Thermoelectric Cooling

[J57ltr]

Brian,

On the efficiency 2 numbers pop into my head 30% and 65%, I'll look it up I have the notes, I think here at work.

You really need a large heatsink on the "hot" side (if you reverse polarity the change sides), they are capable of 70F temperature differentials so the colder you can keep the hot side the colder it gets on the cold side.

Think of this if the air temp is 100F and you can keep the hot side at 135F then you can only get the cold side to 65F. I used large fans or water cooling for my large projects. One thing you have to deal with on electronic applications is condensation. I shorted out a few devices (was able to fix them but that is another story), just because of that. They are really cool I have a dual probe digital temp (K-type) that I tracked temps with. It’s really neat to see the condensation form and then all of a sudden it starts turning to ice all across the plate.

Jeff

[iTz Nicholas]

OK so since they won't work for cooling how about using the heat from the ESC to heat the pad up and create minute amounts of current to increase runtime. I don't think it would be anything substantial but just another idea...

[J57ltr]

It’s all about temperature differential the hot side can be hot but you need another heatsink to at least ambient temp. I had an array that was using the 6 70W units and had a ¼” aluminum plate painted black with a heatsink and muffin fan on the “cold” side and was able to charge a 12V battery at about 2.5amps in sunlight when water cooled I could get about 4.5 amps out of it. But that was with forced cooling and 6 wired in series.

They are fun to play with check out the Ferrotech link I posted there is all the tech you want there.

Jeff

[J57ltr]

Brian I couldn’t get any efficiency numbers because there are so many variables. Things like thermoconductivity of the heating and cooling sinks, losses from insulation between the hot and cold side losses from the cold side to ambient air and so on. From best I can tell from some of the units I have they are capable of pumping about 170 BTU and it takes 76 watts to do that so that comes to about 69.19% using a 127 thermocouple device.

Here are a couple of pics I took of a little setup I did a while back and fired up today on 4S a123. This one was made with a CPU heatsink and the fan attached I added a second larger fan because 1 just wasn't cutting it, the heatsink temp rose too much (about 45F above ambient and now it's about 20F above ambient). The cold side is a chunk of 3/8" aluminum 2"X3"X3/8" thick and a little baby heatsink I just stuck on it for the heck of it, it's about 1" square and 1/2" tall.

Jeff

[aqwut]

Originally Posted by BrianG

Yesp, not worth what you get IMO. There are ultimately three ways to deal with heat:

1: More efficient components.

2: Less current.

3: Heatsink with thermal mass and surface area.

Everything else is just adding complexity.

I have two Quark 125's that use the top plate (connecting the three diffs) as a heatsink and it works wonderfully without the need for any additional active cooling. The chassis is another good heatsink and gets lots of airflow. I just wish more ESCs were built similar to the Quarks in how the FETs are mounted...

You're absolutely correct.. I've tried cooling with the TEC from ebay... it was just a waste... although it's good for cooling the motor after running, but during running... it's just a waste... Although it could be good for a closed system water cooling for the bigger scale stuff.. if you could have the hot side exposed on the outside..

[J57ltr]

Not even for the big scale stuff the more transfers you have the more losses you have a well designed heatsink system will outperform even a watercooled system if you have real long run times (an hour), thermoelectrics add complexity and you are going to loose 30% right off the bat. Run off of a power supply they are great, but batteries even Li-po you are going to run out of power in a hurry. When I charged my 4S1P pack last night I put about 1.4Ah back into it and only ran it for about 25 minutes give or take 5. I was going to graph it but I didn't bother I may later today with the temps and everything so you can see how they can perform.

JEff