arm_fir_decimate_f32.c

00001 /* ----------------------------------------------------------------------  
00002 * Copyright (C) 2010 ARM Limited. All rights reserved.  
00003 *  
00004 * $Date:        29. November 2010  
00005 * $Revision:    V1.0.3  
00006 *  
00007 * Project:      CMSIS DSP Library  
00008 * Title:        arm_fir_decimate_f32.c  
00009 *  
00010 * Description:  FIR decimation for floating-point sequences.  
00011 *  
00012 * Target Processor: Cortex-M4/Cortex-M3
00013 *  
00014 * Version 1.0.3 2010/11/29 
00015 *    Re-organized the CMSIS folders and updated documentation.  
00016 *   
00017 * Version 1.0.2 2010/11/11  
00018 *    Documentation updated.   
00019 *  
00020 * Version 1.0.1 2010/10/05   
00021 *    Production release and review comments incorporated.  
00022 *  
00023 * Version 1.0.0 2010/09/20   
00024 *    Production release and review comments incorporated  
00025 *  
00026 * Version 0.0.7  2010/06/10   
00027 *    Misra-C changes done  
00028 *  
00029 * -------------------------------------------------------------------- */ 
00030  
00031 #include "arm_math.h" 
00032  
00033 /**  
00034  * @ingroup groupFilters  
00035  */ 
00036  
00037 /**  
00038  * @defgroup FIR_decimate Finite Impulse Response (FIR) Decimator  
00039  *  
00040  * These functions combine an FIR filter together with a decimator.  
00041  * They are used in multirate systems for reducing the sample rate of a signal without introducing aliasing distortion.  
00042  * Conceptually, the functions are equivalent to the block diagram below:  
00043  * \image html FIRDecimator.gif "Components included in the FIR Decimator functions"  
00044  * When decimating by a factor of <code>M</code>, the signal should be prefiltered by a lowpass filter with a normalized  
00045  * cutoff frequency of <code>1/M</code> in order to prevent aliasing distortion.  
00046  * The user of the function is responsible for providing the filter coefficients.  
00047  *  
00048  * The FIR decimator functions provided in the CMSIS DSP Library combine the FIR filter and the decimator in an efficient manner.  
00049  * Instead of calculating all of the FIR filter outputs and discarding <code>M-1</code> out of every <code>M</code>, only the  
00050  * samples output by the decimator are computed.  
00051  * The functions operate on blocks of input and output data.  
00052  * <code>pSrc</code> points to an array of <code>blockSize</code> input values and  
00053  * <code>pDst</code> points to an array of <code>blockSize/M</code> output values.  
00054  * In order to have an integer number of output samples <code>blockSize</code>  
00055  * must always be a multiple of the decimation factor <code>M</code>.  
00056  *  
00057  * The library provides separate functions for Q15, Q31 and floating-point data types.  
00058  *  
00059  * \par Algorithm:  
00060  * The FIR portion of the algorithm uses the standard form filter:  
00061  * <pre>  
00062  *    y[n] = b[0] * x[n] + b[1] * x[n-1] + b[2] * x[n-2] + ...+ b[numTaps-1] * x[n-numTaps+1]  
00063  * </pre>  
00064  * where, <code>b[n]</code> are the filter coefficients.  
00065  * \par 
00066  * The <code>pCoeffs</code> points to a coefficient array of size <code>numTaps</code>.  
00067  * Coefficients are stored in time reversed order.  
00068  * \par  
00069  * <pre>  
00070  *    {b[numTaps-1], b[numTaps-2], b[N-2], ..., b[1], b[0]}  
00071  * </pre>  
00072  * \par  
00073  * <code>pState</code> points to a state array of size <code>numTaps + blockSize - 1</code>.  
00074  * Samples in the state buffer are stored in the order:  
00075  * \par  
00076  * <pre>  
00077  *    {x[n-numTaps+1], x[n-numTaps], x[n-numTaps-1], x[n-numTaps-2]....x[0], x[1], ..., x[blockSize-1]}  
00078  * </pre>  
00079  * The state variables are updated after each block of data is processed, the coefficients are untouched.  
00080  *  
00081  * \par Instance Structure  
00082  * The coefficients and state variables for a filter are stored together in an instance data structure.  
00083  * A separate instance structure must be defined for each filter.  
00084  * Coefficient arrays may be shared among several instances while state variable array should be allocated separately.  
00085  * There are separate instance structure declarations for each of the 3 supported data types.  
00086  *  
00087  * \par Initialization Functions  
00088  * There is also an associated initialization function for each data type.  
00089  * The initialization function performs the following operations:  
00090  * - Sets the values of the internal structure fields.  
00091  * - Zeros out the values in the state buffer.  
00092  * - Checks to make sure that the size of the input is a multiple of the decimation factor.  
00093  *  
00094  * \par  
00095  * Use of the initialization function is optional.  
00096  * However, if the initialization function is used, then the instance structure cannot be placed into a const data section.  
00097  * To place an instance structure into a const data section, the instance structure must be manually initialized.  
00098  * The code below statically initializes each of the 3 different data type filter instance structures  
00099  * <pre>  
00100  *arm_fir_decimate_instance_f32 S = {M, numTaps, pCoeffs, pState};  
00101  *arm_fir_decimate_instance_q31 S = {M, numTaps, pCoeffs, pState};  
00102  *arm_fir_decimate_instance_q15 S = {M, numTaps, pCoeffs, pState};  
00103  * </pre>  
00104  * where <code>M</code> is the decimation factor; <code>numTaps</code> is the number of filter coefficients in the filter;  
00105  * <code>pCoeffs</code> is the address of the coefficient buffer;  
00106  * <code>pState</code> is the address of the state buffer.  
00107  * Be sure to set the values in the state buffer to zeros when doing static initialization.  
00108  *  
00109  * \par Fixed-Point Behavior  
00110  * Care must be taken when using the fixed-point versions of the FIR decimate filter functions.  
00111  * In particular, the overflow and saturation behavior of the accumulator used in each function must be considered.  
00112  * Refer to the function specific documentation below for usage guidelines.  
00113  */ 
00114  
00115 /**  
00116  * @addtogroup FIR_decimate  
00117  * @{  
00118  */ 
00119  
00120   /**  
00121    * @brief Processing function for the floating-point FIR decimator.  
00122    * @param[in] *S        points to an instance of the floating-point FIR decimator structure.  
00123    * @param[in] *pSrc     points to the block of input data.  
00124    * @param[out] *pDst    points to the block of output data.  
00125    * @param[in] blockSize number of input samples to process per call.  
00126    * @return none.  
00127    */ 
00128  
00129 void arm_fir_decimate_f32( 
00130   const arm_fir_decimate_instance_f32 * S, 
00131   float32_t * pSrc, 
00132   float32_t * pDst, 
00133   uint32_t blockSize) 
00134 { 
00135   float32_t *pState = S->pState;                 /* State pointer */ 
00136   float32_t *pCoeffs = S->pCoeffs;               /* Coefficient pointer */ 
00137   float32_t *pStateCurnt;                        /* Points to the current sample of the state */ 
00138   float32_t *px, *pb;                            /* Temporary pointers for state and coefficient buffers */ 
00139   float32_t sum0;                                /* Accumulator */ 
00140   float32_t x0, c0;                              /* Temporary variables to hold state and coefficient values */ 
00141   uint32_t numTaps = S->numTaps;                 /* Number of filter coefficients in the filter */ 
00142   uint32_t i, tapCnt, blkCnt, outBlockSize = blockSize / S->M;  /* Loop counters */ 
00143  
00144   /* S->pState buffer contains previous frame (numTaps - 1) samples */ 
00145   /* pStateCurnt points to the location where the new input data should be written */ 
00146   pStateCurnt = S->pState + (numTaps - 1u); 
00147  
00148   /* Total number of output samples to be computed */ 
00149   blkCnt = outBlockSize; 
00150  
00151   while(blkCnt > 0u) 
00152   { 
00153     /* Copy decimation factor number of new input samples into the state buffer */ 
00154     i = S->M; 
00155  
00156     do 
00157     { 
00158       *pStateCurnt++ = *pSrc++; 
00159  
00160     } while(--i); 
00161  
00162     /* Set accumulator to zero */ 
00163     sum0 = 0.0f; 
00164  
00165     /* Initialize state pointer */ 
00166     px = pState; 
00167  
00168     /* Initialize coeff pointer */ 
00169     pb = pCoeffs; 
00170  
00171     /* Loop unrolling.  Process 4 taps at a time. */ 
00172     tapCnt = numTaps >> 2; 
00173  
00174     /* Loop over the number of taps.  Unroll by a factor of 4.  
00175      ** Repeat until we've computed numTaps-4 coefficients. */ 
00176     while(tapCnt > 0u) 
00177     { 
00178       /* Read the b[numTaps-1] coefficient */ 
00179       c0 = *(pb++); 
00180  
00181       /* Read x[n-numTaps-1] sample */ 
00182       x0 = *(px++); 
00183  
00184       /* Perform the multiply-accumulate */ 
00185       sum0 += x0 * c0; 
00186  
00187       /* Read the b[numTaps-2] coefficient */ 
00188       c0 = *(pb++); 
00189  
00190       /* Read x[n-numTaps-2] sample */ 
00191       x0 = *(px++); 
00192  
00193       /* Perform the multiply-accumulate */ 
00194       sum0 += x0 * c0; 
00195  
00196       /* Read the b[numTaps-3] coefficient */ 
00197       c0 = *(pb++); 
00198  
00199       /* Read x[n-numTaps-3] sample */ 
00200       x0 = *(px++); 
00201  
00202       /* Perform the multiply-accumulate */ 
00203       sum0 += x0 * c0; 
00204  
00205       /* Read the b[numTaps-4] coefficient */ 
00206       c0 = *(pb++); 
00207  
00208       /* Read x[n-numTaps-4] sample */ 
00209       x0 = *(px++); 
00210  
00211       /* Perform the multiply-accumulate */ 
00212       sum0 += x0 * c0; 
00213  
00214       /* Decrement the loop counter */ 
00215       tapCnt--; 
00216     } 
00217  
00218     /* If the filter length is not a multiple of 4, compute the remaining filter taps */ 
00219     tapCnt = numTaps % 0x4u; 
00220  
00221     while(tapCnt > 0u) 
00222     { 
00223       /* Read coefficients */ 
00224       c0 = *(pb++); 
00225  
00226       /* Fetch 1 state variable */ 
00227       x0 = *(px++); 
00228  
00229       /* Perform the multiply-accumulate */ 
00230       sum0 += x0 * c0; 
00231  
00232       /* Decrement the loop counter */ 
00233       tapCnt--; 
00234     } 
00235  
00236     /* Advance the state pointer by the decimation factor  
00237      * to process the next group of decimation factor number samples */ 
00238     pState = pState + S->M; 
00239  
00240     /* The result is in the accumulator, store in the destination buffer. */ 
00241     *pDst++ = sum0; 
00242  
00243     /* Decrement the loop counter */ 
00244     blkCnt--; 
00245   } 
00246  
00247   /* Processing is complete.  
00248    ** Now copy the last numTaps - 1 samples to the satrt of the state buffer.  
00249    ** This prepares the state buffer for the next function call. */ 
00250  
00251   /* Points to the start of the state buffer */ 
00252   pStateCurnt = S->pState; 
00253  
00254   i = (numTaps - 1u) >> 2; 
00255  
00256   /* copy data */ 
00257   while(i > 0u) 
00258   { 
00259     *pStateCurnt++ = *pState++; 
00260     *pStateCurnt++ = *pState++; 
00261     *pStateCurnt++ = *pState++; 
00262     *pStateCurnt++ = *pState++; 
00263  
00264     /* Decrement the loop counter */ 
00265     i--; 
00266   } 
00267  
00268   i = (numTaps - 1u) % 0x04u; 
00269  
00270   /* copy data */ 
00271   while(i > 0u) 
00272   { 
00273     *pStateCurnt++ = *pState++; 
00274  
00275     /* Decrement the loop counter */ 
00276     i--; 
00277   } 
00278 } 
00279  
00280 /**  
00281  * @} end of FIR_decimate group  
00282  */