ref: a5b327cd4e11cb97cd0d78763df2e56be3ec00d0
dir: /src/resampl.h/
/* * FILE: resample.h * BY: Julius Smith (at CCRMA, Stanford U) * C BY: translated from SAIL to C by Christopher Lee Fraley * (cf0v@andrew.cmu.edu) * DATE: 7-JUN-88 * VERS: 2.0 (17-JUN-88, 3:00pm) */ /* * October 29, 1999 * Various changes, bugfixes(?), increased precision, by Stan Brooks. * * This source code is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. * */ /* Conversion constants */ #define Lc 7 #define Nc (1<<Lc) #define La 16 #define Na (1<<La) #define Lp (Lc+La) #define Np (1<<Lp) #define Amask (Na-1) #define Pmask (Np-1) #define MAXNWING (80<<Lc) /* Description of constants: * * Nc - is the number of look-up values available for the lowpass filter * between the beginning of its impulse response and the "cutoff time" * of the filter. The cutoff time is defined as the reciprocal of the * lowpass-filter cut off frequence in Hz. For example, if the * lowpass filter were a sinc function, Nc would be the index of the * impulse-response lookup-table corresponding to the first zero- * crossing of the sinc function. (The inverse first zero-crossing * time of a sinc function equals its nominal cutoff frequency in Hz.) * Nc must be a power of 2 due to the details of the current * implementation. The default value of 128 is sufficiently high that * using linear interpolation to fill in between the table entries * gives approximately 16-bit precision, and quadratic interpolation * gives about 23-bit (float) precision in filter coefficients. * * Lc - is log base 2 of Nc. * * La - is the number of bits devoted to linear interpolation of the * filter coefficients. * * Lp - is La + Lc, the number of bits to the right of the binary point * in the integer "time" variable. To the left of the point, it indexes * the input array (X), and to the right, it is interpreted as a number * between 0 and 1 sample of the input X. The default value of 23 is * about right. There is a constraint that the filter window must be * "addressable" in a int32_t, more precisely, if Nmult is the number * of sinc zero-crossings in the right wing of the filter window, then * (Nwing<<Lp) must be expressible in 31 bits. * */