Issues with BLE at higher temperatures

[Jeffrey Haynes]

Let me start by saying we have not ruled out other possible sources of errors at this point.  However, right now all the evidence is pointing to an issue with the nRF52840 when we reach between 70-75C.  The product has a high-powered integrated light, hence the heat source.  At 75C we start dimming the light until we reach 85C at which point the dimming would be complete (in theory, we would never reach that point - in the cases where it works it regulates to around 75C).  Unfortunately, around the 72C mark, we start experiencing issues.  In some cases the issues manifest as the part appearing to completely lock up.  However, much more typical is that we lose the BLE connection, which is constantly reporting out the core temperature from the nRF internal sensor to an app running on a phone.  At no other time do we experience connectivity issues.

Again, we could be misinterpreting what we're seeing, but all parts on the board are rated to a minimum operating temp of 85C.  Right now I'm mostly looking for assurances that the BLE radio is good for constant operation up to around 85C.

[ketiljo]

Issues like this can often the causes by the frequency of the crystal. The radio itself will operate fine up to 85 degrees but it's dependent on a accurate enough clock source. What's the spec of the 32 MHz crystal and the value of the cyrtsl load caps?

[hmolesworth]

Long shot: You might try covering the nRF52840 with an optical shield to guarantee none of the bright light permeates through the die cover or through the pca material to the underside of the die as the die is light-sensitive (similar to most semiconductor dies). Easy to check, hopefully.

[Jeffrey Haynes]

Thanks, I'm actually tracking a possibly-related issue that has nothing to do with temp.  Let me chase that down and then I'll get back to you before we waste any more time with this.  Thanks.

[ketiljo]

That only applies to the CSP, the QFN is not affected by light. 

[hmolesworth]

I would argue same die, thicker case so less affect but not zero; dark leakage is also very temperature sensitive. Metallic module case might be closer to zero.