Lexicon 960L Owner's Manual page 61

960L
ate, with the amount of fluctuation depending on the
direction and strength of the reflections.
When the sound source is continuous – like legato strings,
or pink noise – we perceive these fluctuations as an
enveloping room impression. The time delay of the reflec-
tions does not matter very much, as long as they are
longer than about 10ms. (Below 10ms there are severe
combing effects we will try to avoid in this discussion.) But
most musical sounds (and all speech sounds) are not con-
tinuous.
To understand what happens with speech or music we
must learn how the brain separates sounds into streams.
Streams are the perceptual equivalent of musical lines.
Sentences from a single talker form a stream. A stream
has in general a specific source and a single continuous
semantic content. However the streams themselves are
not continuous at all – in music the streams are composed
of notes, in speech streams are composed of phones – lit-
tle bursts of sound roughly equivalent to syllables. When
we hear a string of phones, our speech apparatus goes
into high gear. First we must separate the phones one
from another, then we must use frequency and time infor-
mation to assign an identity to each phone – at which
point the phone becomes a phoneme, the basic building
block of speech. From phonemes to words, from words to
sentences, from sentences to meaning – all seemingly
effortless and automatic – our brains decode the spoken
word.
The perception of envelopment is a useful by-product of
stream formation. To form a foreground stream the brain
must separate the sound events related to a single source
from the total sonic input. To do this we must be able to
detect when a single phone starts, and when it stops.
Detecting the start of sound events is easy – we just look for
a rapid increase in level. How do we know when one
phone stops and another starts? There are only two ways
possible – we can detect the stop of a phone, or we can
assume it has stopped when we detect the start of anoth-
er. Naturally, we do both. But if we are to hear back-
ground sounds at all, we must detect the stop of phones
before a new phone starts.
How do you know if a phone has stopped? We can do an
experiment – about a 2dB drop in level in a 20ms time peri-
od seems sufficient. What if the level drops more slowly?
Experiment shows that even with a slow drop a 6dB
change is sufficient. What if the sound drops in level by
2dB, and then within 30ms comes back up again? (This
drop could be caused by a low-level reflection.) Its turns
out the level rise – if it occurs within 50ms of the first drop in
level – tends to cancel the effect of the first level drop. The
brain assumes the phone is continuing.
In general, to find the ends of phones the brain looks for a
level drop, and waits for 50ms to be sure the level stays
down. If it does, the sound event – the phone – is assumed
to have ended. Now imagine another simple experiment.
You are listening to someone talk in a noisy room. You can
easily understand the person, but you are aware of the
noise in the room - which is perceived as continuous. How
can this be? It is clear that during the phones of the per-
son who is talking you are unable to hear the room – the
phones are masking the background. Yet you perceive
the background as continuous.
The brain is clearly separating the sound of the room into
a separate stream – the background stream. The neurol-
ogy that detects the background stream works in the
spaces between phones. Thus it cannot work without the
participation of the mechanism that determines the ends
of phones. Again we can experiment. It turns out that the
background detection is inhibited during phones, as we
would expect, and is still inhibited for the first 50ms after
the end of each phone. After this time the inhibition is
gradually released, so the background detector has full
sensitivity within 150ms after the end of each phone. The
loudness of the background is then perceived through a
standard loudness integration, taking about 200ms for full
loudness to develop.
It is the background perception of reverberation that gives
us the sense of envelopment. Clearly it is the reflection
level 150ms and more after the end of sound events that
matters. Note that the relevant time is after the END of
sound events. We are conditioned by years of looking at
impulse responses to think about reflections as always
coming from hand-claps or pistol shots. In speech and
music it is the behavior of reflected energy at the ends of
sound bursts of finite length that is perceived, and the
behavior of the reflections can be quite different when the
sound event is short compared to when the sound event is
long. Notes which are shorter than the time constant of
the reverberation, ~RT/7, will excite the reverberation less
strongly than longer notes.
There is another peculiarity of background envelopment.
It depends on the absolute level of the reverberation, and
not the direct/reverberant ratio. If we play the music loud-
er the reverberation will be louder, and the sound will be
more enveloping. Thus when we mix for envelopment we
must be very careful about our monitor levels, and aware
of how loud the critical customer will play our mix.
The perception of background envelopment depends on
the same fluctuations in interaural time delay that cause
the sense of distance. But the background is only detect-
ed when the direct sound has ceased, and only reflec-
tions are hitting the ears. Thus if the reflections come from
all around us the fluctuations, and the sense of envelop-
ment will be maximum. If reflections come only from the
front we will not get much envelopment at all.
Home Next Page >>
Using the Reverb Programs
<< Go Back
5-5