Getting started with Bluetooth Direction Finding

RE: Getting started with Bluetooth Direction Finding

Andre N — Wed, 28 Feb 2024 11:04:03 GMT

Also the signal during the reference period looks like an amplitude modulated signal and i cant see a clean phase.
Is this due to undersampling or something ?
i dont really follow what is going on with the demodulation.

(this is a CTE without antenna switching)

also what frequency should i expect for the signal?
Here i have 16 samples per period.
The samples are taken 2us apart.
So i have a 32us period and 31.25kHz frequency

My config:
160us CTE signal with 1us slot duration
(= 4us guard perdiod + 8us ref period + (74 x 2 x 1us) sampling period)
which gives me 8 + 74 samples

There is a company called Navigine that published some opensource code for AoA and they use Minew hardware that is surely based on Nordic's PCA20054 antenna board.

When you look at their signal it looks different from mine.
On their reference period they have a normal signal that looks unmodulated and their sampling period looks like it is modulated in amplitude.
Their reference period looks like my sampling period and their sampling period looks like my reference period.

They also mention that they use channels 37 38 and 39 which are the BLE advertising channels.
But as you can see on my last reply, i get the signals on BLE data channels like 0, 17, 28, 29, 31, 35, etc

RE: Getting started with Bluetooth Direction Finding

Andre N — Tue, 27 Feb 2024 12:45:51 GMT

I used the zephyr connectionless locator example

I changed the antenna switch pattern and the cte_recv_cb function to print the data the way i wanted.
I changed the default slot_duration from 2us to 1us and that is it

static void cte_recv_cb(struct bt_le_per_adv_sync *sync,
			struct bt_df_per_adv_sync_iq_samples_report const *report)
{
	printk("CTE, samples, %d, channel_idx, %d, antennas, %d, ", report->sample_count, report->chan_idx, sizeof(ant_patterns)/sizeof(ant_patterns[0]));
	for (int idx = 0; idx < report->sample_count; idx++) {
        printk("%d, %d, ", report->sample[idx].i, report->sample[idx].q);
    }
	printk("\n");

    k_msleep(10);
}

#define ANT_1  0x5
#define ANT_2  0x6
#define ANT_3  0x4
#define ANT_4  0x9
#define ANT_5  0xA
#define ANT_6  0x8
#define ANT_7  0xD
#define ANT_8  0xE
#define ANT_9  0xC
#define ANT_10 0x1
#define ANT_11 0x2
#define ANT_12 0x0
static const uint8_t ant_patterns[] = { ANT_7, ANT_8, ANT_9 };

I noticed this parameter .max_cte_count = 5. (i did not touch it here is the original sample code)
It determines the number of CTEs i get per batch (but they are all the same)
I dont understand the purpose though

UPDATE:
apparently i can set max_cte_count to 0 to receive CTE continuously and if i do so, i still get 5 CTEs
so i went to the beacon code and i found this parameter

/* Number of CTE send in single periodic advertising train */
#define PER_ADV_EVENT_CTE_COUNT 5

so my beacon sends 5 CTEs in a row
for some reason the 1st one is sent on a random channel and the other on channel 0
for some reason (unless i messed up my code) the signals are exactly the same

RE: Getting started with Bluetooth Direction Finding

PaKa — Tue, 27 Feb 2024 12:28:32 GMT

Not sure what this is, have you changed anything else in the sample than the number of antennas?

RE: Getting started with Bluetooth Direction Finding

Andre N — Tue, 27 Feb 2024 11:23:31 GMT

sample count: 82, channel_idx: 29, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 17, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 28, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 31, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 35, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3
sample count: 82, channel_idx: 0, number_of_antennas: 3

Im looking at the signal info and i noticed a weird behaviour.

Every second i get a batch of 5 CTE signals. (probably because in the code i have .max_cte_count = 5)
The first one comes from a random channel and the other ones always come from channel 0.

Is this normal?
Should i discard the 1st signal? or the other 4? or none?

Update: the 5 signals look the same

RE: Getting started with Bluetooth Direction Finding

PaKa — Tue, 20 Feb 2024 11:29:19 GMT

Also note that the further away from the antenna you are, the smaller the phase shift will be. This as you are measuring the difference in length from the transmitter to the different patches. This will basically set the limit where you can reliably measure the angle. I suggest testing with static transmitter at different distances as that should give you a clearer idea on how the angles/distance impacts the measured phase changes. When you have this under control, then you can start looking at moving objects.

RE: Getting started with Bluetooth Direction Finding

Andre N — Tue, 20 Feb 2024 09:57:51 GMT

OK thx
i wasnt using the reference period at all.
i was switching between 3 antennas (ANT7 ANT8 ANT9) and using ANT8 as a phase reference.

i will try what you suggested and see if i get better results

RE: Getting started with Bluetooth Direction Finding

PaKa — Mon, 19 Feb 2024 14:21:44 GMT

I am not sure I understand what you are trying to plot here.

If you want to look at the samples you should do that with time on the X-axis and not the sample number, difficult to say what you are showing. This as the reference period has a different sampling interval than the rest of the samples.

You would then use the reference period samples to predict the phase of the signal on the reference antenna for the entire sampling window. When you then start sampling the different antennas, you compare them to this predicted phase. For the samples with the reference antenna, the error should be 0. If it is not then the prediction is not good enough. The "error" on the other antennas will be the phase shift. If you do this then the phase will be around 0 so any phase shifts will be significantly easier to see than what seems to be in your graphs.

RE: Getting started with Bluetooth Direction Finding

Andre N — Tue, 13 Feb 2024 12:01:45 GMT

Hi again

I'm reviving this ticket because i am still stuck
Here is what i did.

I configured the sample code to work with a 3 antenna switch pattern.
As you can see the 1st signal 'looks' good. We can see the Ref period and some irregularities from antenna switching on the signal.

Then i split the samples from each antenna (2nd row of plots)

I calculate the phase, and adjust to avoid having +/- pi jumps (3rd row of plots)

I interpolate to take into account the time shift caused by the antenna switching (4th row of plots)

But when i plot the phase difference, i get different results event if my beacon doesn't move.
Here you can see the phases of ANT7 ANT8 and ANT9 in two different CTE signals after measuring the position of a static beacon

RE: Getting started with Bluetooth Direction Finding

PaKa — Wed, 06 Dec 2023 12:26:54 GMT

So how does the measured phase shift align with the modeled phase shift?

It's impossible to comment on AoA measurement data if you don't have all the information about how they are taken. The better approach is making a simulation as you can then assess the validity of your measurements.

RE: Getting started with Bluetooth Direction Finding

Andre N — Wed, 06 Dec 2023 10:28:15 GMT

Right now i have my antenna matrix sitting on a table and i am moving around it with my beacon devkit.
i get my Q/I samples;
i separate those samples according my antenna switching pattern;
i calculate the phase with the formula 'phase = atan2(Q, I)';
and finnaly, i calculate the difference between both waves and i display it on these graphs

when i stand right in front of the antennas i get 0 rad phase difference as you can see in this graph which is logical since the signal reaches both antennas at the same time.

in theory, if i am 'aligned' with the antennas, i should get +3.14 or -3.14 phase difference since the waves received by each antenna are in phase opposition

however, when i move around, i cant get past 1rad phase difference (as shown in the other graph)

RE: Getting started with Bluetooth Direction Finding

PaKa — Wed, 06 Dec 2023 08:32:45 GMT

That depends on what you are measuring. If you have one transmitter in a static position compared to the AoA receiver and use 2 antennas then the phase difference between the 2 antennas should be constant forever. So the flatter the line, the better.

When you measure this phase shift between multiple antennas, then you can calculate the angle of the incoming signal.

If the transmitter is moving relative to the AoA receiver antenna, then the phase shift can change with the movement, it will then depend how the movement is in relation to both antennas.

If you set up an excel sheet where you calculate the distance from the transmitter to each of the AoA receiver antenna, then you can calculate the resulting phase for each of the antennas and you can model how the phase will change when you move the transmitter around.

RE: Getting started with Bluetooth Direction Finding

Andre N — Tue, 05 Dec 2023 12:59:32 GMT

Hi
I got some IQ samples switching between two adjacent antennas
I then calculated the phase for each antenna

In the image above,

the blue line represents the phase of the first antenna
the orange line represents the phase of the second antenna
the green line represents the phase difference between the two antennas
the red line is the average phase difference of throughout the whole signal

my problem is that the phase difference is always below 1. Is this normal?
because in theory it should range from -pi to +pi right?

RE: Getting started with Bluetooth Direction Finding

PaKa — Thu, 12 Oct 2023 13:11:57 GMT

Hi, plotting IQ data like this will never provide any information about the incoming signal. You need to transform the data into phase/amplitude and look at the phase data. And do remember that the sampling interval for the reference period is much lower than for the antenna switch period so printing it like above will look incorrect again.

This has been discussed in quite a few cases earlier so I recommend searching and reading up on these.

RE: Getting started with Bluetooth Direction Finding

Andre N — Thu, 12 Oct 2023 09:40:01 GMT

Hi
Thanks for your answer and sorry for the late reply
I'm taking a look into your algorithm.

This is a plot of the 45 samples of the CTE (including the reference period)

It looks 'pretty' apart from the reference period but i wasn't expecting to see a smooth sine function when i plot the signal received by the 12 antennas. Am I not switching antennas or is this normal ?

RE: Getting started with Bluetooth Direction Finding

Elfving — Mon, 02 Oct 2023 12:09:58 GMT

Hello,

Sure thing.

Let me start off by pointing you to the white paper we have available on direction finding. Off course I'll try to answer your questions here, but keep in mind that the whitepaper might be able to answer a lot on its own. It's definitely worth checking out.

[quote user=""]be able to calculate an angle of arrival using nordic's example code[/quote]

Sure thing! The white paper gives you the formulas you would need to use, and I would recommend you to have a look at our samples for this (here and here).

[quote user=""]be able to receive CTE signals from 2 trackers on the same locator[/quote]

Yeah, that might require a bit more work, but is possible.

[quote user=""]implement a CTE beacon on my company's products whose firmware is based on nRF5 sdk v17[/quote]

That however is close to impossible. We have no support for direction finding in our nRF5 SDK. I would recommend you then to move on the nRF Connect SDK if you want this feature. Either way, moving to NCS might be a smart route, as there might not be any more updates coming to nRF 5 SDK.

[quote user=""]

1 - What is the difference between Central/Peripheral and Locator/Beacon example codes ? Which would be the best for my use case?

[/quote]

Whether or not a BLE connection is used. What is best for your use-case depends on what it is. If the two devices are already connected and communicating for instance taking advantage of that already existing connection would be an idea.

[quote user=""]

2 - What is the link between the antennas and the Q/I samples? How does antenna switching work?

[/quote]

This is described in the whitepaper. By switching from one antenna to the other, we are getting another sample. It's the difference between multiple samples that we need, and in addition to that there is noise and multi-path issues that will affect the result, which requires us to have even more samples (and smart algorithms) to counteract.

[quote user=""]But the result I get is random values roughly between -0.05 and 0.05.
[/quote]

What sample are you using to get this? Could you give me a plot of the entire CTE? If you for instance just look at some numbers from the reference period it might not make much sense.

[quote user=""]3 - How can I receive CTE signals from multiple trackers on one locator?
[/quote]

Slightly more work involved with this. You would have to synchronize with multiple devices that do TX periodic advertising. Then you can distinguish provided IQ reports by Sync_Handle

[quote user=""]4 - Lastly I want to know if it is possible to implement a beacon in nRF5 sdk. And is it possible to do this while using the softdevice.
[/quote]

It unfortunately isn't.

Another thing I assume you are wondering about even though you haven't asked, is if we are providing an angle algorithm. We are not, unfortunately. Making a good angle algorithm also requires a bit of work, simply using the equations might not be enough. You can for instance have a look at this.

Regards,

Elfving