This weird trick lowers DAC distortion

[mansr]

10 hours ago, pkane2001 said:

run a water cooler to remove the heat, then run the water through a bucket full of water and ice  add some alcohol to the water to stop it from freezing. Believe it or not, that works for quite a while.

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

[mansr]

10 hours ago, PeterSt said:

Try it. With really low TDP (like measuring 25C or so at the chip's surface) continuous operation on any of the chips, no continuous use would show a higher temperature. Still, running that for 24/7 will degrade THD by the ever so slightest means (like 0.002dB or so). This is audible like crazy (perhaps because of means beyond my knowledge).

You should team up with Rob Watts who "can" hear noise modulation at -300 dB.

[PeterSt]

Yea, cool.

 

13 hours ago, mansr said:

the THD was some 5 dB lower than the starting point.

 

Do you have a use case for this ?

 

Apologies for being serious. Speaking of that, is this the spray you used ?

 

 

 

[mansr]

7 hours ago, gmgraves said:

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. 

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.

[marce]

13 hours ago, mansr said:

I tried that, but I couldn't easily cool the hot side sufficiently.

Device packaging plays a big part...

[Ralf11]

So, SQ on Pluto will be better than here on Earth?  Mercury will have worse SQ...

[mansr]

39 minutes ago, Ralf11 said:

So, SQ on Pluto will be better than here on Earth?  Mercury will have worse SQ...

Mercury is pretty cold on the dark side. The lack of atmosphere might make listening a challenge, though.

[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. 

[fas42]

I just had a thought about how I would do something like this - key would having a tiny evaporative coil in the system, which could be placed near the critical part. Then a small, slow running fan with the air flow funneled as precisely as necessary to what matters, say the top of the DAC chip - this would give one great control on how much cooling was going on.

 

Some of the hardware that would do something like this is inside every, discarded laptop - just run this in the reverse of the normal heat cycle.

[semente]

22 hours ago, mansr said:

I was goofing around with a PCM1794A chip when the idea occurred to me to see what happened if I hit it with some freezer spray. Monitoring with the REW real-time analyser, I gave it a quick blast. The reported THD dropped a little. Another blast lowered it some more. I kept going. The chip got covered in ice crystals. At a temperature of about -25 °C (according to thermal camera), the THD was some 5 dB lower than the starting point.

 

The lesson here is clear. We need silent-running audiophile freezers with stylish looks and convenient cable inlets.

 

Or keep the DAC in the freezer?

[mansr]

26 minutes ago, gmgraves said:

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.

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.

[mansr]

59 minutes ago, gmgraves said:

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. 

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.

[fas42]

5 minutes 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.

 

You got impressive results when all you did was "hit it with some freezer spray" - so all you need to do is replicate that physical process by some means which can maintain it indefinitely.

[mansr]

Just now, fas42 said:

You got impressive results when all you did was "hit it with some freezer spray" - so all you need to do is replicate that physical process by some means which can maintain it indefinitely.

The freezer spray also cooled a large area of the PCB considerably.

[fas42]

Which is why I mentioned using some type of funneling arrangement, to organise air flow between the heat exchanger and the specific part - careful design of this should do the job.

 

I actually used this method for heat treating a graphics chip in a laptop a while ago - normal hair dryer, a crude cardboard cone between it and the part; the heat was concentrated to precisely where it was needed, for about 5 minutes - and did the job.

[sandyk]

[sandyk]

The above was from the DSD1792 Data Sheet.

[esldude]

5 hours ago, mansr said:

Mercury is pretty cold on the dark side. The lack of atmosphere might make listening a challenge, though.

You'll have to go with bone conduction for hearing.

[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.

[jcbenten]

Dunk the entire DAC and not just the chip: https://tinyurl.com/rjgazvl

 

 

[marce]

15 hours ago, fas42 said:

I just had a thought about how I would do something like this - key would having a tiny evaporative coil in the system, which could be placed near the critical part. Then a small, slow running fan with the air flow funneled as precisely as necessary to what matters, say the top of the DAC chip - this would give one great control on how much cooling was going on.

 

Some of the hardware that would do something like this is inside every, discarded laptop - just run this in the reverse of the normal heat cycle.

Look up heat pipes....

Been there done it😀

[marce]

14 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.

This works best with bottom terminated components or components with a heat pad, that the die is on.

[mansr]

1 minute ago, marce said:

This works best with bottom terminated components or components with a heat pad, that the die is on.

Yes, cooling works best on parts designed to be cooled. What a surprise.