Improving heat transfer of ICs

[uncle_leon]

I'm working on a new heat transfer idea for my future amp and I was wondering if anyone can help me with this.

In short, I want to get rid of everything I can, that sits between the heat source and the "heatsinkable" surfaces, on ALL sides of LM3875 except the one with leads. It doesn't matter if there will be voltages present on any of the exposed surfaces.

Can anyone suggest me how much can I lap off the chip before I risk damaging the internals? I would like to also remove the integrated heatsink too, if possible, and replace it with my "thing"

Please don't ask me why I want to do this, just help me if you can I promise to share my crazy solution on the forum if it works out

[CharlieLaub]

Why don't you sacrifice one IC, do some careful lapping of the surfaces, and then report back what you find?

Have you considered using an IC package with exposed metal? I'm not sure if these are available for your chip.

-Charlie

[AndrewT]

dip you chip upside down in a bath of mercury and actively cool the mercury.

[CharlieLaub]

Originally Posted by AndrewT

dip you chip upside down in a bath of mercury and actively cool the mercury.

Great idea, except:
mercury is toxic
mercury has an appreciable vapor pressure, especially when heated, so the container needs to be sealed
mercury dissolves some metals, like gold silver, and copper.

I would NOT try that!

-Charlie

[jcx]

phase transitions carry away heat more effectively - immerse in a FluorInert liquid that boils at the temp you want to hold the device to

I wouldn't cut any metal - there is some effort put into the design of chip bonding/transistion layers to prevent thermal cycling from fatigue cracking the bonding

[CharlieLaub]

Originally Posted by jcx

phase transitions carry away heat more effectively - immerse in a FluorInert liquid that boils at the temp you want to hold the device to

I wouldn't cut any metal - there is some effort put into the design of chip bonding/transistion layers to prevent thermal cycling from fatigue cracking the bonding

This is a very good approach, except you will need to transfer the heat coming from the chip amp away from the fluid or it will all just boil away and there will be no more phase transition to take up heat. In that case, you are essentially describing something like a refrigeration cycle, although if you use a fluid that has a phase transition above room temp, you could do it passively.

-Charlie

[jcx]

sure you have to dump the heat somewhere - a point many who become enthusiatic at the now cheap availability of heat pipes from the PC cooling after market sometimes miss

in fact you could just put a air cooled condenser at the top of the fluid container and have in essence a heat pipe

[CharlieLaub]

Originally Posted by jcx

sure you have to dump the heat somewhere - a point many who become enthusiatic at the now cheap availability of heat pipes from the PC cooling after market sometimes miss

in fact you could just put a air cooled condenser at the top of the fluid container and have in essence a heat pipe

I like the sound of that. What is the fluid again????

-Charlie

[frank1]

Originally Posted by uncle_leon

In short, I want to get rid of everything I can, that sits between the heat source and the "heatsinkable" surfaces, on ALL sides of LM3875 except the one with leads. It doesn't matter if there will be voltages present on any of the exposed surfaces.

Sorry but I don't see the point...
The manufacturers assemble the "die" onto the "cooling" surface contained in the package and this is where the majority of the heat is transfered to. You won't be able to remove it from this without destroying it. If you expose the die without damaging it, you won't be able to transfer any heat from it using conduction as you will damage the bonding wires which will likely corrode and fail.

Are you considering exceeding the rated dissipation of the package?

Frank

[uncle_leon]

Thanks guys, you are giving me lots of new ideas

FluorInert is an interesting one - although it does have a number of really awkward disadvantages, which aren't advertised very widely:
- it dissolves many soft plastics and rubbers, including silicone
- due to its high dielectric strength, it accumulates static like mad - and as we all know, static anywhere near your hifi is a BAD thing
- it will be destroying ozone layer for tens, if not hundreds of years if it gets out, so must be used in a completely sealed system (which is kinda dificult to achieve with silicone and rubber seals eaten away lol)
- it's rather expensive...

After reading posts from jcx and frank1 I'm having second thoughts about the whole lapping thing actually. It seems like it would be a lot more difficult thing than taking the integrated heat-spreader off a Intel CPU (which I have done).

I have lapped about 0.5mm from the bottom and the top of the chip so far and I haven't even cut through the black plastic yet. I will probably lap a bit more anyway, just to see what's really inside my amplifier (maybe a little imp ).