| #include <math.h> |
| #include "Filter.h" |
| |
| Filter::Filter(unsigned int cutoffFrq, unsigned int sampleFrq, unsigned int order) : |
| order_(order + 1), |
| sampleFrq_(sampleFrq), |
| windowTable_(0), |
| sampleHistory_(0), |
| precision_(10) |
| { |
| #if 1 |
| setCutoffFrq(double(cutoffFrq)/double(sampleFrq)); |
| #else |
| setCutoffFrq(.000001); |
| #endif |
| } |
| |
| void Filter::setCutoffFrq(double fc) |
| { |
| fc_ = fc; |
| reCalcWindowTable(); |
| } |
| |
| void Filter::setFilterOrder(unsigned int order) |
| { |
| // must be odd number and 2^x + 1 |
| order_ = (order >> 1) * 2 + 1; |
| reCalcWindowTable(); |
| } |
| |
| Filter::~Filter() |
| { |
| delete [] windowTable_; |
| delete [] sampleHistory_; |
| } |
| |
| |
| void Filter::reCalcWindowTable() |
| { |
| int i; |
| const double pi = 4 * atan(1.0); |
| const double f = fc_; |
| double gain = double(1 << precision_) - 1.0; // 4095.0 |
| double kernelSum = 0; |
| int midorder = (order_ - 1) / 2; |
| double *dblCoeffs = new double[order_]; |
| |
| if (sampleHistory_) delete [] sampleHistory_; |
| sampleHistory_ = new int [order_]; |
| |
| if (windowTable_) delete [] windowTable_; |
| windowTable_ = new int[order_]; |
| |
| // this is used for resampling |
| for (i = 0; i < (int)order_; i++) { |
| |
| double j = i - midorder; |
| double x = pi * j; |
| double c = (j == 0) ? 2.0 * f : sin(2.0 * f * x) / x; |
| |
| // Blackman has better stop-band attenuation |
| double blackman = 0.42 - 0.5 * cos( 2*pi*i/(double)(order_ - 1) ) |
| + 0.08*cos( 4*pi*i/(double)(order_ - 1) ); |
| // Hamming has 20% faster roll-off |
| double hamming = 0.54 - 0.46 * cos( 2*pi*i/(double)(order_ - 1) ) ; |
| // von Hann is in between |
| double vonHann = 0.5 * (1.0 - cos((2*pi*i/(double)(order_ - 1)))); |
| // Blackman–Harris window |
| double a0 = 0.35875, a1 = 0.48229, a2 = 0.14128, a3 = 0.01168; |
| double blackmanHarris = a0 - a1 * cos( 2*pi*i/(double)(order_ - 1) ) |
| + a2 * cos( 4*pi*i/(double)(order_ - 1) ) |
| - a3 * cos( 6*pi*i/(double)(order_ - 1) ); |
| |
| c *= vonHann; |
| dblCoeffs[i] = c; |
| kernelSum += c; |
| } |
| for (i = 0; i < (int)order_; i++) { |
| double rebasedCoeff = dblCoeffs[i] / kernelSum * gain; |
| windowTable_[i] = (int)(rebasedCoeff + 0.5); |
| sampleHistory_[i] = 0; |
| } |
| // Now decrease filter order if last coeffs are zero... |
| /* |
| i = 0; |
| while (0 == windowTable_[i] && i < midorder) i++; |
| if (i) { |
| order_ = i; |
| i = 0; |
| for(int k = order_; k<=midorder; k++, i++) |
| windowTable_[i] = windowTable_[order_ - i] = windowTable_[k]; |
| order_ = (midorder - i) * 2 + 1; |
| midorder = (order_ - 1) / 2; |
| } |
| */ |
| |
| sampleBufPtr_ = 0; |
| sampleBufMask_ = (midorder << 1) + 1; |
| |
| delete [] dblCoeffs; |
| } |
| |
| short Filter::lowPass(short from) |
| { |
| int i; |
| int filteredSample = 0; |
| |
| int ptr = sampleBufPtr_; |
| sampleBufPtr_ = (sampleBufPtr_ + 1) % sampleBufMask_; |
| |
| // store input sample to history ring buffer |
| sampleHistory_[ptr] = int(from); |
| // last coeff not used... |
| i = order_ - 1; |
| do { |
| // convolve sample input with filter kernel |
| filteredSample += sampleHistory_[ptr] * windowTable_[i]; |
| // get previous sample from ring buffer |
| if (!ptr) ptr = sampleBufMask_ - 1; |
| else ptr = (ptr - 1); |
| } while (i--); |
| short output = short(filteredSample >> precision_); |
| return (short) output; |
| } |