– Copied from customer query –
According to page 26 of the UbiSwitch manual, the heatsink is made of aluminum. The TC4-10G data sheet (attached) from the link provided on page 25 of that same manual specifically says that it does not work with Aluminum heat sinks and will corrode the material. Do you have another recommended thermal paste?
We actually moved away from using paste because the airgap between the heatsink and the chip is around 1.5mm, but seems we didn’t update the datasheet. Thank you so much for letting me know that.
We now use a thermal pad, 2mm thick, this is not the exact one we use, but it will work just fine.
As of end of September, all UbiSwitches will ship out with a small pack including the screws, nuts and thermal pad, which will make this all much easier.
The datasheet has been updated.
I notice that the thermal conductivity of this pad (5W/m-K) is considerably lower than the original specification in the UbiSwitch data sheet (79W/m-K). What caused the reduction in conductivity requirement?
I believe the 79W/m-K is ridiculously high for what is actually needed on UbiSwitch.
The chip puts out a maximum of 4 Watts. Let’s do a thermal calculation.
The thickness of the thermal interface pad is 2mm, and the width/length of the pad will be 15mm (same as the chip).
This corresponds to a difference of around 7.2C between the hot side and the cold side of the pad.(link to calculator)
If we used a pad with higher thermal conductivity, (say 10.2W/m-K, like this) we’re at a 3.6C difference
The real truth of the matter is that since the power output of the main chip is pretty low, and the maximum operating temperature is high (105C), you could most likely put anything in there as an interface material, and you’re probably going to be fine. This isn’t an FPGA that outputs 100W, where we have to be more stringent on the thermal design.
That being said, a 10.2W/m-K is going to do a better job of keeping the chip and heatsink at the same temperature compared with a 5W/m-K. Bear in mind our 105C ambient MIL-STD0810H tests were done using a 5W/m-K pad. Using a paste that has a 79W/m-K isn’t really necessary.
Let me know if that makes sense.
Thanks for the calculations. I agree that the 79W/m-K is much higher than I expected, but the large discrepancy was concerning.
No worries. Thanks for pointing out this issue in the datasheet.