NRF24L01+ FCC PART C BAND Edge measurement fail

[tin039]

Our product is using NRF24L01+ and PA to transmitter data. It pass all the test of FCC PART C but the BAND Edge measurement.

The below 2 pictures is NRF24L01+ Band edge measurement in thelaboratory and in our spectrum analyzer.

The RF power of the NRF24L01+ is -6dbm. The RF power after the PA is about 15dbm.

It use frequency from 2403Mhz to 2479Mhz.

It can't pass the band edge measurement.

We have made the test as follow:

1. We removed the PA and measured the signal. The spectrum waveform shape is same between NRF24L01+ output and the PA output. The PA just amplified the output spectrum waveform, not modify its shape. And we have chaged 3 different PA and the result is the same. So we consider it is nothing to do with the PA.

2.We have tested several different product which using NRF24L01+(we bought, not designed by us). The spectrum waveform is the same to us. So we think the circuit design is ok.

3. The pictures above is the product the data as its normal using. If setting the CONT_WAVE = 1(06 register) and sending the carrier, the spectrum wave is very good and it can pass the band edge measurement.

4.We have tested the NRF51822 and NRF52810, the spectrum wave is very good and can pass the test. The picture is as follow. We consider whether is the problem of NRF24L01+?

So, what is the reason that it can't pass the band edge measurement? Is there any way to improve the spectrum wave and pass the test?

Thank you for your help!

[ketiljo]

The reason for this is that the skirt of the modulation envelope is too high at the band edge. The only way to handle this is either to lower the output power on the channels near the band edges, or not use these channels at all. 

There is nothing wrong with the design, it's just the way the PA work. It will lift the whole spectrum and at some point, you violate the band edge power spec. 

The main problem is that the nRF24L01+ can't modulate continuously. When you turn on the PA, you will get a step in power that spreads out in the frequency domain. So the bandwidth of the signal appears to be higher than it actually is.