On Wed, Sep 06, 2017 at 11:49:57PM -0400, Alvin Starr via talk wrote:
There will need to be a real quality DC-DC converter because there is not much point converting the DC output from the panel to AC only to convert it back to DC.
Possibly DC to 12V to some batteries or ultra-capacitors to deal with transients or possibly to provide run time extension. (more on this later)
I believe a lot of telco stuff runs 48V DC. Doesn't need quite as insanely thick wires as 12V and avoids having slight voltage losses drop you below the 12V that is very much required by the GPU and CPU. And there ought to be a lot of 48V to PC power supplies on the market already. Like these: https://www.powerstream.com/DC-PC-48V.htm Given their input range, perhaps one could aim for 60V instead so that dips still stay above the minimum 36V, while still being way below the 72V top allowance.
I worked on passively cooled avionics projects in the past and its doable but somewhat complex when you have to start embedding heat transfer sheets to your PC boards.
The enclosures would need to be sealed and just have a set of wires going to the DC from the panel and some coax going to an antenna.
Sealed is not great for getting rid of heat.
I was not thinking so much about deploying in Canada. We have way too much dark. It feels like 16 hours of dark a day from October to June with the rest of the days being cloudy ;)
Well good, because it does seem crazy here (snow sucks for solar panels too I would imagine).
Initial my thought was pick a nice sunny island in some equatorial area and use the local ocean for cooling. You would not have to pump much and so long as the GPU's are running less than 50C or so life would be good.
Another thought was. Why not float them on a lake or pool of water and let passive cooling take place.
Passive air cooling may not work in truly hot environments.
As a data point Ethereum mining with the latest AMD GPU can earn about $6K a year in ETH if you run 7x24. That's with hardware costing less than $1000. Then you have power, cooling, housing and network access.
Some of these are dumb ideas and some are just engineering challenges the hard part is to figure out which is which.
Remember not long ago there was no way that you could build a rocket to take off and land after putting stuff in orbit. That was just the dreams of SF authors.
Certainly. 16 landings so far I believe. -- Len Sorensen