Filter
Top Products
Additional technical information 37
518 User Guide
518 Operating Background
Transparency in digital systems
Digital audio gives a unique opportunity to deliver exceptional sound
quality in a recording or broadcast. In the first decade of this technology,
considerable progress was made developing the gateways – i.e. A/D and
D/A conversion. The promise of digital audio is only deliverable if the
digital audio signal is handled with complete integrity.
It is has been a common misconception among practising engineers that
the signal is robust once it is in the digital domain. This combined with a
general misunderstanding about dither, has led to working practises that
degrade the potential of the recording. Obviously, it is necessary to
understand the signal processes, and to have a way to process the signal
optimally.
The 518 allows you to achieve a transparent channel with the maximum
subjective dynamic range. This section explains the technical background.
Digital Audio and Quantisation
Digital audio is used because it is possible to transmit the data through
space (broadcast) or time (recording) without any loss of information or
generation degradation. This property of digital audio makes it extremely
attractive. The major development in digital audio has been in extending
our comprehension of the processes of digitisation and reconstruction of
the original analogue signal.
Dither
Dither is an essential lubricant if a digital audio channel that includes
processing is to remain transparent.
Transparency in a digital channel is the result of maintaining linearity at both
high and low levels and not falling prey to signal-processing distortions like,
for example, noise modulation.
It is well established that in order for linearity to be maintained, dither
must be applied at every non-trivial signal processing stage. By non-trivial,
we include all operations that modify a sample. With very few exceptions,
arithmetic operations that modify a sample result in an increase in the
number of bits representing the output result; in a fixed size channel, this
can therefore lead to truncation or quantisation effects.
In a recording process, the typical non-trivial operations will be:
• Truncation and requantisation if the A/D converter has a larger word
size than the recorder.
• Truncation and requantisation if the recorder has a larger word size
than the editor.
• Truncation, requantisation and many potential signal-processing
errors in the editor itself.
• Truncation and requantisation if the editor has a larger word size than
the final carrier.
• Truncation and requantisation mastering house processes the
material (e.g. gain shift, dc filtering etc.).
It is well established that applying dither with triangular probability
distribution of ± 1 LSB to a quantisation process results in perfect
linearity, contributes no noise modulation and permits resolution of signal
elements below the LSB. It is therefore the ideal tool to achieve
transparent processing.