Heatsink Limits

[dknguyen]

Heatsinks don't really have a power limit right? THey just keep getting htoter and hotter right? I got some bolt-on FET modules and a thermal resistance that I need for them, but the only heatisnks I can find with a low enough thermal resistance are light enough and the right size is for BGAs. The FETs themselves dissipate 10W at maximum (I have no idea if BGAs might dissipate more or less power- they might since CPUs might dissipate 100W) and the heatsink's footprint itself is not dissproportionate large to the contact area with the FET.

As long as the heatsink's about the same size as the contact area, and the thermal resistance is low enough, I can pretty much run whateve power I need through it right?

 

[crutschow]

Yes. Heatsinks are usually just a piece of metal with a temperature rise porportional to the power they must dissipate. The only limit is that at some point they will melt. But of course the semiconductor is long fried before that. The thermal resistance is the only thermal parameter that's of interest.

When calculating junction temperatures, remember that the FET has a thermal resistance between junction and case, and there is also some thermal resistance between the case and heat sink. This can be minimized by using thermal grease to mount the FET.

 

[Optikon]

Sometimes the only property of interest is the thermal resistance.

Another thing of interest might be the maximum temperature of the heatsink. If an accessible heatsink can burn some unsuspecting user if touched, that's a problem (maybe a legal one).. so sometimes there are other things of interest.. not to mention the terrible electrical efficiency one probably has with a "hot" heat sink.

 

[dknguyen]

Yeah, I'm aware of the thermal resistance. It's pretty low. Thermal tape has woudl make up 20% of the thermal resistance if I used it right now so I might jsut go with thermal epoxy. I also designed it with 100% headroom with passive airflow because I'm not sure about the airflow. But with a small fan, it's

If someone is close enough to be burned by touching the heatsink in this application they got bigger problems than burning their fingers (rotating helicopter blades!).

 

[crutschow]

Another thing of interest might be the maximum temperature of the heatsink. If an accessible heatsink can burn some unsuspecting user if touched, that's a problem (maybe a legal one).. so sometimes there are other things of interest.. not to mention the terrible electrical efficiency one probably has with a "hot" heat sink.

The temperature of the sink generally has only a small or no effect on efficiency, although it may be an indicator of poor efficiency.

For example, if it's sinking the power from a linear regulator, temperature has no effect on the regulator efficiency. That's determined simply by the ratio of the output to input voltages.

Switching regulator efficiency may be slightly effected since inductor and MOSFET resistance will go up some with temperature.

The main reason to keep the heatsink temperature down, besides the possibility of burning someone, is to keep the device temperature below it's rated limit and/or to minimize the failure rate.

 

[dknguyen]

The temperature of the sink generally has only a small or no effect on efficiency, although it may be an indicator of poor efficiency.

20W out of 3500W total ain't bad.

For example, if it's sinking the power from a linear regulator, temperature has no effect on the regulator efficiency. That's determined simply by the ratio of the output to input voltages.

Switching regulator efficiency may be slightly effected since inductor and MOSFET resistance will go up some with temperature.

I'm using worst case everything in my calculations. Maximum MOSFET resistance at maximum temperature, 50C ambient, 50% slower switching time, zero airflow.

Probably why it's running so hot in my calculations- 100C at full-load.

 

[Boncuk]

Heatsinks

I have made a lot of devices for heavy industrial environment involving high temperatures already at the construction side. A switching cabinet located in a "penthouse" made of iron sheets already has an inside air temperature of 80 deg on sunny days.

So I had to take special care of sufficient cooling. A device with a heatsink temperature of 100 deg points to 270 deg inside the chip, which will melt at 279 deg.

I always try to keep the heatsink temperature low enough not to cause burns.

When dealing with high power devices it is always a good idea to deal with extraordinary cooling too.

A good cooling device is the original CPU-fan cooler supplied with Intel CPUs. It has a copper core and the cooling surface is absolutely straight and even, which does not apply to aluminum heat sinks which require extra treatment for an even surface.

The heat conducting paste delivered with the CPU looks like having aluminum powder in it and the heat conduction abilities are 25% better than the white stuff you can purchase at electronic shops.

As a rule of thumb use heatsinks with a thermal resistance lower than 0.3 deg/W. (The Intel cooler has it!)

If chips could kiss they would do it everytime you give them a "cool life"

Hans

 

[dknguyen]

You are talking about the measured heatsink temperature vs the junction temperature right? So far this is all just calculations and I've derated most things by 50% because I just am not sure how it will all end up. The onyl way for me to really know is to fire it up and measure the increase in temperature of the heatsink.

How does a heatsink temperature of 100C work out to be 270C inside the chip (what chip withstands that in the first place?) unless the power dissipation was really realy high to begin with...like 1W/C for 170W or 1.7W/C for 100W, or 17C/W for 10W, etc.

I would stick big fans and copper, but this is a helicopter and I am limited to about 200 grams of heatsink right now. I'm not sure about the airflow exactly since it's not a plane.

 

[Boncuk]

Heatsinks

I assume that helicopter is a really flying thing. The main rotor causes a strong airstream which could partially be used for extra cooling of the electronics.

Just a small opening in the air downstream will take care of sufficient cooling, provided you take care that the heated air can escape somewhere.

I wouldn't be worried about excessive temperatures in a flying object anyway.

As for the inside temperatur in a chip you might compare the active chip surface with the size of the package. If the package gets as hot as 100 deg/C it takes very high temperatures inside the chip to develop such a tremendeous heat outside it. The mass of the chip is very small (in many cases 1mm2). Logically if there is no great amount of heat the tempeature has to be extremely high.

Try to heat up a cube of copper with the size of 5mm with your soldering iron tip (pen shaped). The copper won't even get warmer than your blood. Now transfer this to an IC and you can imagine the high temperature it has to develop to heat up the package to 100 degrees.

 

[Optikon]

The temperature of the sink generally has only a small or no effect on efficiency, although it may be an indicator of poor efficiency.

For example, if it's sinking the power from a linear regulator, temperature has no effect on the regulator efficiency. That's determined simply by the ratio of the output to input voltages.

Switching regulator efficiency may be slightly effected since inductor and MOSFET resistance will go up some with temperature.

The main reason to keep the heatsink temperature down, besides the possibility of burning someone, is to keep the device temperature below it's rated limit and/or to minimize the failure rate.

I understand this. My point is if you are burning 20W *ANYWHERE* that is 20W's that isnt doing a darn thing useful for you other than contributing to global warming. Now if it is 20W / 3500W (recently revealed by OP) - not bad! if it is 20W/30W - its terrible. So sometimes a hot heatsink is trying to tell you something.


Linear regulators have terrible efficiencies but have other advantages.

Also keep in mind that the heatsink size should only be designed around how well the part can transfer heat from its die to its case. If the thermal resistance is too high, it wont matter how many fins or copper you put on your heatsink - it wont do any good for the part die temp.

I have seen designs where the heatsink is marvelous but then the whole thing is fouled up because the interface has too high of a thermal resistance.. thermal pads can be high. Grease can be high if not done properly. Over tightening / under tightening can cause high thermal resistance.

If the OP has some thermocouples, he can measure how good the interface is.. he might get surprised.

 

[Optikon]

You are talking about the measured heatsink temperature vs the junction temperature right? So far this is all just calculations and I've derated most things by 50% because I just am not sure how it will all end up. The onyl way for me to really know is to fire it up and measure the increase in temperature of the heatsink.

How does a heatsink temperature of 100C work out to be 270C inside the chip (what chip withstands that in the first place?) unless the power dissipation was really realy high to begin with...like 1W/C for 170W or 1.7W/C for 100W, or 17C/W for 10W, etc.

I would stick big fans and copper, but this is a helicopter and I am limited to about 200 grams of heatsink right now. I'm not sure about the airflow exactly since it's not a plane.

No chip will withstand 270C at the die.. I think that's around the glass (SiO2) transition temp (die=liquid)

You should measure lots of stuff. heatsink temp is probably the least useful to know. Case temp of the parts and the delta across theinterface are probably the most revealing pieces of information.

 

[dknguyen]

I've just heard that helicopters can have a hard time cooling because they are hovering rather than always moving through the air. Not really a problem in flight though. I am unsure about the airstream right now because the bulk of the heli is beneath the hub of the of the main rotor (where the blades are moving the slowest). I was debating whether to have a canopy at all to help with that.

The thermal interface from junction to the heatsink contct area is listed as 0.25C/W. There's a 2mm plate of aluminum in the way (the chassis, MOSFET mounts to one side, heatsink mounts on the opposite side), and the heatsink is about 2C/W at 200LFM (so I'm using 4C/W assuming less airflow). And there's thermal epoxy or phase change compound in between all of them- epoxy for sure on the heasink, I am deciding whether I can find room to bolt the MOSFET on and use the phase change compound or to also use epoxy (but then I can't replace MOSFETs). I calculated that about 10-15W (per MOSFET, two are on at any one time) are going to be dissipated at full load.

Attached is my spreadsheet I made (it's actually an excel file, so replace the extension with .xls since I can't attach excel files directly).

 

[dknguyen]

You should measure lots of stuff. heatsink temp is probably the least useful to know. Case temp of the parts and the delta across theinterface are probably the most revealing pieces of information.

So should I be measuring the temperature of the plastic part of the case? Or of the contact area that is supposed to touch the heatsink (which I guess would be close to heatsink temperature if the interface was good)? I could make a test plate and drill small hole through the center of a heatsink and the plate to get a thermocouple to contact the underside of the module.

 

[Boncuk]

Heatsinks

Hi dknguyen,

it really doesn't make any difference if air moves horizontally or vertically.

The airspeed underneath a helicopter main rotor is much higher than the heli can achieve in horizontal flight.

If you leave an opening on top of the canopy the airspeed will still be high enough when hovering with the same effect of two small DC cooling fans.

Inside the helicopter you can mount several guidance plates (aerodynamically shaped and made of lightweight paper enforced with a string of epoxy) to direct the air where you need it. Don't forget the venthole underneath.

Measuring with scientific accuracy is almost impossible on an object like this.

It will be easily possible using a telemetry transmitter and the necessary transducers to connect to it. But then you'll get a gross weight problem I suppose.

Just build your stuff in and fly!

Hans

 

[Mikebits]

So should I be measuring the temperature of the plastic part of the case? Or of the contact area that is supposed to touch the heatsink (which I guess would be close to heatsink temperature if the interface was good)? I could make a test plate and drill small hole through the center of a heatsink and the plate to get a thermocouple to contact the underside of the module.

I think you want T ambient, which would be the plastic case of the part.
**broken link removed**

It is my feeling that you want some degree of confidence in your design prior to taking flight, as mid-flight failures could result in damage to expensive helo parts.

 

[dknguyen]

Yeah, I am testing the ESC inside out before it ever goes into the helicopter.

 

[Mikebits]

Your rig sounds like a lot of weight in the electronics + motor + batt. How big is this thing going to be? I am sure the rotor span will have to be quite large, something like 90 cm diameter maybe?

 

[dknguyen]

Your rig sounds like a lot of weight in the electronics + motor + batt. How big is this thing going to be? I am sure the rotor span will have to be quite large, something like 90 cm diameter maybe?

I am getting the smaller model (1.2m rotor diameter) because it is built better than the larger version which is 1.4m. BUt enough parts are swappable between the two that I can take the parts that "make the bigger on big" and install them. So 1.4m rotor diameter. Mainly because I am uncertain whether the 1.2m rotor diameter can lift my required weight which is a maximum of 13lbs. Mainly because I'm trying to squeeze in a double-sized battery pack.

THen there is the REALLY large version which is 1.8m rotor span, but it's basically a 1.4m one that has been refitted to be such, in the same way I will refit the 1.2m to be 1.4m.

I'd really prefer the 1.8m one...it's just that it's not built as well or as symmetrically as the 1.2m one since that one was designed last.

 

[Mikebits]

I bet those blades set you back a few buck. Make sure you use training sticks on your first flight.

 

[dknguyen]

Yeah, I've been debating that. Depends on how close the simulator is to real life.