This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

AWR1642BOOST: Angle estimation of a target

Part Number: AWR1642BOOST

Respected ,

I am using AWR1642boost +DCA1000 EVM and capturing the ADC raw data . I have seen the training video related to FMCW radar working ( 1DFFT,2DFFT and 3DFFT) ...In that the direction estimation is done on the 2DFFT data and i have a doubt which are mentioned below 

Q1: why can't we find direction by calculating the Phase difference between the 4 RX channels of ADC raw data as shown below figure .  

Q2: What was the reason behind the 3DFFT based angle estimation ? any how both principle of operation is based on Phase so why can't we use the below formula for direction estimation.

Q3 : In Radar studio Post-processing i can see the X coordinate value and back calculated the angle with a formula  angle = sin-1(X/R) *(180/pi)  please tell me if this calculation is wrong .................correct me with the proper formula.

Please reply as soon as possible .

  • Question #1

    This is how the AoA is computed. Please see

    file:///C:/ti/mmwave_sdk_02_01_00_04/packages/ti/demo/xwr16xx/mmw/docs/doxygen/html/index.html

    Question #2

    Please see mmwave training angle estimation section

    https://training.ti.com/intro-mmwave-sensing-fmcw-radars-module-5-angle-estimation

    Question #3

    Please see Question #1

    Thank you

    Cesar

  • Your answers didn't resolved my questions !

    In this

    file:///C:/ti/mmwave_sdk_02_01_00_04/packages/ti/demo/xwr16xx/mmw/docs/doxygen/html/index.html

    Angle estimation is explained as FFT of the detected points in 2DFFT will yeild a peak at angle bin where angles bins are -N/2 : N/2  etc ...... I agree this is how you implement but my question is

    Q1 : ADC raw data of RX1,RX2,RX3,RX4 has its own phase values ......by find the phase difference between RX1-RX2,RX1-RX3,RX1-RX4  and substituting the phase difference value in the equation which is mention in the image angle = sin-1(lambda*phaseDifference/2πs) ...... Why this kind of estimation is not used please justify ( assume you are getting reflect from single static  object)  ............ Because I have done this kind of analysis by keep test source enable and emulated one target at 3 m X , 11 m Y ......... With the formula which i told you to cross verify and conform i got angles as 15 degrees .......  Where by the phase difference difference between Rx antennas and above formula of direction estimating is fail to give angle as 15 degree ...... Now explain me why it is happened like this !  ........

    Please cross verify the equation used to calculating for angle using X coordinate value and try to explain with the similar case which I mention above .......

  • Hi,

    The processing chain in the mmwave SDK out of the box demo was designed to perform detection in the range doppler domain because this processing chain will be able to detect moving objects as well.

    For mostly static objects we have develop the Obstacle Detection processing chain which performs detection in range angle domain.

    http://www.ti.com/tool/TIDEP-0104

    We provide these processing chains as examples that can be further modified.

    It is possible that for your specific use case a different processing chain would have better performance.

    We can only offer limited support for processing chains that we have not developped

    In your example, if you perform angle estimation using directly the raw data, this means that you are not performing any separation in range using fft. This may work for one object but not for more.

    thank you
    Cesar

  • Hey Cesar,

    Can you please tell me that angle = sin-1(X/R) *(180/pi)  or angle = sin-1(X/R)  or angle = tan-1(Y/X) or angle=tan-1(Y/X) *(180/pi) ................ with these many formulas i dont know which formula to use ................... i referred to file:///C:/ti/mmwave_sdk_02_01_00_04/packages/ti/demo/xwr16xx/mmw/docs/doxygen/html/index.html 

    So i dont understand that i was getting angles in radians or degrees ......................so please tell me what formula to use for calculation of angle from X 

  • Hi,

    In file ///C:/ti/mmwave_sdk_02_01_00_04/packages/ti/demo/xwr16xx/mmw/docs/doxygen/html/index.html

    In section "Data Path - (X,Y) Estimation"

    Please see "Figure A: Coordinate Geometry" and "Figure wx"

    Regarding difference between degrees and radians please see a trigonometry text book

    Thank you

    Cesar

  • Hi Cesar ,

    As you told before that angle fft will help us to resolve multiple targets but in the case which i mentioned below link is failed 

    https://e2e.ti.com/support/sensors/f/1023/t/842050

    Please explain the effect why it is occurring or suggest me the answer 

    Your answer resolved many of my doubts so your answer is more valuable and important for me.

    Please reply as soon as possible and thanks in advance .

  • Hi,

    Here is a summary of detection for two static objects. There are dependencies on antenna, algorithm, reflectivity of the objects

    Depending on an antenna used, the radar data can contain the following information: range, velocity, angle.

    If the objects are static, the velocity information does not help to differentiate between objects because both objects have same velocity (they are in the zero doppler bin)

    So, if objects are static we are left with range and angle information.

    If both objects are at the same range, the range information will not help us differentiate them because it is the same for both objects.

    So, we are left with angle information.

    Differentiating the objects with the angle information will depend on several factors

    1) Antenna Angular Resolution (please see other threads defining this)

    If the angle between the two objects is smaller than the antenna angular resolution, then the radar will not be able to detect two objects with traditional fft based processing chain. It may be able possible to detect with higher resolution algorithms such as MUSIC.

    If the angle between the two objects is larger than the antenna angular resolution, then we have to consider the following things

    2) Does the processing chain support multiple peak detection for the same range bin (same distance)?

    Some of the algorithms TI provides detect only one peak (object) for the same range bin. Some other algorithms support multiple peak detection. If the algorithm does not support multiple peak detection it will not be possible to detect the two objects even if the angle between the two objects is larger than the antenna angular resolution.

    If the algorithm supports multiple peak detection, then we need to consider the following

    3) What are the relative RCS values of the two objects?

    The RCS value determines the intensity of the signal reflected by the object. If the two objects we are trying to detect have a very different RCS this means that one of the objects will be much more reflective than the other one. In this case it is possible that in the fft processing the higher RCS side lobes will mask the smaller RCS. In this case it is not possible to detect both objects with this algorithm.

    In order to detect two objects with different RCS we many need to use beam forming algorithms.

    Thank you

    Cesar

  • Cesar ,

    You are way of explanation is awesome ........ I loved this !

    Thank you so much !

    This means a lot  and informing me that lot more things to learn thanks a lot !

  • Cesar ,

    I agree to you points

    But what I observed is when I emulated static targets at

    Target 1 :    x=-10 ,y=10

    Target 2 :    x=10 ,y=10

    It has to show the detection either at x = 10 or -10 because it resolve one target but not two but what I observe is around x =0 at y=10 (results observed in Radarstudio postprocessing GUI ). So I want to know what effect the phase information which resulting x =0

    Similarly this has been done for different target case like

    Target 1 :    x=-20 ,y=20

    Target 2 :    x=20 ,y=20

    And observed X = 0 at Y=20

    Target 1 :    x=-30 ,y=30

    Target 2 :    x=30 ,y=30

    And observed X=0 at Y=30

    The actually result has to be at either of the target position but the result observed at 0 which is strange  ........As I am new to this signal processing so I was seeking for help

    ............ I believe that because  of two equidistant target  phase of the received signal will not get modulated but I don't know why it is resulting at X=0

  • Hi,

    Please use the mmWave SDK 3.3 OOB demo.

    Please enable "multiObjBeamForming" in the configuration. (Check the description of this command in the SDK UG)

    This should enable to detect two objects of similar RCS at different angles.

    thank you

    Cesar