Audio File Format Specifications

File Description: INRS-Telecom file
File Extension: Commonly .aud
File Byte Order: Little-endian

Prof. Peter Kabal, MMSP Lab, ECE, McGill University: Last update: 2017-01-20

INRS-Telecom Audio File Specifications

INRS-Telecom files were in use at INRS-Telecom (now INRS-EMT) and BNR (Nortel Networks). In the 1970's, a recording/playback system was put in place, first running on a Digital PDP-11 machine and later on Digital VAX machines. The first system used low-level calls for disk access and required reading/writing of files which were an integral number of blocks (512 bytes) long. The file header was devised to record the actual number of samples in the file. This file format was used for data bases for speech recognition and speech coding.

The file header is 512 bytes long. INRS-Telecom files contain only single channel 16-bit data in little-endian byte order.

Header Structure

The file header has no explicit file magic. However, INRS-Telecom files used standard sampling frequencies. As such, the bit-pattern of the VAX float representing the sampling frequency can be used as a file magic.

Very old INRS-Telecom audio files (generated on a PDP-11) had \377 bytes for unwritten data. This means that the header will be padded to 512 bytes with an all-ones pattern. As well, any bytes beyond the audio data at the end of the file will contain an all-ones pattern.

Old INRS-Telecom audio files, have an abbreviated date format (2 digit year), 8-DEC-78 16:52:50, padded out with either two zero bytes or two all-ones bytes.

Sample Files

addf8.aud (47 kB)

INRS-Telecom file, 8 kHz sampling rate

tomf8.aud (32 kB)

INRS-Telecom file, 8 kHz sampling rate. This file has 220 unused bytes at the end of the file.