REFRENCE ANTENNA - DIRECTION FINDING CONNECTIONLESS RX CODE

[SRINATH]

Hi, I am trying to implement the AOA using the MUSIC algorithm. I am using direction_finding_connectionless_rx code in zephyr (nordic sdk v2.3.0). 

I am using nRF52833dk (tag) and nRF52833dk with 4x4 coreHW rectangular antenna array (locator) connected with gpio pins 3,4,28,29. I am getting 8 samples from the reference period, 1 to 37 samples with 2 µs antenna switching slots (45 samples totally).

I am trying to toggle only 3 antennas in linear form with reference antenna (7), so I have set the ant_patterns[ ] as { 0x7, 0x6, 0x7, 0x8, 0x7, 0x9, 0x7 } in main.c. I have printed the antenna ID, it's correct as per the pattern I have set in main.c. But according to the BLE's direction finding specification the 1st 8 samples are from the reference antenna (reference period). I am not sure where I should set the reference antenna in the code.  According to the readme.rst file provided in the code, Reference period where single antenna is used for sampling. The samples are spaced 1 µs from each other. The period duration is 8 µs. I am not sure whether it takes reference antenna as a 1st antenna in my antenna switching pattern mentioned above (main.c) 0x7 or It will takes antenna that have been set as dfe-pdu-antenna = <0x7> in (nrf52833dk_nrf52833.overlay) file.

Where should I set the reference antenna?

[Simonr]

Hi again, sorry about the late reply, but the developers didn't get back to me until just now. 

The reference period is done with the first antenna in the antenna switch pattern. The PDU antenna is used to receive PDUs. These two could be the same antennas, it depends on the lds stored in the DTS and provided as the switch pattern from the application.

Our thoughts on your samples and the behavior you see: The samples in the reference period are taken every 1µs. If the CTE frequency is 250kHz and the test conditions are ideal, that should give a phase difference of 90 degrees (when there is no RF noise, no HFXO drift and the carrier frequency is exactly 2.4GHz. Note that the CTE frequency isn't 250/500kHz exactly, but rather offset from the channel center frequency depending on the PHY used. Also, the center channel frequency changes between PDUs, so it's not an ideal 2.4GHz. If the phase shift is constant over time (generates a straight line), your frequency estimation is off. If it moves around 0° then it's noise which is expected in all cases.

How often do you see this issue, does it happen every time and always o the 6th set of IQ samples? What frequency do you pull out of the reference signal on your end? 

It's also worth mentioning that supporting the MUSIC algorithm outside the scope of what we support on our side.

Best regards,

Simon