This document describes a soundfile information chunk that may be added to AIFF, AIFF-C and WAVE soundfiles to indicate to peak amplitude of the soundfile. This information can be used to facilitate soundfile display, quick normalization, and the playback of floating point files.

The PEAK chunk structure:

typedef struct
{
    char          chunkID[4];     /* 'PEAK' */
    unsigned long chunkDataSize;  /* the size of the chunk */
    unsigned long version;        /* version of the PEAK chunk */
    unsigned long timeStamp;      /* secs since 1/1/1970  */
    PositionPeak  peak[CommonChunk.numchannels]; /* the peak info */
}   PeakChunk;

<chunkID> is the character array 'PEAK', for chunk identification.

<chunkDataSize> is the size of the data section of the chunk. It does
 not include the 8 bytes used by chunkID and chunkDataSize.

<version> indicates the version of the PEAK chunk. This allows anything
 after PeakChunk.version to be redefined for future needs. This document
 describes version 1.

<timeStamp> is the number of seconds since 1/1/1970. This is used to
 see if the date of the peak data matches the modification date of
 the file. If not, the file should be rescanned for new peak data.

<peak> is a PositionPeak structure. There is a PositionPeak structure
 for every channel. These are in the same channel order as the samples
 are interleaved.

The PositionPeak structure for version 1:

typedef struct
{
    float         value;    /* signed value of peak */
    unsigned long position; /* the sample frame for the peak */
}   PositionPeak;

<value> is the signed peak value. A value of 1.0 represents 0dBFS for
 non floating point encodings.

<position> is the sample frame number at which the peak occurs. Note
 that the unit for position are sample frames, not sample points nor
 bytes.

Thanks to Richard Dobson, Martin Puryear and Matt Ingalls for their help on this project.
Please send any comments to me at tre@shoko.calarts.edu