ElectricGod said:
You could use PGS graphite thermal pads instead of the ceramic insulators. They have better thermal conductivity than pure copper:
https://docs-emea.rs-online.com/webdocs/1181/0900766b811810f2.pdf
Epic 7kw+ 12 fet controller...or there abouts :)
- Thread starter ElectricGod
- Start date
You could use PGS graphite thermal pads instead of the ceramic insulators. They have better thermal conductivity than pure copper:
https://docs-emea.rs-online.com/webdocs/1181/0900766b811810f2.pdf
ElectricGod
10 MW
MJSfoto1956 said:
Yes please!!!
ElectricGod
10 MW
madin88 said:
Actually you can't in a motor controller...
While great thermal conductors, they are also electrically conductive. They are made from carbon after all. If I wasn't worried about electrical conductivity, I'd mount the mosfets direct to the copper...probably solder them down to it. But the metal tab of every mosfet is also the drain for that mosfet.
Let me know when you can get graphite pads that DON'T conduct current! Until then, ceramic is the best option in a controller.
ElectricGod said:
scroll down the paper and you'll find
I came across this recently in a topic about a big VESC controller where the guy plans to use them to increase the continuous power.
As far as i could understand, the thermal conductivity of those graphite sheets in z-axis (away from the chip) isn't that good, but the conductivity in x and y axis (sidewards from the heatsource) is 3 to 10 times better compared to pure copper.
Check the thermal images. They speak for themself.
I think the idea is to put the graphite sheet under the heat producing part to spread it over a larger area, then use a larger insulator. Having more surface area of insulator can carry more heat. More importantly, the heat producing part has a low thermal resistance at that first interface. Copper might be cheaper and easier.
ElectricGod
10 MW
What am I missing here?
Graphite conducts electricity...that means no bueno for mosfets.
I'd still have to put the mosfets onto a non-conducting material such as ceramic insulators. The only way this could work is to get a custom made ceramic heat spreader to use in place of my copper one. I haven't looked just now, but I bet copper is a better heat conductor than is ceramic. Ceramic works well in thin layers as a thermally conductive and electrically isolating insulator.
I don't see how graphite could be used effectively on the backs of mosfets as long as it conducts electrical current. Hmmm...does a graphite thermal paste exist? Would that work better than the typical white thermal paste we all adore and love? Maybe something like that between the mosfet and the ceramic insulator would be OK?
Graphite conducts electricity...that means no bueno for mosfets.
I'd still have to put the mosfets onto a non-conducting material such as ceramic insulators. The only way this could work is to get a custom made ceramic heat spreader to use in place of my copper one. I haven't looked just now, but I bet copper is a better heat conductor than is ceramic. Ceramic works well in thin layers as a thermally conductive and electrically isolating insulator.
I don't see how graphite could be used effectively on the backs of mosfets as long as it conducts electrical current. Hmmm...does a graphite thermal paste exist? Would that work better than the typical white thermal paste we all adore and love? Maybe something like that between the mosfet and the ceramic insulator would be OK?
Normally you have an insulator directly against the tab on the FETs. This immediately restricts the amount of heat that can flow.
If you place the FETs directly against a copper bar (maybe even solder them) and use an insulator between the bar and the heat sink, you can get more heat flow if the surface area of the insulator is much bigger than the FET tabs. The copper (or graphite) heat spreader has to be split up into 6 sections that are electrically isolated from each other. Parallel FETs can be on the same piece of bar.
It should be possible to model this and see how much difference it would make. I would imagine for short bursts of extreme dissipation it would help a lot. More long term, you would still be limited by the overall thermal path resistance.
If you place the FETs directly against a copper bar (maybe even solder them) and use an insulator between the bar and the heat sink, you can get more heat flow if the surface area of the insulator is much bigger than the FET tabs. The copper (or graphite) heat spreader has to be split up into 6 sections that are electrically isolated from each other. Parallel FETs can be on the same piece of bar.
It should be possible to model this and see how much difference it would make. I would imagine for short bursts of extreme dissipation it would help a lot. More long term, you would still be limited by the overall thermal path resistance.
ElectricGod
10 MW
madin88 said:Nothing, you just need to scroll down in the paper and you'll see that those graphite thermal pads are also available with one or two layers of insulation.ElectricGod said:
Dude!
Ah! Did not see that...
ElectricGod
10 MW
fechter said:
I can see how this would work now. Since this is a 12 fet, every 2 mosfets have the drain tabs electrically connected together. Solder those 2 mosfets down to a single section of copper that's as wide as the 2 mosfets and the full height I have in the mosfet wall. Behind the copper and between it and the mosfet wall that's where a much larger insulator would sit. The copper heat spreader would be cut into 6 electrically isolated sections...each with 2 mosfets soldered to it.
I may do this after a while if I see I need more efficient cooling. It will require isolating the screws from the shell and redrilling the shell a fair bit for more screws. I'll have the temp sensor attached under a screw directly on a mosfet tab. I'll know temps pretty accurately.
I guess the graphite insulator comes in sheets and can be cut to size.
Thanks guys, I can see how I would do this now.
Making a better heat path...
1. Use copper instead of aluminum for all that added metal I put in and on the shell to improve thermal conduction.
2. Use graphite insulators between the layers of added metal instead of thermal paste.
3. Add heat sinks to the added metal so cooling happens all along the heat path to the CPU cooler.
4. Mount the CPU cooler direct to the mosfet wall and/or use a better CPU cooler with more heat pipes and more copper in it.
5. Cut up the copper heat spreader into 6 sections and solder the mosfets to the sections.
6. Use liquid cooling.
Notes:
Number 1: Probably won't do this at all. There's already so much added aluminum, going with copper instead won't matter much.
Number 2: If/when I go with graphite insulators and cut up the copper heat spreader into 6 sections, That's when I would do this too.
Number 3:This is easy and I'll probably do that right up front by mounting a heat sink direct to the mosfet wall.
Number 4: I deliberately didn't do this due to space constraints out the side of the controller.
Number 5: We'll see...maybe later...if heat is significant or not after these other heat path mods are done and tested.
Number 6: Use a couple of liquid CPU coolers so I can cool directly at the mosfet wall and put the radiators where the CPU cooler currently mounts.
ElectricGod
10 MW
A bit more work done...
I ground a point on an M3 screw and then threaded it down each one of the screw holes that secure the mosfets to transfer the screw positions to the copper from the aluminum heat spreader. Those dimples worked perfectly for centering the 2.5mm drill bit.
The completed heat spreader...or so I thought.
I started to mount the mosfets and discovered I forgot about the small overhang on the back edge of the board.
Back to the drill press and the end mill to cut out the lower corner. I'm getting a bit tired of the craptastic finish I get on the drill press. Ah well...good enough for this job I suppose.
Perfect fit! BTW...if you ever take the mosfets loose from the heat spreader, make sure that before you power up the controller you make sure the resistance between the heat spreader and each mosfet tab is several meg ohms. If it's not, you have a problem. One or more of the mosfets is not properly isolated from the heat spreader. In my case, I crunched the edge of a screw insulator and that allowed the screw threads to touch a mosfet. If I had powered up the controller, I would have surely blown some mosfets!
All closed up and finished. I've tested out the controller and it is working perfectly. I'll add the heat sink to the mosfet wall later. That's easy and can be added at any time.
I ground a point on an M3 screw and then threaded it down each one of the screw holes that secure the mosfets to transfer the screw positions to the copper from the aluminum heat spreader. Those dimples worked perfectly for centering the 2.5mm drill bit.
The completed heat spreader...or so I thought.
I started to mount the mosfets and discovered I forgot about the small overhang on the back edge of the board.
Back to the drill press and the end mill to cut out the lower corner. I'm getting a bit tired of the craptastic finish I get on the drill press. Ah well...good enough for this job I suppose.
Perfect fit! BTW...if you ever take the mosfets loose from the heat spreader, make sure that before you power up the controller you make sure the resistance between the heat spreader and each mosfet tab is several meg ohms. If it's not, you have a problem. One or more of the mosfets is not properly isolated from the heat spreader. In my case, I crunched the edge of a screw insulator and that allowed the screw threads to touch a mosfet. If I had powered up the controller, I would have surely blown some mosfets!
All closed up and finished. I've tested out the controller and it is working perfectly. I'll add the heat sink to the mosfet wall later. That's easy and can be added at any time.
district9prawn
1 kW
Wow this is shaping up to be the mother of all controller mods. Though for the time and effort I think you could have built one of the high powered vesc variants.
How is the thermal conductivity of the ceramic to-220 insulators? I like Fechter's idea of the a copper bar directly connected to the drains and then the bar coupled to the heat sink with insulating pad.
I'm going to start building the directfet version of the a200s in a week or so and would like a better thermal interface than the usual silicone pad between fet and aluminium. The directfets really need it as they have quite poor thermal conductivity to the pcb compared to other smd fets.
Though for the time and effort I think you could have built one of the high powered vesc variants.How is the thermal conductivity of the ceramic to-220 insulators? I like Fechter's idea of the a copper bar directly connected to the drains and then the bar coupled to the heat sink with insulating pad.
I'm going to start building the directfet version of the a200s in a week or so and would like a better thermal interface than the usual silicone pad between fet and aluminium. The directfets really need it as they have quite poor thermal conductivity to the pcb compared to other smd fets.
ElectricGod
10 MW
district9prawn said:Wow this is shaping up to be the mother of all controller mods.
How is the thermal conductivity of the ceramic to-220 insulators? I like Fechter's idea of the a copper bar directly connected to the drains and then the bar coupled to the heat sink with insulating pad.
I'm going to start building the directfet version of the a200s in a week or so and would like a better thermal interface than the usual silicone pad between fet and aluminium. The directfets really need it as they have quite poor thermal conductivity to the pcb compared to other smd fets.
Yes...I know...or I have a couple of Lebowsky boards and MCU's and could built that into something epic too.
However, this was based on the PV controller on purpose...can it be done? I still don't know until I do load testing.
Ceramic is pretty good. Ceramic is way better than silicon, mica or kapton at conducting heat. Electrically it's as good as any other insulator. Graphite insulators are the only thing better than ceramic. I don't know yet if I'll need graphite as there is already loads of heat transfer capability here. All that aluminum and copper will soak up lots of heat before the mosfets get too warm.
Can you solder to the tops of each mosfet? I guess it depends on if you use plastic topped mosfets or metal topped ones.If you can solder to them, that's better than anything for heat transfer. Barring that, graphite is best, followed by ceramic. Looking at the board, 3 mosfets in a row are all drain common.
I looked at the A200S board on here:
http://teamtriforceuk.com/a200s-v2-pcb-kit/
district9prawn
1 kW
ElectricGod said:
On the vesc board in my build thread with the hsof-8 fets that is what I'm doing. There are copper bus bars on both sides of the board and the heat sink is attached to the bars under the fets. The fets themselves don't get a metal heat sink, just a crappy cover pcb on top.
ElectricGod
10 MW
district9prawn said:
Put something on top of them too...that is a heat path you are not using. Make some copper "L" shaped parts soldered to the drains and that lay on top of the mosfets. Put thermal paste between the top of each mosfet and the copper L. Now put a ceramic insulator or graphite on top of that to electrically isolate the drains from a much larger heat spreader. That's going to pull away of loads of heat...much more than that the copper busses ever will.
shaman
1 kW
Not sure if this was mentioned already, but Aluminum Nitride (AlN) insulators dominate Aluminum oxide. AlN can have a thermal conductivity of 140–180 W/m•°K.
http://www.surmet.com/technology/aln/index.php
https://www.aliexpress.com/item/AlN...lgo_pvid=bbbcc3ba-67a8-4aeb-8df7-5d5b889bc057
http://www.surmet.com/technology/aln/index.php
https://www.aliexpress.com/item/AlN...lgo_pvid=bbbcc3ba-67a8-4aeb-8df7-5d5b889bc057
ElectricGod
10 MW
shaman said:
I got aluminum oxide insulators. Still...better than anything else that's not ceramic or graphite. They will do...
I wonder how you cut a sheet of this stuff?
https://www.ebay.com/itm/180W-Aluminum-Nitride-Substrate-5-x-7-x-0-025-High-Thermal-Conductivity-LP/292959090458?hash=item4435b8f71a:g:d3kAAOSwepla-mC9:rk:1
f:0shaman
1 kW
ElectricGod said:
Tile saw?
ElectricGod
10 MW
shaman said:
That's funny! I suppose it could work, but this stuff is like 1mm thick, cut with extreme care! I guess a thin diamond wheel in some kind of saw could do the trick. Any amount of wobble in the saw arbor and you'd probably shatter the ceramic. Make sure to support it super well. Ever try to cut glass on a saw? Thermal build up in the glass will cause localized expansion and shatter the glass. I guess you'd have to make sure to keep the ceramic cool at the saw curf for the same reason.
Philaphlous
1 kW
- Joined
- Mar 28, 2017
- Messages
- 431
This controller is a beast!!! I'm about to find out just how much a nearly stock 12 Fet can handle with 2000W.... My battery can now deliver about 2400-2500W easily.
What battery are you using to power this? Hub motor? Have you changed out the hub motor phase wires?
What battery are you using to power this? Hub motor? Have you changed out the hub motor phase wires?
ElectricGod
10 MW
Philaphlous said:
This controller will initially run a C80100 outrunner on a pack capable of 200 amps at 66 volts and 20Ah. Of course the motor won't get close to that much phase amps. I already know 150 phase amps is too much for it. I'm running it at 130 phase amps where it does well without over heating. Once I see this controller runs well at around 5kw on the C80100, I'll put it on a bigger EV. That EV runs on 82v, 32Ah and a max of 320 amps. The controller will have plenty of opportunity to run at 6-7kw on an HLD inrunner. On the 18 fet in that EV, I have phase set to 200 amps continuous. How many 18 fet controllers can do that and it's not even optimized to the level of this controller.
There's no reason why results on a hub would be different from an inrunner or outrunner.
The C80100 other than better bearings and some shaft modding is 100% stock.
The HLD inrunner has the end plates opened up so I can force air through it. Otherwise it's stock.
I've been modding controllers for more power for a good while now. I bet any random 12 fet in my hands can be modded for a lot more power. However, there's no point if it's not also at least programmable. I bet the cheapest 12 fet you can find on ebay, I can double it's wattage. That's actually easy. The mosfets in it will be sooo piss poor, that just swapping in good ones will make a huge improvement in current handling.
John in CR
100 TW
If you're going to the trouble of a gigantic heat pipe heat sink, you might at as well put a fan blowing fresh air through the controller too. Not only do the back sides of the fets produce heat to dissipate, but other stuff in the controller does too. 8 years ago, when high powered controllers didn't exist for a reasonable price, I vented some controllers with great results. I was able to take plain Jane controllers to almost double the current where they blew at with 100% reliability...at least until the time I forgot to turn the fan on and blew a controller on the first slight uphill I came across.
All I'm saying is don't ignore heat from everything but the mosfets if you want reliability. Take it for what it's worth from someone who hasn't done the "walk of shame" back to the house pushing an ebike for over 6 years despite running the highest power through hubbies at with the cheapest possible controllers without all your expert mods.
All I'm saying is don't ignore heat from everything but the mosfets if you want reliability. Take it for what it's worth from someone who hasn't done the "walk of shame" back to the house pushing an ebike for over 6 years despite running the highest power through hubbies at with the cheapest possible controllers without all your expert mods.
ElectricGod
10 MW
John in CR said:
That's actually a great idea. I could put conformal all over the board so that water incursion won't matter. Then I have several 1" fans I could mount in the end plate to push air through the shell. Thanks for the idea. I've had electrolytic caps get pretty hot before and they won't last long like that.
John in CR
100 TW
I used some quite thin centrifugal fans on several vented controllers. I cannibalized them from some Bosch chargers where they were used to suck air through lithium cordless tool battery packs during charging. They didn't move a lot of air 6-10cfm if I remember correctly tracking down similar fans online, but they were tiny (30mm x 30mm x 10mm) with intake on the large face and output on a narrow side, so I was able to mount them inside the controllers with output at the rear of the controller. Then since they could draw a bit of pressure without affecting output flow much, I used some small tubing as intake to direct flow along the mosfet rail to increase flow velocity and turbulence where I thought it was most needed. For the most part, axial fans do a poor job of ventilation due to greatly reduced flow with any kind of restrictions or with components near the intake side.
This is something like what I used. https://www.ebay.com/itm/Ball-Beari...r-Cooling-Cooler-Fan-3cm-/272845545126?_ul=CR
This is something like what I used. https://www.ebay.com/itm/Ball-Beari...r-Cooling-Cooler-Fan-3cm-/272845545126?_ul=CR
ElectricGod
10 MW
John in CR said:
I have several of those fans. Fitting it inside the shell would be impossible, there's just zero room for anything vary large inside.
I'll have to do something like this. The red circles would be openings in the end cap for air to either exit or enter.
Then on the other side mount 3 small flat fans. I have probably 20 small 12v fans that could fit here that came off of North bridge coolers or whatever. They don't produce a lot of air flow, but they also don't take up a lot of space either and they would fit where I can do something. I'll have to hunt for them, but I have a couple of high CFM fans that came from a burnt out PSU. They are LOUD, but they put out some serious air flow. They might fit here, but they are 2"x2"x2".
