Originally Posted by Reikoji
Originally Posted by kaseki
My intended point is that to have a refrigeration based cooler that is practical for a range of circumstances without big flaws like causing physical damage from condensation, some complexity is needed in the refrigeration control system. In an aerospace system of this type, control of refrigerant flow using solenoid valves and of condenser air flow using fan power control would be expected, along with attendant temperature sensors and control processing. Dynamics of CPU heat output changes would have to be accounted for in the control loops to assure that the refrigeration system's reaction time was fast enough.
For a low cost approach, it might be sufficient to maintain a reservoir of liquid that both the CPU coolant loop heat exchanger and the refrigeration heat exchanger (evaporator side) sat in, and a refrigeration loop that attempted to keep this pool at a given temperature using a simple compressor on-off control (like a typical home refrigerator/freezer). The pool volume would have to be large enough (as a thermal mass) that the two loops could operate independently and still achieve reasonable temperature stability.
hmmm, would it also be possible to obtain this result by increasing the rate of coolant flow?
I use to service a piece of equipment that serviced aircraft to keep their systems cool while tests were being performed. It worked on the same concept to cool the propylene glycol going to the aircraft, but I don't quite remember if the coolant lines ever built up condensation. I remember the evaporator froze up quite often...
The rate of coolant flow should be set to achieve the desired delta-T between the cooling block and the fluid. In the static case, no matter what that delta-T is chosen to be, the heat into the pool will be constant. And to pull that heat out of the coolant heat exchanger at a sufficient rate (for its size) to meet the allocated delta-T at the refrigerator end, the refrigerator side will have to do its thing keeping the pool cold enough. This affects the refrigerator design. Given an old refrigerator, though, some parameters are fixed. For example, a given refrigerator may be designed to pull only 50W [number picked out of the air to make a point] of heat out of the refrigerator because it expects the door to be closed most of the time. In such a case, this refrigerator won't be adequate to cool a 150W Ryzen load. Ideally, you would want to sort of match the heat load if using an old-time compressor, because those didn't like to be started and stopped every few minutes. You might need a modern variable compressor such as Perlick uses in their Signature series of bar refrigerators.
Alternatively, using a sufficiently powerful old compressor for the refrigerator side, one might want to cool down a large thermal mass to a low temperature using a long cycle time (low bandwidth loop), and then control the CPU coolant loop flow rate to keep the temperature of the post CPU fluid constant (higher bandwidth loop). All of the possible configurations would need to be evaluated for cost-effectiveness depending on one's goals. I'm imagining a need for a 24/7 folding or whatever PC that needed a solution that didn't require any fiddling once set up, and didn't incur the wrath of a spouse or parent.