From David W. Stebbings
Washington DC
As Barry Fox correctly states, the “spread spectrum” code signal uses
psychoacoustic masking to ensure that the code is inaudible when added to music
signals at low levels. The result of this process is that the code signal is
very similar in character to the accompanying music—so similar, in fact,
that if the code is listened to on its own the associated music is easily
identified.
Consequently, when the encoded music is subject to sound-compression systems
also employing psychoacoustic techniques, the embedded signals are not
distinguished from the audio and are transmitted unharmed.
Communication theory states that the embedded energy is nearly ideally
matched to the channel and is of minimum entropy. It is therefore very difficult
to separate it from the audio signal by any process. Tests with different
encoders have confirmed this result. Furthermore, the code signal survives
processing by multiple codecs.
Advertisement
My second point concerns high-quality analogue-to-digital converters, which
give very low noise levels, such as 20 to 22-bit PCM units or the new Sony
Bit-Stream device. The BBN signals survive such methods because the
code-to-music ratio of minus 19 decibels is maintained down to the noise-floor
of the digital system. One can regard the noise-floor as a random “music” signal
with well-known statistics, and the psychoacoustic coder has the rather easy
task of ensuring that the buried data signal has the same spectral content. The
decoder also has an easy task since the “music” (in this case noise) has a very
low correlation with the wanted embedded signal.
Therefore the decoder can operate with a greater discrimination threshold
than would occur with some forms of music that may intermittently have a higher
correlation with the embedded signal. Far from being a problem, high-definition
or high signal-to-noise ratio systems benefit from the signalling methodology
employed by the proposed RIAA/BBN system without significantly altering the very
low background noise level.
