Direction Finding Phase Shift Between Reference Period And First Slot

RE: Direction Finding Phase Shift Between Reference Period And First Slot

wvfan — Wed, 21 Oct 2020 03:00:31 GMT

Thank you Dmitry.

For the 16us slot, I set it because I found the phase graph is not very stable near the switch slot. I will try your suggestion, the way is much easier than what I'm using now.

On the PCB board, I set all trace lengths to be almost the same. For the phase-to-angle mapping, that's a good idea and I will have a try.

Put more antennas into array is also in the next step's plan. Thank you for providing the algorithm, now I can do some research on it :)

RE: Direction Finding Phase Shift Between Reference Period And First Slot

wvfan — Wed, 21 Oct 2020 02:48:17 GMT

Thank you Øyvind. I designed 3 antennas placed on 3 corners of a square. But now I only use 2 of them to test. I think I have taken the frequency into count, I calculated the average slope of all period and use it as the real frequency, as the black line in the figure shown.

RE: Direction Finding Phase Shift Between Reference Period And First Slot

Dmitry — Tue, 20 Oct 2020 11:16:48 GMT

[quote userid="92250" url="~/f/nordic-q-a/67333/direction-finding-phase-shift-between-reference-period-and-first-slot/275743#275743"]The unstable means the phase shifting and angle will change from time to time. I just did a test and get these phase differences: 93, 93, 90, 94, 79, 95, 103, 103, 103, 99, 95, 110, 104, 98, 106, 87, 100, 94.[/quote]

If you're testing in a noisy enwironment, it's quite expected. You need to filter results, dropping the ones with large difference from the mean value. I would suggest to take an experiment outdoors, away from wifi networks and other sources of noise, to ensure that hardware is working properly.
To get better results, try to shorten the slots as much as response time of RF switch allows (16 us is too much - crystal jitter may influent the results), and take a single sample from each slot close to its end.

[quote userid="92250" url="~/f/nordic-q-a/67333/direction-finding-phase-shift-between-reference-period-and-first-slot/275743#275743"]for two antennas with 6.25cm distance, it's 33 degree and 23 degree ( but actually my angle is set to 75 degree )[/quote]

Did you take into account trace lengths from RF switch to antennas? For evaluation stage, I would not look at absolute values at all - all you need is to figure out the range when results are stable, then create a phase-to-angle mapping.

[quote userid="92250" url="~/f/nordic-q-a/67333/direction-finding-phase-shift-between-reference-period-and-first-slot/275743#275743"]Thank you for the information about the multipath reflection, can it be detected by the magnitude data?[/quote]

Multipath reflections can be mitigated by a multi-element array and spatial spectrum estimation algorithm, such as MUSIC or MVDR. With two antennas, there is not enough information to detect multiple paths.

RE: Direction Finding Phase Shift Between Reference Period And First Slot

Øyvind — Tue, 20 Oct 2020 10:56:08 GMT

What antenna are you using? The figure you have attached does not give much information I'm afraid.

You need to measure the phase on all antennas against the reference slot. The reference phase should not be static, as the frequency on transmitter and receiver are not the same. The phase will drift in a linear fashion.

Kind regards,
Øyvind

RE: Direction Finding Phase Shift Between Reference Period And First Slot

wvfan — Tue, 20 Oct 2020 03:15:58 GMT

Thank you Dmitry. From your answer and the document, the reference period is correct and not a issue.

The unstable means the phase shifting and angle will change from time to time. I just did a test and get these phase differences: 93, 93, 90, 94, 79, 95, 103, 103, 103, 99, 95, 110, 104, 98, 106, 87, 100, 94. The maximum is 110 and the minimum is 79, for two antennas with 6.25cm distance, it's 33 degree and 23 degree ( but actually my angle is set to 75 degree ). I don't know how to calibrate with the data, if I just add a fixed delta to make the average angle to 75 degree ( phase diff 194 ), the 110 and 79 will become around 210 and 179, and now the angle is around 90 degree and 63 degree, which cannot be used.

Thank you for the information about the multipath reflection, can it be detected by the magnitude data? I will have a try. And also for the antenna array, I will reconsider it to avoid degree near 90.

RE: Direction Finding Phase Shift Between Reference Period And First Slot

Dmitry — Mon, 19 Oct 2020 21:12:07 GMT

Hi,

take a look at this section of document provided by Øyvind, the key phrase is "The hardware always aligns the first IQ sample in the first sample slot". After a reference period, there is a gap about 1/2 of TSWITCHSPACING (first switching slot) - 2 usec in your case. It has no impact on accuracy and need not be accounted in calculations.

[quote userid="92250" url="~/f/nordic-q-a/67333/direction-finding-phase-shift-between-reference-period-and-first-slot"] I saw a paper that uses different antennas in reference period and the first slot to calculate the angle[/quote]

Even if it works, it's not the way the hardware is intended to use - Bluetooth Core specification puts a requirement to take samples at same points of all sample windows but nothing is said about reference period.

[quote userid="92250" url="~/f/nordic-q-a/67333/direction-finding-phase-shift-between-reference-period-and-first-slot/275688#275688"]Currently I can get the IQ sample data, but it's not stable.[/quote]

What do you mean exactly? Are IQ data differ from sample to sample in a static environment, or you can't get a stable relation between angle and phase shift? When testing indoors, multipath reflections are the main issue. Also it's crucial to have a well-designed array to get at least useful results, but your graph looks quite accurate.

RE: Direction Finding Phase Shift Between Reference Period And First Slot

wvfan — Mon, 19 Oct 2020 14:59:24 GMT

Hi Øyvind,

Thank you for replying. Here is my setup of radio:

NRF_RADIO->MODE = RADIO_MODE_MODE_Ble_1Mbit;
NRF_RADIO->CTEINLINECONF = RADIO_CTEINLINECONF_CTEINLINECTRLEN_Disabled;
NRF_RADIO->DFEPACKET.MAXCNT = 1250;
NRF_RADIO->PSEL.DFEGPIO[0] = ANT1;
NRF_RADIO->PSEL.DFEGPIO[1] = ANT0;
NRF_RADIO->CLEARPATTERN = 3;
NRF_RADIO->SWITCHPATTERN = 3;
NRF_RADIO->SWITCHPATTERN = 3;
NRF_RADIO->SWITCHPATTERN = 3;
NRF_RADIO->SWITCHPATTERN = 3;
NRF_RADIO->SWITCHPATTERN = 3;
NRF_RADIO->SWITCHPATTERN = 3;
NRF_RADIO->SWITCHPATTERN = 1;
NRF_RADIO->SWITCHPATTERN = 1;
NRF_RADIO->SWITCHPATTERN = 1;
NRF_RADIO->SWITCHPATTERN = 1;
NRF_RADIO->PACKETPTR = (uint32_t)&packet;
NRF_RADIO->DFEMODE = RADIO_DFEMODE_DFEOPMODE_AoA;
NRF_RADIO->DFECTRL1 = 20 << RADIO_DFECTRL1_NUMBEROF8US_Pos
| 1 << RADIO_DFECTRL1_DFEINEXTENSION_Pos
| RADIO_DFECTRL1_SAMPLETYPE_MagPhase << RADIO_DFECTRL1_SAMPLETYPE_Pos
| RADIO_DFECTRL1_TSWITCHSPACING_4us << RADIO_DFECTRL1_TSWITCHSPACING_Pos
| RADIO_DFECTRL1_TSAMPLESPACINGREF_125ns << RADIO_DFECTRL1_TSAMPLESPACINGREF_Pos
| RADIO_DFECTRL1_TSAMPLESPACING_125ns << RADIO_DFECTRL1_TSAMPLESPACING_Pos;
NRF_RADIO->DFEPACKET.PTR = (uint32_t)&rfRaws;

The phase graph is after processed. I made the gap between all adjacent phases less than half the period so the curve will be continuous. And I used the linear regression line as the reference, and got the final curve as the above image shows.

Could you please provide a typical phase and magnitude graph as a reference? Currently I can get the IQ sample data, but it's not stable. There are too many possible issues that can cause this, like RF switch, antenna or programming, so I need some way to debug with it.

Best Regards,

Fan

RE: Direction Finding Phase Shift Between Reference Period And First Slot

Øyvind — Mon, 19 Oct 2020 12:15:08 GMT

Hello Fan,

Can you please elaborate on your setup? What is the reference point? Have a look at our white paper of direction finding.

Unfortunately, this is all we have.

Kind regards,
Øyvind