AOA: Antenna switch pattern and IQ sampling via nrf52833

[Sometimes ever]

Hi！ 

I'm working on direction finding. I'm using a nrf52833dk as a beacon, a nrf52833dk and a Direction Finding Antenna Board (PCA2005) as a locator for a bluetooth direction finding AOA development. as follows:

  

The SDK I am using is NCS 1.7.1, the toolchain is SEGGER Embedded Studio (SES), examples are “direction_finding_connectionless_tx” and “direction_finding_connectionless_rx”.

For the beacon, I added “CONFIG_BT_CTLR_DF_ANT_SWITCH_TX=n” in “prj.conf” to make it work in AOA mode.

For the locator, since a 12-antenna array is used, the GPIOs I use are P0.03, P0.04, P0.28, P0.29, and the antenna pattern is configured in “main.c” as follows:

static const uint8_t ant_patterns[] = { 0x2, 0x0, 0x5, 0x6, 0x1, 0x4,
					0xC, 0x9, 0xE, 0xD, 0x8, 0xA };

And, I made some modifications in the cte_recv_cb() function to print the IQ data.

Now I have the following problems:

    1、I placed the locator obliquely above the beacon to sample several sets of IQ sampling data, and calculated the phase and peak amplitude according to formula

• Amplitude = sqrt(pow(I,2)+pow(Q,2))
• Phase = arctan(Q/I)，

but the I and Q values are different from the sinusoidal signal. The following is one of the sets of data:

        

I would like to ask: Is there a problem with the I and Q values?

If there is a problem, where did it go wrong and how should it be improved? If there is no problem, why is the IQ value far from the sine wave?

    2、In these two examples, are there any parameters corresponding to the current sampling antenna ID? Because I want to know which antenna a pair of IQ values corresponds to, and use this to calculate AOA.

I need a little help to put all this fine.

Regards，

John

[Kenneth]

Do not look at graphs made of the I data and the Q data as that will always give the wrong impression. Look at the calculated phase and amplitude data only. And remember: the first 8 samples are the reference samples with short spacing and the rest are with much longer gaps between them (so printing them with the same distance in a figure will also be incorrect and most likely just confusing). Please check out:
https://infocenter.nordicsemi.com/topic/nwp_036/WP/nwp_036/intro.html  

[Sometimes ever]

Thank you！

I have read your whitepaper carefully, but aside from the first 8 samples, I think the remaining data is still not ideal. What is the reason? Is there a way to improve data quality?

[PaKa]

Why do you believe they are not ideal? Have you run them through an algorithm and gotten bad results? Please have in mind the entire purpose with changing the active antenna is to get a change in phase. Since you only get one value from each antenna you can't look at the resulting graph and see if it bad or not.

[Miguel Roque]

Hello all, I'm exactly at same point as @John. Currently I'm printing the IQ data, however I tried to find any info about the antenna ID that produces each IQ data but I didn't get any success until the moment (looking for the bt_df_conn_iq_samples_report struct, I tried to print the parameter  rssi_ant_id but the value is allways "2", and in fact I'm using 4 linear antennas from the 12 available on the antenna). Another question that maybe you can help me, based on literature people advise to use external algorithms to decode the sampled data and return an angle, many suggest the MUSIC algorithm, but I didn't find many samples on the internet, or at least any sample easy to implement with the returned data from the devkit, where can I find it? 

Best regards, and thank you for your help!

Miguel Soares

[PaKa]

So for the sequence of the antennas, check the table for the antenna pattern and the description in the white paper. The pattern is deterministic so there should be no need to read back the antenna port, this can be calculated instead.