THE ACOUSTICAL DESIGN OF RECORDING STUDIOS
When one considers that the recording industry has been building and
using studios for about 70 years, it is remarkable that so little basic
theory has been published on the subject. To be sure, there are plenty
of "here's how we did it" articles in print as well as a number of
"here's how to do it" examples to be found in books and magazines,
but none of these provide enough of the underlying design principles
to enable a reader to duplicate the performance of such studios unless
he also duplicates the studio. For that reason, while such publications
are interesting and even entertaining, they are of little use to a
studio owner who wants to improve an existing room or build a new one
in a space different from the exemplars given.
The situation grows even more extraordinary in light of the massive
amount of experimentation and research devoted to control rooms over
the past few years.
Those efforts have resulted in enough published material to allow a
studio to select from at least two demonstrably excellent generic
control room designs, both of which spring from the same clearly
expressed theoretical underpinning. While the general case design
will need some cleanup and tuning to achieve optimum results, a
studio owner can use the published theory to modify a given plan,
adapt it to his particular situation, and come up with a fundamentally
decent room. In short, we know how to build good control rooms.
We sure as hell don't know how to build good studios. In fairness,
there are some designers who appear to know something of the subject,
but they don't give away their stock in trade, so a studio owner is
faced with the problem of separating the good designers from the good
talkers. With the near future of his business at stake, that's a
serious problem, made worse by the fact that even very good acousticians
have been known to make very bad mistakes when dealing with recording
rooms.
As an example, the two worst studios the writer has ever encountered
were designed from scratch by a Phd named Sabin. (There were three of
them.) The rooms were retreated within a couple of months, but we
turned out some pretty marginal work in the meantime. Since this
happened in the city's premier recording facility, marginal was a bad
case of egg on face.
Doc Sabin was not at fault in that mess. In fact, nobody was. The whole
thing was a terrible mistake.
The mistake was to confuse a recording studio with a normal acoustical
envronment.
Acousticians ordinarily think of large rooms in terms of theaters and
auditoriums, which have a definite sound source feeding a definite
audience. That applies equally to control rooms, theaters, auditoriums,
and almost everything else acoustical designers get into.
It does not apply to studios, which are completely different from most
other rooms. A studio has any number of sources in the persons of the
musicians, and an audience comprised of those same players. Multiple
scattered sources, ditto listeners. Peculiar room.
Keeping strictly to acoustical performance, the primary function of a
recording studio is to provide adequate isolation between microphones
while allowing the players to hear each other as well as possible.
Acoustical isolation is by far the most often discussed of these two
areas, but since the parameters involved are addressable by acoustical
mathematics, producing satisfactory isolation levels is a fairly
straightforward process. All that's needed is a knowledge of what
constitutes adequate isolation and several pages of mathematical
computations. Happily enough, there is a way round that last item.
Treating a room for multidirectional listenability is a good deal more
difficult, as it is not a direct function of the room's global
characteristics and therefore cannot be treated mathematically. General
solutions are available, and they work nicely, but they have more to do
with old fashioned intuitive acoustics than with the glitzy new computer
aided stuff.
Among other things, this means that the merits of a suggested treatment
cannot be readily confirmed by punching up one's handy dandy number
cruncher.
Getting on with it, the specific design parameters are:
1: What is a reasonable isolation level and how much treatment is needed
to get it?
2: What constitutes acceptable listenability and how is that managed?
3: What's the catch? (There's a bear in every woods.) ((Sometimes several
bears.))
Item one. Acoustical isolation between instruments is a function of the
degree to which the sound of one dies away before getting to the next.
When the die off is inadequate the sound of one instrument falls through
the mike of the next and trashes it.
If it is excessive the musicians can't hear each other properly, which
makes group playing difficult and ruins section sound. Everything in the
real world is a compromise, and acoustical isolation is no exception.
The amount of acoustical attenuation for a given instrument in a room
depends on the room's global characteristics. As with any radiated
field, sound pressure levels diminish as the square of the distance
from the source. Double the distance, lose 6 Db SPL. The equation holds
for any distance in a perfectly dead room or out of doors.
In a normal room, however, the walls reflect some of the sound. Since the
source sound level diminishes with distance, at some point the reflections
from the walls will equal the source level. Beyond that point the sound
no longer dies away, and the level becomes constant at any further
distance. The distance at which this transition takes place is currently
called the Critical Distance. It has been called other things in other
centuries, as it's existence has been known for a very long time. It is
easy to observe, easy to measure, and a remarkably accurate indicator of
a room's acoustical performance.
The Critical Distance (Dc) of a sound source depends on the reflectivity
of the walls and how much wall surface the sound hits. As an example, a
firecracker hung on a string in the middle of a room produces a spherical
sound field which will bounce off all six walls of the room. Six, because
sound has no sense of direction, and can't tell a floor or ceiling from
any other surface. This spherical source is assigned a "Q" (figure of
merit) of 1, meaning it has no directionality at all. Hang the cracker
against a wall, and it radiates a hemispherical pattern. That's a Q of 2.
Halfway up the wall and in a corner it's a half hemisphere, and the Q is
4. On the floor and in a corner, Q =8. Q represents the beam width of the
sound source. The higher the number, the narrower the beam.
The narrower the beam, the less wall surface is struck by a source's
sound. Therefore, the higher the Q, the longer the Dc. And the higher
the surface reflectivity, the shorter the Dc.
It follows from the above that low Q instruments will have the shortest
Dc's, and the poorest isolation. As it happens, low Q describes both the
human voice and the entire rhythm section. A moment's thought will
explain that. Bass, piano, guitar and drums were used to accompany the
human voice for several centuries before mikes and such were invented,
and were designed to match it. They match quite well, which leaves us
with a kit of Q 2.5 instruments as the basis of isolation design.
The most difficult instrument in terms of isolation is the voice. Not
because it's so soft, but because of limiting. Unless a studio wants
to turn out 1940's records, there is no choice but to limit vocals,
and the limiter costs about 12 Db of isolation as it pulls up the
consonants in the singer's words. What this amounts to is that the vocal
channel should show something approaching minus 20 Db when the vocalist
is quiet; 12 db for limiting, and 10 to 14 to clear the consonants and
allow a little dynamic range for the singer. Since other instruments
work nicely with a clearance of 6 to 10Db adequate vocal isolation
becomes the criterion for acoustic design in studios.
Vocal isolation is made a little easier by the small size of the
instrument, which allows miking at a half foot without running into
serious proximity effects, and generally presents about 86 Db SPL on
mike. Hardly thunderous, but the peak level differences between voice
and the rhythm instruments are not as great as commonly assumed. It's
limiting up the minus 12 Db consonants that give rise to vocal iso
problems. Still, since other instruments can be 6Db or more over the
voice's 86 Db, the room characteristics have to lay for about 26 Db
of acoustical loss from a vocal to mike distance of six inches to any
other mike twenty acoustical feet away, keeping in mind that the 20
feet may simply be a few feet in front of the vocal mike. It is not a
straight line measurement.
The problem is made harder by a simple but nasty fact. A SOURCE GOES
CONSTANT VOLUME AT IT'S Dc, AND THE VOLUME IS THE SAME EVERYWHERE IN
THE ROOM. Distance beyond Dc makes no difference in fallthrough, and
directionality has no effect, except to mud up the fallthrough.
Dead flats don't work. Hyper cardioid mikes don't work. Nothing works.
The levels of the rhythm instruments have to fall about 26 Db before
going constant volume, or you can't work a vocal anywhere in the room.
Item one is 26 Db. More is nice, but getting much over 26 in a small
room involves so much treatment that the studio turns into an anechoic
chamber, with fuzz covering every wall.
Which brings up item two; listenability of the room.
Totally fuzzed walls return no sound to the players, who respond by
playing louder. And worse.
It's very difficult for a group of musicians to work in concert (pun
intentional) unless they can hear each other. Outdoors, with nothing
otherwise coming back to the players, stage monitors are used to supply
the sound of the group to the group. An engineer can use the studio
playback speakers for the same purpose, and it works surprisingly well,
but in both cases the players hear the mixer's balance, not their own.
They play in one balance and hear another, which creates some subtile
but nasty musical corruptions.
As an example, no mixer will let a solo ride at too low a level. It's
the mixer's job to maintain a proper balance, and mixers do their jobs.
So if a musician plays a tentative first solo, the mixer raises it's
level as needed, and it plays back in proper balance. After a few takes,
the soloist gets used to the idea that his solos will come out right no
matter what he does, and lays back on all of them. It's easier to play
soft. The mixer also adjusts to the situation, and bumps the level for
each solo. All this sounds pretty good at the time, but after a few days
both parties discover that the solos don't sound like solos. They sound
like lifted fills. That's because a solo is generally a high energy item,
and when a player lays back rather than putting out the energy, the solos
lack drive and intensity.
Technically speaking, this is a matter of harmonic content. When an
instrument is played hard or loud, the energy shows up as an increase
in harmonics, and the result is a loud sound. When not, not, and
artificially boosted soft solos just don't make it.
While this is one of the less obvious problems involved, musicians who
can't hear their overall sound well enough to maintain solo and section
balances during performance are very unlikely to play at their full
potential, and they need to hear themselves directly. That's especially
true if the mixer is tricking up the sound as it goes through the console
to the tape.
Since the usual studio setup points the musicians and their instruments
at the control room, the obvious (and normal) way to supply direct
feedback is to bounce the player's sound off the control room wall.
Control room walls are left reflective as a matter of conventional wisdom,
and are even somewhat optimized by stacking the musicians cases against
the wall under the control room window. The cases offer a fair degree of
dispersion to the strong boundary layer sound traveling along the floor
to the control room wall, and integrate it before reflecting it back to
the players. That won't work with a rug on the floor, but improves things
quite a lot otherwise.
Given that a primary function of the control wall is to supply a live
surface to the musicians, it can be made more effective by using some of
the techniques employed in the backs of control rooms. These involve
substituting RPGs for the stacked cases, retreating the wall for maximum
reflection, moving the control room window to the upright position, and
installing a reflector above the window angled to bounce even more sound
back to the rhythm section. The combination of flat and dispersed
reflections has been shown to be optimum for critical listening, and if
it's good enough for engineers, why not supply it to the people who are
doing the actual work in a studio?
In any case, the amount of acoustical treatment in a studio is limited
by the need to leave the major part of the control room wall reflective.
And there advantages to a live floor in allowing solid boundary layer
sound at the control wall, in addition to making it easier to move things
around in the studio.
The side walls are far less critical. Because of that, they are commonly
either left untreated or given some kind of uniform treatment. Neither
is a good idea.
Flat, straight walls have been known to be acoustically unacceptable for
centuries. That's partly because sound reflects off such walls as a flat
smack, which sounds bad, and partly because it bounces so strongly. If
the side walls are either untreated or evenly treated the sound will
ricochet around the room like a ball on a billiard table until it finds
an open mike to get into. That was the problem with Doc Sabins' room.
It is probably possible to control the results by putting an absorbent
flat behind every mike in the room, but it's a tedious process, and
interferes with player communication. Much better to clean up the bounce.
Since the villains in the piece are flat, evenly treated walls, the
obvious remedy lies in knobbing up the walls and installing absorptive
treatment in patches.
Both objectives can be accomplished by hanging live sided boxes filled
with Fiberglas on the walls. (See drawing.) Floor to ceiling treatment
is unnecessary, as mikes rarely point up. The boxes should start high
enough off the floor to clear chairs and other clutter leaned against
the wall, and will generally top out at eight or nine feet above the
floor.
A box with reflective sides will act as a disperser, and at a foot or
so deep will disperse down to about 550 Hz. Not ideal, but not bad,
and at a foot the boxes are pretty manageable. They are normally spaced
at three to six feet apart, leaving the walls reflective between them.
This presents a combination of dispersion, absorption and reflection to
both the musicians and the mikes, and cleans up the billiard ball
syndrome quite nicely while presenting an optimum listening environment
to the hard working types in the studio.
The back wall can be treated in the same way in small rooms, although
it is best to leave the back as live as possible, as reflections from
it give the players a sense of being in a room rather than working
with their backs to a vacuum.
In cases where a great deal of absorption is needed, the wall area
above eight feet and below about two can be totally treated without
ruining the generally live sound of the room, as the ear only needs
a little encouragement to think it's in a normal environment.
The ceiling is another matter, and needs be almost entirely dead,
because it is almost never high enough to establish a decent modes
structure. The standard literature lists acceptable room proportions
of up to two to one as an extreme case, and the vast majority of
ceilings are well over that. As always, the best way to deal with an
unsolvable problem is to eliminate it, and since a non-reflective
surface generates no modes structure one way or another, dead ceilings
are the norm in most studios.
The ceiling also presents the largest area available for serious
treatment, especially as it can be totally absorbent without making
a room sound dead. Short, yes. Dead, no.
Unless a ceiling is extremely low the ear ignores it, preferring to
take it's cues more or less horizontally.
It is critical that the ceiling treatment be acoustically flat in it's
absorption. Given an ordinary grid hung 16 inches below the structural
ceiling, flat response can be accomplished with 1-1/2 inch Fiberglas
ceiling panels or with thinner panels and a Fiberglas batt overlay.
It is wise to check manufacturer's literature for exact specifications,
as the low end absorption of the ceiling must extend far enough into
the bass range to avoid the common fault of acoustical treatment that
soaks out the top end of the room while leaving the low end live. That
kind of treatment results in a muddy room with terrible isolation probs
in terms of bass, floor tom, and bass drum.
The need for flat low end response applies to all room treatment unless
the studio has big windows or it's walls are so flimsy as to transmit or
absorb bass by vibrating to it. Even then, bass attenuation will seldom
exceed 30%, leaving 70% to be supplied by other means. While there are
any number of bass absorbing devices which can be built or purchased,
they are inconsistent in operation, inefficient except in corners, and
very difficult to analyze as to the number and size required.
On balance it's more practical to install the general treatment in such
way as to absorb uniformly from bass to cymbals.
Controlling high frequency reflections is easy, but bass absorption is
largely a matter of absorber depth, and it takes considerable thickness
to get flat down to 60 Hz. Hung ceilings manage it with thin panels and
the 16 inches between panels and the real ceiling, but a wall mounted
absorber needs a minimum depth of 6 inches for Fiberglas (703) board,
and a foot for glass wool.
DON'T USE THIN TREATMENT! Carpeting and drapes absorb 2 to 14% of bass
while soaking out 60 to 70% of the top end, yielding a room with no
presence and extreme boominess. Bad for playing, worse for recording.
At 70% efficiency, they also require an excessive amount of treatment
and wall area. Interestingly enough, both products cost far more than
proper acoustical materials, and are not necessarily more attractive.
By and large, Fiberglas in one form and another is probably the most
practical treatment available, and it can be covered in any number of
handsome fabrics or in Tectum if a durable wall is needed.
With the type and location of studio treatment in hand, we can finally
address the question of how much absorption is needed.
The following data are not hypothetical. The Dc figures were determined
during extensive reality testing of a newly written acoustical design
computer program.
The test method consisted of retro engineering a number of recording
studios, in each of which the writer had done some hundreds of sessions.
The majority of the studios were acceptable, a few were marginal, two
were bad, and two superb.
The object of the exercise was to find a common parameter that related
to actual studio performance, and the voice Dc proved to be a figure of
merit for isolation in properly treated rooms. Other correlations became
evident over several years of repeated computer runs on these and other
studios in an acoustics course taught by the author at a local college.
Designing for isolation is both simpler and more difficult than it first
appears. The simple part is very simple indeed, as voice Dc and therefore
isolation turns out to be function of the amount of absorption in a room
regardless of room size.
The absorption required for 26 Db of acoustical loss from 6 inches to 20
feet (a voice Dc of 11 1/2 feet) is about 2700 Sabins. Sounds easy.
If the practice were as straightforward as the theory, one could stuff
2700 square feet of Fiberglas into a studio and open for business without
further ado.
Unfortunately, what's wanted is 2700 Sabins of absorption, and the actual
amount of treatment for that figure can vary from less than 1500 to just
over 2500 depending on the size of the room. First bear in the woods.
The reason for a difference between actual treatment and effective
absorption is that the standard Sabin formula is linear, and absorption
in highly treated rooms is not.
In fact, when 80% of the wall surface absorbs at 1 Sabin per square foot,
the effective absorption of the treatment is doubled. There is a formula
for this effect, (Norris Eyring) which is reasonably accurate, but since
it involves the use of natural logarithms, it is tedious to use.
Second bear.
The third bear is the well-populated acoustical forest is the difficulty
of accurately assigning absorption values to various materials already
in the room. Most standard materials can be looked up in tables printed
for the purpose, but there are always a few things that aren't listed.
Additionally, it is very easy to mistake one kind of acoustical material
for another and come up with significant errors in calculations.
Calculations are a pain anyway, so it's best to circumvent the bears by
measuring the acoustical performance of the room.
There are several thoroughly scientific ways to do this, and any number
of manufacturers eager to sell equipment for the purpose, but as a
practical matter such measurements are of little or no use to the studio
owner. Cheap equipment yields cheap results, and the data gleaned from
upscale equipment require expert (and costly) interpretation.
In the first instance, the figures aren't completely trustworthy, and in
the second routinely repeating the tests will cost a fortune.
Following the KISS (keep it simple, stupid) rule, the writer prefers to
measure a room by determining it's voice Dc. The equipment costs nothing,
it takes about two minutes, and the results are more than accurate enough
for real world use.
Better still, being a simple-minded test, it reports simple-minded
figures with no interpretation, no ambiguity. Best of all, a Dc check
makes it's measurement at about 100 Hz, where improper treatments cause
a the majority of isolation problems.
Measuring a voice Dc is child's play provided one keeps in mind that the
purpose is to determine the global characteristics of the space. Toward
that end, it is essential to make the measurement in the acoustical
center of the room. Given normal treatment, that will be in the physical
center as well.
In cases where the absorption is considerably greater on one wall than
another, the acoustical center will have to be found.
Again, dead easy. Using the incredibly sensitive instruments found on
either side of the human head, one sidesteps away from one wall toward
another until the reflected sound from the two are equal in each ear.
If the reader has not done this in past, he may find it useful to
calibrate his ears to wall sound by stepping up to a live wall and
varying his wall to head distance from a couple of feet to a couple of
inches until the wall sound is firmly fixed in mind. It is usually
perceived as a kind of pressure on the ear, and will very reliably
inform the listener of his position in a space. No sound other than the
room's random noise is needed, and once the listener knows the sound of
a close wall he will find that he can walk to within a foot or so of any
live wall with his eyes closed, This is simply a case of practicing a
normal human ability into a skill. The blind do it all the time. So do
the rest of us, but unconsciously.
The writer once deadened one wall of a hallway, and sighted people veered
into it to the point of wearing out the treatment.
Having determined the acoustical center of the room, Dc is measured by
two people more or less astride the room's center starting at a distance
of 15 to 20 feet. One of them walks toward the other droning one, one,
one as the other waits for the sound of the talker's voice to suddenly
get louder. The process works both ways, with the talker's voice abruptly
going constant volume as he retreats, but the writer's experience with
several hundred students indicates that toward is easier to hear than
away, particularly in the learning stage. It is also easier to hear if
the talker walks briskly at first. He can slow down for greater accuracy
once the listener has the sound of the transition in mind.
While rare, there is one case in which it is nearly impossible to get
a decent Dc measurment, i.e. a room with a very high ceiling and a
dead floor. Short a couple of tall ladders and considerable time, it is
not possible to get to the vertical center of the room, so it's back to
plan B. (drop back 10 and punt)
The pair can also check room's frequency response by measuring the Dc
using the word six, leaning on the s and x and suppressing the vowel,
so that most of the sound is at 3 to 5 Khz. This is a pretty rough test,
but if the Dc's are wildly disparate, they indicate a room with more
absorption in the midrange than at the low end.
While Dc is a square root function of a room's global characteristics
and therefore a rather short ruler, the breakover is sufficiently abrupt
to make measurements to within a few inches quick, easy, and repeatable
by any number of talker-listener pairs. Other than the one above, the
only conditions under which it doesn't work properly are rooms in which
the Dc is greater than the wall spacing, (rare) and huge rooms which
appear to divide themselves into several acoustical areas due to
extreme losses between one wall and another.
In the first instance the room will be too small and dead to be of any
practical use, and in the second the room volume will be well in excess
of a million cubic feet. The writer knows of one at 6 million that acts
funny, but it's in no danger of being used for studio work.
Once the Dc of a room has been measured, some acoustical modifications
may seem in order. If so, a few cautionary notes should be kept in mind.
First, the Dc varies as the square root of the room absorption, so
doubling the effective treatment and thereby halving the reverberation
time will extend the Dc to only 1.4 times it's previous figure. This
presents no problem in a medium to large room, but good isolation in a
30x20x10 foot studio would require some 1550 square feet of Fiberglas
scattered over only 2400 square feet of surface area.
Even with the floor thickly carpeted, leaving a 20x10 foot control
room wall reflective would require a 75% treatment of the other walls,
and result in a reverberation time of just over one tenth second.
Some rooms are simply too small to treat for live studio work, as they
get too dead. The 6000 cubic foot case in point is probably the workable
minimum.
Second, a big studio is rarely allowed more that about one second of
reverberation time, which results in a voice Dc approaching 20 feet.
Obviously, such a room needs no help in isolation, and is best left
alone. It is a general rule in acoustics that big rooms are easy. It's
the little ones that give you the pip.
Third, professional engineers commonly do good work in bad conditions.
The writer has done any number of sessions in studios with 7 to 8 foot
Dc's which turned out well enough to sell bags of records. It's not
impossible to record in a room with poor isolation, it's just damn hard
work.
The point of proper treatment is that it allows one to get decent sound
with any reasonable setup, and it eliminates time lost in fooling around
trying to correct the room's faults.
Fourth, none of the figures given are engraved in stone. A twelve foot
Dc is better than ten, and less good than sixteen, but acoustics are
inherently inexact, and there is no sharp point at which rooms switch
from bad to good; they just glide from exasperating to no problem, with
the latter occurring and something around a 12 foot voice Dc for the
bulk of studio work.
In summary, a few minutes spent in measuring the real world acoustical
characteristics of a recording studio may reveal unnecessarily poor
isolation, and some of the treatment methods suggested herein may
improve it's general usefulness.
Since the measurement involves no expense and the treatment is designed
to make experimentation easy, these techniques offer a practical way for
a studio to confirm or optimize it's recording rooms.