Routing of COAX antenna cable across chip seems to impact RF performance

[garydrule]

All,

We are in product on a product using the nRF52832 and are noticing in manufacturing that we aren't able to make it through the self test part of fab because of shoddy RF communication. We sell many other products with the same chop that do pass so in looking for the differences we have found that routing a coax antenna cable across only certain parts of the chip have an effect.  I didn't believe it when first brought to my attention but I have seen (at least small sample size) evidence that it may be an issue.  Routing the COAX cable not over the chip seems to fix the issue, I want to know if this fix has a real hardware answer behind it.

An email from my team..

"Attached are the two photos, we seem to be able to duplicate our failure with the "noise" path.  The failure of course shows as a CSMA collision avoidance failure.  We have measured the noise floor on the clear path and it is about -95 dB; measured on the noise path and it can be up to -80 dB.  The CSMA backoff level is set at -85 dB.  We also were able to insert a piece of copper tape under to coax cable in the noise path and that "fixed" the problem.  Routing the cable more across the center of the chip also seems to fix the problem.  It seems to only exist when the cable is in the path shown. Best we can guess is we have some sort of ground or noise coupling going on. 

1. Any insight it to what is happening, or is there something critical about that corner of the chip?

2. Do they recommend a minimum distance from a coax antenna cable to the chip, either in the X-Y plane or vertical spacing in the Z axis up off the chip?

3. Do they think noise is being coupled through the board or specific components or traces near the chip?

4. Are we completely off base and something else is going on?"

Any insight would be great!

We are using a flex PIFA LSR antenna with a U.FL pcb side connector.  Please see attached photos

This routing is OK. 

This routing causes an issue 

[haakonsh]

Hey Gary,

This might be PAN [138] RADIO: Spurious emission on GPIO exceeds limits in radiated tests. The workaraound is to tie p0.25 and p0.26 to GND, or decouple them with 12pF if they are in use. 

The issue is that the internal bonding wires for those two pins run too close to the radios ground pin and picks up noise from the receiver, which they in turn radiate. 

Therefore I'm guessing that the receiver is picking up its own noise through the coax when measuring the RSSI. This noise is not present when transmitting. 

Cheers,

Håkon.

[garydrule]

That has been implemented on this design from a note we had a year ago or so.

Thanks for the feedback!

[AmbystomaLabs]

The radio is just picking up its own LO or RF noise as you have surmised. Running the coax directly over the radio is never a good idea for this reason. 

Local ground noise which contains the LO/noise is likely the culprit. Start first by improving the local bypass caps around the radio. Also take a good look at your ground plane to make sure you haven’t created a poor RF path off the nRF that is contributing to local ground currents that can ride on the coax shield.

I suppose you can try a double shielded coax. But since this is an RX event there shouldn’t be any outbound signal on the coax that the radio doesn’t see. Most likely it’s a ground or bypass cap problem.

[AmbystomaLabs]

Also if you want a quick verification that you are dealing with a ground noise problem, clamp a ferrite around the coax just after the u.fl connector. Problem should go away when you do this.