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

DaveDog said:
Here is a great article about thermal conductivity.

https://www.electronics-cooling.com/2016/08/design-considerations-when-using-heat-pipes/#

As the article notes, heat pipes are a couple magnitudes better than solids for heat conduction.

It would seem that getting a couple heat pipes and integrating them into the 8mm copper plate that you will mount to
the fets may be ideal. As has been observed, the shortest possible heat path is optimum.

I won't be using heat pipes. But, the copper heat spreader will mount to the side wall and the top of the internal add-on heat spreader. IE: screwed down to both and have heat sink paste to both. That's about the best I can do. Heat pipes into the copper are not doable in any kind of a way I can think of.
 
drc said:
Dear god! You are very serious about this ! As normal I'm lost a little with the detail of your build but I fully expect you to have an amazing controller! I'm keen to know how how well this performs on your build and what it can successfully drive (motor wise).

I think you should set a goal .......

CAN AN ELETRIC STAND UP SCOOTER HIT 100mph. At this rate I think so! You just need someone willing to try it out!

You should set up shop and build custom eletric scooters! The EV Elon Musk

FML! 100mph on a stand-up...I've done 60 on my blue stand-up and that was pretty intense. I think anything more than that you really need to plant your butt on a seat. Wind resistance gets to be a big issue too.

The more I do, the more I think about what else to do.

Goals...well...I have motors of various wattages...we'll see.

A side project is building an inline phase current meter so I can see all 3 phases at the same time and do logging. I think I have an inexpensive option for that and some of the parts are ordered already. I'll need to be able to reliably measure 200 phase amps at least.
 
This is the original mosfets after I tried to beef up the legs...pretty inconsistent results.

Mosfet%20legs%201.jpg


The replacement TI mosfets got the legs pre-tinned before installing them in the controller. Once they were soldered in I was able to flow solder onto the legs much more consistently. There was no real reason to do the gate legs, but oh well.

TI%20mosfets%201.jpg

TI%20mosfets%202.jpg

TI%20mosfets%203.jpg
 
ElectricGod said:
What do you recommend for the ceramics? I think you mean these specifically right? They are 104's or .1uF right now. There's some SMT caps in each gate control circuit too.
The ones on the power stage i would swap out all for rubycon ZLJ. Those are one of the best low ESR caps you can find today.
Between the power traces on the backside i would solder as many ceramic caps as possible.

SMD X7R, size 2220 or 3025, 100V or 250V should be a good choice.

https://www.digikey.at/product-deta...2Vkx2MkNncGtqU1FNczRxckIyaW9cL1VwRktNeVI1In0=

https://www.digikey.at/product-deta...2Vkx2MkNncGtqU1FNczRxckIyaW9cL1VwRktNeVI1In0=

these would be optimal, but $$$:
https://www.digikey.at/product-deta...2Vkx2MkNncGtqU1FNczRxckIyaW9cL1VwRktNeVI1In0=

thats what i did on same controller when i had to replace the FET's:
https://endless-sphere.com/forums/viewtopic.php?f=4&t=90536&start=275#p1400763

I don't know if it would make sense to change the caps on the gate drivers.
You could check with an osci if the FET's are switch properly on and off (quick and without too much oscillation). Then swap the caps for larger or better ones and compare the switching properties. It might be the case that different FET's have differnt requirements on the drivers if they have differnt parameters like gate charge Qg for instance.

Very clean work and good soldering skills :thumb:
 
madin88 said:
ElectricGod said:
What do you recommend for the ceramics? I think you mean these specifically right? They are 104's or .1uF right now. There's some SMT caps in each gate control circuit too.
The ones on the power stage i would swap out all for rubycon ZLJ. Those are one of the best low ESR caps you can find today.
Between the power traces on the backside i would solder as many ceramic caps as possible.

SMD X7R, size 2220 or 3025, 100V or 250V should be a good choice.

https://www.digikey.at/product-deta...2Vkx2MkNncGtqU1FNczRxckIyaW9cL1VwRktNeVI1In0=

https://www.digikey.at/product-deta...2Vkx2MkNncGtqU1FNczRxckIyaW9cL1VwRktNeVI1In0=

these would be optimal, but $$$:
https://www.digikey.at/product-deta...2Vkx2MkNncGtqU1FNczRxckIyaW9cL1VwRktNeVI1In0=

thats what i did on same controller when i had to replace the FET's:
https://endless-sphere.com/forums/viewtopic.php?f=4&t=90536&start=275#p1400763

I don't know if it would make sense to change the caps on the gate drivers.
You could check with an osci if the FET's are switch properly on and off (quick and without too much oscillation). Then swap the caps for larger or better ones and compare the switching properties. It might be the case that different FET's have differnt requirements on the drivers if they have differnt parameters like gate charge Qg for instance.

Very clean work and good soldering skills :thumb:

Thank you for the recommendations!

Looking at what's available in X7R, looks like 10uF gets the price a bit lower and Mouser is cheaper. I can solder up a few of these in parallel and add wires to replace the 1uF caps already in the board. I'll also add 2 more like you did. That will get me 20uF in place of the 1 uF's already there and add another 20uF like you did with 6.8uF caps.

https://www.mouser.com/ProductDetail/Murata-Electronics/KCM55QR72A106KH01L?qs=sGAEpiMZZMs0AnBnWHyRQN7%2fAA2D2lPPVu7Pf%2fpi5A8J0YCSL7JvGg%3d%3d

While I'm at it, I looked at my selection of 100v electrolytic caps, most are not very good quality. I'm looking at 470uF caps. The existing caps are 16mm x 26mm. I can use longer caps in place of the 2 laid down ones, but the vertical ones, no way. They will go mostly external.

This looks like a reasonable option. 32 mOhms and 1850mA ripple current.
https://www.mouser.com/ProductDetail/661-EKZE101L471MLN3S

If you see a flaw in my selections, please point it out.

I bought some 8 awg PTFE wire. Since the strands are so much thicker that what is used in silicon wire, it's really stiff. It was a thought to use it in this controller, but NVM.
 
Alan B said:
Every surface in the heat path adds thermal resistance to the heat path. As does thermal paste.

There were some problems with the ceramic caps, precipitated some major failures when methods did that a few years ago. Might be good to understand what went wrong there. Was discussed here on ES.

Yup...can't help that so make sure to use decent thermal paste and make lots of surface area for thermal conduction.

Thanks for the hint about Methods. I looked around at his posts, but didn't find what you are referring to. Perhaps you could post the URL or give me more info...like thread title please?
 
I pulled all the big wires out to redo them.

This is a 4 step process...takes a looong time!

1. I tin the wire end and then once it is completely saturated with solder, I then "flick out" all the solder I can get out of the tinned end.
2. I compress all the strands together with a crimper. This closes up all the small gaps left behind between the wire strands. The tinned wire will retain this compressed state. The red wire end below is pressed together, but not reduced. Before doing this the untinned end is probably 20% larger than the tinned and compressed end.
3, Using a file, I slowly reduce the wire end to just barely fit in the through hole. A tight fit is what I'm looking for so that I preserve as much of the 8 awg wire as possible. Filing the end down could be fast if I didn't care about getting a very tight fit in the hole. The wire end is larger than a 10 awg wire.
4. Once I have the wire end to the right diameter, I put heat shrink around the silicon insulation and over lapping the end a little so it shrinks around any loose strands that might come apart when the wire is soldered in place.

Reduced%208awg%20wire%20end.jpg


I think the above process while taking longer to do will also hold up better than this first method. I've completely eliminated the crimp-on berrol and no longer have a short section of 10 awg wire anymore. This allows for more wire flexibility closer to the board and keeps the wires from pressing on the roof of the controller so much.

8%20AWG%20wires%203.jpg


While writing the above stuff, I thought of compressing down the end of a crimp berrol. I have several sizes of crimpers so I kept compressing the end together tighter and tighter just now. I was able to get the end to exactly the size of the through holes in the board. Now I don't need to file on the wires anymore, just tin them and solder them into the berrol. Push the compressed end of the berrol into the through hole and solder it in place. This is not as close to the board as reducing the wire end, but it's a lot faster and I'm still reducing about 1/4 of vertical length. Compressing that end down took a minute at most. Reducing down the wire end takes 20 minutes to get a perfect fit. Time to redo the 3 phase wires like this. BATT+ and BATT- have not been reduced yet.

Compressed%20berrol%20end.jpg


Getting the berrol end crunched down like the left part is the hardest to do. It takes going back and forth around the diameter to get it to crush together.

first%20crimp.jpg


I move to a smaller crimper to get to this slightly more compressed left part. There's still air gaps and voids in there. To get to that final size on one of my crimpers on the jaws it says ".1". That's the final compressing that gets the end to final size.

second%20crimp.jpg


This took just a few minutes to compress down 5 berrols. While I was at it, I compressed half the length of the berrol so they would be a bit shorter when installed in the board. Extra length will get snipped off after they are soldered in. I wish I would have thought of this sooner!

5%20compressed%20ends.jpg
 
I found some decent 200v electrolytics in a couple of dead power supplies. Down the middle is all the negative poles with the positive legs around the outside. The 5.5mm bullets lets me insert the capacitor bank inline with any controller. The wires are 10 awg. It's wrapped in several layers of Kapton and the opening where the wires come out is sealed in hot glue. It should stay water tight.

Capacitor%20bank%201.jpg

Capacitor%20bank%202.jpg

Capacitor%20bank%203.jpg

Capacitor%20bank%204.jpg

Capacitor%20bank%205.jpg
 
ElectricGod said:
New wire ends in place with the crimped down berrols.
why you don't just cut off single strands until the wire fits into the bore? :D

ElectricGod said:
The 5.5mm bullets lets me insert the capacitor bank inline with any controller. The wires are 10 awg.
The caps will not help much if they are not placed colse to the FET's where the voltage spikes occur. You are shooting on sparrows with cannons.
 
madin88 said:
The caps will not help much if they are not placed colse to the FET's where the voltage spikes occur. You are shooting on sparrows with cannons.

The resistance in the wires will be much less than the ESR of the caps, so that's not really a problem. The inductance of the wires is much more of an issue, so they should be kept tight against each other and as short as possible. At really high frequencies, the aluminum electrolytics don't perform well anyway and that's not their purpose. Some SMD MLCCs across the rails are good for the really fast stuff.

This approach is pretty common in RC controllers and it seems to work OK.
 
fechter said:
madin88 said:
The caps will not help much if they are not placed colse to the FET's where the voltage spikes occur. You are shooting on sparrows with cannons.

The resistance in the wires will be much less than the ESR of the caps, so that's not really a problem. The inductance of the wires is much more of an issue, so they should be kept tight against each other and as short as possible. At really high frequencies, the aluminum electrolytics don't perform well anyway and that's not their purpose. Some SMD MLCCs across the rails are good for the really fast stuff.

This approach is pretty common in RC controllers and it seems to work OK.

The external electrolytics are there to help with the low frequency stuff. No room in the shell for more big caps...gotta do what I can and the external bank is all I can do. I'll get some ceramics for adding inside the shell for the high frequency stuff.
 
I'm still waiting on the copper and ceramic insulators to arrive. A few more days to go...

In between time, I finished up the shell.

I got some copper sheet and cut a couple of squares. The heat pipes are not particularly flat on the bottom so I coated the surface with flux. Then the squares, got coated in solder so that there would be lots of extra. Then I put it all back together and headed for the glass top stove. The glass surface is brilliant for reflowing. I was doing some LED mods a couple of days ago and normally I use an old aluminum frying pan, but now that I gave a glass topped stove, why bother? Once the copper squares were in place, I heated the whole thing on the stove top until the solder started flowing. I then needed to tighten the screws a little since the solder was now liquid and I could squeeze it out. This ought work much better for heat transfer than thermal tape.

Outer%20shell%20reenforcing%201.jpg


Since I was adding an external heat spreader, I needed to remove a few heat sink fins where that thinner strip of aluminum is and I also needed to add a section under the CPU coolers. LAter on, once I have the copper, I'll drill 4 more holes in the thinner strip to secure into the copper.

Outer%20shell%20reenforcing%203.jpg


Now that there was flat areas for an external aluminum angle, I made this part to use all the existing screw holes. I bought a an M3 screw kit a while back. It has lots of lengths in the kit. They have come in quite useful for this build. Everything has allen head screws now.

Outer%20shell%20reenforcing%204.jpg


Outer%20shell%20reenforcing%205.jpg


Outer%20shell%20reenforcing%206.jpg


That's a lot of metal for a heat path.

Outer%20shell%20reenforcing%202.jpg


A little thermal grease and I can screw down the mounting plate.

Outer%20shell%20reenforcing%207.jpg


Outer%20shell%20reenforcing%208.jpg


I'll need to take the CPU cooler section back off again once I have the copper for the heat spreader. I needed to make sure it all fit together OK so I fully assembled it.

Outer%20shell%20reenforcing%209.jpg


Outer%20shell%20reenforcing%2010.jpg


For now I need the factory heat spreader in place to fit the mosfet wall screws properly, but this will get replaced soon enough with copper.

Outer%20shell%20reenforcing%2011.jpg
 
It's probably overkill, but it sure looks cool. That should be able to dissipate 1kW easily. Let's hope the total amount of heat it needs to dissipate is much less than that. The limiting factor should no longer be heat dissipation.

It would be nice to know exactly how much heat it really needs to dump. It could possibly be estimated from measuring the heat sink temperature of a given controller and cover it with insulation so all the heat goes into raising the temperature. Then weigh the heat sink and look up the heat capacity of the material and do some math on the temperature rise. Sort of like a calorimeter.
 
fechter said:
s
It's probably overkill, but it sure looks cool. That should be able to dissipate 1kW easily. Let's hope the total amount of heat it needs to dissipate is much less than that. The limiting factor should no longer be heat dissipation.

It would be nice to know exactly how much heat it really needs to dump. It could possibly be estimated from measuring the heat sink temperature of a given controller and cover it with insulation so all the heat goes into raising the temperature. Then weigh the heat sink and look up the heat capacity of the material and do some math on the temperature rise. Sort of like a calorimeter.

The PV telemetry module has 2 temp sensor inputs. I'm going to mount one on a mosfet and the other on the copper heat spreader. That should give some interesting information. Ideally they will both read the same and that will be at most 3-5 degrees F warmer than ambient.

I think with this much metal, it will take a LOT to heat soak it and worst case the CPU coolers may need a harder blowing fan. This fan is really quiet so you know it's not moving much air. I have a box with fans and controllers in it. I bet I have a single fan controller I can add to this. There's no reason for the fan to run if it's cool.

If I need to dissipate 1kw...geez...that's a LOT. The power busses are beefed up, I have external electrolytic caps and I''ll be adding ceramics to the board soon.

I wonder if I can reach the 180 amp limits of the silicon which are well above the leg limits?
 
The copper arrived last night and the ceramic insulators will arrive tomorrow at the latest.

This is one hell of an expensive controller!
new mosfets
lots more aluminum
copper heat spreader
ceramic insulators
CPU coolers
8 awg wires
bigger shunts


The copper needs to be slimmed down a little as it's too thick. It's 8mm and I need 7mm. I'll probably try a mill end in my drill press to take off the thickness. Clamp the copper in my cross vice which is mounted to the drill table. Might try my 3 axis rotary table.

The insulators need one side ground down so the side by side mosfets can fit together. I bought some diamond wheels that arrived last night for taking down one side of the insulators.

Cap%20close-up%201.jpg
 
200 ceramic insulators arrived yesterday. 2 arrived broken.

Ceramic%20insulators.jpg


I put a diamond grinding disc on my drill press and used my cross vice to take thin grinds off the stack of insulators.

Drill%20press%20setip.jpg


I started with a stack of 30 of them held together with a screw and nut and then super glued the edges so they could not come apart.

Ceramic%20insulators%202.jpg


This is one edge ground off. I really should have done both long sides at the same time.

Ceramic%20insulators%20-%201%20side%20ground%201.jpg


Ceramic%20insulators%20-%201%20side%20ground%205.jpg


Heating them up on the stove makes the super glue release. Just be careful of the fumes!

Ceramic%20insulators%20-%201%20side%20ground%207.jpg


The ceramic gets a little discolored by the semi-burn super glue, but it is otherwise clean. As you can see, I've trimmed off about 1mm of width.

Ceramic%20insulators%20-%201%20side%20ground%208.jpg


Once I had them all apart, I tried a couple of the insulators and they were still too wide. I needed to slim down the other side too. They all got clamped together with the screw and nut and then the edges super glued again. I then used the same set up in the drill press to grind off the other edge. between both sets of grinding, I lost 2 insulators. This is one of the two insulators that broke while grinding.

Ceramic%20insulators%20-%201%20side%20ground%209.jpg


Once done, I reflattened all the insulators on this 800 grit diamond wheel. Not all of the discoloration would come off as the ceramics are not perfectly flat. As long as I got any residual super glue off the flat surfaces, that's probably sufficient and any very small variations in thickness will fill with thermal paste.

diamond%20wheel.jpg


I trimmed off a little length too. This is going to fit to the mosfet spacing and not hit the board.

Ceramic%20insulators%20-%20final%202.jpg
 
On hold for a bit to get a couple other projects cleared out for a friend.

I was also waiting on new band saw blades. All my old ones are dull from trying to cut hardened motor shaft. Cutting hardened shaft is more a case of burning through it than it is chipping out bits of steel with the teeth on the typical band saw blade. To drill hardened shaft pretty much requires carbide drills. HSS drills tend to start cutting and quickly become dull. Depending on the grade of the HSS, I've had them become dull immediately. Resharpen the bit and they dull instantly again. Others last a bit longer, but rarely can they drill all the way through a 12mm shaft. Maybe I'm not buying high quality HSS bits? Carbide is a totally different matter. It will cit just about anything but it sure is brittle! I have a selection of carbide stuff specifically for motor shaft work.

Anyway, new blades arrived a couple of days ago. Now I can cut down the copper block to fit inside the shell and get it drilled and tapped. I think this is the last item on the list!
 
I've been finishing up my XB-502 build. I think that's done-done now...with the exception of something going wrong.

Time to get back to this controller build.
The copper for the heat spreader has been sitting on my bench for a could of weeks now. I'll get that cut and drilled over the next few days.
 
Yyyeeessss dude! I love the project! Freaking awesome on a 12fet controller. Makes me feel a little better about my 12fet running about 2000W max or around 48A. The only thing I've modded on my controller is all new waterproof wires for the sensing wires and the shunt mod. I'll soon be changing out the 14awg phase and battery wires for 10awg silicone wires. I wanted something beefier for the phase wires like 6awg but that'd just be total overkill for a bike that'll only pull 2000w max... Im more so limited by the phase wires going through the shaft into the hub motor. I've heard you can squeeze 10awg ptfe wire through there but I'm just not going to go that crazy into modding it. It works, I'll upgrade the easy components and that'll be that.

I'm mostly stop and go traffic on my bike and I routinely pull 1800-1900w on accelerating and regen on braking. The controller barely gets warm! It's open to air flow on the bike so I'm sure that helps alot.
 
Philaphlous said:
Yyyeeessss dude! I love the project! Freaking awesome on a 12fet controller. Makes me feel a little better about my 12fet running about 2000W max or around 48A. The only thing I've modded on my controller is all new waterproof wires for the sensing wires and the shunt mod. I'll soon be changing out the 14awg phase and battery wires for 10awg silicone wires. I wanted something beefier for the phase wires like 6awg but that'd just be total overkill for a bike that'll only pull 2000w max... Im more so limited by the phase wires going through the shaft into the hub motor. I've heard you can squeeze 10awg ptfe wire through there but I'm just not going to go that crazy into modding it. It works, I'll upgrade the easy components and that'll be that.

I'm mostly stop and go traffic on my bike and I routinely pull 1800-1900w on accelerating and regen on braking. The controller barely gets warm! It's open to air flow on the bike so I'm sure that helps alot.

Skinny wires are losses. They heat up instead of conducting more current. I've opened up several motors and replaced the phase wires. It's worth it. I had a 500 watt hub a while back. It was totally limited by it's phase wires. Once I replaced them, it ran fine at 2kw. i have a long board that will get 4 hover-board motors on it. I'll at least take apart one of the motors and see if the wires are sufficient or not. Most likely they are too small and are limiting the motors. AKA...mod your motor with BIGGER wires. It will help. Also, If you fix the controller for more power and you don't do anything to the motor, what's the point?

5kw with 10 awg on everything is fine. You won't need bigger wires than that. Since this 12 fet project is theoretically going to do 7kw or so, I used 8 awg wire everywhere and that will do the job.

With minor modding...NOTHING as extensive as I have done here...4kw is a piece of cake. I've been running a 12 fet at 4-5kw continuous for well over a year now. I modded a 12 fet for a friend in the UK and he had tons of cooling direct to the mosfet wall and was doing 6kw. If you are getting 2kw with a decent controller, then depending on your mosfets and a few other things, you should be able to double that or more.

This is open air and running 5kw with no extra cooling. There is an internal added heat spreader. It runs warm at 90F.

Currie%2010181117%2012.jpg


Heat%20spreader%202_zpsm8emddvg.jpg
 
Yea. I'm battery limited as my new panasonic ncr18650 would be able to deliver about 60 maybe 65amps burst, the bms is limited to 45A continuous...but I'd rather treat my batteries nicely than push them too hard just to get a little more acceleration.

I'd rather not open up the hub motor as I don't want to ruin the hall sensor wires and I think 14awg should be OK going through the shaft then once the wires are out, switch them to 10awg along with xt150 connectors for the phase wires.


I've been working hard on my new battery especially through grad school too and I'm hoping to have that completed in the next month or two.
 
I've been busy with other projects and finally got back to the copper heat spreader.

I drilled the 5 holes securing the spreader to the shell wall and then relaized I also needed to reduce the thickness by about 1.5mm. I don't have a mill and my drill press isn't particularly new or precise. I locked the quill in the retracted position to help make it as stable as possible. I then raised the table to just clear the jaws of my cross vice. I have several carbide end mills that are in decent condition. I'm sure if I had a mill...even a cheap one...I would get better results than this, but I don't so I "made do" with what I had. The Wilton cross vice is their best model and after some tweaking I've gotten it to be free of slop and silky smooth. The drill press is the biggest source of slop. I took a 1mm deep cut and 2mm wide each pass. It took a while to mill down the surface. I found that I got a better finish cutting across the copper in one direction than the other direction. As a result, I cut in one direction only which made this take doubly long. I tried several motor speeds to see if that would help the cut quality and ended up using the fasted RPM I could get from the drill press. I'm pretty sure I really needed more RPM's to get a good finish. Never mind doing this on a drill press.

Copper%20heat%20spreader%201.jpg


Copper%20heat%20spreader%202.jpg


I made a second pass over the copper at .5mm deep and that got me a bit better finish. Still it's pretty lousy compared to what I've done on an actual mill. There's just too much slop in my drill press and that creates vibration and a rough and inconsistent finish.

Copper%20heat%20spreader%203.jpg


Copper%20heat%20spreader%204.jpg


I then put a sheet of sand paper on some tempered glass and sanded the surface flat and smooth. As you can see there's some low spots. I mostly sanded them out, but that still left a few spots that were too deep to sand out.

Copper%20heat%20spreader%205.jpg


Copper%20heat%20spreader%206.jpg


I took the piece of copper over to my glass topped stove and heated it up. A little solder melted into the low spots took care of that problem. The solder is maybe .2mm thick in the deepest spot after sanding it flat again. This is quite usable and flat. There were a few deep spiral cuts into the copper to fill in the right area. I'm not sure what went wrong in that larger area to get the slightly deeper cut. I'll move my cross vice some and mount it differently in the future. That will help some, but it won't fix the slop in the drill press.

Copper%20heat%20spreader%207.jpg


In the controller shell. 100% coverage of the mosfet wall. I haven't done it yet, but I'll also put 4 screws in the top edge to pull it up to roof of the shell to add a bit more heat path.

Copper%20heat%20spreader%208.jpg


The new heat spreader is 1.5mm thinner, 10mm taller and 6mm wider than the original one. I'll be using ceramic insulators behind the mosfets. There's not much else I can do to improve pulling heat away from the mosfets after this.

Copper%20heat%20spreader%209.jpg
 
Back
Top