fft.c

TMS2812系列DSP源代码程序包：C28x 快速傅立叶变换程序库

C28x 快速傅立叶变换程序库 
            C28x 快速傅立叶变换程序库PCWIN95/98/2000/ME/NT/XPv0.9b08/17/2004

      ;*************************************************************************************
      ;*********************** SECTION 1: FFT LIBRARY 
      *********************************** 
      ;*************************************************************************************
      Thank you for trying C28x Software Collateral. 
      FFT Library is installed in C:\TIDCS\C28\DSP_TBOX\FFT directory.
      Fast Fourier Transforms are an efficient class of algorithms for the 
      digital computation 
      of the N-point Fourier transform (DFT). In general, their input sequence 
      are assumed to 
      be complex.  In many real applications, the data sequences to be processed 
      are real valued. 
      Even though the data is real, complex-valued DFT algorithm can still be 
      used. One simple 
      approach creates a complex sequence from the real sequence; that is, real 
      data for the real 
      components and zeros for the imaginary components, The complex FFT can 
      then be =applied directly. 

      However, this method is not efficient as it consumes 2N memory locations 
      (Real & Imaginary) 
      for N point sequence. When input is purely real, their symmetric 
      properties compute DFT very 
      efficiently. One such optimized real FFT algorithm for 2N-point real data 
      sequence is packing 
      algorithm. The original 2N-point sequence is packed as N-point complex 
      sequence and N -point 
      complex FFT is performed on the complex sequence. Finally the resulting N 
      -point complex output 
      is unpacked into another N+1 point complex sequence, which corresponds to 
      spectral bin 0 to N 
      of 2N-point real input sequence. Spectral bin 0 to N is sufficient, as the 
      remaining bins N+1 
      to 2N-1 are complex conjugates of spectral bins N-1 to 1. 
      The real FFT requires 2N+2 memory locations to compute the FFT for 
      2N-point real valued sequence, 
      which is highly preferable in contrast to the complex FFT that consumes 
      4N-locations for 2N-point 
      real valued sequence. Moreover using this strategy, the complex FFT size 
      can be reduced by half, 
      at the FFT cost function of O(N) operations to pack the input and unpack 
      the output. Hence, 
      the real FFT algorithm computes the FFT of a real input sequence almost 
      twice as fast as the 
      general FFT algorithm.
      This FFT library contains generic FFT module (32-bit implementation) for 
      real/complex FFT.

      FFT LIBRARY
      |===============|=====================================================|
      |  Module Name | Description                     |
      |===============|=====================================================| 
      | CFFT32  | 32-bit complex FFT module              |
      |---------------|----------------------------------------------------------------------|
      | RFFT32  | 32-bit Real FFT module            |
      |=====================================================================|
       
      DOCUMENTATION:
      |===============|======================================================|
      |  DOC  | DIRECTORY LOCATION                    |
      |===============|======================================================| 
      |  MODULE DOC | C:\TIDCS\C28\DSP_TBOX\FFT\DOC\FFT_MDL.PDF                  
       |
      |=====================================================================|
       

      Version History:
      ================
      version 0.9b dated 8/17/2004
      1. Updated the FFT module documentation and used FFT size "N" instead of 
      "2N"
         in real FFT documentation
      2. In Real FFT modules, FFT size constant "TWO_N" is replaced with "N"

      Version 0.9a dated 12/06/2003 
      1. Some the math function had .ref being used in the *.asm, it is changed 
      to .def 
         Because of this bug, compiler was reporting that the function name is 
      not available.
      2. Most of the functions did not clear the OVM flag at the end, it is 
      modified.
      3. ASP & NASP is been used within functions, we should not disturb the ASP 
      bit.
         It is meant for only interrupt handler. Called function should not 
      change this bit.
      4. RFFT_acq function routine required on more call to clear the "acqflag" 
      after the "count"
         reaches "0" and all the data is stored in the "ipcb" buffer. Fixed this 
      bug such that "acqflag" 
         is cleared when the last data is aquired.