BGT60TR13C - RX phase shift

[f_raymackers]

Hello all,

I'm developing low-level processing on a BGT60TR13C demo board with the help of the SDK and the Python Wrapper. Here are the steps I actually did:

1. Retrievement of raw-data
2. Range-Doppler FFT for each RX antenna
3. CA-CFAR
4. Angle of Arrival based on complex-angle difference between Range-Doppler map of RX antennas.

From the step 1 to 3, all look good. But I have a problem on the the Angle-Of-Arrival. I have an angle shift of arround 30° in the E-plane. Because Range-Doppler map is coherent, I think the problem comes from a bad phase shift between RX antennas. So the question is, is there an factory calibration of RX antenna?

Thanks in advance,

François

Radar SDK Version: v3.2.0+92.a81b2f3db

RadarBaseboardMCU7_v253

[Honey_D]

Hello @f_raymackers ,

The range angle estimation algorithm can be explained as shown in the attached figure.

The range doppler map provides information of range and velocity of the detected object. The range and velocity are determined by the frequency change and the phase shift in the IF signal respectively. Thus, the 2D matrix formed represents range and velocity values. These values are evaluated to form the range angle matrix.

DBF is a system of uniform linear antenna array (ULA) which will receive phase shifted signal, thus the different antennas are weighted individually in their phase and amplitude so that they are summed up to zero and the signals are in-phase. It precise and simultaneous processing of different angles and frequencies and can be implemented on FFT processed spectral and range doppler maps.

For further information on the algorithm please refer to the Digital Beamforming document from the Radar development Kit located at C:\Infineon\Tools\Radar Development Kit\3.2.1.202205180921\application\RDK-Package\RDK_Package

Best Regards,

Honey

[f_raymackers]

Hello @Honey_D,

Thanks for your reply. I just checked your Python script about DBF and tested it in the situation but I have too an angle shift of around 20° in the E-plane (angle determined with RX2-RX3 - demo board on landscape mode) when I walk from 0 to 3m. This angle shift is nearly canceled when I put the board on the other orientation (portrait) and I did angle estimation with RX1 and 3.

Can you check about that? What's about antenna calibration?

About my estimation of angle, I simply do the complex angle calculation of the range-doppler bin concerned between RX2-3 (dRX23) and then calculate: Angle-of-Arrival = arscin(λ * dRX23 / (2*π*d))

Thanks for your help.

Best Regards,

François

[Honey_D]

Hello @f_raymackers ,

The antennas of DBF systems are set up of the uniform linear array thus at least two antennas should be positioned in plane of detection to perform beamforming. Thus, the minimum requirement for detecting azimuth and elevation is a ULA with three antennas in an L shape.

Therefore the orientation of your sensor depends on the direction of detection so as it gives you uniform linear antenna array and there will no phase shift in the signal.

About the angle estimation, BGT60TR13C uses the concept of DBF algorithm, where the angle is estimated from range doppler map which is formed of the matrix of range and velocity calculated from received IF signal.

I would suggest you to go through the DBF algorithm document, to understand the antenna radiation pattern you can refer application note and refer to range-angle-map python file (C:\Infineon\Tools\Radar Development Kit\3.2.1.202205180921\application\RDK-TOOLS-SW\RDK-TOOLS-SW\SDK\apps\py\examples) for code reference.

Best Regards,

Honey

[f_raymackers]

Hello @Honey_D  and thank you for your reply,

The problem is not about Range-Doppler map and ULA understanding but about physical phase shift between RX2 and RX3.

I'l explain you: for my use case I have to positionnate the BGT60TR13C with the orientation below. I use the RX2 and RX3 to determine the azimuth (E-plane). But the azimuth angle is shifted of about 20-30°. This shift is not present on the elevation angle (H-plane).I tried both with my AoA and with your DBF algorithm and the shift is the same (in your DBF algorithm I just changed the RX mask to use only RX2 and RX3 instead of the RX1 and RX3 by default in your code).

So my answer is: is it normal? Are the RX antennas tilted? Have you saw this angle shift in your tests?

The RX radiation pattern below shows a tilt between antenna RX2 and RX3. Is there the same tilt in the phase shift between RX2-3?

Thank you in advance.

Best Regards,

[Honey_D · Accepted solution]

Hello @f_raymackers ,

The phase shift in the E-Plane pattern is observed due to the close proximity of the antennas in package and it is stronger for electric field than for magnetic fields.

We would recommend you to use Rx3-Rx1 pair which has better phase linearity.

Best Regards,

Honey

[f_raymackers]

Hello @Honey_D,

Thanks for your reply, so this problem is known. Unfortunately for my application, I have to use Rx2-Rx3 pair. I'll do a simple calibration but do you have a calibration method to preconize me?

Best Regards,

François

[Honey_D]

Hello @f_raymackers ,

Apologies for the delay in response!

Since the phase shift is known, and it is a fixed phase offset between receiver antenna 2 and 3. 

The offset can be fixed by using zero degree calibration method:

Place a corner reflector at boresight (zero degree azimuth & zero degree elevation aligned) at a known distance and calculate phase difference between receiver channels 2 & 3. This phase offset can be used as a calibration matrix.

Hope this helps!

Best regards,

Honey

[f_raymackers]

Hi @Honey_D,

No problem with the delay. Following my previous message, I did exactly the same solution as you proposed and it works fine.

Bests Regards,

François