Controller Cooling - Why Not?

[mwkeefer]

Hello All,

I'm about to test a theory I have about seriously cooling controllers for eBikes (and probably killing an IRFB4110 EB809 at the same time).

After testing the dielectric properties (don't ask) of several synthetic motor oils, then testing them against things like epoxy I've used to anchor some components (100v caps, etc) for corrosive action and after all has gone surprising well... I have replaced the end gaskets which these units come equipped with with (of all things) blank Header Gasket Metal / Cork cut to fit and then some to ensure a fair seal.

Finally I didn't want to mold the wiring passing through the controller yet to water proof and make it leak proof so I've just run it upside down - all the OIL I'm using (off the shelf, expensive synthetic) was added through a tapped screw hole /. filler hole next to the wire loom and I've filled it to about 96% which en covers the entire PCB, the FET RAIL, etc.

So far (it's RAINING, Damn IT) things seem well, I've applied power, turned on ignition and thrown it on a GNG Gen2 BB Drive Dahon on the Bench and spun things up but I can't go out and load test it.... Even when I can, the G2 BB drive can't really handle too much power but the next step is to run it at 18S (10G loop on second 330 resistor to engage it - this enabled dynamic change from 10-12S into 13S-20S operations...)

Thought I'd test it on the BB Mount first because I use quite a lot of partial throttle and have some base temperature readings for them in various ambient temperatures so I can compare new model (enhanced) to existing specs - I would hope to see the controller appear to run hotter at the case indicating better thermal dissipation.

If this all goes well I will be having CNC machined, re-designed enclosures for all the popular xie chang controllers, lyen's PCBs, Methods controllers. First batch will be 6S and 12S and available for EB8 PCB, EB2 PCB and the newest EB3 series PCBs.

With how I have my CNC machine work done, if you need a custom enclosure (providing this hold up as intended) I will take shipment of custom controllers and design custom enclosures for the PCB / Heat Transfer Bus and return the controller with the PCB mounted and tested in the new CNC liquid cooled controller.

Anyone have some thoughts?

-Mike

 

[bowlofsalad]

Thermodynamics is a subject that interests me greatly. I have done some prototyping with some concepts that sound similar to your own (involving dielectric fluids). There is a lot to consider with various fluids, one idea is how various fluids will effect things like various kinds of rubbers. This is why a lot of the time in this types of applications they used silicone transformer oil.

While there are pros and cons to all things we consider, I think that in a mobile application, filling the controller, especially with the number of wires coming out of the controller, would be less than ideal. If I really needed to design a controller with substantially higher cooling capacity, I would look to the PC industry. I think your best bet would be to create as close to solid surface contact area as possible between the areas of heat and the surface of the controller. Between those two surfaces, place what is known as thermal grease, thermal compound, thermal adhesive or thermal pads. Look into CPU coolers, imagine using one of those without the fans, maybe place it upon the controller case.

Some add some kind of oil to their hub motor for a similar effect that you are going for, but they don't fill the hub full of oil, maybe only 1/3rd of the way full. Perhaps you could take a similar approach, I imagine the controller sitting upside down and flat or parallel with the ground (most of the time), the heat would go from the caps to the surface of the controller, on the top of the controller, as I suggested, you could place a CPU cooler. The benefit of not filling the controller to the brim is obvious, but worrying about having absolutely perfect seals everywhere on the controller seems like it might be difficult. There is also another idea, I think the term used was wicking. Somehow, the fluid makes it's was through or over the cables coming out of whatever fluid or oil it sits in and ends up slowly leaking all over. I imagine this making an unpleasant mess on your bike.

Even with perfect heat conduction to the controller case, I imagine that in some situations the surface area on a controller won't be enough. The many fins on a CPU cooler will increase your surface area for cooling by lots. Aiming to place the cpu cooler around the center of the heat would probably suck most of the heat out from the controller.

I don't know what kind of heat conduction goes on in a controller, but if the ideas I suggest aren't being implemented, these should help lots.

 

[Sancho's Horse]

I believe MWKeefer was one of the first...if not the first...to fill a hub motor with oil on this forum. So...

Thanks for all the testing you are doing. I am really interested in what you find. I have been working on some oil cooling of a hub motor, and the one detail I cannot seem to pin down is pressure equalization. I found acceptable valves, but no acceptable prices...so, I was leaning towards routing the oil overflow/pressure equalization through the controller. I probably would not seal the controller as well as you, and I am leaning towards synthetic atf, but I am really looking forward to what you are doing.

 

[Chalo]

Anyone have some thoughts?

It seems pretty obvious to me that the upper bound on power transmitted through a FET is not limited by heat evacuation from the heat sink, but by the rate at which heat can be moved from the transistor junction to its backing. Exceed that rate, and it won't matter whether the heat sink is embedded in a block of ice. You'll still burn the FET.

If heat buildup is a problem, then improved cooling could surely help. But forced air seems like a much better, more direct approach than liquid-to-air cooling. To me, all this liquid cooling nonsense folks try to impose on electrical systems looks like yet another manifestation of moto-cult mental illness.

 

[Hillhater]

Remember this....
http://endless-sphere.com/forums/viewtopic.php?f=3&t=40549&hilit=oil+cooled+controller

 

[full-throttle]

Mike,

Thermal conductivity of synthetic oil is similar to Kapton tape ~0.12W/(m*K) so I wouldn't expect much gain from using it.

Once I've got some spare time I'm going to experiment with Loctite 315 - a self shimming adhesive. Not only it's thermal conductivity is nearly 7x the one of Kapton, it also fills all the microscopic gaps between the FET and the heatsink. The downside of 315 is the 5-6mil thickness comp to 2.7mil (total) of the thinnest Kapton tape.

 

[bowlofsalad]

Mike,

Thermal conductivity of synthetic oil is similar to Kapton tape ~0.12W/(m*K) so I wouldn't expect much gain from using it.

Once I've got some spare time I'm going to experiment with Loctite 315 - a self shimming adhesive. Not only it's thermal conductivity is nearly 7x the one of Kapton, it also fills all the microscopic gaps between the FET and the heatsink. The downside of 315 is the 5-6mil thickness comp to 2.7mil (total) of the thinnest Kapton tape.

Why would you use loctite over the thermal adhesive they use in the PC industry?

 

[full-throttle]

Dielectric strength of 500V

 

[John in CR]

Why oil instead of a potting compound? The only benefit I can think of for oil would be the lack of stress on parts from different rates of thermal expansion. Are you going to eliminate the wires or use solid core to keep oil from seeping through the wires?

 

[cal3thousand]

Seems like liquid is a bad idea. Look at the high power PC guys for the answers.

Air cooling with proper thermal paths are the ticket. If you do get into liquid cooling, consider running tubes and not just fluid sloshing about. I have a hard time seeing uncontrolled fluid doing well with sensitive/high power electronics.