01-17-2020 10:43 AM
I'm working with the Kintex-7 FFG676 package and am looking for a specific physical dimension. Please see attached diagram
I can find the major dimensions in the "Xilinx UG475 7 Series FPGAs Packaging and Pinout Specification, User Guide," but the distance from the underside of the lid to the bottom of the BGA is not listed.
I also see that you do not provide 3D models to customers, so I am unable to find an answer to this questions myself.
Can you help me find this demension (with tolerances)? Please let me know if you have any questions on my request, and when I should expect to hear back from you.
Thank you for the help!
01-31-2020 03:36 AM
01-17-2020 10:50 AM
01-21-2020 06:08 AM
Thank you for the clarifying question. I am looking to mount a heat sink to the top of the FPGA, hooking into this portion of the lid.
My goal is to get official dimensions from Xilinx, and these forums are the support path they have directed me to.
01-21-2020 07:47 AM
01-30-2020 11:42 AM
Thank you for the reply, Dr. John.
I understand that heat sinks are often mounted to the PCB or adhered to the top. These are both non-ideal options for my application, and I would thus like to explore others - including the "clip-on" option I mentioned. One example of such a solution is in figure 3-12 of the document you linked (which was useful, thank you).
With this type of retention, the heat sink mouting bracket clips to FPGA itself. I was assuming this clips to the lid of the package, but it could be clipping to the substrate.
My question is: what is the thickness of the portion of the lid this would clip to? (illustration in my initial post) Is information provided officially somewhere?
01-31-2020 12:00 AM
01-31-2020 02:56 AM
Some BGA heatsinks do actually clip on between the PCB and the FPGA body - example. There's not much gap there, but most FPGAs don't require a very big heatsink and so thin plastic wedges are sufficient. Obviously if you've got a big chip that needs to dissipate 20W+ heat (or withstand substantial vibration) you'll need to provide appropriate mounting holes on the PCB, but for lab use on small chips these heatsinks do a nice job.
01-31-2020 03:36 AM
01-31-2020 03:40 AM
The distance you mention depends on how the solder balls melt down and reshape, that depends on many factors and I believe it's "highly" variable (maybe plus/minus 0.1 mm).
my advice: you don't need that distance. At least a precise figure. Merge with mechanical engineers. They talk of "nominal" measures and tolerances, so everything has a max and a min. Your heatsink will attach to the FPGA lid and will be at some (min, max) distance to the PCB. Use the min to check is not touching any component. The heatsink will probably be held by screws tightened a bit more or less depending on the distance. You may want some o-rings with the spacer to accomodate for the tolerance. I think that's it.
01-31-2020 03:43 AM
01-31-2020 04:45 AM - edited 01-31-2020 04:54 AM
So is your plan to decapsulate packages for alternative heatsink? Good luck. why don't you use lidless packages? The dimension you want may be in lidless packages documentation. Have you stumbled upon UG475 Appendix B? I hope you are aware of the implications of attaching heatsinks to the bare silicon. Lidded packages exist for a reason. Do you think a fraction of a mm of metal will add too much thermal resistance, keeping in mind "thermal adhesive" is not that much conductive, is just thin? Have you pondered the risks and associated cost of heatsinking the silicon? Have you done a cost-benefit analysis? My take on that is you should have serious reasons for doing that.
I was in a company doing some tests on chip cards. One of the operations was decapsulation: tear the contacts away (that was easy), then swell the card with acetone and scratch *carefully* with an xacto knife till there was a thin layer above the chip, then rub *very gently* with a cotton bud. The silicon side exposed is actually the back, where there are no transistors, just bulk silicon, nevertheless it's amazing how easily the chip is damaged. And I guess the bigger the more fragile. a card chip is some 3 mm, an FPGA die can be 10, 20, 30 mm (?) with similar thickness I guess. Again: good luck.
01-31-2020 05:03 AM
@drjohnsmith I suspect that he's just trying to clip the heatsink on anywhere it'll fit, and is under the (mistaken) understanding that it should be clipped onto the heatspreader.
As above, the standard way to attach a heatsink (when no PCB mounting points are available and adhesive is not acceptable) is to clip onto the FPGA package directly.
01-31-2020 07:35 AM
Thank you for the input. One heat sink manufacturer had specified that the "clips" should be mounted to the heat spreader, hence my original line of questioning.
From this discussion, it sounds like this is not recommended (due to shock/vibration concerns), and the preferred approach it to mount to the substrate. As @drjohnsmith mentioned, I do have those dimensions so should be able to get what I need.
I appreciate the discussion and the help you've provided.