Compiler/AWR1642BOOST: algorithm、TIDEP-0092

Hi Lin,

Answers are inline.

1. We did road testing and found a problem as following: when a car moved toward our implemented radar and got close enough, with velocity approximated our car’s velocity (the relative velocity approximated zero), then the tracking output on GUI started flashing. Is there any mechanism in the algorithm which eliminate target with small R or with v approximates zero?

Ans. There is no specific mechanism that eliminates targets as they come closer.  It could be that the target's SNR did not meet the detection threshold. What you can do in such a case, is to modify the SNR as a function of range. The structure that controls the SNR as a function of range is the 'SNRThresholds' structure. You can see it being populated in dss_main.c::MmwDemo_populateSRR. Look for the line saying

 /*! @brief CFAR thresholds are varied as a function of range */
    obj->SNRThresholds[0].rangelim = (uint16_t) (10.0f * (float)(1 << obj->xyzOutputQFormat));
    obj->SNRThresholds[0].threshold = convertSNRdBtoThreshold(1, 18.0f,CFARTHRESHOLD_N_BIT_FRAC);

What this says is that objects closer than 10m should have an SNR of 18 dB or higher. You can reduce the SNR to 16 or 15 dB. 

2. We see the 15 degrees angle resolution in TI DEP-0092 documents and wonder how it is calculated. If we do 32 points FFT on antennas’ outputs, shouldn’t we get a better angle resolution to be 180/32=5.625 degrees? Why’s the resolution gets worse? If using chinese remainder theory to extend max velocity, we only find the first peak when doing angle estimation, so when two objects has same range and velocity, we can not distinguish two objects in angle domain, right ?(usually call "overlapping issue" in DSP)

Ans. We define 'angle resolution' as thus, if the angle resolution is x degrees, then if there are two objects with the same velocity and the same range, and their azimuth angles are closer than x degrees, then the radar cannot distingush between them. 

3. The same in the documents, we see that the time duration of SRR subframe to be 30 ms, but a set of slow and fast chirps should only takes 8.3 ms. Are there multiple sets of slow and fast chirps transmitted in a SRR subframe? If so, how many of them, and how do they work?

Ans. The SRR subframe is transmitted completely in 8.3 ms, with ~ 4.5ms for the slow chirps, and 3.7ms for the fast chirps. The rest of the time (i.e 21.7 ms), the RF part of the device is idle.

4. Is there hard detection interval, ex: In range 0~80m with peakgrouping only find peak to range bins up to (224)?

Ans. Yes. We only detect up to the 224th bin. This is because the band beyond 90 % of the 'IF sampling frequency' is used as the transition band for the various filters in the DFE.

4b. What is the interval of angle ? Is positive and negative 60 degree of y axis (broadside)?

Ans. It is +/- 60 degrees.

Regards

Anil