SDIF: Sound Description Interface Format: Original Home Page

This is the original SDIF home page, retained for historical interest. If you want the latest on SDIF, visit the current home page.

Welcome to the home page for the SDIF standards effort. We look forward to your feedback and participation.

What's in a Name?

The "S" in SDIF now stands for "sound" rather than "spectral", reflecting changes to make the format more general and able to represent sound in non-spectral as well as spectral domains.

SDIF used to be called SDIFF (Interface File Format). The word "file" was removed because Internet applications demand sound descriptions as "streams". The SDIF standard now addresses three contexts in which sound descriptions are commonly used:

• Files
• Streams
• Embedded in other objects

SDIF Documents and Resources

Narrative

History

This effort was suggested by Xavier Rodet of IRCAM. Discussions with scholars and vendors at the 1995 ICMC confirmed the need for this standard in the academic and commercial communities . Adrian Freed coined the name and invited ICMC delegates to participate in the development work. Discussions over the months led to new requirements for the standard particularly in the area of internet applications. Also, although originally conceived of as a portable (and not necessarily efficient) interchange format, CNMAT has decided to use it as the primary working format for its sound analysis and synthesis tools. This has resulted in a standard grounded in real needs and field tested.

Definition

Example Sound Descriptions

• STFT
• phase vocoder
• tracks a la McAulay and Quatieri
• Picked spectral Peaks and their Frequency/phase/amplitude estimates
• noise
• HMM tracks
• pitch estimates
• cepstral coefficients
• resonant filter coefficients
• lpc
• formants
• diphones
• spectral envelopes
• "note lists"
• wavelets
• sampled sounds

Interchange

SDIF is an interchange format so portability is very important. To facilitate portability SDIF adopts the following existing conventions and standards:

• Data is IEEE 754 floating point. Most data is 32-bit, the most important exception being time tags which are 64-bit floating point numbers.
• Strings are null-terminated ASCII.
• SDIF data streams contain a token to identify them as such.
• Big-endian

Embedding

To faciliated embedding of SDIF data in other files or data streams, an IFF chunk is used to encapsulate SDIF data. IFF is the parent of several important standards AIFF and RIFF.

Streaming

To facilitate streaming applications, SDIF files consist of a contiguous sequence of time stamped data frames sorted in ascending temporal order.

A departure from IFF conventions has been made to support streaming applications. The size value after the "FORM" id may contain the value hex ffffffff when the size of succeeding data is unknown when the header is generated. Programs reading such SDIF streams or files are expected to use the "out of band" end of file indication provided by the operating system or network services to signal the end of SDIF data.

Efficiency

The IFF standard requires padding so that data elements are on short word boundaries (16-bit). SDIF continues in this tradition requiring padding to (32-bit) word boundaries. Padding is rarely required since most SDIF elements are 32 or 64 bits.

File Name Convention

UNIX, MAC, NT, etc. :*.sdif

DOS: *.sdf

Syntax

An SDIF chunk contains four elements:

• The 32-bit ID "FORM"
• A 32-bit size: the number of bytes in the SDIF chunk not including "FORM" or this size count.
• The 32-bit ID "SDIF"
• The SDIF frames.

SDIF frames are themselves chunks. Every frame consists of

• A frame type name (from the table below)
• A 32-bit size: the number of bytes in this frame not including the frame type name or this size count
• A 64-bit time tag
• A 64-bit ID. When frames with different times have the same ID number, it means that they are part of the same series of data. This ID number allows a single SDIF chunk to have multiple interleaved series of frame data of the same type.
• The frame data.

Arrangement and interpretation of frame data is specific to each frame type. However, most frame data will consist of a single matrix of data.

Matrix data consists of

• A 32-bit integer row count
• 32-bit integer column count
• The matrix as a vector of single-precision 32-bit, floating point values in row-major order.

Semantics

This table describes the interpretation of matrix frame data for some SDIF frame types:

Function	Tag ID	Row Interpretation	Col 1	Col 2	Col 3	Col 4
Pitch	'1PCH '	Pitches	Pitch (Hz)	Confidence (%)
Peaks	'1PIC'	Peaks	Frequency (Hz)	Phase (radians)	Amplitude	Confidence (%)
Tracks	'1TRC'	Tracks	Index (1++)	Frequency (Hz)	Phase(radians)	Amplitude
STFT	'1STF''	Bins	Real	Imaginary
Resonances	'1RES'	Resonances	Frequency (Hz)	Amplitude	Bandwidth (Hz)	Saliance (0..1)
Sound	'1SND'	Sample	Channels (ordered:left right, front, back)

Note that SDIF frame type tags all start with 1. This corresponds to the version number of the SDIF spec. and will be used to create future derivatives.

The 1++ notation indicates the tracks are numbered from 1. The notation 0..1 indicates the closed unit interval.

Example

Under Construction

Resources

Audio file formats: ftp://ftp.cwi.nl:/pub/audio/