Can an oil cooled controller take to-220s to 75a per fet?

Karma did it, and I'm wondering why no one else has done it.

I mean, if you could push a 6 fet or 12 fet controller to it's full potential of 75a per fet (more if you do the zombiess mod) you'd be looking at as much power as a 18 fet or even a 36 fet in a much smaller package. And you could put whatever size caps you want into your custom case.

My 18 fet died, so I'm really thinking hard about doing this mod. I'm just looking for some input on calculating heating. I figure the oil will wick away the heat from the fet cases and especially from the heatsink. And then the box itself will give a much larger surface area to radiate heat. Greater surface area + great heat capacity = monster power?

i think the FET's will heat up very quickly and warm the oil around them. If the warm oil doesn't
move away and is replaced by cool oil very very quickly, the FETs will burn out.

I'm going to try a heatsink, encased in a square metal tube with a blower on one end. I think the oil
route is messy and you still need some sort of flow to carry the heat away. I don't think convection
type circulation is enough for the power level you're talking about, you really need to actively (pump or
blower) generate a flow.

You don't think the movement of the bike would be enough to stir it around (assuming it's not completely full)?

Another option would be to put a tiny fan at the bottom of the case to spurt oil everywhere.

Your idea is very good too. Allows forced air cooling, while still being waterproof. Do you have any pictures of your design?

Let's do some math and see what happens:

TO220 FETs have a thermal resistance from junction to heatsink of around 0.9 deg C per watt in total (usually they are around 0.4 deg C/W junction to case and another 0.5 deg C/W case to heatsink). Let's say the FET is an IRFB4110, with an Rdson of around 4.5 mohm at 20 deg C junction temp. Rdson increases with junction temperature at approximately 0.8% per deg C. Let's assume we start off with the FETs at 20 deg C and that the oil filled case provides an infinitely good heatsink, one that conducts all the heat away without increasing in temperature. This means that the only thing limiting the FET will be its own internal thermal resistance, together with the Rdson vs temperature characteristic.

At 75 A, the FET will be dissipating 75² x 0.0045 = 25.3 W when current starts to flow and the junction is at 20 deg C.

At this dissipation, the FET junction temperature will very quickly rise to about 20 + (25.3 x 0.9) ~ 43 deg C. At 43 deg C Rdson will have increased by about 18%, so will now be around 5.3 mohms.

With Rdson at 5.3 mohms, the FET dissipation will increase, to 75² x 0.0053 = 29.8 W.

At this new, higher dissipation, the junction temperature will rise to about 20 + (29.8 x 0.9) ~ 47 deg C. At 47 deg C Rdson will have increased by around 21.6%, so will now be around 5.5 mohms

This runaway, where temperature carries on rising as Rdson rises and increases dissipation, continues until the FET cooks internally and blows...................

Remember the above is with an infinitely good heatsink, one that can hold the case of the FET at 20 deg C. No real world heatsink, oil filling or anything else will be able to do this, or even come close, so the temperature rise at the junction will be higher in less time.

The bottom line is that these small package FETs will never be able to handle high current for more than a very short time, because their internal structure is such that they simply can't get the heat out of the package. By way of comparison, the big FET package I have sat on my desk in front of me (an IXYS VMM 650-01F) has a junction to heatsink thermal resistance of 0.12 deg C/W and an Rdson of 1.8 mohm, so can handle many times the current of a TO220 package safely.

Jeremy, I'm honored to have you respond to my post. I pay special attention to your posts as they always deliver! As is the case with this!

I'm a little bit fuzzy about the math though! The way your calculating that it seems that any amount of current would show an increasing resistance from heat that would snowball until the fet overheated, but I know this is not the case. Is there anyway to calculate max continuous current using the perfect heatsink? This way I can get a good estimate of what is the theoretical maximum continuous current I could get with a perfect cooling system.

And I guess the follow up to this is (which is a big question) why aren't people just buying tiny 6fet controllers and outfitting them with giant mosfets?

There's an equilibrium point in practice, but this varies with ambient temperature and heatsink thermal resistance and the key is to keep that equilibrium point below the maximum junction temperature.

Jeremy my man, you forgot the thermal interface pad for electrical isolation. Gotta add about 1degC/W for the thermal pad.


To test for equilibrium possibility, I like to start from the end and see if it climbs or falls (with a magical theoretical perfect heatsink that never increases in temperature).

For a 4110, on a perfect heatsink that never gets warm it looks something like this.

2.3 (for 160deg die temp) x 3.7RdsOn, 6mOhm *75x75 = 33.75W Combined Rth 0.9C/W+1C/W for 1.9C/W = 64degC deltaT between heatsink to die with a imaginary perfect heatsink that never rises a degree.

Realistically in this situation with the FET actually in a little thin aluminum ebike controller case, you're looking at 6 of these FETs shedding 34w each, 204W, you're going to have the inside surface of the case where the thermal pad contacts it reaching over 100degC, which means your FETs die is over 164degC, and you're going to make magic smoke.

This also doesn't compensate for all the conductor heat from the FET legs and traces and cap ripple V*esr heating, voltage divider heat, etc.

I see what I was missing. The squaring of current here makes a big deal in heat generation. Double the number of fets and you get half the I^2R losses.
I guess I just thought it was weird that people are running huge 18fet, 24fet, and 36fet controllers and yet only running the FETs themselves at 25% of their packages rated capacity. Seems like their is a huge room in performance. Like maybe these fets are not what we should be using, or perhaps some sort of cooling should become standard.

If we could find a way to bump that up to 50% (37.5a), that would mean a 12 fet controller could do 75a continuous. That'd be a tremendous gain.
Can anyone show the math (I believe it involves calculus) to show how the heating in this system will work at 37.5a? I get as far as 37.5^2*0.0045 = ~6w per fet * 12 fets = 72w of heat

What happens with large amounts of small FETs in parallel is current imbalances. The FET with the lowest miller voltage and lowest gate charge will accept the full current load first, the one with the largest gate charge will hold it the longest. The FET with the lowest body diode Vf will get hit with most of the flyback. The FETs located nearest to the phase lead leaving the PCB will have higher average current than those further away. At some point, the linear distance required to put say a line of 6 FETs in parallel means the ouside 4 FETs may be carrying say 30% lower average current than the inside two FETs.

For these reasons, a 36 fet is unlikely to have the same current handling of say 3 x 12 fet controllers, though the sum of it's parts on paper would indicate otherwise.

You guys are teasing me with info I don't know how to push through to an end result.

How much do you think I can push through a 4110 in a fresh air ventilated controller with the peaks only being 3-4 seconds long? I've got 6 fets on each side of each phase and successfully pushing 53A through each, though the it's really more like 40A for most of the few seconds and only 1 second at 53. I'm finding the controller case isn't getting any hotter at 316A battery side limit than at 220A, I guess because it spends so much less time at high current. Those are all battery side numbers and the phase side peak is 30% higher. The way my controller is set, current on the CA climbs to peak at peak power, so I think what each fet actually sees in motor current is roughly 53A throughout acceleration until current starts tapering off above peak power. Is 70A each realistic without too much risk (I have no equivalent controller), so I can check out 30kw input?

John

With peaks of only 3-4seconds long, likely >150amps per device.

I have the tools to test these things right now, I should do it.

liveforphysics said:

With peaks of only 3-4seconds long, likely >150amps per device.

I have the tools to test these things right now, I should do it.

Click to expand...

:lol: You just made my day. So even with some imperfect sharing of current 70-75A is realistic for the short WOT blasts as long as I give it time to cool down. The junction won't be anywhere near ambient though, even on the first blast. I'll see about getting a temperature on the heat spreading bar, but it seems like SteveO did a good job of thermal attachment, because the heat gets out to the cover at the heat sink bars so quickly. I need to measure, but I don't think I've seen much more than 80°C at the case, if that high, and the fan helped case temps tremendously once I channeled the air flow along the fets.

John

Yep, stuff that is super high powered and only needs little bursts of power can survive stupid over-driving abuse.

The previous motor on Deathbike could only sustain ~7-8hp on the dyno continuous. You could burst it about 8x higher for 5second bursts.

liveforphysics said:

Yep, stuff that is super high powered and only needs little bursts of power can survive stupid over-driving abuse.

The previous motor on Deathbike could only sustain ~7-8hp on the dyno continuous. You could burst it about 8x higher for 5second bursts.

Click to expand...

I don't think I'm to saturation yet, or at least it's only the very early stages if it is. I definitely felt the increase going from 280A to 316A. I really don't want to blow the controller though.

On the flipside, I think 60mph is above the efficient limit of these high pole/slot count motors. My ventilated Hubmonster gets it's hottest on the highway at WOT at 58mph only pulling 4-5kw, yet I was able to blast up the mountain at a much lower average speed due to the curves using much higher power, and the motor wat at exactly body temp at the top, with no warm air I could feel rising up from stator. I think there's are real place for this same motor layout made with much better and thinner lamination steel.

I really need to get one of the large diameter hubbies from the company that makes my high efficiency mini-monster...good steel and less than half the pole/slot count. They're claiming a peak efficiency of 94% at 60V. The 6 phase wires worry me though due to no controller options outside of the factory controller. Maybe we'd get lucky and it's just dual windings and dual halls, so two controllers would work. It's the same weight as your GM and bolts to the rim instead of made onto one like most of the big scooter hubbies.

John

John in CR said:

... it seems like SteveO did a good job of thermal attachment, because the heat gets out to the cover at the heat sink bars so quickly.,,

John

Click to expand...

I don't have a cue about what he does, but he modded my 12 fet controllers shunt to 100A battery current and the two of them are still alive after 2k miles of mountain trails despite getting too hot to touch almost everyday. He also made my five 18 fets pull 150A and they rock, but three of them are fried since the beginning of the year. Nevertheless, it is a little price to pay for so much power and small size, to replace or repair a controller once in a while. I'm hard on controllers, but tires cost me even more thanks to them... and Steveo

To put a word on topic. I am planning motor and controller liquid cooling with my next build, just a cooling line of Prestone running inside both and a small radiator on the fork.

Riding high power in short bursts sure help a rig to last, but cooling should make possible some longer short bursts :wink:

What happens with large amounts of small FETs in parallel is current imbalances.

Click to expand...

I've read that before and I understand it, but I don't think I fully understand the implications. Basically, it depends on "large amount." Is it 36 fets? 24 fets? 18 fets? If an 18 fet controller can't sure current well, then you might as well do smaller controller and maximize what the fets are capable of. That's kinda what I had in mind here. Too many fets, and you can't get them matched enough. Too little fets and the I^2R losses become a problem. Seems like a balance could be struck between the most number of fets that will share current well, and not produce too much heating. The advantage of cooling is it can mitigate some of the heating problems. The advantage of oil bath cooling is the huge thermal capacity it will offer. Not sure of the specific heat capacity of mineral oil compared to aluminum, but it's gotta be at least in the same range. Translation; your heatsink is now 10x the size with 5x the surface area to radiate heat.

[youtube]G_KYIhDHHMg[/youtube]
more info here: http://www.karmabike.info/Projects.htm

He say's it's a cheap $20 ebay controller rated at 65v but ran at 90v. He doesn't say how much power he's throwing at it though. It's a 1606 hub, and the simulator says it could be showing up to 5kw peaks if he put enough solder on the shunt

Hi auraslip great idea I wanted to try the same need a few of these,

400W 26A Thermoelectric Cooler, Peltier Plate.

http://www.aliexpress.com/product-fm/499981999-Free-Shipping-Brand-New-400W-26A-Thermoelectric-Cooler-Peltier-Plate-TEC-Guaranteed-100--wholesalers.html

You would reach sub zero temps with good heat sinks works for pc's.(I am going to try it with my 24 FET 4115 MOSFET 84-132V Freeway Speed LYEN Edition Controller and wouldn't mind trying with my Cromotor too big heat sink disk brake side but that's off subject) any way looks promising.

liveforphysics said:

Jeremy my man, you forgot the thermal interface pad for electrical isolation. Gotta add about 1degC/W for the thermal pad.

Click to expand...

You're right, I should have added that I'd assumed an infinitely good thermal contact between the case the the infinite heatsink.

liveforphysics said:

To test for equilibrium possibility, I like to start from the end and see if it climbs or falls (with a magical theoretical perfect heatsink that never increases in temperature).

For a 4110, on a perfect heatsink that never gets warm it looks something like this.

2.3 (for 160deg die temp) x 3.7RdsOn, 6mOhm *75x75 = 33.75W Combined Rth 0.9C/W+1C/W for 1.9C/W = 64degC deltaT between heatsink to die with a imaginary perfect heatsink that never rises a degree.

Realistically in this situation with the FET actually in a little thin aluminum ebike controller case, you're looking at 6 of these FETs shedding 34w each, 204W, you're going to have the inside surface of the case where the thermal pad contacts it reaching over 100degC, which means your FETs die is over 164degC, and you're going to make magic smoke.

This also doesn't compensate for all the conductor heat from the FET legs and traces and cap ripple V*esr heating, voltage divider heat, etc.

Click to expand...

Looks good to me.

One other thing here is that the duty cycle is pretty important practically. Pulling maximum current from the controller isn't usually a long duration event for most people, so the average current the controller handles may well be less than half the maximum. Cars and motorcycles driven "normally" on the road, for example, often average only around 10 to 20% of their maximum power.

They would suck down a lot of juice, but one of my 36fet irfb4110 controllers has a heat sink bar that would fit 3-4 of those. Thx for the link. I had no idea peltier's had gotten so powerful, much less under $30ea including shipping.
John

Olly3012 said:

Hi auraslip great idea I wanted to try the same need a few of these,

400W 26A Thermoelectric Cooler, Peltier Plate.

http://www.aliexpress.com/product-fm/499981999-Free-Shipping-Brand-New-400W-26A-Thermoelectric-Cooler-Peltier-Plate-TEC-Guaranteed-100--wholesalers.html

You would reach sub zero temps with good heat sinks works for pc's.(I am going to try it with my 24 FET 4115 MOSFET 84-132V Freeway Speed LYEN Edition Controller and wouldn't mind trying with my Cromotor too big heat sink disk brake side but that's off subject) any way looks promising.

Click to expand...

Worth remembering that Peltier heat pumps are very inefficient and make the problem of getting rid of heat far, far worse. If you look at the specs for these you'll find that to get a low temperature on the cool side you need to dump the heat that's being pumped plus the heat the device itself creates, out of the hot side. They typically run with a Carnot efficiency of only around 5 to 10%. If you use them at around their optimum working point then they typically put out around three to four times the heat on the hot side as they draw in on the cold side, and they need to work at a low delta T (the difference between the hot and cold side temperatures) to even get this good.

This means you need a very, very large heatsink, and more importantly a very good thermal path, on the hot side to get rid of the heat. It also means you are going to use a heck of a lot of battery power running the Peltier device - they suck current like there's no tomorrow.

I've experimented with them and even made a coolbox for the car with one around 20 years ago. To give you an idea of the size of heatsink needed, my DIY coolbox uses a finned bit of extruded heatsink that's around 200mm x 100mm x 40mm high, with a fan blowing air over it. Even with this I can only run the 40mm square Peltier at around 8 A before the heatsink gets too hot, as it's trying to lose around 120 W, of which only around 10 to 15 watts is actually heat being pumped from the cool box. The cold side can get to close to freezing point, but it's only shifting a small amount of heat to try and keep the inside of the well-insulated box cool. If I were to put more heat on the inside then the heatsink would quickly get too hot to work well.

If you were to fit four of the 50mm square Peltiers to a controller, then you would need to chuck nearly 1.6 kW in just to drive them. That's 1.6 kW of extra heat you have to get out of the vicinity of the controller somehow...................

Just for some real-world results, I want to point out that a couple days ago I blew up an 18-FET xie-chang that had had it's shunt pretty much turned into a regular 10g wire.
http://www.endless-sphere.com/forums/viewtopic.php?p=590587#p590587
I didn't even get 200A battery current into it, on a 9C 2807 on the back of CrazyBike2, starting it moving. I'm sure everything in the controller was already pretty warm in there, after a couple of minutes of pulse-and-glide on the street in front of my house, and a couple of pulses of >150A battery current.

Unfortunately I have no idea how much phase current went thru it, but that unequal sharing certainly showed it's ugly head in the failure: A single FET in one half of a bridge in one phase failed shorted gate-source-drain, about an ohm, and and kept the controller from being able to do anything else. I guess that failure mode saved the rsst of it. :lol:

Anyhow, the controller's shunt has been returned to something close to normal, so now it only pulls about 80A battery current with a stalled motor. Not yet road tested.

Would like to be able to measure phase currents, but haven't worked out how I could do that with the equipment I already have, and without writing code to do it. (i r not a programmer, though I do have some MCU devkit stuff for STM8 and STM32)


Regarding getting rid of the heat, I think one of the simplest ways to help a lot is simply forcing air down the "tunnel" of the typical ebike controller. Perhaps even better is a tunnel heatsink like Lowracer used. If the FETs can all be mounted on separate heatsinks (per half of each bridge) so that you need no electrical insulation between the FET and the heatsink, and do this heatsinking directly to the "outside" so that there are no "mounting bars" that then have to be coupled to a case for a heatsink, you've got a pretty good thermal path that takes away two levels of thermal barriers.

Beyond that, getting into larger packages with larger dies in them and better/bigger bonding wires is probably the next step up. But that would require an entirely new power section design, and couldn't be easily done or modified on an existing controller by the average ebike "overclocker", unlike the heatsink creation/modification, which requires only minimal mechanical skills and following basic directions from whoever does it (successfully) first.

amberwolf said:

the heatsink creation/modification, which requires only minimal mechanical skills and following basic directions from whoever does it (successfully) first.

Click to expand...

I will try this in about a month (when I get my controller PCB). I have in my head plans for a heatsink which can be build
with simple tools, for use with a blower. I want to try to get rid of about 130W of dissipation...

I'll post pictures of the build of the complete controller in a few weeks

My idea is to have the peltiers hooked up to the battery charger so they only run before a ride and don't waste battery power. With the increased thermal capacity of the oil, starting a ride at 0 Celsius might be a big help. Especially if it's only a short race.

You could also do the same thing with an oil filled hub motor.

AW, the problem with using air is that it's very complicated to do for someone like me if you want to keep it water proof. Where as oil is as simple installing the wire junction in the box, bolting the controller down, and filling it up with oil. Maybe an hours worth of work.

auraslip said:

My idea is to have the peltiers hooked up to the battery charger so they only run before a ride and don't waste battery power. With the increased thermal capacity of the oil, starting a ride at 0 Celsius might be a big help. Especially if it's only a short race.

You could also do the same thing with an oil filled hub motor.

AW, the problem with using air is that it's very complicated to do for someone like me if you want to keep it water proof. Where as oil is as simple installing the wire junction in the box, bolting the controller down, and filling it up with oil. Maybe an hours worth of work.

Click to expand...

You still have to have a very, very large heatsink on the Peltiers. If you used four of these things as John suggested, then you've got to dump 1.6 kW of heat just from the Peltiers, plus whatever they pump out of the controller when cooling it down. That heat has to go somewhere, preferably somewhere a long way from the controller because you don't want the waste heat warming it up as you're trying to cool it down.

If you want to pre-cool a controller it's be far simpler to just put an ice pack on it.

A cpu water block and radiator would get rid of heat and them self don't use much extra energy like here

http://www.youtube.com/watch?v=RcKiEUd2zDc&feature=related

and can be quiet small and compact that's what I will use for my controller.