Page 1 of 1

Big Heatsinks

PostPosted: Wed Aug 13, 2014 3:11 pm
by theover
Already pretty early on after I finally received my Parallella 7010 board and accessories, I tried to put on more cooling area for the Zynq. I took a little thermal grease from my CNPS9900DS accessory tube, and put some extra aluminum on the supplied little heat sink.

Some may guess what can happen, as Murphy's law states: it did happen. The extra heat sink got a bit shifted to one side and somehow connected so supply pins on the board. So the main (smd, IIRC it was called Z1 in the schematics) fuse blew and the board was dark.

Luckily only the fuse had blown, without any effect of the board or how Linux and the Epiphany functioned, after I had replaced the SMD 3 amp fuse, I put a little arc of one strand of medium current wire over it's place, so heavy short circuits should still be fused. So the board is luckily exactly the same as it was before, but it's a warning when playing with large (electrically conductive) heatsinks: put some isolation tape on the edges that are likely to make contact with the board and it's parts to prevent trouble !


Re: Big Heatsinks

PostPosted: Thu Aug 14, 2014 12:18 am
by aolofsson
Amazing story! Impressed that you were able to bring the board back from the dead! (and glad that not all shorts kill the board permanently). is your view that we should have a better fuse solution for the board?

Re: Big Heatsinks

PostPosted: Fri Aug 15, 2014 1:26 pm
by theover
It must have been a good design!

Seriously, I was planning to use the board in my 1HE 19 inch rack unit mixer/processor with nice ethernet connector (where I now have a ACME and spartan 3 board in), but with the USB and occasional SD card failures, it isn't the industrial half-product yet that I wanted.

It's a lot of cheap power however, and I think it is very cool the parallel compiler is decent and on-board, and the whole Linux connection can work, so let's say I had sweat a little when testing my simple fix. Besides the problems it had run fine before, and behaved the same after, so the fuse must have been a right one.

Maybe I should order a replacement fuse SMD, it should be reasonable possibly to solder in with a fine tip leadless soldering station, but for the moment, it runs. I've contemplated on adding a big capacitor together with my de-bounced startup/stop circuit, and then it may be a better solution to add a current source limit with short timer to switch off (to no fry something when hitting max current.

Nobody can call computer design easy!


Re: Big Heatsinks

PostPosted: Fri Aug 11, 2017 11:09 am
by frankbuss
I have one of the Kickstarter boards, which usually needs a fan to cool it. But I mounted it on a really big heatsink which I bought some time ago for cheap from a remnant sale, no fan needed anymore 8-)


I mounted it on one side, to have some room for additional components for a planned project, like linear power supply ICs. For some kind of heat pipe I used 14 mm copper cubes (I got some from a craft store, gold plated, which I sanded on two opposite sides, for better contact), glued with "silverbead thermal glue", which is viscous like toothpaste, so the slightly different height of the Epiphany and the Zynq didn't matter, and then screwed the board to the heatsink:


I used a screwplate to create threads for M3 screws in the heatsink (a 2.5 mm drill before using the screwplate works very good). The screws are 20 mm long. Maybe a bit too short, 25 mm would be better, because I attached the Porcupine on top as well, with 5 mm hollow standoffs, and maybe I could use some nuts at the heatsink side for more stable connection, but works for me, doesn't need to withstand strong vibration like in a car or something. And don't fasten it too tight, to avoid bending your board, just a bit to secure the cubes.

Don't forget to bend the serial port pin headers before gluing it, if you need the serial port.

Running xtemp for half an hour looks good (ambient temperature is about 20°C, running the headless version, xtemp output on my X server on my PC) :


For some stress the sysbench benchmark program is running since boot, which causes 50% CPU load. I don't think it will ever reach 50°C, which is far below the max recommended temperature of 70°C.

The voltages on the pin headers of the Porcupine are good, so I guess the 5 mm standoffs are the right height.

Re: Big Heatsinks

PostPosted: Sun Aug 20, 2017 6:06 pm
by olajep

That's one BIIIG heatsink :D

frankbuss wrote:For some stress the sysbench benchmark program is running since boot, which causes 50% CPU load. I don't think it will ever reach 50°C, which is far below the max recommended temperature of 70°C.

Quite impressive! IIRC I get ~55C with heatsink + fan running cpuburn and the matmul example in a loop.
Would be interesting to see your numbers when you add some load on the Epiphany (generates ~2W)

Code: Select all
$ cd ~/epiphany-examples/apps/matmul-16
$ ./
$ ./

// Ola

Re: Big Heatsinks

PostPosted: Mon Aug 21, 2017 3:14 pm
by frankbuss
The Ephipany app doesn't work well. Sometimes it runs for 3 loops, then just stops (the system is still running, I can login to another shell), sometimes it crashes:

parallella@parallella:~/epiphany-examples/apps/matmul-16$ ./
/home/parallella/epiphany-examples/apps/matmul-16/ PASSED
./ line 12: 2400 Segmentation fault ./matmul.elf $@ ../../device/Release/e_matmul.elf > $LOG
/home/parallella/epiphany-examples/apps/matmul-16/ FAILED!!!!

After a crash, or when it hangs, it hangs always when I start it again. I used a fresh image, all original Parabuntu for the Zynq 7020 with HDMI support. Do I need to configure something or is this another know problem? It is one of the Kickstarter boards.

For cpuburn I started "burnCortexA9" twice, "top" shows 100% load. Running for an hour now and no more than 48°C.