This paper presents a low-cost variable-length FFT processor for digital video broadcasting -terrestrial / handheld (DVB-T/H) systems employing pipelined shared-memory architecture, in which a radix-2/2 3 /2 4 FFT algorithm, multi-path delay commutator (MDC), a novel data scaling approach are exploited. Based on this architecture, novel low cost index block scaling approach was proposed to increase area efficiency. Also, an elaborate memory configuration scheme applied to make single-port SRAM without degrading throughput rate. The SQNR performance of this FFT processor has signal-toquantization noise ratio (SQNR) of 8K-point FFT is about 46.8 dB at 11bit internal word length for QPSK/16QAM modulation.
The maximum clock frequency of proposed design is 110 MHz.Keywords-DVB, FFT, variable length, scaling, pipelined shared memory.
I. INTRODUCTIONOrthogonal frequency division multiplexing (OFDM) attracts much attention in wireless communication systems, because it is feasible to many diverse requirements. In the typical OFDM systems, the complex symbols are modulated by means of inverse fast Fourier transform (IFFT). After guard interval insertion, the resulting OFDM symbol is serially transmitted over various channels, and at the receiver end the data stream is recovered by the inverse operation. Therefore, the FFT and IFFT blocks are the key computational blocks in OFDM systems. The long-size FFT is commonly adopted in OFDM system to increase transmission bandwidth or efficiency in many wireless applications such as digital video broadcasting (DVB), digital audio broadcasting (DAB), digital media broadcasting (DMB), very high-speed digital subscriber line (VDSL) and other mobile applications.The 2k/4k/8k point FFT should be performed in 224µ/448µs/896µs, respectively, according to DVB standards. On designing a long-size FFT processor, one still has to consider its power consumption and hardware cost. Furthermore, the power consumption of both data access in memory and operation of complex multipliers is more than 75% of total power consumption in an FFT processor [1]. To reduce the power consumption of the FFT memories and complex multipliers, the useful method is reducing the memory access times, internal word length and the number of operations in complex multipliers.The memory-based architecture provides lower hardware complexity and high flexibility. However, it has disadvantages