ST STA013 Application Note

®
STA013 MPEG 2.5 LAYER III SOURCE DECODER
This Application Note helps the user to work with STA013 Evaluation Board, installing the PC Soft-
ware Driver, understanding how STA013 MPEG 2.5 Layer III Source Decoder works, and explaining
how to control the device functions.
The document is composed by 5 main points:
- MPEG 2.5 Layer III Technical Introduction
- STA013 Device Description
- STA013 Evaluation Board Description
- STA013 PC Software Driver Description
- General Information on STA013 Evaluation Environment
MPEG 2.5 LAYER III TECHNICAL INTRODUCTION
An Overview
Since 1988 MPEG group has been working on the standardisation of high-quality low-bitrate audio cod-
ing techniques. Two standards have been completed: MPEG 1 (coding of mono and stereo signals at
sampling rates of 32, 44.1, and 48 KHz) and MPEG 2 (backward-compatible coding of 5+1 multichannel
sound signals and low-bitrate coding of mono and stereo at sampling rates of 16, 22.05, 24 KHz).
A non standardized evolution of MPEG 2 has been developed by industry, called MPEG 2.5.
It is practically the extension to the sampling rates 8, 11.025, and 12 KHz, of the current MPEG 2 stand-
ard.
The basic idea behind perceptual coding, the general audio coding theory implemented by MPEG, is to
hide the quantisation noise below the signal-dependent tresholds of hearing. In this view, the most im-
portant question in perceptual coding is: "how much noise can be introduced to the signal without being
audible?". The most prominent feature in psychoacoustics is masking in the frequency domain. A fainter
signal is completely masked by a masker that is louder and has a similar frequency content.
The overall processing chain of an MPEG encoder can summarised in the following four steps:
•
A filter bank is used to decompose the input signal into subsampled spectral components (time to
frequency domain). Together with the corresponding filter bank in the decoder it forms an analysis-
synthesis system.
•
Using either the time-domain input signal or the output of the analysis filter bank, an estimate of the
actual (time-dependent) masked threshold is computed using rules known from psychoacoustic.
This is called the perceptual encoding system.
•
The spectral components are quantized and coded with the aim of keeping the noise, which is intro-
duced by quantizing, below the masking treshold. Depending on the algorithm, this step is done in
different ways, from simple block companding to analysis-by-synthesis system using additional
noiseless compression.
•
A bitstream formatter is used to assemble the bitstream, which typically consists of the quantized
and coded spectral coefficients and some side information, such as bit allocation information.
March 1999
APPLICATION NOTE
by Ruggero DE LUCA
AN1090
1/17