Hi,
I studied ZoomFFT algorithm in lab0004-high-accuracy-14xx and I'm very confused about the process of slow DFT .I read High Accuracy Lab User's Guide、data_path.c code and Radar Hardware Accelerator User's Guide,but still can't completely understand the algorithm.I read 'Frequency shifter with auto-increment mode (a slow DFT mode)' part in Radar Hardware Accelerator User's Guide very carefully and following is my understanding to the algorithm.
1、Input 1024 samples（set 1024 samples per chirp） into FFT subblock performs FFT of 1024 points，it will generate 1024 bins.Store both 1024 sample and 1024 bin into memory.
2、Find peak index(corresponding to coarse peak frequence) with Statistics.(coarsePeakIndexBin)
3、As SAMPLES_TO_ZOOM_IN_ONE_SIDE is 2 in data_path.c code，I should pick up 2 bins on the left side of coarsePeakIndexBin and 2 bins on the right side of coarsePeakIndexBin,so I got 5 bins index total.
4、As described in Radar Hardware Accelerator User's Guide,“The starting frequency for the DFT computation is specified in the TWIDINCR register (similar to the frequency shifter mode)”，I think then we should calculate TWIDINCR according to previous 5 bins index?
5、Then we calculate fine Frequency according to the formula below where x(n) are values of 1024 samples,SRCACNT=1024 in this case，FFTSIZE=14, K means iteration times (as many as the number of DFT bins required)so k=0-4.
6、Find fine frequency peak after calculated.
Where am I wrong above and what's the principle of that？Please tell me.OR Can you provide some document of this algorithm besides High Accuracy Lab User's Guide、data_path.c code and Radar Hardware Accelerator User's Guide?
Thank you so much!!
