Sound on Sound, August 2011
input signal, or listen to the difference between the input
and encoded signals (in other words, the audio that the data
compression is chucking out) by clicking on the Diff button.
Although part of what audio data‑compression schemes
do is completely lossless (in other words, removing data
redundancy on a purely mathematical level), this typically only
cuts file sizes roughly in half. To carve anything else away, you
need to start reducing the resolution of the audio information,
which inevitably adds noise and/or distortion to the signal.
The purpose of psychoacoustic encoding is to funnel these
potentially undesirable artifacts into areas of the frequency
spectrum where the human hearing system is least likely to
perceive them, and it’s because this bit of smoke and mirrors
is so effective that typical MP3 files can be 10‑12 times smaller
than their uncompressed PCM source files. However, the more
you try to squish the data, the more the processing artifacts,
as showcased in the Difference signal, start to impinge
unmusically on the listening experience. The aim of this
plug‑in is to help you decide which codec produces the most
acceptable trade‑off between audio enjoyment and file size.
While final quality judgements should clearly be made
by ear, Fraunhofer Pro‑Codec does, nonetheless, provide
a great deal of supplementary visual information to support
the auditioning process. Central to this is the multi‑function
display area that dominates the centre of the GUI. By default,
this shows a zoomable, real‑time FFT frequency analysis (the
yellow line) of either a single selected input channel or the
mono sum of all of them. The red shaded region beneath the
yellow line represents the spectral content of the Difference
signal, while a set of green lines above it indicates the Noise
to Mask Ratio (NMR), a theoretical measurement of how
effectively any data‑compression nasties are being concealed
or ‘masked’ by the spectral energy of the music itself. The
higher up the display the NMR line reaches in any given
spectral band, the more likely it is that the listener will detect
processing artifacts in that frequency region. Although Sonnox
admit that a given individual’s subjective impression of artifact
levels inevitably won’t correlate exactly to the calculated NMR
readings, the dancing green line is still useful for focusing your
attention towards the most likely spectral ‘danger areas’, as far
as encoding quality is concerned.
When you’re working with a stereo input, the Display
Type button accesses no fewer than three circular meters.
The left‑hand one is a Lissajous display with pink and yellow
traces, allowing comparison of the input and output stereo
Not sure you can tell the difference? Pro‑Codec can set your
mind at rest with its two blind‑testing modes: AB and ABX.
The first simply lets you toggle between auditioning any pair
of codecs (or any codec and the input signal) from a single
button. Normally the plug‑in’s GUI makes it visually clear
which signal you’re listening to, but if you don’t want that
distraction, the Hide button banishes all visual cues. The ABX
mode provides a more formal extension of this idea, offering
a proper ABX testing scheme which will inform you in concrete
terms whether your preference for one of any pair of the
available signals is statistically significant. I love this feature
— so much so, in fact, that I wish Sonnox would issue it on its
own as a stand‑alone plug‑in, so that I could test other plug‑in
processes with it too!
Blind Testing
69
www.soundonsound.com / August 2011