set up for 5 screws so I can mount the side heat sink.

Anyway, I plugged it in and turned the throttle. The motor moved just a tiny bit and that was it. Turning the bell was stiff like a phase was shorted. As soon as I unplugged the phases the motor turned freely again. I also noticed some kind of scraping sound in the motor. I'm not sure what that is about. Did I break a magnet or worse a phase is shorting on the stator or armature? Something about this motor fried those mosfets. I don't blame the controller or the mosfets at all. I've run 2 other motors (HLD inrunner and AsStroFlight 3220) without issues. Now I need to take apart my 12090 and find out what is going on to cause this to happen. There was no external short. All phase ends are heat shrunk to prevent this.

The up-side here is now I can test all 30 replacement mosfets and find the best 12 for this controller instead of assuming I have 12 that are similar.

I don't know about your controller. I can say this much however. They come into the USA without any mosfets and Vadym installs them himself. Early on he was getting them direct from the factory with IRFP4110's in them. I quickly convinced him to use better parts. I think at that point...was the last time he got pre-populated controllers done in China. If you had AOT290's in yours, there was a stretch of time where AOT had some significant quality issues. I mentioned earlier seeing mosfet variation. That was mostly in AOT290's from about a year to 18 months ago. It may be you got caught by their quality issues. I doubt Vadym was at fault here.

My assumption is he buys his mosfets from Digikey or Arrow or Mouser like the rest of us do that want legit parts. However, if he can get legit parts from a Chinese seller for lots less money, well more power to him! But I really don't know that. China isn't all "fakes and knock-offs". I buy legit 18650's from a Chinese seller. To get the exact same cells from any place in the USA costs a good bit more. I've tested them...they provide the capacities and specs the manufacturers say they should. That's about at "legit" as I can test for.

A beefed up a PV 12 fet for a friend in the UK knowing he would go crazy, but didn't think he would do more than 5kw. He has had it for a couple of months now and rides his scooter daily. It has the mosfets in it provided by Vadym...Digichip MDP10N027. He's 100% the reason why I'm doing this silly build. I want to duplicate and destroy his results. He doesn't have an amp meter that reads phase amps well or else he doesn't know how to use it so we are guessing on actual phase amps. However, he has his 12 fet set to 100 battery amps and 150 phase amps and is running at 82 volts. Yeah...just crazy insane settings! He sees 40C peaks in the controller and usually sees 32-35C. He has a blower on his outrunner to keep it at 60C. I'm sure his battery amps are correct. His clamp on meter at least does DC amps well. It's all the harmonics and well over 60hz that keeps it from doing AC phase amps reliably.

I have a decent AC amp meter that ought to read the motor phases. I should see what my phase amps are. I'm running a 12 fet at 66 volts, 60 battery amps and 135 phase amps on hand picked AOT290's. It came with IRFP4110's in it. It runs at 5kw all day long. I know my battery amps are right...just never checked the phase amps.

Anyway, this controller failure was (99% sure) a motor issue, not the controller. My pal in the UK is doing insane things with the mosfets provided by Vadym. I think your failure was a fluke. Sometimes stuff dies for no good reason.

Again...can't blame Vadym for this...100% me being a moron. I have a PV 18 fet. It had legit IRFP4115's in it. I was doing a test and like an idiot I didn't have the phase connections protected. 2 touched while running the motor and I killed 16 of the 18 mosfets. Now it has 18 hand picked AOT290's in it. I run it at about 6kw, but only becasue that's about all the more the HLD inrunner can do.

 

[ElectricGod]

Work from last night...

CPU cooler mounting plate. This will screw down to the top of the controller via 6 M3 screws. I need to transfer those 6 screw positions to the top of the controller shell yet.

Mosfet wall and heat spreader ready for 5 screws instead of 3.

This heat sink will get slimmed down some and then mount to the mosfet wall heat spreader via those 5 screws.

 

[ElectricGod]

A bit more work done on the shell and cooling...

This is the CPU cooler adapter plate. I got the top of the controller shell drilled and tapped for M3 to mate to the adapter plate. The close up shot is showing 2 dots in the aluminum...registration marks so I know how the plate orients to the top of the controller shell. The mosfet wall heat spreader has them too. This plate was the last thing to need drilling and tapping to mount to the controller shell. Once done I cleaned up all the holes so there were no barbs or high spots anywhere. A good flat fit between everything is paramount! 12 screws holding the internal heat spreader and 6 holding the adapter ought to make for a really good tight fit too.

Now that it has all been drilled, tapped and cleaned up I could put thermal paste on it all. The internal heat spreader took quite a lot.

The adapter plate got 3 layers of thermal tape between it and the bottoms of the CPU coolers. The aren't particularly flat on the bottom. I guess I could have used thermal gap filler, but that stuff doesn't take being compressed tightly very well. The tape seemed like a better option since it still fills small gaps and can be squeezed hard. Those 8 screws get tightened down incrementally so it all goes together flat. Once it was all flat, I would loosen a single screw at a time and put locktite on its threads so they can't come loose again.

These 2 button head screws are dead center in the adapter plate and right under the CPU heat sink fins. It took a while to tighten them...a 1/3 turn at a time with an allen wrench.

The CPU coolers are held to the adapter plate by 8 M3 screws and 6 more hold the plate to the top of the controller. The 2 CPU coolers have a single fan between them with both fan brackets on either side of it. Zip ties squeeze that all together top and bottom so there is no movement between anything. The 6 screws that hold the CPU cooler assembly in place have threads in the top of the controller and into the internal heat spreader. It's just slightly less than 1/4" of thread per screw. I think it will hold long term and provide good clamping force to the controller shell. I can't put locktite on these screws, they are coated in thermal grease.

This heat sink will mount here. I intend to make is narrower and it's about 6mm longer than the controller shell. I'll also knock off the corners some since it sticks out the side of the controller pretty far and the corners are sharp. What more can I do to cool this controller?

 

[ElectricGod]

Changed my mind...

This is just ridiculous with that heat sink hanging off the side of the controller. The 2 CPU coolers will be plenty despite not being mounted directly to the mosfet wall. I don't know yet, but the dual CPU coolers may be overkill and I can use this heat sink in place of them.

I'll take apart the CPU cooler assembly and put this under it instead. That will triple the heat path to the CPU coolers and provide lots of metal for short term heat soaking. I'll probably trim it down to cover the entire side and most of the top. It takes up far less space, no heat sink fins to snag on stuff and it looks better. The mosfet wall, not including the original heat spreader is now 12mm thick! I think the heat spreader screwed to the mosfets is 10mm thick. LOADS of metal right here!

 

[district9prawn]

Surely the one big heatsink or one cpu cooler should be plenty.

 

[ElectricGod]

Surely the one big heatsink or one cpu cooler should be plenty.

Heat will be the biggest enemy to longevity in this 12 fet.

Keeping the mosfets cool will matter a lot.

I am going for 7kw+++ in a 12 fet running TO-220 mosfets. I'm looking to get pretty darn close to the leg limits of the TO-220 package.

That means LOTS of cooling.