An efficient memory compression scheme for 8 k FFT in a DVB-T receiver and the corresponding error model

This paper presents a methodology for the efficient compression of the memory needed for the storage of twiddle factors of long FFTs. Such FFTs are used for the demodulation of the received signal in DVB-T systems, where the minimization of size and power consumption, hence of cost, are of great importance. This minimization can be achieved by reducing the storage requirements. This paper shows that there is no need to store all the coefficients of the first and second stages, but only a fraction of them. The appropriate number of coefficients that have to be stored can be obtained by an error model which is also described and verified by extensive simulations. The proposed approach achieves a total of 94% memory compression at a cost of 1.3 dB in SQNR.