Many
of today's digital effects processors offer you considerable
control over the creation of artificial ambiences for your
music, and if you know how reverberation works in real spaces,
you'll be better equipped for designing fake ones
Reverberation
is something that few people are consciously aware of, yet
it is one of the most fundamental aspects of a room's sound
character. If you were to blindfold someone, take them to
an unfamiliar building and lead them through a succession
of rooms, clapping or shouting in each one, they would almost
certainly be able to give a pretty accurate description of
the size of each room. If they were being particularly perceptive,
they might even be able to suggest where they were standing
in each room and probably even give some idea of what there
was in terms of wall coverings, curtains, soft furnishings
and so on! In other words, it is reverberation that gives
your brain most of the information it needs to create an aural
picture of your immediate environment.
Every
room has its own sound or 'acoustic', and part of the job
of a recording engineer is to assess whether a room's characteristic
sound is worth using in a recording. If the dimensions, layout
and fabric of the room enhance the recorded sound quality,
all well and good, but if not, the microphone technique used
should minimise any room sound so that an artificial acoustic
can be added later from a reverb processor.
In
this age of digital technology, artificial reverberation is
not only more affordable than ever before, but can also be
stunningly realistic and very controllable. With a good understanding
of the physics of natural reverberation, and the fundamental
operational principles of reverb processors, it is possible
to quickly create the illusion of any acoustic environment
you can imagine.
Remember,
though, that if you choose to use artificial reverb, it is
essential that the recording has the absolute minimum of the
recording venue's room sound. If the original environment
can be heard, adding extra 'fake' reverbs will just result
in a cluttered sound and the mix will often become confused
and indistinct.
TIMING
To
understand what a reverberant sound actually is and what information
our hearing system is able to extract from it, you need to
think about how sound waves travel and what happens when they
encounter various surfaces.
The
first thing to consider is how fast sound travels in air.
My old school physics books said 760mph and the new ones probably
say 340 metres per second, but I find it hard to relate these
numbers to anything meaningful (other than the average speed
of traffic around the Evesham bypass in the morning....).
A
much more useful figure to tuck away in the dark and dusty
recesses of your mind is that sound travels roughly a foot
each millisecond. Assuming that you're not too young to relate
to feet as a valid dimensional measurement, this rule of thumb
will allow you to calculate and set one of the most critical
parameters of any reverb processor. One quick side note: the
speed of sound varies with the condition of the air. Temperature,
humidity and pressure (ie. altitude) all have significant
effects on the speed of sound, and in certain applications
the 1ft/ms guide is not sufficiently accurate. However, as
far as dialling up room sounds on a reverb unit is concerned,
it's close enough. Imagine that you're standing in the middle
of a very large, brick-walled barn with a deep covering of
straw on the ground. You have a spontaneous urge to clap your
hands: what happens? Well, the very first thing you'll hear
is the direct sound of your hand clap, and it's this direct
sound that the brain uses to pinpoint the direction of the
sound source.
Assuming
that the hand-clap radiates sound waves in all directions
simultaneously, the next thing you'll hear will be reflected
sound from the various room boundaries or nearby objects.
As the floor of this imaginary barn is covered with a deep
layer of straw, there will be no significant reflection from
the ground (although in practice this is often a dominant
source), so the first 'room sounds' will be reflections from
the side walls and ceiling. If the barn measures 40ft by 60ft
and is 20ft high, and assuming we're standing in the middle
of it, the very first reflection will come from the roof,
after about 30 milliseconds.
This
time delay can be estimated by working out the distance the
sound wave has to travel -- if you're a man standing up in
the barn, your hands and ears are likely to be about four
and six feet respectively above the floor, so (20-4) + (20-6)
= 30. Sound travels roughly 1 foot a millisecond, so it will
take 30ms for the sound of the clap to reach the ceiling and
return to your ears. Similar calculations reveal that the
side-wall reflections occur 40 and 60ms after the initial
direct sound.
"...artificial
reverberation is not only more affordable than ever before,
but can also be stunningly realistic and very controllable."
The
time taken for these first reflections to be heard goes a
long way towards defining the perceived acoustic size of the
room. Short delays imply small rooms, and long delays large
rooms. To mimic this natural characteristic, artificial reverberation
units normally allow the user to set the time delay between
the direct sound and the very first reflection with a parameter
called Pre-Delay. This is generally adjustable in millisecond
increments (or finer) over an astonishingly wide range.
So
the first reflection comes from the closest surface, and is
followed by others from the adjacent room boundaries, the
whole ensemble being known as Early Reflections. The timing,
relative amplitudes and timbre of these individual reflections
are determined by three things: the placing, angles and nature
of reflective surfaces; the location of the sound source;
and the position of the listener. Moving any of these will
change the relative timing of the early reflections quite
significantly, but our hearing system is remarkably good at
extracting even the most subtle information. Consequently,
these initial reflections and their relative timing are very
important in defining an imaginary acoustic space.
The
better reverb units allow the user to alter not only the value
of the pre-delay, but also the number, grouping, timing, amplitudes
and tonal qualities of the first reflections. In some cases,
these parameters are preset by the manufacturer and are simply
selected from a list of programmed options, such as Hall,
Chamber, or Plate, although occasionally they are derived
from measurements taken in genuine acoustic environments.
The most sophisticated machines allow the user to specify
the cubic volume of the imaginary room, or even its precise
dimensions, together with the source and the listening positions!
TAILS
So first we hear the direct sound; then, a short time later,
a number of discrete reflections return from the various surfaces
in the room. However, these reflections don't just stop when
they reach the listener -- they continue until they reach
other surfaces, where they instigate more reflections. These
reflections start even more reflections and the sound density
becomes too great to allow us to distinguish the reflections
as separate events.
At
each reflective surface, some of the sound energy is absorbed,
and more is lost as the sound travels through the air -- which
is why reverberation gradually dies away. This reverberation
'tail' may last for anything from, say, 0.3 seconds (for a
dead-sounding room), through to several seconds in a church
or big concert hall.
The
length of the reverberation tail is usually specified in terms
of its 'RT60'. This is defined as the time taken for the reverberation
to fall by 60dB in level below the original direct sound.
Every reverb unit allows this time to be adjusted, normally
through a parameter called Decay Time.
It's
important to note that the reverberation tail lasts for different
durations at different frequencies. High-frequency sound waves
have a lot of trouble persuading air molecules to vibrate
quickly enough to pass the sound energy onwards. Consequently,
high-frequency sounds tend to die away, as they travel, much
faster than mid-frequency sounds. On top of that, high-frequency
sounds are absorbed by soft furnishings (which includes people
and even wallpaper!). On the other hand, high frequencies
reflect strongly from a wide range of surfaces, such as windows,
sound desks, equipment racks, and so on. At the other frequency
extreme, low frequencies are only reflected by large and very
solid objects, so there may be little LF in the reverberation
at all in some circumstances, but a definite bass 'bloom'
in cave-like rooms!
To
help provide this level of realism, most reverb units allow
you to adjust the reverberation time for high (and sometimes
low) frequencies relative to middle frequencies, and introduce
some kind of overall equalisation to the reverberation tail.
KNOBS
Let's
recap on the parameters that today's digital reverb processors
are likely to offer for the simulation of real acoustic spaces.
Firstly, although I haven't previously mentioned it, there
is usually a means of balancing the direct sound against the
reverb. The direct sound is often referred to as 'dry' and
the reverberation as 'wet', so a wet/dry control will probably
be in there somewhere. Some of you will know how unpleasant
it sounds when an analogue-to-digital (A/D) converter is overloaded,
so a critical control on digital reverbs is the input level
control, and its associated headroom meter.
The
first control which defines the reverb character is pre-delay,
which effectively defines the distance of the first reflective
surface. This will be followed by one or more parameters for
controlling the number, timing, amplitude, and timbre of the
other early reflections. Some machines provide controls called
'Pattern', 'Level' and 'Room Size'; others might simply offer
preset venue simulations ('Hall', 'Chamber', 'Jazz Club',
and so on).
After
the early reflections, the reverberation tail is set by a
control for overall decay time. This is normally accompanied
by a parameter that adjusts the relative decay time at high
(and perhaps) low frequencies. There's usually also some means
of setting overall tonal characteristics, although this may
be little more than a simple bass and treble equaliser.
Finally,
having skilfully set all of these parameters to create a wonderfully
believable artificial acoustic(!), you can usually store your
fake room in one of a number of user memories.
CREATING
SPACES
You cannot create an artificial acoustic space if you don't
know what a real one sounds like to start with, and the only
way you can find out is to actively listen to sounds in as
wide a range of environments and circumstances as possible.
Everyone has a very detailed subconscious knowledge of how
different rooms sound, and although few are able to analyse
the reverberation structure, most spot incongruities in artificial
reverberation very easily.
It's
very revealing and informative to consciously listen to the
sound of different rooms as you move around in a building
-- even in places with which you thought you were familiar.
Try to analyse in your own mind what sort of pre-delay, decay
time, early reflections and high-/low-frequency decays naturally
occur to create the 'sound' of that room. Don't just listen
to indoor reverberations, either -- try to assess the reverberant
features of the local high street, the great outdoors, a wood
or forest, or wherever you happen to be. You will find reverberation
in places you didn't expect it, and may be surprised to discover
that places you assumed to be reverberant actually are not!
MONO
AND STEREO
The very nature of genuine reverberation is that it tends
to come at you from all over the place, but particularly from
the sides of the room. This has significant effects on the
compatibility between stereo and mono versions of your mix,
since the mono listener is effectively denied any information
from the sides of the stereo image.
To
see how this happens, consider a simple M&S (Middle &
Side) stereo microphone technique being used to record something
in a reverberant room: the stereo listener hears the full
acoustic in all its glory, but the mono listener hears only
the forward-facing 'M' microphone, not the sideways-facing
'S' microphone -- and guess which one picks up the bulk of
the room sound? This absence of reverberation in mono afflicts
artificial reverb processors as well as natural acoustics.
In practice, the amount of reverberation heard in mono may
be substantially less than that in the stereo balance, and
if mono listeners are likely to be an important part of your
music audience, always check for mono compatibility. In general,
you almost always have to compromise the balance in some way
because either the mono will be too dry, or the stereo will
be too wet!
"You
cannot create an artificial acoustic space if you don't know
what a real one sounds like to start with...""You
will find reverberation in places you didn't expect it."
Tricks
worth trying include reducing the stereo width of the reverb
(turn the pan-pots in a bit towards the centre instead of
having the reverb returns running out to full left and right),
or mix in a small amount of reverb from another reverb processor,
panned centrally. The extra reverb should be set up with the
same parameter values as the stereo reverb, although a slightly
shorter pre-delay and longer HF decay time often work well.
The balance between the dry sound, the mono reverb, and the
stereo reverb needs to be adjusted carefully, while you continuously
switch between mono and stereo listening to find the most
uniform results in the two modes.
In
matrix surround systems (such as Dolby Surround), real or
artificial stereo reverb tends to spread across the rear channel
quite naturally as a result of the way in which the rear-channel
information is encoded and decoded. Altering its stereo width
controls the front-back balance, narrowing the reverb pulls
it to the front, and increasing the width pushes more to the
sides and rear.
Many
stereo digital reverb units have a single input and a stereo
output, and this often causes people to wonder how the reverberation
can be 'true stereo' with only a mono input. The answer is
simple if you consider the real situation of a sound source
within a reverberant space.
If
someone claps, there's only one sound source, yet the reverberation
will come from all directions and could be captured by a simple
stereo microphone array -- a mono input to the room and a
stereo output from it. Of course, in a more complex situation
with, say, a string quartet in the room, there are multiple
sound sources and each will have slightly different pre-delays
and early reflection patterns, but this is usually a very
subtle distinction, and in practice the mono-in, stereo-out
system of most digital reverb units works perfectly adequately.
Something
few people ever check is the line-up of a stereo reverb unit.
However, it is a stereo source and should be treated in just
the same way as any other stereo signal, which means making
sure that the left and right reverb outputs have the same
gain and equalisation through your mixer. I find that a quick,
easy and reliable method of doing this is to simply dial up
a 3- or 4-second decay time and send a brief burst of signal
into the machine. Listen carefully to the dying reverb tail:
it should decay centrally, possibly even becoming narrower
in width as it goes (although this depends on the particular
algorithm). If the reverb tail appears to collapse towards
one side or the other, your return channels have different
gains and should be adjusted.
CHOOSING
AND USING REVERB
In general, two reverb units will meet the needs of pretty
much every recording situation. One machine would normally
be set for a short, bright sound (perhaps a plate setting)
for percussive sounds, whilst the other would be set to a
longer, warmer patch, providing a 'lush' quality for vocals
and solo instruments. You could also try passing some instruments
through both reverbs (percussive one first) for a third alternative.
Some
engineers like to use several reverbs to create a layering
effect, but I generally find that this approach causes a loss
of definition and adds confusion to the overall sound. Going
back to the idea that the artificial reverb is merely replacing
the poor acoustics of a less-than-ideal recording venue, it
could be argued that there should only be one reverberation
sound for everything, as would be the case if the musicians
all played live in the same reverberant room!
The
next issue to address is how much reverb to use. The classic
mistakes of the novice are using too much reverb return on
everything and allowing reverb tails to be too long. Reverb
generally needs to be subtle, and ideally only the loudest
musical peaks should cause obvious tails. Even the biggest
halls rarely have a reverb time in excess of four seconds,
and often a two-second decay time is easily long enough.
The
choice of reverb parameters is dependent on both personal
taste and the nature of the programme material, so it's impossible
to give specific recommendations, but try to create life-like
environments wherever possible. Most reverb units offer a
number of special effects, such as gated or reverse reverbs,
and these are best used sparingly, so that they keep their
impact. While we're on the subject of special effects, it's
worth trying out the pseudo-reverberation programs too. Algorithms
such as 'Ambience' or 'Alive' can often add extra definition
and life to dull vocals, or spice up closely-mic'd solo instruments
without your having to resort to using those horrible exciters
(just a personal opinion, of course...).
Normally,
reverb sends are taken post-fader, so that direct signal level
adjustments are reflected in their reverb returns. However,
it can often be useful to send pre-fader, and not allow any
direct signal into the final mix at all. This is particularly
effective with sustained keyboard string sounds and the like,
where it helps to make less-than-ideal synth sounds blend
a lot more smoothly.
Another
useful trick is to set up a reverb specifically for the keyboard
sounds, and route the reverb returns through a chorus unit.
This provides a completely different kind of sound to chorusing
the keyboards directly and adds an interesting 'swirling'
quality which can be very effective if used discreetly.
Although
reverb processors are most important during mixdown, they're
also vital during recording, especially when recording vocals.
Many singers have enormous trouble pitching properly without
reverberation and it's essential to have the ability to route
reverb returns to the headphone monitor mix. The reverb setting
for the cue monitor is not particularly critical to the performance
(provided it is broadly appropriate) and need not be recorded,
although some engineers do like to record voice and reverb
together (occasionally as a complete mix but more usually
on adjacent channels on the multitrack machine). This is particularly
useful if the reverb plays a part in the performance (through
timing or percussive vocal effects, for example).
As
artificial reverb becomes more and more elaborate, there's
a trap which many engineers find themselves falling into.
It's possible to become so engrossed in adjusting each parameter
minutely, trying every possible combination along the way,
that you lose sight of the original idea. The best way of
getting the sound you want, quickly, is to understand the
nature of real reverberation and apply that knowledge to creating
the acoustic space you've imagined. It's far better to think
for a minute or two, and then dial the right numbers in, than
to sequentially try every preset on the machine, hoping to
stumble across something that sounds OK.
ONE
ALGORITHM OR TWO?
A word of warning -- not every reverb processor is as flexible
as it might seem. Particularly with multi-effects units, it
is quite common to find that there is actually only one reverberation
algorithm. The wide range of supplied preset environments
(Hall, Room, Plate, and so on) is actually composed of variations
in the delay, decay and EQ settings of a single algorithm.
In these cases, you'll find that no matter how you adjust
the reverb parameters, all settings sound very similar: the
overall character of the room does not seem to change, and
this is because the pattern of the early reflections remains
fixed. The better machines have a number of different algorithms
and a variety of early-reflection patterns, which allow a
larger range of different room types to be created, each with
distinct and individual sonic characters.
Fortunately,
there is an easy way to find out which category a particular
machine falls into. Select two, theoretically diverse, programs
-- perhaps a Hall and Plate. Set the delay, decay, EQ and
any other parameters to identical values and store the new
settings in a couple of user memories so that they can be
recalled easily. Next, listen critically to the quality of
the reverberation while switching between the two presets.
There should be an obvious difference in the character of
the room acoustic if the machine uses different algorithms,
with different early reflection patterns. (Try closing your
eyes and imagining the dimensions and furnishings of the fake
room.) If you cannot spot any differences, the chances are
that the machine uses the same algorithm for all its reverb
programs.
THE
MUSEUM OF ARTIFICIAL REVERBERATION
Artificial
reverb has been around for more or less as long as people
have been performing in non-ideal acoustic environments. In
more recent times, however, various electronic and electro-mechanical
methods have been developed, although none was as effective
as the current generation of digital designs.
One
of the simplest and most obvious systems was the echo room
-- literally a room, often tiled and full of ceramic sewer
pipes to provide a wealth of reflective surfaces. A loudspeaker
generated the direct sound, and one or more microphones collected
the resulting reverberation. The echo room has the advantage
that the reverb is naturally very complex, but it is also
difficult to adjust, and requires a large and quiet room!.
One
of the first electro-mechanical systems (and one which remains
popular to this day) is the plate. This employs a large sheet
of metal (typically 6ft by 4ft) suspended on springs within
a sound-deadening case as a reverberant space. A vibrating
transducer feeds the direct sound into the metal plate and
a pair of pick-ups extract the reverberation as the vibrations
bounce off the plate's edges. A motorised damping plate parallel
to the main one can be remotely positioned at varying distances
to control the duration of the reverb. The plate has a characteristic
metallic, bright, sound quality which has become intimately
associated with pop music. Virtually every digital reverb
I have used includes a simulation of the humble plate -- which
is a very good indicator of just how popular this mechanical
system remains!
Another
enormously popular electro-mechanical system is the spring-line
reverb. This technique has been around for a very long time,
and I'm sure everyone has come across guitar amps with spring
reverbs installed. The operating principle is similar to the
plate, in that a transducer sets up vibrations in a spring,
which rattle back and forth, to be extracted by a pick-up
at the other end. The character of a particular spring reverb
unit is fixed (other than the wet/dry balance), but can be
optimised for the sound source at the design stage by careful
choice of the number, length, diameter and compliance of the
spring(s).
All
manner of record-replay systems have been developed to provide
a reverb effect, but none have survived the digital revolution.
The earliest ideas simply used a three-head tape machine,
where the direct signal was recorded onto tape and the replay
signal provided the reverberation. The tape speed and head
spacing determined the pre-delay and if some of the replay
signal was mixed with the direct signal, a pseudo-reverb could
be created. The results are hardly realistic, but the system
was popular at a time when the alternatives were too expensive
or impractical.
The
record-replay theme was further developed into machines like
the WEM Copycat and the Roland Space Echo, which used tape
loops and multiple replay heads, with the ability to adjust
the contribution and feedback of each head -- but then solid-state
technology arrived...
Bucket
Brigade systems became popular for a brief time (fortunately)
just before the first true digital reverb machines hit the
market. Bucket Brigade delays were a halfway house between
analogue and digital systems, but were no more realistic or
flexible than the earlier tape-loop products -- and were often
a lot noisier!
The
advent of digital technology really revolutionised artificial
reverb, basically because the time-domain signal processing
of digital audio lends itself very well to the kind of sound
manipulation needed to create realistic reverb. Creating a
pre-delay is simply a case of storing sound in a memory until
the required time has passed. The early reflections are created
by replaying the direct sound repeatedly at suitable moments,
with level and equalisation changes as necessary. The main
body of the decay is created by cycling the direct sound through
a complex set of short feed-back and feed-forward delays,
configured to introduce the desired equalisation characteristics.
Digital
reverbs are available to suit all pockets from a large number
of manufacturers, including Yamaha, Sony, Digitech, Klark
Teknik, ART, Ensoniq and Alesis, to name but a few. However,
if you ask any professional sound engineer to name their favourite
reverb machine, chances are you'll hear one name above all
others. Lexicon are probably the most popular manufacturer
of digital reverbs, and their product line extends from push-button
preset units to the most complex state-of-the art systems.
Published
in SOS October 1997. |