Original question:
IWR1443BOOST: Phase discontinuity between RX channels from each TX antenna
Hi Joe Quintal:
we have the same question with "Santi Ortega" , about phase discontinuity on 1642 board in MIMO mode with the sdk demo program.
the phase of tx1-(rx1~rx4) is discontinue with tx2-(rx5~rx8) and the phase difference between the two sets(rx1~4 and rx5~8) is
change with the target angle changing. the original post is locked, we want to know have this problem been solved?
thanks
Hi,
I am not aware whether the previous problem has been solved. But I have done quite some antenna calibration and angle estimation myself.
To start with on IWR1443BOOST, the antenna pattern TX1-(RX1~RX4) and TX3 - (RX1~RX4) should be continuous (instead of total 8 visual antenna from TX1 and TX2 listed in your message).
We would expect a continuous phase among these 8 antennas. If not, we would expect a continuous phase after applying the phase/gain calibration. When you do phase/gain calibration, you need to be really careful about the setup. Basically, there should be a single strong target at 0 degree at azimuth direction. The target should be the only target in the nearby range. If you are using a corner reflector, it needs to be big enough to get enough SNR and be placed far enough to assume it is a point target.
Sometimes, it is hard to get the target placed at exactly 0 degree during calibration time, then your phase estimation will introduce a bias. In that case, when you look at your DOA spectrum, you should still see an improvement on the DOA spectrum, i.e., the side lobe is much lower. Depend on how clean is your setup, you may not see perfect phase, but I would expect some improvement.
If you have done all these procedure, and still did not see the improvement on the DOA spectrum. Please provide us the results and plot to get further help from us.
Best,
Zigang
HI,
Based on your complex data. I did some calibration based on the first set of data (8 degree), and then used it for your second set of data (9 degree). The code is listed at the end. And here is the result after compensation. So, the peak is located at 6 degree for both 8 degree and 9 degree cases. So, it is not perfect different by 1 degree. But I would say phase estimation error at several degrees is normal. I may also agree that a simple phase compensation between TX1 and TX2 maybe be enough, there maybe other leakages from RX1 to RX2 and etc. But I would say if the angle estimation error is within several degrees cross angles, it is acceptable, and that is what we see as well.
The MATLAB code is attached here:
a = [874325 - 2296796i
205170 - 2403574i
-472478 - 2213569i
-1294669 - 1950097i
2513575 + 410691i
2257925 - 490783i
2293441 - 968409i
1510409 - 1593431i].';
angleDelta = mean(angle(a(6:8)./a(5:7)));
jump1 = angle(a(5) / (a(4)*exp(1j*angleDelta)));
b = [a(1:4), a(5:8)*exp(-1j*jump1)];
a1=exp(-1j*pi*((0:7).'*sin((-90:90)*pi/180)));
figure; plot(-90:90, abs(b*a1).^2)
a = [-2387530 + 205527i
-2014882 + 918611i
-1605891 + 1574635i
-815417 + 2038463i
-317777 - 2407704i
-1266327 - 2044338i
-1622046 - 1619236i
-2134822 - 609021i].';
b = [a(1:4), a(5:8)*exp(-1j*jump1)]
grid
hold on; plot(-90:90, abs(b*a1).^2,'r')
|
|
