WRID>RIFF-WAVE>fmt >

Note: all chunks are four bytes. fmt is "fmt ". The last character is a space (ASCII 32).

TODO: overview

This chunk is fiddly and it's one of two required chunks. The interpretation of data in this chunk depends on the value of the first 2 bytes, the format_tag:

A number indicating the WAVE format category of the file. The content of the format-specific-fields [ed: everything after block_align] portion of the fmt chunk, and the interpretation of the waveform data, depend on this value.

[RIFF1991]

The first several fields are consistent, followed by format specific fields. The most common layout is format_tag = 0x0001 (WAVE_FORMAT_PCM).

Several formats are documented in detail below. Additional formats are documented in [RIFF1994] . I prioritiezed documenting formats currently in use, but I probably missed some. If run across files in a format not documented below, please let me know.

WAVE_FORMAT_PCM (0x0001)

Chunk	Name	Bytes	Type	condition	Description	WRID
fmt	id	4	u8[4]	id = "fmt "		...>id
fmt	size	4	u32			...>size
fmt	format_tag	2	u16	= 0x0001	A number indicating the WAVE format category of the file. The content of the portion of the ‘fmt’ chunk, and the interpretation of the waveform data, depend on this value.	...>format_tag
fmt	channels	2	u16		The number of channels represented in the waveform data, such as 1 for mono or 2 for stereo.	...>channels
fmt	samples_per_sec	4	u32		The sampling rate (in samples per second) at which each channel should be played.	...>samples_per_sec
fmt	avg_bytes_per_sec	4	u32		The average number of bytes per second at which the waveform data should be transferred. Playback software can estimate the buffer size using this value.	...>avg_bytes_per_sec
fmt	block_align	2	u16		The block alignment (in bytes) of the waveform data. Playback software needs to process a multiple of block_align bytes of data at a time, so the value of block_align can be used for buffer alignment. The block_align field should be equal to the following formula, rounded to the next whole number: channels x ( bits_per_sample / 8 )	...>block_align
fmt	bits_per_sample	2	u16		The bits_per_sample field specifies the number of bits of data used to represent each sample of each channel. If there are multiple channels, the sample size is the same for each channel. The block_align field should be equal to the following formula, rounded to the next whole number: channels x ( bits_per_sample / 8 )	...>bits_per_sample

WAVE_FORMAT_UNKNOWN (0x0000 or default to this if unknown)

All other format_tags should follow the following structure (known as WAVEFORMATEX), which is also the structure for the WAVE_FORMAT_UNKNOWN format_tag.

Chunk	Name	Bytes	Type	condition	Description	WRID
fmt	id	4	u8[4]	id = "fmt "		...>id
fmt	size	4	u32			...>size
fmt	format_tag	2	u16		A number indicating the WAVE format category of the file. The content of the portion of the ‘fmt’ chunk, and the interpretation of the waveform data, depend on this value.	...>format_tag
fmt	channels	2	u16		The number of channels represented in the waveform data, such as 1 for mono or 2 for stereo.	...>channels
fmt	samples_per_sec	4	u32		The sampling rate (in samples per second) at which each channel should be played.	...>samples_per_sec
fmt	avg_bytes_per_sec	4	u32		The average number of bytes per second at which the waveform data should be transferred. Playback software can estimate the buffer size using this value.	...>avg_bytes_per_sec
fmt	block_align	2	u16		The block alignment (in bytes) of the waveform data. Playback software needs to process a multiple of block_align bytes of data at a time, so the value of block_align can be used for buffer alignment. The block_align field should be equal to the following formula, rounded to the next whole number: channels x ( bits_per_sample / 8 )	...>block_align
fmt	bits_per_sample	2	u16		The bits_per_sample field specifies the number of bits of data used to represent each sample of each channel. If there are multiple channels, the sample size is the same for each channel. The block_align field should be equal to the following formula, rounded to the next whole number: channels x ( bits_per_sample / 8 )	...>bits_per_sample
fmt	extra_size	2	u16		The count in bytes of the extra extensible data. The size in bytes of the extra information in the WAVE format header not including the size of the WAVEFORMATEX structure. (size of fields from format_tag through extra_size inclusive (all fields except id, size and the extra_bytes))	...>extra_size
fmt	extra_bytes	size-18	u8[size-18]		The extra information as bytes. For use when format_tag is unknown, so this portion can't be parsed deteriministically.	...>extra_bytes

WAVE_FORMAT_ADPCM (0x0002)

ADPCM is a lossy audio compression format which compresses to approximately 1/4 the size of raw PCM data. It was used in telephony and early consumer multimedia systems, and some more recent embedded devices because very little CPU time is needed to decode (compared to other lossy formats).

This variation of ADPCM is also known as Microsoft ADPCM.

The [RIFF1994] specification includes an algorithm for encoding and decoding ADPCM data.

Chunk	Name	Bytes	Type	condition	Description	WRID
fmt	id	4	u8[4]	id = "fmt "		...>id
fmt	size	4	u32			...>size
fmt	format_tag	2	u16	= 0x0002	A number indicating the WAVE format category of the file. The content of the portion of the ‘fmt’ chunk, and the interpretation of the waveform data, depend on this value. This must be set to WAVE_FORMAT_ADPCM.	...>format_tag
fmt	channels	2	u16		The number of channels represented in the waveform data, such as 1 for mono or 2 for stereo.	...>channels
fmt	samples_per_sec	4	u32		Frequency of the sample rate of the wave file. This should be 11025, 22050, or 44100. Other sample rates are allowed, but not encouraged.	...>samples_per_sec
fmt	avg_bytes_per_sec	4	u32		The average number of bytes per second at which the waveform data should be transferred. Playback software can estimate the buffer size using this value. ((samples_per_sec / samples_per_block) * block_align).	...>avg_bytes_per_sec
fmt	block_align	2	u16		The block alignment (in bytes) of the waveform data. Playback software needs to process a multiple of block_align bytes of data at a time, so the value of block_align can be used for buffer alignment. (samples_per_sec x channels) block_align 8k 256 11k 256 22k 512 44k 1024	...>block_align
fmt	bits_per_sample	2	u16		This is the number of bits per sample of ADPCM. Currently only 4 bits per sample is defined. Other values are reserved.	...>bits_per_sample
fmt	extra_size	2	u16		The size in bytes of the extended information after the WAVEFORMATEX structure. For the standard WAVE_FORMAT_ADPCM using the standard seven coefficient pairs, this is 32. If extra coefficients are added, then this value will increase.	...>extra_size
fmt	samples_per_block	2	u16		Count of number of samples per block. (((block_align - (7 * channels)) * 8) / (bits_per_sample * channels)) + 2.	...>samples_per_block
fmt	coefficient_count	2	u16		Count of the number of coefficient sets defined in coefficients.	...>coefficient_count
fmt	coefficients	4 * coef_count	(i16, i16)[coef_count]		These are the coefficients used by the wave to play. They may be interpreted as fixed point 8.8 signed values. Currently there are 7 preset coefficient sets. They must appear in the following order. Coef1 Coef2 256 0 512 -256 0 0 192 64 240 0 460 -208 392 -232 Note that if even only 1 coefficient set was used to encode the file then all coefficient sets are still included. More coefficients may be added by the encoding software, but the first 7 must always be the same.	...>coefficients

WAVE_FORMAT_IMA_ADPCM and WAVE_FORMAT_DVI_ADPCM (0x0011)

ADPCM is a lossy audio compression format which compresses to approximately 1/4 the size of raw PCM data. It was used in telephony and early consumer multimedia systems, and some more recent embedded devices because very little CPU time is needed to decode (compared to other lossy formats).

This variation of ADPCM is also known as IMA ADPCM, or DVI ADPCM. It is documented in more detail in its own specification: Recommended Practices for Enhancing Digital Audio Compatibility in Multimedia Systems

The [RIFF1994] specification includes an algorithm for encoding and decoding DVI ADPCM data.

Chunk	Name	Bytes	Type	condition	Description	WRID
fmt	id	4	u8[4]	id = "fmt "		...>id
fmt	size	4	u32			...>size
fmt	format_tag	2	u16	= 0x0011	A number indicating the WAVE format category of the file. The content of the portion of the ‘fmt’ chunk, and the interpretation of the waveform data, depend on this value. This must be set to WAVE_FORMAT_DVI_ADPCM.	...>format_tag
fmt	channels	2	u16		The number of channels represented in the waveform data, such as 1 for mono or 2 for stereo.	...>channels
fmt	samples_per_sec	4	u32		Sample rate of the WAVE file. This should be 8000, 11025, 22050 or 44100. Other sample rates are allowed.	...>samples_per_sec
fmt	avg_bytes_per_sec	4	u32		The average number of bytes per second at which the waveform data should be transferred. Playback software can estimate the buffer size using this value. ((samples_per_sec / samples_per_block) * block_align).	...>avg_bytes_per_sec
fmt	block_align	2	u16		The block alignment (in bytes) of the waveform data. Playback software needs to process a multiple of wBlockAlign bytes of data at a time, so the value of wBlockAlign can be used for buffer alignment. bits_per_sample block_align 3 (( N * 3 ) + 1 ) * 4 * channels 4 (N + 1) * 4 * channels where N = 0, 1, 2, 3 . . . The recommended block size for coding is 256 * bytes* min(1, (<samples_per_second>/ 11 kHz)) Smaller values cause the block header to become a more significant storage overhead. But, it is up to the implementation of the coding portion of the algorithm to decide the optimal value for <block_align> within the given constraints (see above). The decoding portion of the algorithm must be able to handle any valid block size. Playback software needs to process a multiple of <block_align> bytes of data at a time, so the value of <block_align> can be used for allocating buffers.	...>block_align
fmt	bits_per_sample	2	u16		This is the number of bits per sample of data. DVI ADPCM supports 3 or 4 bits per sample.	...>bits_per_sample
fmt	extra_size	2	u16		The size in bytes of the extended information after the WAVEFORMATEX structure. This should be 2.	...>extra_size
fmt	samples_per_block	2	u16		Count of number of samples per block. (((block_align - (4 * channels)) * 8) / (bits_per)sample * channels)) + 1.	...>samples_per_block

WAVE Format Tags (Categories)

Most of the fmt chunk definitions are for proprietary compressed formats from the early 90s. With a few exceptions (see above) I haven't documented them here, because I think they're extremely rare. If you're aware usage of a particlar format_tag that isn't already documented please let me know, and I'll add it to the book.

From [RIFF1991] , Registered FOURCC Codes and WAVE Formats and mmreg.h:

Learning References

Specification

• [RIFF1991] Multimedia Programming Interface and Data Specifications 1.0(1991). See pages 56-65.
• [RIFF1994] New Multimedia Data Types and Data Techniques 3.0(1994). See pages 12-22.

• WAVE and AVI Codec Registries RFC2361. This is historical, not updated after WavFormatEx was introduced in 1994
• Registered FOURCC Codes and WAVE Formats Similar to above, but from Microsoft docs.
• mmreg.h online version of MS SDK file which is referenced in the [RIFF1991] spec.