Bill Schottstaedt (

sndlib is a collection of sound file and sound synthesis function written in C and running currently in various Unices via OSS or ALSA, Mac OSX, and on old Windows systems. To build sndlib ( if possible, and sndlib.a):


To install it, 'make install' — I've tested this process in Linux.

The following files make up sndlib:

The naming scheme is more as less as follows: the sndlib prefix is "mus" so function names start with "mus_" and constants start with "MUS_". Functions involving sound files referenced through the file name start with "mus_sound_", functions involving files at a lower level with "mus_file_", functions involving header access with "mus_header_", functions involving audio hardware access with "mus_audio_", and various others just with "mus_" (number translations, etc). Conversions use the word "to" as in "mus_samples_to_bytes".


Sound files have built-in descriptors known as headers. The following functions return the information in the header. In each case the argument to the function is the full file name of the sound file.

  mus_long_t mus_sound_samples (const char *arg)        /* samples of sound according to header */
  mus_long_t mus_sound_framples (const char *arg)         /* samples per channel */
  float mus_sound_duration (const char *arg)       /* sound duration in seconds */
  mus_long_t mus_sound_length (const char *arg)         /* true file length in bytes */

  int mus_sound_datum_size (const char *arg)       /* bytes per sample */
  mus_long_t mus_sound_data_location (const char *arg)  /* location of first sample (bytes) */
  int mus_sound_bits_per_sample(const char *arg)   /* bits per sample */
  int mus_bytes_per_sample(int format)             /* bytes per sample */

  int mus_sound_chans (const char *arg)            /* number of channels (samples are interleaved) */
  int mus_sound_srate (const char *arg)            /* sampling rate */

  mus_header_t mus_sound_header_type (const char *arg)      /* header type (aiff etc) */
  mus_sample_t mus_sound_sample_type (const char *arg)      /* sample type (alaw etc) */
  int mus_sound_original_format (const char *arg)  /* unmodified sample type specifier */
  int mus_sound_type_specifier (const char *arg)   /* original header type identifier */

  char *mus_sound_comment (const char *arg)        /* comment if any */
  mus_long_t mus_sound_comment_start (const char *arg)  /* comment start (bytes) if any */
  mus_long_t mus_sound_comment_end (const char *arg)    /* comment end (bytes) */
  int *mus_sound_loop_info(const char *arg)        /* 8 loop vals (mode,start,end) then base-detune and base-note  (empty list if no loop info found) */

  int mus_sound_write_date (const char *arg)       /* bare (uninterpreted) file write date */
  int mus_sound_initialize(void)                   /* initialize everything */

The following can be used to provide user-understandable descriptions of the header type and the sample type:

  char *mus_header_type_name(mus_header_t type)             /* "AIFF" etc */
  char *mus_sample_type_name(mus_sample_t samp_type)        /* "16-bit big endian linear" etc */
  char *mus_header_type_to_string(mus_header_t type)
  char *mus_sample_type_to_string(mus_sample_t samp_type)
  const char *mus_sample_type_short_name(mus_sample_t samp_type)

In all cases if an error occurs, -1 (MUS_ERROR) is returned, and some sort of error message is printed; to customize error handling, use mus_set_error_handler and mus_set_print_handler.

  mus_error_handler_t *mus_error_set_handler(mus_error_handler_t *new_error_handler);
  mus_print_handler_t *mus_print_set_handler(mus_print_handler_t *new_print_handler);

To decode the error indication, use:

  char *mus_error_to_string(int err);

Header data is cached internally, so the actual header is read only if it hasn't already been read, or the write date has changed. Loop points are also available, if there's interest. To go below the "sound" level, see headers.c — once a header has been read, all the components that have been found can be read via functions such as mus_header_srate.


The following functions provide access to sound file data:

  int mus_sound_open_input (const char *arg) 
  int mus_sound_open_output (const char *arg, int srate, int chans, mus_sample_t sample_type, mus_header_t header_type, const char *comment)
  int mus_sound_reopen_output (const char *arg, mus_header_t type, mus_sample_t format, mus_long_t data_loc)
  int mus_sound_close_input (int fd) 
  int mus_sound_close_output (int fd, mus_long_t bytes_of_data) 
  int mus_sound_read (int fd, int beg, int end, int chans, mus_float_t **bufs) 
  int mus_sound_write (int fd, int beg, int end, int chans, mus_float_t **bufs) 
  mus_long_t mus_sound_seek_frample (int fd, mus_long_t frample)

mus_float_t defaults to double. It is set when sndlib is built, and refers to Sndlib's internal representation of sample values.

mus_sound_open_input opens arg for reading. Most standard uncompressed formats are readable. This function returns the associated file number, or -1 upon failure.

mus_sound_close_input closes an open sound file. Its argument is the integer returned by mus_sound_open_input.

mus_sound_open_output opens (creates) the file arg, setting its sampling rate to be srate, number of channels to chans, sample type to sample_type (see sndlib.h for these types: MUS_BSHORT, means 16-bit 2's complement big endian fractions), header type to header_type (AIFF for example; the available writable header types are MUS_AIFC (or AIFF), MUS_RIFF ('wave'), MUS_RF64, MUS_NEXT, MUS_NIST, MUS_CAFF, and MUS_IRCAM), and comment (if any) to comment. The header is not considered complete without an indication of the data size, but since this is rarely known in advance, it is supplied when the sound file is closed. mus_sound_open_output function returns the associated file number.

mus_sound_close_output first updates the file's header to reflect the final data size bytes_of_data, then closes the file. The argument fd is the integer returned by mus_sound_open_output.

mus_sound_read reads data from the file indicated by fd, placing data in the array obufs as mus_float_t values (floats normally). chans determines how many arrays of samples are in obufs, which is filled by mus_sound_read from its index beg to end with zero padding if necessary.

mus_sound_write writes samples to the file indicated by fd, starting for each of chans channels in obufs at beg and ending at end.

mus_sound_seek_frample moves the read or write position for the file indicated by fd to the desired frample.


The following functions provide access to audio harware. If an error occurs, they return -1 (MUS_ERROR).

  int mus_audio_initialize(void)
  int mus_audio_open_output(int dev, int srate, int chans, mus_sample_t format, int size)
  int mus_audio_open_input(int dev, int srate, int chans, mus_sample_t format, int size)
  int mus_audio_write(int line, char *buf, int bytes)
  int mus_audio_close(int line)
  int mus_audio_read(int line, char *buf, int bytes)

mus_audio_initialize takes care of any necessary initialization.

mus_audio_open_input opens an audio port to read sound data (i.e. a microphone, line in, etc). The input device is dev (see sndlib.h for details; when in doubt, use MUS_AUDIO_DEFAULT). The input sampling rate is srate or as close as we can get to it. The number of input channels (if available) is chans. The input sample type is format (when in doubt, use the macro MUS_AUDIO_COMPATIBLE_FORMAT). And the input buffer size (if settable at all) is size (bytes). This function returns an integer to distinguish its port from others that might be in use.

mus_audio_open_output opens an audio port to write data (i.e. speakers, line out, etc). The output device is dev (see sndlib.h). Its sampling rate is srate, number of channels chans, sample type format, and buffer size size. This function returns the associated line number of the output port.

mus_audio_close closes the port (input or output) associated with line.

mus_audio_read reads sound data from line. The incoming 'bytes' bytes of data are placed in buf. If no error was returned from mus_audio_open_input, the data is in the format requested by that function with channels interleaved.

mus_audio_write writes 'bytes' bytes of data in buf to the output port associated with line. This data is assumed to be in the format requested by mus_audio_open_output with channels interleaved.

Music V

clm.c and friends implement all the generators found in CLM, a music V implementation, and clm2xen.c ties these into the languages supported by the xen package (currently s7, Ruby, and Forth). The primary clm documentation (which describes both the Scheme and Common Lisp implementations) is clm.html found in clm-5.tar.gz or sndclm.html in snd-15.tar.gz alongside sndlib at ccrma-ftp. The simplest way to try these out is to load them into Snd; see extsnd.html, examp.scm, and snd-test.scm in snd-15.tar.gz for more details. The following briefly describes the C calls (see clm.h).

clm.c implements a bunch of generators and sound IO handlers. Each generator has three associated functions, make-gen, gen, and gen_p; the first creates the generator (if needed), the second gets the next sample from the generator, and the last examines some pointer to determine if it is that kind of generator. In addition, there are a variety of generic functions that generators respond to: mus_free, for example, frees a generator, and mus_frequency returns its current frequency, if relevant. All generators are pointers to mus_any structs.

The other generators are:

Some useful functions provided by clm.c are:

and various others: see clm.h.

The more useful generic functions are:

Errors are reported through mus_error which can be redirected or muffled. See clm2xen.c for an example.


This program prints out a description of a sound file (sndinfo.c).

int main(int argc, char *argv[])
  int fd, chans, srate;
  mus_long_t samples;
  float length;
  time_t date;
  char *comment;
  char timestr[64];
  mus_sound_initialize();	    /* initialize sndlib */
  fd = mus_file_open_read(argv[1]); /* see if it exists */
  if (fd != -1)
      date = mus_sound_write_date(argv[1]);
      srate = mus_sound_srate(argv[1]);
      chans = mus_sound_chans(argv[1]);
      samples = mus_sound_samples(argv[1]);
      comment = mus_sound_comment(argv[1]); 
      length = (double)samples / (float)(chans * srate);
      strftime(timestr, 64, "%a %d-%b-%y %H:%M %Z", localtime(&date));
      fprintf(stdout, "%s:\n  srate: %d\n  chans: %d\n  length: %f\n", 
	      argv[1], srate, chans, length);
      fprintf(stdout, "  header: %s\n  sample type: %s\n  written: %s\n  comment: %s\n", 
	      timestr, comment);
    fprintf(stderr, "%s: %s\n", argv[1], strerror(errno));

This code plays a sound file (sndplay.c):

int main(int argc, char *argv[])
  int fd, afd, i, j, n, k, chans, srate, outbytes;
  mus_long_t framples;
  mus_float_t **bufs;
  short *obuf;
  fd = mus_sound_open_input(argv[1]);
  if (fd != -1)
      chans = mus_sound_chans(argv[1]);
      srate = mus_sound_srate(argv[1]);
      framples = mus_sound_framples(argv[1]);
      outbytes = BUFFER_SIZE * chans * 2;
      bufs = (mus_float_t **)calloc(chans, sizeof(mus_float_t *));
      for (i=0;i<chans;i++) 
        bufs[i] = (mus_float_t *)calloc(BUFFER_SIZE, sizeof(mus_float_t));
      obuf = (short *)calloc(BUFFER_SIZE * chans, sizeof(short));
      afd = mus_audio_open_output(MUS_AUDIO_DEFAULT, srate, chans, MUS_AUDIO_COMPATIBLE_FORMAT, outbytes);
      if (afd != -1)
	  for (i = 0; i < framples; i += BUFFER_SIZE)
	      mus_sound_read(fd, 0, BUFFER_SIZE - 1, chans, bufs);
	      for (k = 0, j = 0; k < BUFFER_SIZE; k++, j += chans)
		for (n = 0; n < chans; n++) 
                  obuf[j + n] = MUS_SAMPLE_TO_SHORT(bufs[n][k]);
	      mus_audio_write(afd, (char *)obuf, outbytes);
      for (i = 0; i < chans; i++) free(bufs[i]);


This program writes a one channel NeXT/Sun sound file containing a sine wave at 440 Hz.

int main(int argc, char *argv[])
  int fd, i, k, framples;
  float phase, incr;
  mus_float_t *obuf[1];
  fd = mus_sound_open_output(argv[1], 22050, 1, MUS_BSHORT, MUS_NEXT, "created by sndsine");
  if (fd != -1)
      framples = 22050;
      phase = 0.0;
      incr = 2 * M_PI * 440.0 / 22050.0;
      obuf[0] = (mus_float_t *)calloc(BUFFER_SIZE, sizeof(mus_float_t));
      k = 0;
      for (i = 0; i < framples; i++)
	  obuf[0][k] = MUS_FLOAT_TO_SAMPLE(0.1 * sin(phase)); /* amp = .1 */
	  phase += incr;
	  if (k == BUFFER_SIZE)
	      mus_sound_write(fd, 0, BUFFER_SIZE-1, 1, obuf);
      if (k > 0) mus_sound_write(fd, 0, k - 1, 1, obuf);
      mus_sound_close_output(fd, 22050 * mus_bytes_per_sample(MUS_BSHORT));

This is program uses the clm.c oscillator and output functions to write the same sine wave as we wrote in SndSine.

int main(int argc, char *argv[])
  int i;
  mus_any *osc, *op;
  osc = mus_make_oscil(440.0, 0.0);
  op = mus_make_sample_to_file("test.snd", 1, MUS_BSHORT, MUS_NEXT);
  if (op) 
    for (i = 0; i < 22050; i++) 
      mus_sample_to_file(op, i, 0, .1 * mus_oscil(osc, 0.0, 0.0));
  if (op) mus_free(op);

Here is the fm-violin and a sample with-sound call:

static int feq(float x, int i) {return(fabs(x-i)<.00001);}

void fm_violin(float start, float dur, float frequency, float amplitude, float fm_index, mus_any *op)
 float pervibfrq = 5.0,
   ranvibfrq = 16.0,
   pervibamp = .0025,
   ranvibamp = .005,
   noise_amount = 0.0,
   noise_frq = 1000.0,
   gliss_amp = 0.0,
   fm1_rat = 1.0,
   fm2_rat = 3.0,
   fm3_rat = 4.0,
   reverb_amount = 0.0,
   degree = 0.0, 
   distance = 1.0;
  float fm_env[] = {0.0, 1.0, 25.0, 0.4, 75.0, 0.6, 100.0, 0.0};
  float amp_env[] = {0.0, 0.0,  25.0, 1.0, 75.0, 1.0, 100.0, 0.0};
  float frq_env[] = {0.0, -1.0, 15.0, 1.0, 25.0, 0.0, 100.0, 0.0};
  int beg = 0, end, easy_case = 0, npartials, i;
  float *coeffs, *partials;
  float frq_scl, maxdev, logfrq, sqrtfrq, index1, index2, index3, norm;
  float vib = 0.0, modulation = 0.0, fuzz = 0.0, indfuzz = 1.0;
  mus_any *carrier, *fmosc1, *fmosc2, *fmosc3, *ampf;
  mus_any *indf1, *indf2, *indf3, *fmnoi = NULL, *pervib, *ranvib, *frqf = NULL, *loc;
  beg = start * mus_srate();
  end = beg + dur * mus_srate();
  frq_scl = mus_hz_to_radians(frequency);
  maxdev = frq_scl * fm_index;
  if ((noise_amount == 0.0) && 
      (feq(fm1_rat, floor(fm1_rat))) && 
      (feq(fm2_rat, floor(fm2_rat))) && 
      (feq(fm3_rat, floor(fm3_rat)))) 
    easy_case = 1;
  logfrq = log(frequency);
  sqrtfrq = sqrt(frequency);
  index1 = maxdev * 5.0 / logfrq; 
  if (index1 > M_PI) index1 = M_PI;
  index2 = maxdev * 3.0 * (8.5 - logfrq) / (3.0 + frequency * .001); 
  if (index2 > M_PI) index2 = M_PI;
  index3 = maxdev * 4.0 / sqrtfrq; 
  if (index3 > M_PI) index3 = M_PI;
  if (easy_case)
      npartials = floor(fm1_rat);
      if ((floor(fm2_rat)) > npartials) npartials = floor(fm2_rat);
      if ((floor(fm3_rat)) > npartials) npartials = floor(fm3_rat);
      partials = (float *)calloc(npartials, sizeof(float));
      partials[(int)(fm1_rat)] = index1;
      partials[(int)(fm2_rat)] = index2;
      partials[(int)(fm3_rat)] = index3;
      coeffs = mus_partials_to_polynomial(npartials, partials, 1);
      norm = 1.0;
  else norm = index1;
  carrier = mus_make_oscil(frequency, 0.0);
  if (easy_case == 0)
      fmosc1 = mus_make_oscil(frequency * fm1_rat, 0.0);
      fmosc2 = mus_make_oscil(frequency * fm2_rat, 0.0);
      fmosc3 = mus_make_oscil(frequency * fm3_rat, 0.0);
  else fmosc1 = mus_make_oscil(frequency, 0.0);
  ampf = mus_make_env(amp_env, 4, amplitude, 0.0, 1.0, dur, 0, NULL);
  indf1 = mus_make_env(fm_env, 4, norm, 0.0, 1.0, dur, 0, NULL);
  if (gliss_amp != 0.0) 
    frqf = mus_make_env(frq_env, 4, gliss_amp * frq_scl, 0.0, 1.0, dur, 0, NULL);
  if (easy_case == 0)
      indf2 = mus_make_env(fm_env, 4, index2, 0.0, 1.0, dur, 0, NULL);
      indf3 = mus_make_env(fm_env, 4, index3, 0.0, 1.0, dur, 0, NULL);
  pervib = mus_make_triangle_wave(pervibfrq, frq_scl * pervibamp, 0.0);
  ranvib = mus_make_rand_interp(ranvibfrq, frq_scl * ranvibamp);
  if (noise_amount != 0.0) fmnoi = mus_make_rand(noise_frq, noise_amount * M_PI);
  loc = mus_make_locsig(degree, distance, reverb_amount, 1, (mus_any *)op, 0, NULL, MUS_INTERP_LINEAR);
  for (i = beg; i < end; i++)
      if (noise_amount != 0.0) fuzz = mus_rand(fmnoi, 0.0);
      if (frqf) vib = mus_env(frqf); else vib = 0.0;
      vib += mus_triangle_wave(pervib, 0.0) + mus_rand_interp(ranvib, 0.0);
      if (easy_case)
	modulation = mus_env(indf1) * 
                     mus_polynomial(coeffs, mus_oscil(fmosc1, vib, 0.0), npartials);
	modulation = mus_env(indf1) * mus_oscil(fmosc1, (fuzz + fm1_rat * vib), 0.0) + 
	             mus_env(indf2) * mus_oscil(fmosc2, (fuzz + fm2_rat * vib), 0.0) + 
	             mus_env(indf3) * mus_oscil(fmosc3, (fuzz + fm3_rat * vib), 0.0);
      mus_locsig(loc, i, mus_env(ampf) * mus_oscil(carrier, vib + indfuzz * modulation, 0.0));
  if (fmnoi) mus_free(fmnoi);
  if (frqf) mus_free(frqf);
  if (!(easy_case))

int main(int argc, char *argv[])
  mus_any *op = NULL;
  op = mus_make_sample_to_file("test.snd", 1, MUS_BSHORT, MUS_NEXT);
  if (op)
      fm_violin(0.0, 20.0, 440.0, .3, 1.0, op);

The CLM version is v.ins, the Scheme version can be found in v.scm, and the Ruby version is in v.rb. This code can be run:

cc v.c -o vc -O3 -lm io.o headers.o audio.o sound.o clm.o -DLINUX

For generators such as src that take a function for "as-needed" input, you can use something like:

static mus_float_t input_as_needed(void *arg, int dir) {/* get input here — arg is "sf" passed below */}

static SCM call_phase-vocoder(void)
  mus_any *pv;
  int sf; /* file channel or whatever */
  pv = mus_make_phase_vocoder(NULL, 512, 4, 128, 0.5, NULL, NULL, NULL, (void *)sf);
  mus_phase_vocoder(pv, &input_as_needed);
  /* etc */

Michael Scholz has written a package using these functions, and several CLM instruments: see the sndins directory, and in particular the README file, for details.

Other examples

The primary impetus for the sound library was the development of Snd and CLM, both of which are freely available.

Extension Languages

Much of sndlib is accessible at run time in any program that has one of the languages supported by the xen package (s7, Ruby, Forth); the modules sndlib2xen and clm2xen tie most of the library into that language making it possible to call the library functions from its interpreter. The documentation is scattered around, unfortunately: the clm side is in sndclm.html and extsnd.html with many examples in Snd's examp.scm. Most of these are obvious translations of the constants and functions described above into Scheme. To initialize sndlib, call Init_sndlib, or, at run time, use s7's loader and s7_init_sndlib:

(let ((sndlib (load "" 
                (inlet (curlet)
                  (cons 'init_func 's7_init_sndlib)))))

Init_sndlib ties most of the functions mentioned above into the extension language (s7, Forth, or Ruby).

  mus-next mus-aifc mus-rf64 mus-riff mus-nist mus-raw mus-ircam mus-aiff mus-bicsf mus-soundfont mus-voc mus-svx mus-caff

  mus-bshort mus-lshort mus-mulaw mus-alaw mus-byte mus-ubyte mus-bfloat
  mus-lfloat mus-bint mus-lint mus-b24int mus-l24int mus-bdouble mus-ldouble
  mus-ubshort mus-ulshort

  mus-sound-samples (filename)             samples of sound according to header (can be incorrect)
  mus-sound-framples (filename)              framples of sound according to header (can be incorrect)
  mus-sound-duration (filename)            duration of sound in seconds
  mus-sound-datum-size (filename)          bytes per sample
  mus-sound-data-location (filename)       location of first sample (bytes)
  mus-sound-chans (filename)               number of channels (samples are interleaved)
  mus-sound-srate (filename)               sampling rate
  mus-sound-header-type (filename)         header type (e.g. mus-aiff)
  mus-sound-sample-type (filename)         sample type (e.g. mus-bshort)
  mus-sound-length (filename)              true file length (bytes)
  mus-sound-type-specifier (filename)      original header type identifier
  mus-sound-maxamp(filename)               returns a list of max amps and locations thereof
  mus-sound-loop-info(filename)            returns list of 4 loop values (the actual mark positions here, not
                                           the so-called id's), then base-note and base-detune
  mus-header-type-name (type)              e.g. "AIFF"
  mus-sample-type-name (format)            e.g. "16-bit big endian linear"
  mus-sound-comment (filename)             header comment, if any
  mus-sound-write-date (filename)          sound write date
  sample-type-bytes-per-sample (format)    bytes per sample

  mus-sound-open-input (filename)          open filename (a sound file) returning an integer ("fd" below)
  mus-sound-open-output (filename srate chans sample-type header-type comment)
                                           create a new sound file with the indicated attributes, return "fd"
  mus-sound-reopen-output (filename chans sample-type header-type data-location)
                                           reopen (without disturbing) filename, ready to be written
  mus-sound-close-input (fd)               close sound file
  mus-sound-close-output (fd bytes)        close sound file and update its length indication, if any
  mus-sound-read (fd beg end chans sdata)  read data from sound file fd loading the data array from beg to end
                                           sdata is a float-vector that should be able to accommodate the read
  mus-sound-write (fd beg end chans sdata) write data to sound file fd
  mus-sound-seek-frample (fd frample)          move to frample in sound file fd
  mus-file-clipping (fd)                   whether output is clipped in file 'fd'
  mus-clipping ()                          global clipping choice

  mus-oss-set-buffers (num size)           in Linux (OSS) sets the number and size of the OSS "fragments"

;;; this function prints header information
(define info
  (lambda (file)
     ": chans: " (number->string (mus-sound-chans file))
     ", srate: " (number->string (mus-sound-srate file))
     ", " (mus-header-type-name (mus-sound-header-type file))
     ", " (mus-sample-type-name (mus-sound-sample-type file))
     ", len: " (number->string
                (/ (mus-sound-samples file)
                   (* (mus-sound-chans file) (mus-sound-srate file)))))))
s7 repl and sndlib
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>

#include "mus-config.h"
#include "s7.h"
#include "xen.h"
#include "clm.h"
#include "clm2xen.h"

static void mus_error_to_s7(int type, char *msg)
  s7_error(s7,                               /* s7 is declared in xen.h, defined in xen.c */
	   s7_make_symbol(s7, "mus-error"),
	   s7_cons(s7, s7_make_string(s7, msg), s7_nil(s7)));

int main(int argc, char **argv)
  s7 = s7_init();

  mus_error_set_handler(mus_error_to_s7); /* catch low-level errors and pass them to s7-error */

  if (argc == 2)
      fprintf(stderr, "load %s\n", argv[1]);
      s7_load(s7, argv[1]);
      s7_load(s7, "repl.scm");
      s7_eval_c_string(s7, "((*repl* 'run))");


/* gcc -o sl sl.c /home/bil/test/sndlib/libsndlib.a -Wl,-export-dynamic -lasound -lm -I. -ldl -lgsl -lgslcblas -lfftw3
 * (load "sndlib-ws.scm")
 * (load "v.scm")
 * (set! *clm-player* (lambda (file) (system (format #f "sndplay ~A" file))))
 * (with-sound (:play #t) (fm-violin 0 1 330 .1))