Cool running MOSFETS

[ding]

While I was supposed to be concentrating on something entirely different (work) I had a revelation/really stupid idea, you be the judge.

I am having problems with sinking my FETS and after taking measurements with a thermocouple surface temp probe I did the calcs and decided the problem was not the heatsink itself but rather the case/sink interface.

MY IDEA:

Solder the case to a thin plate (1-3mm) of aluminium/copper then mount the copper plate complete with FETS to the heatsink. The case plate would effectively become part of the case with little to no difference in thermal resistance.

This would have 2 benefits:
1) The area over which the conduction between case and sink is much larger. This is of benefit because heat dissipation (W) is inversely proportional to surface area. Therefore a doubling of contact area of the case would double the heat dissipation ability of the case. This is why TO-3 cases are better than TO-220 cases.

2) Secondly the clamping force between plate and sink can be greater than the maximum clamping force of the mosfet and only limited by the method of fastening. Increasing the clamping force also decreases the thermal resistance. Note: mica/kapton insulating material and silicon grease is still required between the plate and the sink.

The only problem is will the heat of soldering destroy the FET? I have a couple of FETS which I am willing to sacrifice in the name of audio science and will experiment this weekend. I think I will get the plate up to soldering temperature in the oven and sink the FET as much as possiblem, then apply the solder and place the FET in a pool of solder. After the bond has taken i will cool the whole thing as quickly as possible.

Does anyone have any thoughts/suggestions?

Cheers

Dan

[Edited by ding on 06-07-2001 at 11:59 PM]

 

[Petter]

Many people mount on to a piece of copper so that the electrically isolating interface can be made bigger. Soldering should not be necessary.

Disadvantage: Increased input capacitance

Petter

 

[Eric]

Output transistor capacitance

Petter: I've been wondering about the capacitance of mounting transistors to heatsinks, and it doesn't really seem to be a big deal. Consider the following:

If the capacitance turned out to be 100 picofarads (pF) and the effective impedance at the case of the transistor (this is determined mostly by the circuit) is 1 ohm, the RC time constant would be 100 picoseconds. That would translate to a
"cutoff " frequency of 1.6 GHz. IS this really worth being concerned about? Or have I missed something else?

 

[GRollins]

Dan,
Plan on using copper for your experiments. It's well nigh impossible to solder to aluminum. I would advise against cooling too quickly, as mechanical stress could pop your transistor loose.
Good luck.

Grey

 

[Asen]

May be the solution is a bigger ceramic plate instead of copper. I've seen such plates which contact with the uper side of TO220 too (not only the bottom).

 

[Gregs]

Increasing heat sinking of power devices

Hi,

Off subject a bit, but I have been trying to add to the thermal safety margin of my MOSFETS by putting a small (T0220) heatsinks on the front of the devices in addition to the usual heatsinks on the back side. Said another way, devices have the usual large aluminum heat sink in the usual location on the back side of the device, And has the smaller heat sink acting as a very large washer for the screw on the front side.

I realize devices are made so best heat transfer is from the back side. However the devices get hot on the front side as well (finger test proves this beyond doubt!). Believe this helps some, have no idea how much. As small heat sinks are avaible for next to nothing from surplus sources, costs little and makes me feel beter. If anyone has any ideas how much, or little, this adds to the cooling, would be interested.

Personally my motto for DIY projects is: If it is worth doing, it is worth overdoing.

Regards,
Greg

 

[Petter]

Importance of input capacitance

I don't know how much mounting the device on a piece of metal without isolation has in terms of numbers. I only know it becomes higher.

Now, if you have a Pass X1000 with 40 devices being driven by each half of the circuit, you take the current capacity of the input stage and construct an RC circuit. Remember you have to drive every device rail to rail (and then some). In this case it is important, in other cases, it may have noe effect, it might even help if you can use fewer devices.

Petter

 

[ding]

It works

I tried my idea without the soldering (petters idea). I ended up using aluminium instead of copper as it is rather difficult to get (I am still working on getting some). The arrangement I used was like this

- transister
- silicone grease
- aluminium plate
- silicone grease
- mica insulator
- silicone grease
- heatsink

This arrangement (not optimised yet) yields a Rcs of between 0.5 and 1 degC/W this is a vast improvement over what I had before (greater than 2 degC/W). I intend to find some copper (the thinner the better) and mount the transisters and sheet copper with a channel section of aluminium and 2 tensioning fasteners to give an even pressure distribution (as shown on elliot sounds web site).

I believe these steps will further reduce the thermal resistance of the case to sink interface hopefully to about 0.5 degC/W. This will give me a much greater safety margin on my mosfets and hence solve the problem.

Thanks for the help everyone

Cheers

Dan

 

[Petter]

Thinner is not better

There will be an optimal thickness of your piece of copper based on the application (transistor type, thermal conductivity of "mica" -- remember there needs to be lateral heatflow for this to function).

The good news, however is that erring on the thick side is unlikely to cause much of a negative effect -- the thicker the piece, the larger the contact area you can use effectively.

Petter