This weird trick lowers DAC distortion

[gmgraves]

4 hours ago, pkane2001 said:

 

Here you go: https://hothardware.com/news/evga-closed-loop-liquid-nitrogen-cooling-system

 

I don’t doubt it, as a former semiconductor engineer, I have seen experiments where microprocessors were subjected to cryogenic temperatures. They always ran faster, and more efficiently, the colder the chips became. In all the experiments to which I was savvy, the cold was applied to the bare die, not to the encapsulated IC. While a cooling system using a refrigerant is somewhat impractical, we could use one of those semiconductor “heat pumps”, the kind used in portable electric coolers, and which are cold on one side and hot on the other. I don’t believe that it would work unless the cold side of the semiconductor heat pump were in direct contact with the DAC die. 

[gmgraves]

8 hours ago, mansr said:

This case is a bit different since the chip is only producing a small amount of heat, less than 0.5 W. The problem is that to cool it below ambient temperature, you're fighting against inflow of heat from the PCB, which is made of copper and readily conducts heat from the air around it. In the end, you're just cycling a bunch of heat through the DAC chip without actually making anything much colder.

But you are cooling the die below ambient and certainly below it’s normal operating temperature. So we’re still, essentially talking about the same phenomenon.

[gmgraves]

9 hours ago, mansr said:

I have a boxful of PC water cooling stuff. The hard part is mounting the water block on the 10x6 mm chip.

Processor water cooling will not be cold enough to make any appreciable change in the distortion of a DAC chip I’m afraid. In the case of the water-cooled  microprocessor, the cooling is used to keep an over-clocked chip from self destructing due to self-generated heat. 

[gmgraves]

7 hours ago, mansr said:

It would work to cool the hot side of the Peltier element. Alternatively, it could be attached to the backside of the PCB, giving the Peltier cooler a better starting point.

That’s true, the backside of the Peltier element does need to dump the heat removed from the cold side, the liquid cooling would do that. Thanks, I’d forgotten that those solid state coolers are called Peltier elements.

[gmgraves]

7 hours ago, mansr said:

Yes, there will be an area of lower temperature by the cold side of the Peltier element. How much lower and how far it extends depends on the thermal resistance of the various parts. To achieve any significant effect, you need to have high thermal resistance between the die and everything other than the cooling element. That isn't the case here, so cooling the chip is a bit like trying to create a vacuum in a leaky container.

Yes, that is correct.