wavexwrite.m

function wavexwrite(y,Fs,nbits,wavefile,speakers)
%WAVEXWRITE Write WAVE_FORMAT_EXTENSIBLE sound file.
%   WAVEXWRITE(Y,FS,NBITS,WAVEFILE, SPEAKERS) writes data Y to a WAVEX
%   file specified by the file name WAVEFILE, with a sample rate
%   of FS Hz and with NBITS number of bits.  NBITS must be 8, 16,
%   24, or 32.  Stereo data should be specified as a matrix with two 
%   columns. For NBITS < 32, amplitude values outside the range 
%   [-1,+1] are clipped.
%   
%   SPEAKERS is a vector of speaker numbers
%            e.g. [1,2] for stereo
%                 [1:5] for 5.1
%                   [3] for center channel
%
%   WAVEXWRITE(Y,FS,WAVEFILE) assumes NBITS=16 bits.
%   WAVEXWRITE(Y,WAVEFILE) assumes NBITS=16 bits and FS=44100 Hz.
%
%   8-, 16-, and 24-bit files are type 1 integer PCM.  32-bit files 
%   are written as type 3 normalized floating point.
%
%   Modified 2004/07/27 by Sylvain Choisel
%   to handle WAVE-FORMAT-EXTENSIBLE
%   email: sc@acoustics.aau.dk

% Parse inputs:
error(nargchk(2,5,nargin));
if nargin < 3,
  wavefile = Fs;
  Fs       = 44100;
  nbits    = 16;
elseif nargin < 4,
  wavefile = nbits;
  nbits    = 16;
end

% If input is a vector, force it to be a column:
if ndims(y) > 2,
  error('Data array cannot be an N-D array.');
end
if size(y,1)==1,
   y = y(:);
end
[samples, channels] = size(y);
if nargin < 5,
  speakerMask=2^channels-1;
else
  if (length(speakers)~=channels)
      error ('Speaker mask length must be equal to the number of channels');
  end
  speakerMask=sum(2.^(speakers-1));
end

% Determine number of bytes in chunks
% (not including pad bytes, if needed):
% ----------------------------------
%  'RIFF'           4 bytes
%  size             4 bytes (ulong)
%  'WAVE'           4 bytes
%  'fmt '           4 bytes
%  size             4 bytes (ulong)
% <wave-format>     14 bytes
% <format_specific> 2 bytes (PCM)
%  'data'           4 bytes
%  size             4 bytes (ulong)
% <wave-data>       N bytes
% ----------------------------------
bytes_per_sample = ceil(nbits/8);
total_samples    = samples * channels;
total_bytes      = total_samples * bytes_per_sample;

riff_cksize = 4+48+total_bytes;   % Don't include 'RIFF' or its size field
fmt_cksize  = 40;               % Don't include 'fmt ' or its size field
data_cksize = total_bytes;      % Don't include 'data' or its size field

% Determine pad bytes:
data_pad    = rem(data_cksize,2);
riff_cksize = riff_cksize + data_pad; % + fmt_pad, always 0

% Open file for output:
[fid,err] = OpenWaveWrite(wavefile);
error(err);

% Prepare basic chunk structure fields:
ck=[]; ck.fid=fid; ck.filename = wavefile;

% Write RIFF chunk:
ck.ID   = 'RIFF';
ck.Size = riff_cksize;
error(write_ckinfo(ck));

% Write WAVE subchunk:
ck.ID   = 'WAVE';
ck.Size = [];  % Indicate a subchunk (no chunk size)
error(write_ckinfo(ck));

% Write <fmt-ck>:
ck.ID   = 'fmt ';
ck.Size = fmt_cksize;
error(write_ckinfo(ck));

% Write <wave-format>:
fmt.filename        = wavefile;
fmt.wFormatTag  = hex2dec('FFFE');            
fmt.nChannels       = channels;     % Number of channels
fmt.nSamplesPerSec  = Fs;           % Samples per second
fmt.nAvgBytesPerSec = channels*bytes_per_sample*Fs; % Avg transfer rate
fmt.nBlockAlign     = channels*bytes_per_sample;    % Block alignment
fmt.nBitsPerSample  = 8*bytes_per_sample;
fmt.cbSize          = 22;
fmt.wValidBitsPerSample = nbits;
fmt.dwSpkMask       = speakerMask;
fmt.subFormat       = [1 0 0 0 0 0 16 0 8*16 0 0 hex2dec('AA') 0 hex2dec('38') hex2dec('9B') hex2dec('71')];
error(write_wavefmt(fid,fmt));

% Write <data-ck>:
ck.ID   = 'data';
ck.Size = data_cksize;
error(write_ckinfo(ck));

% Write <wave-data>, and its pad byte if needed:
error(write_wavedat(fid,fmt,y));

% Close file:
fclose(fid);

% end of wavwrite()


% ------------------------------------------------------------------------
% Private functions:
% ------------------------------------------------------------------------


% ------------------------------------------------------------------------
function [fid,err] = OpenWaveWrite(wavefile)
% OpenWaveWrite
%   Open WAV file for writing.
%   If filename does not contain an extension, add ".wav"

fid = [];
err = '';
if ~isstr(wavefile),
   err='Wave file name must be a string.'; return;
end

% modified 27/07/2004 sc
%if isempty(findstr(wavefile,'.')),
%  wavefile=[wavefile '.wav'];
%end
[pathstr,name,ext,versn]=fileparts(wavefile);
if (~strcmp(lower(ext),'.wav'))
    wavefile=[wavefile '.wav'];
end

% Open file, little-endian:
[fid,err] = fopen(wavefile,'wb','l');

return


% ------------------------------------------------------------------------
function err = write_ckinfo(ck)
% WRITE_CKINFO: Writes next RIFF chunk, but not the chunk data.
%   Assumes the following fields in ck:
%         .fid   File ID to an open file
%         .ID    4-character string chunk identifier
%         .Size  Size of chunk (empty if subchunk)
%
%
%   Expects an open FID pointing to first byte of chunk header,
%   and a chunk structure.
%   ck.fid, ck.ID, ck.Size, ck.Data

errmsg = ['Failed to write ' ck.ID ' chunk to WAVE file: ' ck.filename];
err    = '';

if (fwrite(ck.fid, ck.ID, 'char') ~= 4),
   err=errmsg; return;
end

if ~isempty(ck.Size),
  % Write chunk size:
  if (fwrite(ck.fid, ck.Size, 'ulong') ~= 1),
     err=errmsg; return;
  end
end

return

% ------------------------------------------------------------------------
function err = write_wavefmt(fid, fmt)
% WRITE_WAVEFMT: Write WAVE format chunk.
%   Assumes fid points to the wave-format subchunk.
%   Requires chunk structure to be passed, indicating
%   the length of the chunk.

errmsg = ['Failed to write WAVE format chunk to file' fmt.filename];
err    = '';

% Create <wave-format> data:
if (fwrite(fid, fmt.wFormatTag,      'ushort') ~= 1) | ...
   (fwrite(fid, fmt.nChannels,       'ushort') ~= 1) | ...
   (fwrite(fid, fmt.nSamplesPerSec,  'ulong' ) ~= 1) | ...
   (fwrite(fid, fmt.nAvgBytesPerSec, 'ulong' ) ~= 1) | ...
   (fwrite(fid, fmt.nBlockAlign,     'ushort') ~= 1) | ...
   (fwrite(fid, fmt.nBitsPerSample,  'ushort') ~= 1) | ...
   (fwrite(fid, fmt.cbSize,          'ushort') ~= 1) | ...
   (fwrite(fid, fmt.wValidBitsPerSample, 'ushort') ~= 1) | ...
   (fwrite(fid, fmt.dwSpkMask,       'ulong' ) ~= 1) | ...
   (fwrite(fid, fmt.subFormat,       'uchar' ) ~= 16)
   err=errmsg;
end

return


% -----------------------------------------------------------------------
function y = PCM_Quantize(x, fmt)
% PCM_Quantize:
%   Scale and quantize input data, from [-1, +1] range to
%   either an 8-, 16-, or 24-bit data range.

% Clip data to normalized range [-1,+1]:
ClipMsg  = ['Data clipped during write to file:' fmt.filename];
ClipWarn = 0;

% Determine slope (m) and bias (b) for data scaling:
nbits = fmt.nBitsPerSample;
m = 2.^(nbits-1);

switch nbits
case 8,
   b=128;
case {16,24},
   b=0;
otherwise,
   error('Invalid number of bits specified.');
end

y = round(m .* x + b);

% Determine quantized data limits, based on the
% presumed input data limits of [-1, +1]:
ylim = [-1 +1];
qlim = m * ylim + b;
qlim(2) = qlim(2)-1;

% Clip data to quantizer limits:
i = find(y < qlim(1));
if ~isempty(i),
   warning(ClipMsg); ClipWarn=1;
   y(i) = qlim(1);
end

i = find(y > qlim(2));
if ~isempty(i),
   if ~ClipWarn, warning(ClipMsg); end
   y(i) = qlim(2);
end

return


% -----------------------------------------------------------------------
function err = write_wavedat(fid,fmt,data)
% WRITE_WAVEDAT: Write WAVE data chunk
%   Assumes fid points to the wave-data chunk
%   Requires <wave-format> structure to be passed.

err = '';

if fmt.wFormatTag==1 | fmt.wFormatTag==3 | fmt.wFormatTag==hex2dec('FFFE'),
   % PCM Format
   
   % 32-bit Type 3 is normalized, so no scaling needed.
   if fmt.nBitsPerSample ~= 32,
       data = PCM_Quantize(data, fmt);
   end
   
   switch fmt.nBitsPerSample
   case 8,
      dtype='uchar'; % unsigned 8-bit
   case 16,
      dtype='short'; % signed 16-bit
   case 24,
	  dtype='bit24'; % signed 24-bit
   case 32,
      dtype='float'; % normalized 32-bit floating point
   otherwise,
      err = 'Invalid number of bits specified.'; return;
   end
   
   % Write data, one row at a time (one sample from each channel):
   [samples,channels] = size(data);
   total_samples = samples*channels;
   
   if (fwrite(fid, reshape(data',total_samples,1), dtype) ~= total_samples),
      err = 'Failed to write PCM data samples.'; return;
   end
   
   % Determine # bytes/sample - format requires rounding
   %  to next integer number of bytes:
   BytesPerSample = ceil(fmt.nBitsPerSample/8);
   
   % Determine if a pad-byte must be appended to data chunk:
   if rem(total_samples*BytesPerSample, 2) ~= 0,
      fwrite(fid,0,'uchar');
   end
   
else
  % Unknown wave-format for data.
  err = 'Unsupported data format.';
end

return

% end of wavwrite.m