Using a heatsink on my Zynq-7000, not sure what kind of thermal thing to use

[jeremy2002]

I've stopped using my Zynq-7000 for a while, because I needed a heatsink due to how concerningly hot it gets when the ARM9 CPU is in use (ex. Linux-based applications (especially it's preinstalled Linux)).

I finally found a MOSFET 17mm x 17mm x 9mm heatsink that would work perfectly for this exact purpose, however, it seems it was supposed to come with a adhesive thermal pad, yet I did not get any with it when I got it in the mail this afternoon.

So, where can I find a suitable adhesive thermal pad? And what thickness should I use?

EDIT:

Ok, now I'm confused (and it doesn't help that I keep encountering the half or the internet that are jerkmouthes ?).

Adhesive thermal PADS or Adhesive thermal PASTE?

[MIke Vine]

On my arty spartan s7 I used 'SEKISUI #5760' which is an cheap adhesive thermal pad. And it seems to work great- it keeps the heatsink firmly connected to the chip and the chip runs a lot cooler. I dont know how thermally efficient it is though compared to the pastes.

Something like:

https://www.ebay.co.uk/itm/Japan-SEKISUI-5760-Double-sided-Thermal-conductive-Adhesive-Tape-for-Heatsink-/171112683534

[jeremy2002]

On 7/27/2018 at 5:46 PM, MIke Vine said:

On my arty spartan s7 I used 'SEKISUI #5760' which is an cheap adhesive thermal pad. And it seems to work great- it keeps the heatsink firmly connected to the chip and the chip runs a lot cooler. I dont know how thermally efficient it is though compared to the pastes.

Something like:

https://www.ebay.co.uk/itm/Japan-SEKISUI-5760-Double-sided-Thermal-conductive-Adhesive-Tape-for-Heatsink-/171112683534

Thank you for your suggestion. I have bought it and I will try it out when it arrives.

[xc6lx45]

I think your question isn't as well defined as it could be, that may affect the quality of the answers :-)

Honestly, I'd worry about other things. I'm not aware that Xilinx has a reliability problem, expect it to crash long before they break (a possible exception to a general rule: modern GPUs, but that's a different story). And most likely, the regulators will break before the FPGA.

If I were selling your board at volume for an industrial application, I'd think about long term reliability. Apparently Digilent considers the risk low enough to ship it without a heatsink. Does this tell us something?

But, you've got a board and you aren't using it for fear of damaging it. That is bad so go and fix it :-)

 

[jeremy2002]

22 minutes ago, xc6lx45 said:

I think your question isn't as well defined as it could be, that may affect the quality of the answers ?

Honestly, I'd worry about other things. I'm not aware that Xilinx has a reliability problem, expect it to crash long before they break (a possible exception to a general rule: modern GPUs, but that's a different story). And most likely, the regulators will break before the FPGA.

If I were selling your board at volume for an industrial application, I'd think about long term reliability. Apparently Digilent considers the risk low enough to ship it without a heatsink. Does this tell us something?

But, you've got a board and you aren't using it for fear of damaging it. That is bad so go and fix it ?

 

The regulators will break over what? Heat or time?

[xc6lx45]

Just by numbers (read through the forum), and it's neither surprising nor reason for concern, given that the typical board operates in an uncontrolled environment.

My point is simply that, the FPGA / Zynq is a very robust component.

If you've ever been involved in an industrial HTOL qualification, it may give you a different view on those things... square meters of circuit boards in a real oven, and failures are still rare exceptions. The idea of avoiding a one-in-a-million chance of your device failing by lowering T a couple degrees (your cooling sink is much less effective than the HTOL oven in the other direction...) seems misplaced. Judge the cost of your board divided by a million (or pick any large number) against the cost of lost opportunity if you don't use the board you've bought.

[xc6lx45]

Speaking of which, here is some actual public data from Xilinx.

See page 16, 125 degrees Celsius (which is 257 degrees Fahrenheid), 1000 hours...
Even 85 degrees is not a temperature I'd keep my finger on for long. And Arrhenius is exponential.

[jeremy2002]

28 minutes ago, xc6lx45 said:

Just by numbers (read through the forum), and it's neither surprising nor reason for concern, given that the typical board operates in an uncontrolled environment.

My point is simply that, the FPGA / Zynq is a very robust component.

If you've ever been involved in an industrial HTOL qualification, it may give you a different view on those things... square meters of circuit boards in a real oven, and failures are still rare exceptions. The idea of avoiding a one-in-a-million chance of your device failing by lowering T a couple degrees (your cooling sink is much less effective than the HTOL oven in the other direction...) seems misplaced. Judge the cost of your board divided by a million (or pick any large number) against the cost of lost opportunity if you don't use the board you've bought.

 

17 minutes ago, xc6lx45 said:

Speaking of which, here is some actual public data from Xilinx.

See page 16, 125 degrees Celsius (which is 257 degrees Fahrenheid), 1000 hours...
Even 85 degrees is not a temperature I'd keep my finger on for long. And Arrhenius is exponential.

Interesting, but I'm pretty sure temperature-triggered throttling (especially on the ARM9 part of the FPGA) still has a role there. Presumably.

[xc6lx45]

on a Raspberry Pi :-)