AWR6843AOPEVM: Antenna Geometry definition when data processing chain is not used

Hi, Tim:

Eventually, the antenna definition will be needed during angle estimation.   You can follow the code to see how the antenna definition is used in angle estimation module. 

Best,

Zigang

I am looking at specific examples where the Angle of Arrival is not calculated. For example the vital signs example.

The example also does not make use of the compensation in rx channel phase and range bias. Is this because the application is independent of the antenna geometry and the on-chip antenna manufacturing tolerances for the awr6843AoP?

I have found this code in the   multi_gesture_68xx_aop_dss project

int16_t rxPhaseRotVector[NUM_USED_CHANNELS] = {1,1,1,1, -1,-1,-1,-1};

/* auto-scale the input so as to prevent overflow in the 32-bit FFT. */
AutoScale((int16_t*)DOAIn);

/* Rotate the RX channels according to the ODS RX phase rotation vector */
for (antIndx = 0; antIndx < NUM_USED_CHANNELS; antIndx++)
{
    DOAIn[antIndx].real *= rxPhaseRotVector[antIndx];
    DOAIn[antIndx].imag *= rxPhaseRotVector[antIndx];
}

#elif defined(MMW_6843_AOP) && defined(SOC_XWR68XX_ES2)

// Arrange the DOAIn in 2D array format (below is based on 6843 AOP ES2 antenna, long axis of EVM is horizontal)
/* Channels are arranged as (because the chirps are sent in this order: TX3, TX2)
*
*
* ch-2 ch-1 ---> (TX-1 row: Not used)
* ch-3 ch-0 ---> (TX-1 row: Not used)
* ch-11 ch-9 ch-7 ch-5
* ch-10 ch-8 ch-6 ch-4
*/
memset((void *)&gDOA2D[0][0], 0, NUM_ANGLE_BINS_GESTURE * NUM_ANGLE_BINS_GESTURE *sizeof(cmplx16ReIm_t));

gDOA2D[1][0] = DOAIn[7];
gDOA2D[1][1] = DOAIn[5];
gDOA2D[1][2] = DOAIn[3];
gDOA2D[1][3] = DOAIn[1];

gDOA2D[0][0] = DOAIn[6];
gDOA2D[0][1] = DOAIn[4];
gDOA2D[0][2] = DOAIn[2];
gDOA2D[0][3] = DOAIn[0];

which i think is performing compensation for the differences in Tx1 and Tx3 virtual channel configurations on the AoP. compared to a default antenna configuration (this code is doing so for the IWR6843)

I am trying to apply this code to the vital signs example so that I can improve the accuracy of the application.

However I am not entirely understanding what the input and output parameters for this compensation should be. And also in what sequence should this code be placed within the program flow. (Before 1D FFT? After 1D FFT?)

The question being asked is not application specific however. I am asking in a more general sense on how I can implement compensation for antenna geometry differences and compensation of rx gain phase offset for any application that does not make use of the data processing chain. 

Thanks again 

Tim 

HI, Tim:

For the vital sign demo, only one antenna pair is used.  Therefore, there is no need to understand the antenna pattern or compensate for the phase cross different antenna pairs. 

Best,

Zigang

HI, Tim:

In the out of box demo, the phase compensation is only needed during angle estimation.  If your signal processing chain does not need angle estimation, then it will be totally different chain.  You will need to understand how you going to use the signal cross antenna to figure out where did you need the phase compensation.  In general, if you need to use phase information cross different antenna, that is where you will need the phase compensation.  

Ideally, when the target is at bore sight (0 degree in azimuth and 0 degree in elevation), the phase should be all the same from all antenna pair.  However, there can be off due to the non-perfectness in routing between different antenna.  So the phase compensation is to fix this.  

Best,

Zigang

Hi Zigang Yang,

I have added the other antennae to allow for elevation readings. The current vital signs demonstration only allows for readings when the passenger's chest would be at azimuth level. Different passengers will have different torso lengths, therefore an application would sometimes require two elevation angles if a single sensor is to be used for front row vital signs detection.

Elevation is still required even in the case where only the driver's vitals would be required given that their will be different torso length between different drivers. 

Please don't consider this question as application specific, given that I know that you don't offer support for applications.  This is only mentioned as a response to "Therefore, there is no need to understand the antenna pattern or compensate for the phase cross different antenna pairs. "

My questions which I think are still unresolved are general questions as this information can be utilised in any application we decide to develop.

The questions that are unanswered are the following:

1. How can I include range and phase compensation for manufacturing tolerances when the data processing chain is not utilized?

           -What data needs to have range and phase compensation on it?

                           adc data? 1D FFT output data?  

           - What would be the point where I add range and phase compensation?  (Where in the program flow?)

                          During interframe processing? During Interchirp processing? After windowing? After 1D FFT? Before sending my processed outputs? 

2. The same questions above, but for antenna geometry. 

The questions arose after observing the significant improvement in the quality of results obtained in the single row occupation demo, after antenna geometry and

compRangeBiasAndRxChanPhase were defined. This application however makes use of the data processing chain. The data processing chain is a black box. Other than some documentation on 

file:///C:/ti/mmwave_sdk_03_05_00_04/packages/ti/datapath/dpc/objectdetection/objdetdsp/docs/doxygen/html/index.html
and
file:///C:/ti/mmwave_sdk_03_05_00_04/packages/ti/datapath/dpc/dpu/aoa2dproc/docs/doxygen/html/index.html

The documentation itself doesn't indicate the process right?

Thanks again and I look forward for your replies! :)
Tim

Hi Zigang,

Thanks for your reply.

By question 2 with regards to the antenna geometry definition, I am asking about this assignment of virtual channels, as shown in this diagram obtained from file:///C:/ti/mmwave_sdk_03_05_00_04/packages/ti/datapath/dpc/dpu/aoa2dproc/docs/doxygen/html/index.html

I have the same questions I had for gain, phase and range compensation (which you explained quite clearly) but for these antenna geometry definitions instead. I would like to have the understanding of what data needs to have Tx1 and Tx3 -180 degree phase shift operated on,  and where in my program flow should this be done.

With respect to your vital signs reply  I'm not sure if I understood you correctly.  Given that the provided vital signs demo uses a horizontal linear array, how does the application cater for different torso lengths, i.e. different chest heights (heart outside of horizontal plane) ?

I find this information interesting.

Eager to hear from you

Tim

Thanks Zigang, 

It would be greatly appreciated!