-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Heat must be removed by liquid coolant. There's no point in going through 2 conversion steps, when 1 would have done just fine. I'd imagine the cooling block(s) in the inverter is barely adequate. It's probably just single pieces of aluminum with a single channel through them. Even lowering the input temperature won't help the heat capacity that much without redesigning the cooling block. The quickest solution here is probably
1. Order large block of copper, get time on a mill.
2. Source some part that will fit like http://www.customthermoelectric.com/Water_blocks/images/WBA-3.0-0.85-CU-01_open_1000.jpg
The thing is, a heat pipe can move something like 80 times the energy that a piece of copper of similar size can - so the two stage conversion lets me build a much bigger heat exchanger to transfer heat into the water, safely separated from the chips themselves.
The real question is whether you can get more heat out of the chip into the pipe - maybe you need something exotic like graphene to get the energy off the chips faster? I think this approach will be limited by the rate of heat transfer from the silicon itself into the cooling system rather than by any part of the system itself.
Immersion Cooling - 3M Novec United States
Using something like this could help to solve it. Essentially you seal the inverter after applying a boiling enhancement coating to the chips the place a coil pack inside if you want to be able to actively recondense the fluid or you could integrate a heatsink on the inside and chill the heatsink without penetrating the case.
This type of cooling is used in military applications like compact radar systems and the like. Heat pipes are using this same principle but you have the conductive losses of the copper along with the more limited surface area for heat transfer.
Of course any of these additions or added cooling solutions add cost and complexity in order to add a capability that is only really useful if you are tracking or racing. So in other words highly unlikely to be implemented.
Using something like this could help to solve it. Essentially you seal the inverter after applying a boiling enhancement coating to the chips the place a coil pack inside if you want to be able to actively recondense the fluid or you could integrate a heatsink on the inside and chill the heatsink without penetrating the case.
This type of cooling is used in military applications like compact radar systems and the like. Heat pipes are using this same principle but you have the conductive losses of the copper along with the more limited surface area for heat transfer.
Of course any of these additions or added cooling solutions add cost and complexity in order to add a capability that is only really useful if you are tracking or racing. So in other words highly unlikely to be implemented.
A heat pipe works by evaporating liquid at the hot side. Might as well make that liquid moving at high rate and circulating into a radiator.
Why reach for something exotic when the capacity of what's there just needs to be improved?
Won't work for that much heat. Like I said above: Don't need anything exotic. Just optimize the existing path. Not like there's been a large community of people try to extreme cool integrated circuits for 20 years or anything
Won't work for that much heat. Like I said above: Don't need anything exotic. Just optimize the existing path. Not like there's been a large community of people try to extreme cool integrated circuits for 20 years or anything
It's not all that exotic... It's been used in closed systems, zero moving parts, for decades. The AWACS systems for instance utilize this. That link is just an example and happens to be a newer non flouronized blend that is meant to be more eco friendly and used in data centers. All the fluid does is increase the amount of surface area on the chip that can be used to remove heat. How you deal with it after that could be anything.
Question is how do we know the existing path is not already optimized? Seems like an oversight I would not expect from Tesla. Wk may know if he has had it open.
In any case it's unlikely to change because the current design does exactly what Tesla wants it to do which is meet nearly all uses of the car. It's more likely to see a new high performance inverter integrated into the next roadster.
SomeJoe7777
Marginally-Known Member
The evaporation is the key. For most fluids the specific heat capacity (this is the number of Joules of energy that 1 gram of the substance can absorb by raising the temperature of it by 1 degree C) is far less than the heat of vaporization (this is the number of Joules of energy that 1 gram of the substance can absorb when it changes phase from liquid to vapor).
For water, the specific heat capacity is 4.186 J/gC. e.g. If you flow 100 g of water over a surface and the 100g of water is raised by the heat of that surface by 1 degree C, the water carried away 418.6 J of energy.
Water's heat of vaporization is 2260 J/g. If you allow 100 g of water to boil to steam, the water carried away 226,000 J of energy, over 500 times more.
In a cooling system, high liquid flow rates preclude phase changes. Pumping the coolant through faster ensures the coolant stays in a liquid subcooled state and minimizes temperature gradients across the object being cooled, but you will never be able to take advantage of the heat of vaporization if the cooling system is designed this way.
For this application (inverter cooling), heat pipes would actually be ideal because a big limitation is the packaging. Having heat pipes carry the heat away very efficiently to a place where a much larger surface area can be exposed to the forced circulation coolant would significantly increase the ability to cool the inverter.
You're claiming that a liquid flow can't move 10, 100, or even 500 times as much mass of water by the heat spreader as could possibly be cycled into vapor and back within a heatpipe? Not to mention the fact that the Model S inverter has the peak capacity that it does already necessitates conduction into the heatspreader that is good enough, or they would quickly release their magic smoke.
Your argument has just about the same merit as space solar power. Yeah great it's super efficient in space, but just not a bright idea. har har.
SomeJoe7777
Marginally-Known Member
Ummm ... I made no such claim.
Of course you can remove the same heat with forced coolant flow ... provided you have enough flow rate and enough surface area to reduce the differential temperature required. And that is exactly the point ... the piping size, differential pressures, and differential temperatures required to remove the same heat as a heat pipe are too large to fit inside the drive unit packaging.
Thus, use heat pipes to move the heat outside of the volume-constrained interior of the drive unit where you can then have a forced coolant heat exchanger that is larger.
Also, you seem to believe that there is some sort of barrier to the deployment of a heat pipe in this situation. Is it because you think they're rare or expensive or theoretical? You do know that there are dozens of aftermarket CPU coolers for sale for home-built PCs that all use heat pipes and are under $100, right?
I'm not at all seeing how your "not a bright idea har har" comment applies here. Are we serious in talking about thermodynamics, or are you just blowing off the reality of the science because you feel like it?
You sound like an ICE driver who doesn't want to listen to the Tesla owner.
You're missing the point entirely, being fixated on a single mechanism instead of the big picture. Show me a commercial example of a heatpipe being used as an intermediary heat spreader between a heat source and liquid cooling.
Low thermal mass TO-247 package survives something as long as a run to 155 mph, heat transfer into the heat spreader seems to be working well enough. If you removed all cooling those things would literally evaporate in a fraction of a second. Even at 98% efficiency, there's over a 100 watts coming out of each package. Given they are TO-247's that's actually less heat flux than out of many modern CPU's. Actually you could go to something on the order of 2-4 times that while being equivalent to retail CPU's that people are using normal liquid cooling to keep near ambient temp.
artsci
Sponsor
discrete IGBT's in ms/x are liquid cooled. can only move so much heat with the surface area of the existing heat exchangers. Changing the pump speed will actually reduce heat transfer. May be possible to replace the nose cone with a larger radiator to dissipate the heat.
Then again the IR IGBT's that is on the Model S are several years old and you could likely get the same current throughput with a lower switching igbt.
Then again the IR IGBT's that is on the Model S are several years old and you could likely get the same current throughput with a lower switching igbt.
Similar threads
