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: Chirp Integration for SNR improvement, Object detection algorithm for Range Estimation

Part Number: AWR1642BOOST
Other Parts Discussed in Thread: AWR1642

Greetings Forum!

I hope this thread finds all of you in the best of health.

I am working with AWR1642 for altitude measurement from surface of earth. I have reached a certain level in my work and now face a couple of challenges for which your guidance is solicited.

I would like to share my working upto this point :

As i am only working with range estimation so my antenna configuration is 1 Tx and 1 Rx.

Previously, i was targetting 300m range but i was apprised by applications manager Mr Cesar on your forum that this range might be too ambitious to achieve given the RF power budget available so I did some deliberation and now targeting 150m range. I have done some surface clutter rcs survey using the MAT*** you know what tool and found that i at 150m height with given antenna beamwidths of 1642 antenna, i can get -108dBm of signal power with a grazing angle of 90 degrees looking down. The surface type is Farms/grass with standarad deviation in height of around 3m. 

My working on noise floor calculation yielded -139dBm of noise power with 16 chirps of 300us each,hence an observation duration of 4.8ms per frame. Since i am not interested in doppler bins, the radar cube contains 2D FFT output in a contiguous manner for 16 doppler bins one after the other. My FFT size is 1024 and frame periodicity is 20ms. I have turned off the UART output for now because i am generating 32 times more data for simple range profile than the maximum throughput of UART~1Mbps. I have observed FFT values being updated every frame using CCS debugger on the OOB Demo SDK 3.06. As per my understanding if the UART throughput allowed, the data being generated is calculated as follows: ADC Sample size(4 byte, 2 byte I & 2 byte Q)* num of doppler bins(16)* num of range bins(1024)*frames per second(50 frames)*uart bits per data byte(8 data bits, 1 stop 1 start=10 bits)= 32.7 Mbps. Since this uart writing is being executed using a blocking context implemented using semaphores to mantain process synchronization between MSS and DSS domain, application is not designed to cater such high baud rate let alone the bandwidth of UART peripheral will not allow it either. 

1. At this point, I have observed that for the case 1D FFT data for 16 chirps makes it 2D FFT data that i dont need for doppler estimation, I want to integrate the data from 16 chirps using non-coherent integration method to benefit from SNR improvement as stated in literature books and your application notes. If i reduce the frame rate from 50Hz to 1 Hz, this data is manageable by UART but why is the application sending 16*1024 FFT data to demo visualizer GUI for range profile of zeroth doppler? Is the GUI integrating this 1D FFT data in PC before displaying? I would like to know more about this implementation in GUI if it is happening. For now i dont want to loose focus by messing with the java script source code of the GUI, i want to save that effort for later once i am done editing the demo visualizer software for my application. What neccesary changes do i need to make in the embedded target software to integrate the 16x 1DFFT values on the processor before sending in on UART?

2. The second part of my question lies with the algorithm being used in the OOB demo software to detect the object, I have noticed that the results buffer parsed and sent by DSS domain to MSS domain contain the number of detected objects and based on whether peak grouping, CFAR-CA is turned on and off, the value changes from say 15 objects to 3 objects in a lab environment. Where is this algorithm implemented? in which file and is it user editable? As per my understanding and you may validate please, that for the range only estimation the Quadrature(phase) information is redundant, so once the FFT has been processed for a frame, the algorithm sweeps from 1st range bin to FFT size-1 range bin looking for sufficient difference between noise floor and received power to determine the presence of object in that range bin with CFAR turned off. It parses this information by multiplying the range bin index with range resolution in the objout buffer and passes the memory location handle to MSS.

For my application, I only need SNR values for each detected object, i shall turn on CFAR-CA and send all the detected objects to MSS, the MSS may decide or the same may be decided in DSS before sending that say 3 objects were detected in range bin 15, 40 and 63 each having SNR value of 30, 26 and 22 dB respectively.  it only makes sense that the object with max SNR value is the surface of earth.

Please guide me through the process. I have in depth knowledge on what i am doing here. I have read almost all the articles on your website apropos what i am saying and i speak with experience. Pleae my post be treated given my hardwork and attention to detail here.

I intend to contribute in the form of an application note for texas instruments once i am done with my findings covering all aspects of application. I shall be using simplelink or similar wireless link as telemetry radio for sending altitude data of minidrone in real time and display on GUI. I shall cover various terrain types and summarize my performance results. I beleive in knowledge sharing and improvement through collaboration

I want to know more about these intricacies, please guide me on the mentioned issues.

Best Regards