Omnia Audio Omnia.9/XE - Manual (Page 101)

The Effects of Audio Processing on Encoded Audio
Streaming Isn’t FM
Processing for streaming audio is very different than processing for FM. Unfortunately, many
broadcasters fail to realize just how different it is, and the results can easily be observed in arguably the
majority of streams delivered by traditional terrestrial broadcasters.
Unlike FM transmission where loudness is realized at least in part by clipping, the input to a streaming
encoder must NEVER be clipped.
At first glance, this might seem like an easy enough goal to achieve: Simply set the output of the
streaming processor at such a level as to avoid clipping the input of the encoder. But upon closer
examination and with an understanding of how bit rate-reducing (lossy) codecs to their job, this simply
isn’t enough.
What Encoders Like
Lossy encoders (including the MP2, MP3, AAC, and WMA encoders in Omnia.9/XE) perform best when
the incoming audio has short term dynamics, aka “peaks”. This is where they tend to hide any audible
encoding artifacts. When the content lacks short term dynamics, the artifacts are more readily revealed.
Therefore, it is not only unnecessary to make sacrifices in your processing for the sake of loudness, it is
unwise as well.
With all processing - including FM - there is a point of diminishing return where a small amount of
increased loudness will come at the price of a large drop in openness and quality. Depending upon the
market, it it sometimes necessary for an FM station to walk right on that fine line to stay competitively
loud. Thankfully it is possible to step back from it with streaming audio deliver a product that is plenty loud
and significantly more enjoyable to listen to than FM radio.
How Listeners Listen
We are not suggesting that you feed a quiet stream. Not at all! But unlike the habits of FM radio listeners
who are able to quickly punch up stations with presets on their car stereos, changing streams requires
more time. By the time the new stream buffers and loads, any comparative loudness opportunities have
been lost (assuming the new stream is within a couple of dB’s of the previous one).
The Benefits of Undo
The very widespread problem of source material that has been dynamically over-processed and even
clipped during the mastering process is well known.
We have already established that content with short term dynamics is better received by streaming
encoders than content with little peak content. We can control the amount of additional processing that we
apply, but what can be done about the source material?
Omnia.9/XE’s solution uses the same “Undo” process found in the Omnia.9 hardware processor. It is a
two-stage process involving de-clipping and multiband expansion. The audio is first examined for clipping.
If it is present, the de-clipper intelligently recreates a real peak in place of the flat-topped signal. Next, a
multiband expander restores some dynamic range to the signal.
The Effects of Audio Processing on Encoded Audio
101