> Thanks for everyone's suggestions. > > For the CPUs on our old payload, we used heat straps straight to the > gondola frame (a very large heatsink). > A giant metal (or graphite/carbon foam) plate that attaches to the ROACH > and is then attached to the gondola frame would work. > Another thing to note is that if only the FPGA and PowerPC are dissipating > large power loads, and everything else can be passively cooled through the > ground plane of the ROACH, a much smaller heat strap can be made to both. > I'm unsure if there's anything else that needs a large amount of cooling > at > float altitude (125,000 ft). I think there must be a list of power > dissipation in each component.
At a minimum, the memories, the network PHY chips, and the power supply modules. There may be a few other bits as well. You can take the Bill of Materials and look through the data sheets. I've seen some of the liquid-cooling schemes used in computers lately, and the ones that fully immerse the board in cooling fluid seem to work pretty well, so that's certainly an option. Folks who sell pre-packaged liquid-cooled servers use standard Supermicro motherboards and other off the shelf parts, obviously without spinning disks... A roach should survive in there. I think Jason's passive cooling plate is also a good idea. You can use chunks of thermal foam (I forget who makes it) to thermally bond and stress-relieve between the plate and the parts. The power supply boards might be difficult to make contact with, with the tall inductors and the surface mount power transistors... You could cool the plate with liquid from the back, especially where the FPGA, power supplies and the memories are. But as Steve suggested, it may well be easier and lighter to build a nice aluminum pressure vessel to hold the computing equipment than to try to design conduction cooling for all of it. The pressure vessel itself could be heat-sinked to the gondola and cooling fins put on the inside. It wouldn't need to hold much pressure, so it could be made pretty thin and light. John > > > On Tue, Apr 16, 2013 at 7:28 AM, dana whitlow <[email protected]> wrote: > >> It seems to me there are two problems here: >> >> 1. Getting rid of the heat altogether, and from the comments I've seen, >> that will >> apparently need to be by radiation. "In space, no one can..." >> >> Merely dunking things in liquid will just get you hot liquid. And >> I'd >> be concerned >> about the chip packages' leak integrity. >> >> 2. Also you need to carry the heat from source to the radiator without >> excessive >> temperature difference. >> >> My point is, don't forget the radiator. Then ask, perhaps, "Can I just >> attach the >> chips to the radiator?". Perhaps this is already what Jason is getting >> at, but he >> didn't say how his "heatsink" was going to dispose of the heat. >> >> Dana Whitlow >> Arecibo Observatory. >> >> >> >> On 4/16/2013 7:03 AM, Francois Kapp wrote: >> >> Hi Brad, >> >> Given it's a balloon payload, what are your mass constraints? >> Conductive cooling is going to involve throwing metal at the problem, >> possibly a lot of metal... >> >> -Francois >> >> >> On Tue, Apr 16, 2013 at 12:57 PM, Steve Maher >> <[email protected]>wrote: >> >>> Heatpipes and dunking <http://www.pugetsystems.com/submerged.php>in >>> liquid (e.g. mineral oil) are options we've explored (with PCs), >>> although >>> we seem to be favoring pressure vessels at this point. >>> >>> Steve >>> >>> >>> On Tue, Apr 16, 2013 at 2:34 AM, Jason Manley <[email protected]> >>> wrote: >>> >>>> I don't know of anyone who's actually done this, but we do also have a >>>> passive-cooled concept for our new digitiser board (loosely based on >>>> ROACH2). We've been thinking about a prototype that'd essentially be >>>> turning a ROACH2 board upside-down and bolting it onto a giant metal >>>> plate >>>> that's been appropriately milled out of a solid chunk of alu. The >>>> FPGA, >>>> PPC, QDR and other hot bits would make contact with the plate and use >>>> it as >>>> heatsink. The other high-profile parts (RAM DIMM, for example) needs >>>> holes >>>> or cavities milled out the alu. >>>> >>>> I think Vaughan (cc'd here) might have some mechanical drawings of >>>> ROACH2 with the holes and things in the appropriate place. It might >>>> help >>>> you get started. He could also provide some CFD models and heat load >>>> measurements of the various components. >>>> >>>> Jason Manley >>>> DSP Specialist >>>> SKA-SA >>>> >>>> Cell: +27 82 662 7726 <%2B27%2082%20662%207726> >>>> Work: +27 21 506 7300 <%2B27%2021%20506%207300> >>>> >>>> On 15 Apr 2013, at 20:29, Brad Dober wrote: >>>> >>>> > Dear Casper Collaboration Members, >>>> > >>>> > My name is Brad Dober, and I'm a grad student at UPenn working on a >>>> ROACH-based MKID readout for a balloon payload. At our float altitude, >>>> air-cooling the ROACH and ADC/DACs isn't feasible Has any other group >>>> looked into or even designed a conduction-cooled system for .either >>>> the >>>> ROACH1 or ROACH2? >>>> > >>>> > Any leads would be extremely helpful. >>>> > >>>> > Sincerely, >>>> > >>>> > Brad Dober >>>> > Ph.D. Candidate >>>> > Department of Physics and Astronomy >>>> > University of Pennsylvania >>>> > Cell: 262-949-4668 >>>> >>>> >>>> >>> >> >> >> -- >> Francois Kapp >> >> Sub-system Manager >> Digital Back End >> meerKAT >> >> SKA South Africa >> Third Floor >> The Park >> Park Road (off Alexandra Road) >> Pinelands >> 7405 >> Western Cape >> South Africa >> >> Latitude: -33.94329 (South); Longitude: 18.48945 (East). >> >> (p) +27 (0)21 506 7300 >> (p) +27 (0)21 506 7360 (direct) >> (f) +27 (0)21 506 7375 >> (m) +27 (0)82 787 8407 >> >> >> > > > -- > Brad Dober > Ph.D. Candidate > Department of Physics and Astronomy > University of Pennsylvania > Cell: 262-949-4668 >
