Radiated immunity test problem

[Agoff]

 I’ve  provided  radiated immunity test for our devise based on the nRF52832.

The test has been provided by electromagnetic pulse with a power of 10V/m ,duration 1sec and the same test with a duration 3sec,  the frequency range from 2GHz to 2.7GHz  with a frequency grid 1% between each hopping. 

During the test in the frequency range from 2.4GHz to 2.7GHz wasn’t  observed any interference of the device .

But in the lower frequency range from 2GHz  to 2.345GHz  with a  duration pulse  of 1s, were observed difficulty  in signal transmission.

For the  pulse duration of 3s, the connection  between device and computer was completely interrupted and stop pairing .

The problematic frequencies are 2.165,2.231,2.253,2.298,2.345GHz.

This test has been provided In accordance with the standard for medical home use , the minimum pulse duration must be at least 1s.

Have to be taken into consideration that harmful frequencies are not  in range of the BLE frequency range. I want to know   what affects to  processor ,at low frequencies, stop to work correctly?

Thanks in advance. 

[AmbystomaLabs]

It is unlikely the burst is causing processor issues.  Most likely is the low frequency burst is just aliasing inband in the receiver. 

With most ETSI (RED) testing you are allowed to just spec how you fail during immunity and as long as your public documentation (ie, user manual) declares it that is sufficient to pass.  However, you mentioned a medical home use requirement so you may not get this luxury to just spec and it call it a day.

If it truly is aliasing in-band your only solution will be a tighter bandpass filter on the front end to remove the 2.165-2.345GHz band. If BLE allows it, you could also just map out the problematic frequencies in software.  As long as they aren't associated with advertising channels, the link won't care if it can't use 5 channels. Essentially for every frequency you fail on there should be a corresponding 'bad' ble channel.

Another thought is to look at how the CCA routine is running. Offhand I don't know how often ble determines channel assessment but this could be a way to mitigate the problem.