Add on T0247 water cooling upgrade board for a 12 FET

[zombiess]

I've been working on an add on board to allow me to swap to some IRFP4568 150V TO-247 FETs because they can handle a lot more power than the IRFP4115 FETs can. This is just my initial test run to see how it will perform. I'm going to try it without any water first and see how hot things get, then see if adding the water will help. Everything is off the shelf for the most part, just customized by myself. The water cooler has been modded by me to accept the FETs and with some tweaks to hopefully aid in thermal transfer.

It's all aluminum with the top block being nickel plated copper where the water makes contact with it. The TO-247 FETs are clamped between the shells quite firmly with some Bergquist Q-Pad 3 insulators after the insides were lapped smooth to deburr after drilling the holes and remove any texture. I then applied thermal paste to the front side of the FET body to make contact with the front face of the cooler. There was a gap where it originally mounted together with 2 screws since the TO-247's are thicker than what this was designed for, so the gap in the shell was filled up with a thermal transfer material supplied with the kit that seems pretty good, but I had to use two layers of it, but it's better than having an air gap. Screws and mounting hardware are brass to help promote heat transfer between the surfaces. I wanted to use aluminum hardware but was unable to find any in the size I needed so I went with brass as the next best alternative; I have no idea if it will help or not, but figure it can't hurt since it adds thermal mass and has a thermal conductivity of about 100 W/(m.K).

Circuit board layout might not not be ideal, but I was going for maximum trace width (which can be beefed up with wire and solder) and still be compact. There appears to be enough room to run double the amount of caps compared to a normal board. The brown caps are 200V 330uF Nippon KXG series caps rated for 1.4A ripple @ 120hz and 3.2A @ 100khz. Dissipation factor is supposed to be below 0.20@120hz according to the data sheet. The main power cap is a 1000uF 160V Panasonic ECG series with a 3A ripple current at 120Hz and a ESR of 200 mOhms @ 120hz.

Battery positive and Phase wires are setup to be 8 gauge directly to the board with jumpers to supply 12 gauge back to the factory board. All caps, phase and power wires should fit inside the standard 12 FET case. The case will need to be modified by having a slot cut out of it to allow the extension board to stick out so it's not going to be water proof. It's also going to require short circuit protection which I'll probably handle with several layers of tape and foam.


******Please take the following with a grain of salt as I'm unsure if I've done this math correctly******* Tell me if I'm wrong please!
I only started learning about MOSFET power design about a month ago (I'm a digital guy) so I hope I don't have any giant red flags in this design. I'm far from an electronics expert but I feel I have a grasp on most concepts. I've done many many hours of reading this past month before trying to take on this project. This forum is full of some good designers and always comes up in my Google searches. Thanks for sharing your info everyone.

Using this calculator http://www.electronics2000.co.uk/calc/reactance-calculator.php to figure out the reactance for a frequency of 10khz (I'm not really sure what speed our controller switch at) I'm getting a ESR of 48 mOhm per cap. It looks like I'm able to fit two per location between phases on the board for 660uF and 24 mOhm total.

This looks to match the spec sheet which says they have a impedance of around 3-6 ohms @ 120hz and the calculator says they have 4 Ohms ESR @ 120 Hz.

I'm using some 1uF 275VAC polypropylene caps in place of what ever normally comes there in parallel with the aluminum electrolytics. The spec sheet for them states they have 0.1% dissipation factor at 1khz so they should also bet quite low ESR. At 1kHz the reactance calculator says they have an Xc of 159 ohms. 159X0.001 = 159mOhms.

This is my very first attempt at doing power design, so I'm sure some of you guys who are way smarter than me will be able to pick this apart and tell me where I've screwed up, please do so because I'd like to learn more. I am going to try this out and see how it performs since it seems many things are a compromise in design and this is no exception and amateur at that.

My biggest worry with this is the gate driver traces being long. It was suggested to me to use twisted pair wire to connect them but I don't know where to make the connections. If you do, please add your info to my thread here http://endless-sphere.com/forums/viewtopic.php?f=2&t=34577 so I can add the proper holes to the circuit board.

Here are the pictures of what I have so far which shows the layout and a printout of the circuit board.

The 1uF poly caps might need to lay on top of the electrolytics to fit.

It's going to be snug, but I think I can squeeze in two 330uF caps. I also have the option of using a single 1000uF cap like on the battery feed.

Straight down shot giving you an idea of how it's going to all fit and what it will interfere with. One small cap on the board will have to be laid over on it's side for the left most 330uF cap to fit.

Circuit board layout showing how it's going to fit. I will be joining it to the control board with header pins

Bottom shot of the water cooling block all assembled with my IRFP4568 FET's, 6 per section. The screw heads just barely clear each other.

Front shot showing the bolt patter and water hookups. Water hookups can swivel.

End view showing how the FET sandwich is put together. The white stuff is ceramic thermal paste that squeezed out from the front of the FETs.

If this works out well, I'll move on to an 18 FET version. I might also just try to redesign an entire power stage using gate driver IC's and make a 24 FET controller based off a 6 or 12 FET board since the drivers will no longer be an issue, I'll just take the PWM signal and feed it to my own FET drivers. I'm not sure if this water blcok will handle the load I'm going to put on it, but I can always add a small fan to blow over it as well, water + forced air I also have a 6 peltier 2 pass water chiller from an old computer project that can handle quite a few watts that could be used to chill some water in a small reservoir or slow the temp rise if I really want to get nutty about this. It worked well on my old water cooled computer setup never letting the CPU core go above 55C with me putting way too much voltage into it. At stock settings it kept the CPU temps down around 25C under full load and used a controller to keep it from getting too cold and causing condensation. At 100 amps I should only have about 180 watts of heat to get rid of if my math is right. 100A^2*(0.006ohm/2)) = 30 watts heat produced per pair of FETs ignoring switching losses. 30 watts * 6 pairs = 180 Watts of waste heat, maybe 220W with switch losses?.

P.S.
Big thank you to Edward Lyen for selling me the board and some parts I needed to build this experiment. I was just going to rebuild my 12 FET IRFB4110 controller that popped, but decided to get experimental instead.

 

[zombiess]

No interest in this? Thought someone might have a comment or two.

 

[ohzee]

Looks pretty sweet to me tho way above my pay grade or rather ability level.

Still I find it a pretty cool way to get more out of these controllers. Like to hear how it works on the road.

 

[zombiess]

Looks pretty sweet to me tho way above my pay grade or rather ability level.

Still I find it a pretty cool way to get more out of these controllers. Like to hear how it works on the road.

Should have the board / controller built this week. Hoping to test it out out by the weekend on my Croatian Hubzilla in a 20" wheel.

It's really not that difficult or complex, I only just started learning power electronics myself, this is just an extension board. If this works I'll start working to produce these if the specs come out good. I'm hoping for some really high amp ability out of it. I just ordered up some IR2110 FET drivers. Hoping if this works to expand these out to 24 FET's. That should be good for a nice compact 20KW continuous 150V controller.

IRFP4668 TO-247 200V FETs could be a possibility too for those with low KV motors or those like me who want a suicidal speeds. IRPF4568's seem to be the best going though since the IRFP4468's max out at slightly more than IRFP4568's.

 

[amberwolf]

Interest, yes, but no relevant comments yet.

 

[etard]

Frickin cool Zombiess! I want to see this thing succeed, if only for 20 seconds. :lol: Why not just cool the whole case with non conductive fluid like, mineral oil or the like and just pump it through an external radiator. I have actually been trying to get around to this mod after looking at Karma's Space Heater controller mod. Thud has me running these 12 Fet's at the ragged edge, so some sort of cooling mod is in order.

 

[parabellum]

Those are comparably expensive fets, right? I also hope this tinny mounting walls can get the heat quickly enough to the the sink, they do not appear having potential and cross section of squeezing 200w of heat. You may consider to make 1 solid cooper block instead.

 

[zombiess]

Those are comparably expensive fets, right? I also hope this tinny mounting walls can get the heat quickly enough to the the sink, they do not appear having potential and cross section of squeezing 200w of heat. You may consider to make 1 solid cooper block instead.

I paid around $5.50 for each FET, they aren't that much and they blow IRFB4115s out of the water. It's basically a IRFB4110 with a max 150V operating voltage and more power handling capability.

The goal of this is to avoid custom machining and use off the shelf hardware that only require slight modifications. Sure a copper block with cooling holes run through it would be better, but it wouldn't be this compact or cheap. If this thing produced 200W of heat, that's 50W of heat per plate, that's really not too bad, just adding a fan would keep them pretty cool.

I see many people on here talk about max effort setups and the threads are one year or older and most have no prototype even built. I'm hoping to have a running controller this week and I started this project last week. Two weeks of design/build time, now lets just hope it works.

This is what testing is for. This is a proof of concept. After so long of reading and theory I can only take so much before going to the build it an try it method after I feel satisfied the theory is sound. I do strongly believe this is going to work, just not sure how well. The stock heat sinking design on the controllers leaves a lot to be desired and isn't very efficient, but yet works and to some pretty impressive power levels. I'm pushing almost 9KW into my 18 FET after some thermal and power mods on the not so great IRFB4115 FETs and they are asking for more. I would not be surprised if I end up putting 12KW into it and it doesn't have anywhere near the cooling this thing does.

TO-247 FETs have a much larger body and a pad on the back that they use to transfer the heat out of. The body of the FET also gets quite hot as anyone who has worked with them knows, that's also why I decided to use thermal paste on the front of the FET body as well. In addition the brass hardware helps to thermally link the two sides together.

I will be monitoring FET temps with my thermocouple logger. First without a water cooling setup, then with water cooling running through, then possibly with a peltier chiller setup. This way I'll know how effect the water will have.

 

[parabellum]

Right, price difference is not that big. I like your approaches and quick handling and strongly believe is success of your design, just pointing on potential bottle necks. If it gets on the way to desired performance it can be upgraded easily later replacing those 2 wings with 1 solid block. Prove of concept is more important for now. :wink:

 

[zombiess]

Right, price difference is not that big. I like your approaches and quick handling and strongly believe is success of your design, just pointing on potential bottle necks. If it gets on the way to desired performance it can be upgraded easily later replacing those 2 wings with 1 solid block. Prove of concept is more important for now. :wink:

You hit the nail on the head with that one. Can always upgrade later, this is version 1.0, build it, see if it works, then start tweaking. I'm a big fan of off the shelf hardware where possible because I'm not very good mechanically and have zero machining ability. First run machining is also pretty expensive and usually time consuming regardless of material cost.

If this whole thing turns out to be crap I'll have wasted at most $300 if I pop every single FET, if not, the 12 FETs will go onto a different expansion board and be placed inside a standard 18 FET case. This was my original plan until I decided to try water cooling first because it can be much more compact. For reference that 12 FET module is much heavier than it looks due to the top plate being a copper bar. It weighs about the same as a fully built 12 FET controller, so there is quite a bit of mass.

 

[Lebowski]

ESR doesn't work like that.

What you calculated for the 330uF @ 10 kHz is the frequency dependent part of the impedance.

I think to make clear what I mean it's easier to talk about inductors. With an inductor it's
easy to understand that the impedance is made up of 2 parts. On the one hand there is the
bit with the magnetic field and such, this is the actual wanted inductive behavior. On the
other hand the wire that makes up the inductor has a resistance. So the total impedance
is made up of a reactance (the magnetic bit) and a resistance (the ESR).

Caps have the same. There's a reactance part which has to do with 2 plates and the electric
field inbetween them. And then there are the dielectric losses, wire resistance etc which is
all modelled by the ESR (equivalent series resistance).

The 48 mOhm you calculated is only the reactance part and is not the ESR. Look at the
datasheet of your caps, ESR is probably much higher (I got a good 1800 uF cap at home which
has an ESR of 80 mOhm, extrapolating would mean a good 330uF cap has 0.44 Ohm ESR).

Build an active snubber circuit would be my advice
http://www.endless-sphere.com/forums/viewtopic.php?f=30&t=34436

 

[zombiess]

ESR doesn't work like that.

What you calculated for the 330uF @ 10 kHz is the frequency dependent part of the impedance.

I think to make clear what I mean it's easier to talk about inductors. With an inductor it's
easy to understand that the impedance is made up of 2 parts. On the one hand there is the
bit with the magnetic field and such, this is the actual wanted inductive behavior. On the
other hand the wire that makes up the inductor has a resistance. So the total impedance
is made up of a reactance (the magnetic bit) and a resistance (the ESR).

Caps have the same. There's a reactance part which has to do with 2 plates and the electric
field inbetween them. And then there are the dielectric losses, wire resistance etc which is
all modelled by the ESR (equivalent series resistance).

The 48 mOhm you calculated is only the reactance part and is not the ESR. Look at the
datasheet of your caps, ESR is probably much higher (I got a good 1800 uF cap at home which
has an ESR of 80 mOhm, extrapolating would mean a good 330uF cap has 0.44 Ohm ESR).

Build an active snubber circuit would be my advice
http://www.endless-sphere.com/forums/viewtopic.php?f=30&t=34436

I got an Capacitor ESR meter tonight and it's saying the ESR of my 330uF caps is around 220 mOhm so if I can trust the meter, I guess that is the ESR of the caps. Either way it's pretty much semantics as these are the caps I'm using, hopefully two in parallel with a 1 uF Polypropylene cap for double the capacitance and even lower ESR. How important is the ESR anyways?

What is the switching frequency these controllers operate at? 10-20khz? I've never scoped one in operation.

 

[rhitee05]

I also hope this tinny mounting walls can get the heat quickly enough to the the sink, they do not appear having potential and cross section of squeezing 200w of heat. You may consider to make 1 solid cooper block instead.

I'm also a little concerned about this. I don't think you necessarily need a full-custom waterblock, but it could be much better if you mounted the FETs back-to-back onto a larger piece of bar stock, maybe something that was the same width as the water block. I think those little wings could end up being a limiting factor. It's not quite clear if you did this or not, but for mounting I would suggest just using a flat bar across the front of the FETs as a clamp to hold them tightly against the surface. This is better than using the mounting hole, and it also avoids any insulation issues with the screws. Don't bother with thermal paste on the front of the cases - they will get warm, but there's virtually zero benefit as the package epoxy is not very conductive.

 

[etard]

I completely understand what you are saying Zombieman, I am probably in that category of talking about ideas and methods and never following through. I like your style, do a bit of research, come up with a simple plan and execute! If you ever need a solid block of aluminum milled or machined for these FETs let me know, I'll do it for free and send it up to you no prob man. I really want to see this succeed in the simplest package possible, watercooling adds complexity and if we can shed the same heat with a small fan and a solid block of aluminum, I'm on board. By the way, can you link a source to these FETs?

 

[zombiess]

I completely understand what you are saying Zombieman, I am probably in that category of talking about ideas and methods and never following through. I like your style, do a bit of research, come up with a simple plan and execute! If you ever need a solid block of aluminum milled or machined for these FETs let me know, I'll do it for free and send it up to you no prob man. I really want to see this succeed in the simplest package possible, watercooling adds complexity and if we can shed the same heat with a small fan and a solid block of aluminum, I'm on board. By the way, can you link a source to these FETs?

http://search.digikey.com/scripts/DkSearch/dksus.dll?x=0&y=0&lang=en&site=us&KeyWords=irfp4568

I originally bought mine from Allied electronics for about $5.40 each, but they have now jacked up the price http://www.alliedelec.com/search/productdetail.aspx?SKU=2732785

Water cooling really isn't that complex and is reliable, the main reason I'm interested is two fold. It's VERY good at controlling thermal load and it allows you to make things into more compact packages (ever seen a 24FET or 36 FET controller on a bike, it's not exactly small).

If you are offering machining to help me prototype, count me in. I'm going on vacation, but if I have down time I'm going to be doing some mechanical design and circuit board layout. I too like the back to back approach on a nice sized chunk with something like 2-4 1/4" holes bored through it that can be tapped for NPT fittings. I've already got a good lead for manufacturing in copper on a final design, but would like to start off with aluminum due to cost and testing. I already have several ideas in mind, just need to draw them up.

I've actually already started the research and am starting the new design with separate Hi/Lo FET driver IC's and most likely a buffer which should be quite a bit better than this one as long as you'll do the machining of the alum as you stated. I can't draw 3d, but I can give you a very simple 2D mechanical drawing of it. It's not exactly rocket science, just some bar stock with holes bored through it and pilot holes ready for mounting the FETs after everything is tapped to 6-32.

I'm also a little concerned about this. I don't think you necessarily need a full-custom waterblock, but it could be much better if you mounted the FETs back-to-back onto a larger piece of bar stock, maybe something that was the same width as the water block. I think those little wings could end up being a limiting factor. It's not quite clear if you did this or not, but for mounting I would suggest just using a flat bar across the front of the FETs as a clamp to hold them tightly against the surface. This is better than using the mounting hole, and it also avoids any insulation issues with the screws. Don't bother with thermal paste on the front of the cases - they will get warm, but there's virtually zero benefit as the package epoxy is not very conductive.

I'll find out in testing how good it works, just remember, this whole assembly has the mass of an entire fully built 12 FET controller, it's not exactly light. Now that I have a machinist willing to help, back to back will be the next revision most likely if etard can get me something by early Feburary/March. I'm starting on the design tonight.

I will be using the mounting holes and also clamping. There are no electrical isolation issue with screws on TO-247 packages as the hole is insulated from the thermal pad on the back. This way we get the best of both worlds, just like what I did in the pictures. Mounting holes + clamping the FETs.

I want to publicly thank each and every one of you for helping me out by contributing to this thread and helping me. I sincerely appreciate it because I'm a little over my head on some things and I'm not afraid to admit it. Your suggestions are only going to help improve this

If anyone could help me locate the source return on the circuit board in this thread http://endless-sphere.com/forums/viewtopic.php?f=2&t=34577 with high res pics I'd really appreciate it. This way if I experience ringing in the gate drive I can most likely solve it pretty quick with twisted pair and a resistor.

Thanks,

Jeremy

 

[etard]

Draw it up man! I got plenty of THICK aluminum sitting on my shelf ATM and I'm not afraid to use it! If you want to get crazy we can also make it water-cooled and use two pieces sandwiched and mill in pockets behind each FET for quick heat dispersion into the water. 2d mechanical drawing will be fine.

 

[liveforphysics]

This is really a fun thread to follow. I enjoy seeing your work with controllers.


Any cap has an ESR dependent on the freq that it's being tested. At very high freq's, like clamping transients, the little poly caps really shine, and have a lower ESR than the big electrolytics. When the freq drops lower (like motor commutation freq), then it's in the range were the electrolytics shine, and provide the stability you need.

When using an ESR meter, you need to set the freq to the area you're looking at protecting. For the little poly caps, this may mean 500khz (higher order harmonic transients you're trying to clamp), for the electrolytics, it may mean testing at 200-500hz (around the commutation freq you're dealing with).

 

[zombiess]

This is really a fun thread to follow. I enjoy seeing your work with controllers.


Any cap has an ESR dependent on the freq that it's being tested. At very high freq's, like clamping transients, the little poly caps really shine, and have a lower ESR than the big electrolytics. When the freq drops lower (like motor commutation freq), then it's in the range were the electrolytics shine, and provide the stability you need.

When using an ESR meter, you need to set the freq to the area you're looking at protecting. For the little poly caps, this may mean 500khz (higher order harmonic transients you're trying to clamp), for the electrolytics, it may mean testing at 200-500hz (around the commutation freq you're dealing with).

Thanks for the info. Unfortunately my test is a cheap one and I can't control the frequency independently. But it's still good to get an idea of what a cap is capable of and weeding out bad ones / less desirable ones since the ESR scales.

I had to spend an entire day reading on caps just to figure out wtf to buy LOL. Went with what I thought were good choices and affordable, I hope you approve of the design. I read that polypropylene is one of the best to act as a snubber so I figgured 1uF should be quite an upgrade from what ever tiny little cap is on the factory boards, having a 275VAC working voltage doesn't hurt either.

 

[rhitee05]

Ceramic caps are also a good option for the higher-frequency range. At those 100s of kHz-and-up kind of frequencies you really care more about ESL than ESR. SMT ceramic caps are perfect for this, and they're small and cheap so you can put a whole pile of them on the board.

 

[nieles]

Bottom shot of the water cooling block all assembled with my IRFP4568 FET's, 6 per section. The screw heads just barely clear each other.

could be the picture, but i looks like the aluminium plate the fets are mounted to is a bit warped.

If anyone could help me locate the source return on the circuit board in this thread viewtopic.php?f=2&t=34577 with high res pics I'd really appreciate it. This way if I experience ringing in the gate drive I can most likely solve it pretty quick with twisted pair and a resistor.

http://www.hordsoffun.com/ebike/pics/6fet.pdf this looks like the the same circuit as used on your controller. only this is a 6 fet and yours is a 12 fet.

looks like the source return is at the collector of Q2b?

not my schematics btw.

 

[zombiess]

Bottom shot of the water cooling block all assembled with my IRFP4568 FET's, 6 per section. The screw heads just barely clear each other.

could be the picture, but i looks like the aluminium plate the fets are mounted to is a bit warped.

If anyone could help me locate the source return on the circuit board in this thread viewtopic.php?f=2&t=34577 with high res pics I'd really appreciate it. This way if I experience ringing in the gate drive I can most likely solve it pretty quick with twisted pair and a resistor.

http://www.hordsoffun.com/ebike/pics/6fet.pdf this looks like the the same circuit as used on your controller. only this is a 6 fet and yours is a 12 fet.

looks like the source return is at the collector of Q2b?

not my schematics btw.

Thanks for that schematic, that looks like the same gate driver that's on my board, the resistor values look the same as what I measured. Thank's a ton for posting that because now I understand the driver setup better.

Yes, the ends did slightly warp as i got a little too zealous with the screwdriver when tightening it down, I backed it off a little and it's good now. There isn't enough room between everything to use washers which helped prevent the warping, had to get rid of them to make it fit which made it easier to warp as you see in this pic, I went back and fixed it by loosening up the nuts/bolts. It was only warped on the ends a tiny bit, noting by the FETs so it's got a solid flat contact. I even lapped the plates with sand paper over glass to make sure it was as flat as I could get it since it had a texture and I wanted it as smooth as possible.

 

[ZOMGVTEK]

Is that water cooler made for two RAM sticks? I'd question how good of a thermal path that is, assuming its unlapped aluminum made to dissipate only a few watts.
I'm wondering if liquid cooling is worth it on a bike controller. The idea is to make this thing stupid reliable, and having little to no thermal mass tends to not accomplish that.

That said, I like your 'add-on' board approach. It's great to see someone doing controller mods like this.

 

[zombiess]

Is that water cooler made for two RAM sticks? I'd question how good of a thermal path that is, assuming its unlapped aluminum made to dissipate only a few watts.
I'm wondering if liquid cooling is worth it on a bike controller. The idea is to make this thing stupid reliable, and having little to no thermal mass tends to not accomplish that.
That said, I like your 'add-on' board approach. It's great to see someone doing controller mods like this.

I did lap the plates and it's got some mass to it. Yes it is a ram cooler I've repurposed for a quick test setup.

I'm building / testing and collecting data, it's who I am. Too many people talk about doing this and that and having it perfect / no compromise and 1-2 years later their threads are still going with little to no progress or data. How many DIY controller projects do we have going on the forums? How many can we buy now? I want data and I want it now! This is only a quick test to see how it does in what's might be a worst case scenario with a not so good heat sink. I've already received two offers for machining to be done, just one of the reasons I love this site!

If you have ever opened up a 12 FET controller, you will notice there is a spreader bar with a thick electrical insulator which I don't believe is the greatest for thermal transfer. The spreader bar is 135mm x 19.8mm x 7.8mm. It then interfaces to the aluminum enclosure but not in a very flat solid manner (could be fixed by using screws along the top of the bar) in most of the controllers I've opened up. So the thermal transfer path is FET Junction to FET case to mystery insulator, to not so good spreader bar mounted not so well to the enclosure to ambient air. Leaves quite a bit to be desired one would think, but yet it works. These spreader bars are also not lapped flat and are usually pretty rough with pits and grooves. Even worse, they have a polished finish(polish is bad because it reflects heat back). That stock spreader bar has a volume of 13.5cm x 1.98cm x 0.78cm = 20.8^3 cm.

My hacked together water cooling block:
At the top it uses a piece of nickel plated copper. It's size is 125mm long x 19 mm wide x 5mm thick. The FET sandwiches are two matte black anodized plates 1.30mm thick x 125mm long x 32 mm high each giving each one a surface area of 40^2cm with two per 6 FETs for a total surface area of 80^2cm per sandwich or a total of 160^2cm (24.8^2”) of surface area to radiate heat from which is about the same as putting a flat 1x2” piece of .05” thick alum on each TO-247 FET. The plates are lapped smooth and use Bergquist Q-Pad 3 insulators which have a thermal conductivity of 2.0 W/m-K with a thermal impedance of at worst 0.65 C-in^2/W if my clamping force is only 10 psi, each pad is 0.75^2” or smaller, the front side of the TO-247 body is also ceramic thermal pasted to the other heat sink side of which both 6 FET sandwiches have 10 brass bolts/nuts holding everything together adding further mass/surface area and thermal conductivity. They should help spread the heat between the plates a little since brass has a thermal conductivity of 109 W/(m.K) which is pretty good.
Each sandwich is thermally interfaced to the copper water block by an area 125mm long x 7.1mm wide for a total contact area of 125x7.1 = 8.8^2cm and there are two of them so the water block has a surface area contact of 17.6^2 cm to the plates holding the FETs.

The total heat sink has a volume of about 19.3^3 cm (based on my measurements) with 11.9^3cm being the copper water block which is heavier than all the alum plates combined and this is all sitting out in the open air with the ability to run water through it. Let's not forget my circuit board either. It will most likely be beefed up with a few parallel 12 gauge copper house wire strips on the traces + solder, that's even more heat sinking ability.

Now compare the two and think about it. From a thermal stand point my setup seems inadequate on paper, but when you look at what we have been using and how poor it is thermally, it’s amazing that we aren’t vaporizing more FETs than we already are.

I've already started on a new and improved version of this due to the machining offers, but don't look for anything from me until sometime around march because I've got a busy travel schedule from late Dec to Early February, and I need to test this thing out to see how good/bad it is.

Failure is always an option and I'm not afraid to admit failure and will do so if this sucks, but it’s still a valid result and data point.

P.S. If you don’t like the idea of water cooling the controllers, just wait to you see what the future holds…. Mwahahahahaha!!! I’ve said too much…

 

[ZOMGVTEK]

I'm all for trying out new ways to get stuff done. It's just an interesting choice.

I decided to make an 18 FET TO-247 controller about a week ago, and managed to blow it up already. The remains are sitting in a pile, along with the 4 other 15 FET controllers I destroyed by doing dumb things. I still have another 15 FET with 250V FET's that I need to blow up. It's also possible it could actually work at 48S, but unlikely.

I'm armed with some new knowledge, and I'm going to try and make a snazzy 18 FET 150V power stage over the next few days. From everything I've come across people on here doing, it would be the best 150V controller available, and I'm fairly sure I can get it done for about $40-50. I have $20 invested in snubbers, the rest I managed to acquire at no cost.

This is somehow the only picture I managed to take. I have 18 hand picked TO-247's, matched by Rds(on), and a .75 x 1" copper bar to mount them on. I have the PCB mostly designed, I will try to etch it tomorrow. If you sip the datasheet koolaid, it will be good for about 500A phase current. I plan to run no less than 300A phase through it. I'm torn between making a copper slab based waterblock, and just using a big damn heatsink and a thermally switched ducted fan. This thing is probably going to weigh 10 lbs, but it will hopefully be robust.

 

[parabellum]

I'm wondering if liquid cooling is worth it on a bike controller. The idea is to make this thing stupid reliable, and having little to no thermal mass tends to not accomplish that.

Have water cooled aquastarmonster esc (mounted on scooter, powering sk2 80-100), size about 2 matchboxes weight 250g, it does not even get warm (it means heat is dissipated effectively) on 9KWh continuous unlike 18 fet infineon size of 24 math boxes over 1kg, 8kwh + 3minutes and you can not touch it anymore or 9fet infineon burned after 20sec on 6KWh climbing.
Yes I prefer reasonable complexity and "real reliability" before "stupid reliable" which is not reliable at all. Have 3 burned Infineons waiting for fet replacement, just sick of it.
Go for water cooling, big damn heat sinks sucks.