How Recordings Are Made II:
Digital Hard-Disk-Based Recording
Jay Kadis and Daniel J. Levitin
1
The Brave New (Digital)World
Th e last five years or so have seen what may be th e biggest cha nge in
recordin g te chnol ogy since the introduction of multi -t rack recording in the
1960s-the move t o hard disk-b ased digit al recording. This may repr esent
an even mor e pr ofound chan ge than the introdu ction of digital recording
in the 1980s because hard-di sk-based recor ding allows for editing and ma­
nipu lati on of the signal in ways th at ar e fundam entall y different from that
which came befor e, even with digital t ape. Because developm ents in this
domain have been so rapid , th ere is a dan ger th at anythin g we writ e to­
day (summ er of 2005) may becom e quickly outd ated ; neverthel ess, we will
at te mpt to discuss principl es, t echniqu es , and tec hnologies as the y exist
t oday. Whil e some details may change, we beli eve that th e fundamental
principl es will apply for some t ime (as t he pri nciples an d techniqu es of
t ape-b ased recording , discussed in th e pr eviou s article, still apply). We
will exa mine the many advant ages that computer -based recordin g systems
pr ovide alon g with th e related implic ation s and difficulti es.
Record ing engineers and compo sers have long eyed the comput er' s
potenti al for making and manipulatin g sound recordings, but early com­
pu ters were too slow and expensive to be pr actic al. As personal comput ­
ers became fast er and cheaper , deskt op computers acquired th e ability to
perform fun ction s th at previou sly necessit at ed a building full of expensive
elect ronic devices. Thi s fueled a rac e am ong software develop ers to find
15
16
How Recordings Are Made II: Digital Hard-Disk-Based Recording
new ways of making and pr ocessin g music . The now-widespr ead avail­
ability of t ools pr eviou sly availabl e on ly t o professionals has facilit at ed
an explosion of music create d by musi cians pr eviou sly un able t o realize
their ideas . Th e demo cratiz ation of the record ing pro cess also brings
the pr edict able result whenev er powerful tool s are placed in th e hands of
less-skilled ope rat ors: overuse of gimm icky pr oduct ion tricks and po orly
composed and performed mus ic in abunda nce. Simply using pr ofessional
to ols does not guarant ee the production of pro fessiona l-qualit y record­
ings, and providing access to a profes sional recordin g environment does
noth ing to improv e musical composition . Wh ether boon or ba ne, we will
see how the personal comput er has alt ered t he relation ship betwe en music
creat ors, th eir too ls, and th e list ener .
Th e fundam ental differen ce between the older and newer syst ems for
record ing and manipulating sound is th e manner in which t he signa ls are
encoded and store d . P rior to th e development of digita l audio recording ,
sound recordi ngs were ma de by processing continu ously vary ing volt ages
genera t ed by microphon es or electronic instrum ent s. Th ese volta ges were
converted into pr oportional magn etic field s in ana log tape recorders and
st ored on magn eti c tap e. Th e term anal og indi cate s that the signal volt­
age is directly proportional to the original sound pres sure level and is
cont inuous in nature (t he term "ana log" comes from th e same root as th e
word "an alogy" ). Any manipu lation of the analog sound repr esent ation
had to be mad e in what is called real-tim e, meaning simulta neously as
the musicians or tape record er played.
In contrast , digital audio devic es first convert t he continuou sly varyin g
volta ge signals into a series of numbers that repr esent the signal am pli­
tude. This proc ess, called sampling, requi res t hat the measurements be
made very frequent ly so t hat t he digital repr esentation closely reflect s th e
analog signal . To und erstand sampling, imagine that you wanted to ob­
tain an estimat e of how much tr affic passes by your livin g room window ,
the room in which you plan to set up your new recordin g system . If th e
cars are going by slowly enough, you can see a car , look down to your
notebook, and write it down. If they're going by quickly , in the time it
takes you to look down in your notebook and writ e it down, anot her car
or two may have passed and you will have miss ed counti ng th em. You
can see int uit ively that th e amount of tim e you spend looking out th e
window (sam pling) the t raffic has to be relat ed to how long it ta kes a car
t o pass by and how long it takes to write down each entry . In the case of
an audio signa l, the rate of voltag e change is somewhat analogous to th e
speed of t he car s: the higher th e sign al frequenc y, the faster the voltag e
is changin g.
1. Th e Brave New (Digital) World
17
In aud io sam pling , we st ore only t he values t hat we meas ure at th e
sample t imes an d do not save inform atio n ab ou t what happ ens bet ween
sampl es; thu s, sampling a signal result s in a discrete repr esenta tion , one
th at is not cont inuous. Wh en we choose the sam pling rat e prop erly (ac­
cording to t he Nyquist th eorem of 1928) , we don 't have to worry abou t
missing impo rtan t inform ati on. Anot he r ana logy comes from th e world of
film-m ak ing. A film camera does not record cont inuo us images; it sam ples
th em , putt ing each image into a porti on of a conti nu ously moving piece
of film , called a fram e. Th e st andard frame ra te for prof essional film
cam eras is 24 fra mes per second (fps) , st andar d NTSC vid eo in Nor th
Ame rica is 30 fps , and fast er frame rates exist also. T his sampling rate
is sufficient to give t he flicker- free illusion of smoo t h moti on when the
proj ect ed image is viewed (t he proj ector doubl e shutte rs t o displ ay 48
images pe r second , which is beyond t he human eye's fusion ra t e). This
syst em works well as long as no element of the pict ure moves t oo quick ly
with respe ct t o t he frame ra te. For example a quickl y spinning wagon
wheel, whose rot at ion t akes less than two frame ti mes to comple te, will
not appear to spin at t he correc t ra te when proj ect ed and might even
ap pea r to ro ta te slowly backwards. Thi s phenome non is kn own as alias­
ing and can be prevente d by limiti ng the frequ ency of th e sign al b e­
ing sampl ed t o less t han half of t he sam pling ra te (t hat 's t he Nyqui st
th eorem again ) .
The very high sp eed of modern comput ers allows the digit al sa mples
represe nt ing t he recordi ng to und er go significant signa l pro cessin g in t he
short ti me be tween wh en new sampl es are acq uired or playe d. Effects m ay
also be applied to exist ing digit al recording s. This is especially useful for
effects t hat requi re t oo many resour ces to ru n in r eal-tim e on tod ay's
proc essors . Many effect s th at we could not eas ily accomp lish using ana ­
log te chniqu es are possib le, from edit ing perform an ces in very sma ll pieces
to shifting ind ividu al not es in t ime by t iny fracti ons of a second or cor­
rect ing th e pit ch of a singer's p erform an ce. We can crea te sound s dire ctly
by comput er synt hesis and use sampled sounds and loops of music to
creat e new comp osition s. We can make math ematic al mod els of instru­
ments t hat allow compu te r pr ogram s t o sim ulate th e physical beha vior
of real and imagin ar y inst rume nts , and t he sounds of th ose instrum ents
can be played by t he compute r . We can cut an d past e musical soun ds as
easily as we do text in a word pr ocessing program , moving th em freely
in t ime or pit ch spac e to crea t e spec ial effects or to improv e a flawed
performance.
Th e advance s menti oned t hus far ap ply t o digit al recordings: music
st ored as a series of digital sam ples of acous tic pressur e over tim e as ac­
18
How Recordings Are Made II: Digital Hard-Disk-Based Recording
quired by microphones fed into analog-to-digital converters . In addition,
computers enable the capture and manipulation of the gestural informa­
tion pertaining to how a musician actually manipulated an instrument
to create the performance. Specially outfitted instruments must be used,
which encode the musician's performance as a stream of instructions per­
taining to which note was struck, when it was struck , how hard it was
struck, and when it was released. This stream of information may then
be edited and played back through a synthesizer or even an actual instru­
ment such as a modern player piano. This entire approach is reminiscent
of the player piano rolls of old, but it retains more gestural information
than piano rolls did.
This is the result of a technology known as MIDI (Musical Instrument
Digital Interface), providing a digital instruction language and hardware
connection standard that links electronic instruments to each other and
to computers. MIDI sends data that contain instructions on when to
play notes (note-on), how dynamically the notes sound (velocity), how
loud it should play (volume), and when the note should stop (note-off or
velocity 0) as well as many special controller values and device-specific
commands (sys-ex). MIDI allows synthesizers to play back digital scores
stored in computer memory as sequences of events . MIDI permits se­
lecting and altering the sounds to use for playback and any effects like
reverberation that may be produced by the devices. MIDI networks con­
nect racks of different synthesizers so that they operate as one instrument,
allowing composers and arrangers to hear a symphony played by sampled
instruments, for example. Popular music may be produced entirely in
this manner without the creator being able to play any instruments at all!
MIDI may also be used to connect the devices in a studio so that their
operation is controlled from a central location, often the mixing console,
using MIDI machine .cont rol instructions. MIDI time code can be used
to synchronize playback from different sources such as tape recorders and
synthesizer sequencers. While MIDI allows the fledgling arranger to try
out ideas and textures before bringing them to live musicians, it will not
prevent the arranger from writing notes that cannot be played within
the range of a real instrument or requiring playing in other impossible
ways .
Not everything in the production of music has changed fundamen­
tally as a result of moving to digital audio. Many techniques of mixing
and creating effects are still accomplished in a fashion similar to that
of the analog studio. For instance, a mixing console is as central to a
digital studio as it was to an analog studio, although it may now exist
completely in software with virtual knobs and sliders on the computer
1. The Brave New (Digital)World
19
display which may be moved with a mouse instead of a hand. The engi­
neer still needs to have the ability to change several parameters quickly
and perhaps more than one at a time. The mouse is a poor substitute
for a mixer full of controls; consequently, digital studios often interface
devices with physical knobs and sliders reminiscent of the analog mixing
panel to their software mixers. The digital audio workstation, or DAW,
is modeled on the recording studio of old because the methods developed
in the traditional recording studio are still sensible and adequate ways
to record and assemble multitrack recordings, albeit the DAW does have
some advantages over studios of yore.
One of the major advantages of digital audio is the nondestructive
nature of editing and recording. Now we can have as many tracks as the
speed and memory of the computer will allow rather than the fixed num­
ber of tracks provided by a dedicated tape recorder. If we wish to add
material, we can punch in without overwriting the previously recorded
tracks. We may select parts of tracks to be combined in a playlist, a list­
ing of pieces of sound that are to be played back in the order we specify
regardless of when the pieces were originally recorded. This process is
automated and recallable, so we can make small changes and listen to
multiple versions easily. The ability to independently manipulate pieces
of data is known as nonlinear editing. Using analog systems, tracks
recorded synchronously to tape cannot easily be moved in time relative
to each other whereas digitally recorded tracks can be accessed freely and
independently.
Another advantage of the computer-based studio is the cost savings
of using software for simulating the hardware devices found in the tradi­
tional studio. Signal processors like compressors, limiters , delays, rever­
berators, and other special-effects devices were plentiful but expensive in
the traditional studio. These hardware devices are now being modeled
in software (often called plug-ins because they are add-ons to the basic
software package), creating programs that process the digital information
much like the analog gear of old. The advantage is that rather than buy­
ing a new compressor when w,e exhaust our supply, we need only click a
button to install another copy of the software emulation and we're ready
to go on with the mix. The number of copies that we may use is limited
only by the power and speed of our computer. Add-on signal-processing
cards are available to increase the amount of processing power a computer
may provide, including both complete recording and mixing systems like
Digidesign 's ProTools TDM and add-on digital signal processing (DSP)
systems like the Universal Audio UAD-I, designed to augment host-based
recording systems like Apple's Logic .
20
2
How Recordings Are Made II: Digital Hard-Disk-Based Recording
Trouble in (Digital)Paradise
Digital audio continues to evolve rapidly, t akin g advan ta ge of. the con­
stant stream of faster computers and new interface form ats that allow
more data to be transferred and stored. The sp eed of this evolution is
both an advantage and one of th e potential problems pla guin g users of
the new technologies. Issu es of incompatibility are exacerbated wh en the
elements of the syst em ar e all changing quickly , and often indep endentl y.
While analog t echnologies were also constantly improving, an analo g tape
recorder (like Studer or Ampex) or analog consoles (like Neve or SSL) were
usable for decad es. Digital hardware and software change quite rapidl y
and complet ely, requiring frequent upgr ad es t o remain compatible. Thi s
state of constant change mak es it hard er to train new eng ineers and the
need to continually learn new softwar e and hardwar e tak es time away
from the job of recording musi c for exper ienced engin eers . Th e softwar e
is complicat ed and may contain bugs that cause system failures in special
circum stanc es not discovered by th e pr ogramm ers, leadin g to comput er
crashes , lost information , and angry custom ers. Selecting add-on audio
an d MIDI interf aces for th e comput er is complicate d by differin g require­
ments of th e various comput er syst ems and recording software package s.
Recording engin eers now also need to be comput er techni cians to be as­
sur ed of th e abilit y to keep th eir studio s ope ra t iona l.
For all its advant ages, digital audio introdu ces some probl ems t hat
did not exist with the an alog approach . A friend of ours, a famou s pro­
du cer / engin eer who has a studio in his home , spent sever al month s get­
ting th e various comp onent s of his system running: the computer hard­
ware , oper ating syste m, recording softwar e, sequencin g software, various
input/output devices, plug-ins , et c. During this time, a minor upgrad e to
the compute r' s oper atin g system was released, but since it wasn 't com­
patible with the plug-ins , our friend decided (of cour se) not to do th e
up grad e.
He began to record an album , and after four months , his comp ute r' s
motherboard failed . The manufacturer offered to send our friend a br and
new comput er, with a fast er processor and more hard disk space, and the
whole package was sent by next day air . Unfortunat ely, t he new com­
put er came with a newer version of th e operatin g syst em (one that was
incomp atibl e with the plug-ins) and the new computer's logic board was
incapabl e of operating und er the previous system . This meant that the
album production had t o stop complet ely. Wi th t he plug-ins and various
oth er I/O devices not working , the producer could not maintain continu ­
ity betw een wha t he had done before and aft er the hardware chan ge. This
2 . Trouble in (Digital) Paradise
21
demonstrates th at t he comple x int erd ependen cy of hardware, periph erals,
ope rat ing syst ems, drivers, and applicat ions demands a thorou gh under­
sta nding on th e par t of tho se depending on such syste ms for pr odu ction.
Software and hard ware incompat ibiliti es aren 't t he only sour ce of po­
ten tia l t rouble. T ime syn chro ny is requi red when sam pling, since we must
gua ra ntee th at the t iming of t he sample measur ement s is corr ect for every
tr ack we record . While most digital syst ems are able to autom at ically syn­
chr onize with th emselves and with each other, th e accur acy of th e clock
th at determin es th e sample tim e must be consist ent throughou t th e ent ire
syste m.
For ins t an ce, if devices clock each sampl ed word at different t imes,
clicks may be genera te d in t he au dio data str eam . Or, if t he reference clock
is not perfectly regular, digit al dat a may be output with slightl y different
int ervals betw een words, alterin g th e audio dat a produced and pot entially
affecting th e perceived st ereo image. Ther e is also an issue of delays
create d in th e digit al syste m as data are moved arou nd insid e th e devices .
In order for many tracks of recorded music to be played simultaneo usly,
each must have und ergone exac tly t he sam e intern al delays, or lat ency.
If simultan eously-record ed , live tracks are played back with different
delays , they combine to produ ce pea ks and dip s in the frequ ency response
as certa in frequ encies reinfor ce an d others can cel t o produ ce an unwanted
comb filt er effect . Thi s is a parti cular pr oblem with st ereo t racks where
the sound is simultane ous ly record ed bu t played back slightly apart, caus­
ing the ster eo image to shift . (Wh en track s are overdubb ed , t his is not so
much of an issue sin ce the sounds were not tim e-locked to st art except by
th e musician' s t iming accur acy, which is nothin g like the microscopic res­
olut ion of th e sample clock.) Each process, ana log-t o-digita l conversion,
data stor age, da ta ma nip ulat ion, and digital-to-an alog conversion must
del ay all t rac ks t he exac t sam e amo unt in order for th em to be played
back synchr onously.
Loss of syn chronization is not just an abstract problem but one that
occur s frequ entl y in practic e. In th e analog world, t ape machin es typically
have t wo heads over which the t ap e pas ses and which read or write ma­
te rial onto t he tape . One head is use d for high qu ality playb ack, and th e
oth er is used to record . Thi s is becau se t he ph ysical pr ocesses of recording
and reprodu cing magneti c signals are different and opti mization requir es
different head designs for each process. If one wan ts to overdub (add an
inst rument to one that has alre ad y been record ed) , a p otenti al problem
exists.
_
Suppose that you've recorded drums an d now you want to add bass .
If the record head and t he play back head are not in t he same physi cal
22
How Recordings Are Made II: Digital Hard -Disk-Based Recording
tape
t
i
I
I
Record head
Playback head
Tape direction
Rgure 1.
location along th e tape's path , there will exist an asynchrony . You'll hear
t he drums playing the moment the tap e passe s the playb ack head, of
cours e. As you play bass along with those drums, the record head will
put the signa l on tape, but if th e record head is an inch or so away from
th e playback head, your bass part will be out of synchronization with
th e drum part by the distan ce between th e two heads. At 15 inches per
second (IPS) with an inch between the heads , t his one inch would cause
th e bas s part to be recorded 66 ms later, or th e equival ent of 1/16th note
at 64 beats per minute (bpm) .
This probl em was solved in analog recording by combining lower fi­
delity playba ck electronics on th e same physical component as the record­
ing head, in what is often called a sync head. This permitt ed th e musician
to hear ba ck the previou sly record ed track (at somewhat lower fidelity )
and add something to it virtually instan taneously, and in (virtually) per­
fect synchrony .
In the analog world, no matter what happens to the m ast er tape, th e
performances on it will remain synchroniz ed with one an oth er so long as
they started out synchronized ; not necessarily so in th e comput er-b ased
digital world. The two of us had to mix a song once for which the session
file had become corrupted. We still had access to the individual sound
files of the performance-the
basic tracks and all the overdubs-but
the
Pr oTools session file that contain ed instru cti ons for how t he tracks were
to line up in tim e had been destr oyed . We had to imp ort each individu al
sound file-for th e kick drum , snare drum , bas s guitar, vocals, etc .-in t o
2 . Trouble in (Digital)Paradise
23
a newly created session file. Because some of the drum files were edited
after recording to get rid of noise at the beginning, and because some of
the files were from overdubs not recorded with the basic tracks, lining up
all the sound files in the mix/edit window did not succeed in synchronizing
them. In fact, there was no information available at all to tell us where
to place the files with respect to one another. Lining up the snare drum
and kick drum was time-consuming but not complicated because it. was
obvious how those parts were intended to fit together (and bleed-through
in the mics gave clues). Lining up the vocals and the guitar solo, however,
was extremely difficult . In some cases we knew roughly (within half a
second or so) when they were supposed to occur, but the vocals and solo
had been performed with a particular feel, a very specific relationship to
the beat that we were unable to recapture . We spent hours nudging the
parts around by 10 and 20 milliseconds to get what sounded good to us,
with no objective information about how the singer had intended to place
his vocals with respect to the beat. To make matters worse, there were
pieces of vocal and guitar performances -a few notes here and there­
that had been recorded on separate tracks as repairs or fixes, and we had
no idea where they were supposed to go or what they were intending to
repair or fix.
The ease with which files are created by computer recording often leads
to sessions with huge numbers of individual sound files. This demands
careful attention to file naming and record keeping. Every time you create
a new track, give it a name that indicates what it is. Good names: Lead
Vocal I, Snare Drum, or Replacement Rhythm Guitar. Bad names: Audio
6 (the default name the computer software might assign), RE20 (the name
of the microphone you used-but what instrument did you record?), or
July 7. Take as many notes as possible about the track and write them
directly to the computer or in a project journal: the microphone used,
mic pre-amp, time of day, compression settings, and what the part was
intended to do (that is, how it was intended to fit into the final mix) .
The microphone and signal processing notes will help you if you want to
go back and recreate that sound. Be sure to distinguish tracks that were
intended as retakes or replacements as opposed to primary parts, so that
you or some poor mixing engineer doesn 't waste days trying to figure out
if two parts are redundant or not.
In the digital domain, it takes a small but often noticeable amount of
time for a digitally-recorded signal to be converted to analog. Suppose
that it takes 20 milliseconds for a signal to wind its way through an A-to­
D or D-to-A converter, and you want to play bass along with a previously
recorded drum part. The drums take 20 ms to get through the converters,
24
How Recording s Are Made II: Digital Hard-Disk-Based Recording
so your speakers get th em 20 ms aft er the comput er "plays" them . Now,
if you 're sit ting 8 feet from your spea kers, assum e that it t akes anothe r 7
ms for t he sound to reach your ear s. (Wh ile we usua lly th ink of overdub s
empl oying headphon e monit oring, som e perform ers, repo rte dly inclu ding
Frank Sinatr a , dislike performin g wit h headphon es and favor loudspeaker
monitorin g for overdub s, allowing a mor e natural performance.) You p lay
alon g as best you can , but it tak es th e computer anot her 20 ms for t he
soun d from your bass to get thr ough the A-to-D converte r on t he way
back in. Your soun d is now record ed 47 ms aft er t he dr um tr ack (not
unlik e the case with ana log overdub s th at we just spo ke abo ut) . For more
infor mati on see Der ek DiFilippo's "Pe rceivable Audi o Lat encies" in A udio
Anecd otes 1.
The software design ers, in theor y, know how long th e hardwar e con­
vert ers ta ke t o pro cess sound, and part of th eir job is to build in a syn­
chronizat ion funct ion t hat sho uld syn chr onize your overdu b wit h t he pre­
viously recorded tr ack auto mat ically. Th e only thin g t ha t they can 't t ake
into account is the sm all delay from th e spe akers to your ear becaus e t ha t
will vary from room to room. The hyp othetical 7 ms in our exampl e is
truly insignificant when it is t he only delay, but it can make a not iceable
difference when it adds up wit h oth er sources of delay. Most semipr ofes­
siona l and pr ofession al software record ers have a way to minimize, th ough
not elimina te , convert er delay. Known by names such as low laten cy or
overdub mode, this options shr inks the bu ffer size durin g playback t o give
you th e fast est playb ack possib le. Thi s is usually at th e expense of be­
ing ab le to use larg e amo unts of built -in pro cessing , such as compressors ,
equal izers, reverbs , etc., which ta ke ti me to employ but which can be
add ed bac k in dur ing mixdown. Alt ern atively, you can monitor input s
dir ectl y and not list en th rough the digital device, elimin ating the digit al
monit orin g delays entir ely.
3
Summation for the Defense?
As we are beginning to realize, every technological revision comes with a
cost . When fuel injection replaced carbur et ors in au to mobiles (new cars
haven 't had carbur eto rs since the early 1990s) , it pr ovided a more reli­
able system for delivering fuel. Gon e are th e cold morn ings when t he
car wouldn 't t ur n over at all, the sudd en sput t ering dur ing high alti tu de
climbs , an d much of th e pollution cau sed by imprecis e air/ fuel mixtures.
Fuel inj ecti on works well without req uiri ng maintenan ce for many tim es
the miles bet ween carb orator tune-up s, but when the fuel inj ection com­
3. Summation for the Defense?
25
puter fails, it fails all at once and no amount of tin kering will get you back
on th e road . A carbure tor failed gradually, gracefully-you had warning
coughs and sputt ers as gasket s decayed or springs lost their tension , and
if you kn ew wh at you were doing you could enrich or lean out t he mixture
to accomm od at e cha nging climat e, elevat ion, or wear.
Wh en ana log t ap e, or for th at matt er tub e electronics (amps, com­
pr essors) , fail , th ey ofte n do so gra dua lly; th e tub es begin to leak and
p erform an ce is comp romised, but th e hardwa re is still useable . Old ana­
log tap es eventu ally wear out as a funct ion of age, poor st ora ge conditions ,
or t oo many playbacks, but on t heir way to wearin g out th ey ar e st ill use­
able. T he loss of high-frequency informa t ion or occasional drop-outs are
often t he first clue t hat a tap e is deterio ratin g. But , with digit al record­
ing as found on digital ta pe , har d-disk recordin g, or even CDs and DVDs ,
t he dete riorat ion is masked by err or-correc t ion schem es. Error corr ect ion
ensures t hat even imp erfect media (ever hold a CD up t o a light source
and noti ce th e pinho les?) can play back bit -far -bit accur ate sound -t hat
is, until th e det eriorat ion exceed s a thr eshold where the erro r corre ction
can no longer cop e, pot enti ally rend erin g th e recordin g utt erly and in­
sta nt ly unu sabl e. Whil e som e professional gear will report digita l media 's
bit error rat e to allow an enginee r to monito r pot entia l deter ioration of
media, t his featu re is far from un iversa l. In a pin ch, we could always use
a partl y compromised analog t ap e (as was done for many CD reissues­
th at 's why some of your favorit e albums don' t sound all that great on
CD), bu t a digita l recordi ng t ha t has b een corr upte d cann ot be used
at all.
What advantages does digit al recordin g techn ology offer to the mu­
sician? With MIDI-b ased samplin g syste ms like Tascam's GigaStudio,
a compos er can hear a close approximat ion of a symphony playing her
composition using only th e comput er. A"musi cian can record music in a
bedroom studio and expe riment with different arra ngements, perfecting
songs without requirin g the parti cipati on of ot her musicians. Recordings
may be edite d t o produ ce near-p erfect perfor man ces, and slight timing
and pitch prob lems may be elimina te d . Bands can record their own al­
bums, taking as mu ch tim e as the y desir e without spe nding a fortun e for
studio time. Musi cian s in far-flun g areas of t he world can collaborate
in recording by sendin g mu sic over th e Int ernet or thr ough the mail for
others to contribut e to and return . Musician s can now record , master ,
duplicate, and offer for sale at p erforman ces th eir own CDs. Digital mu­
sic is easily distributed over th e Int ern et , so unknown band s can find an
audience that they could never meet ph ysically. This represent s a t rue
democratization of t he recordin g and distrib ution process.
26
How Recordings Are Made II: Digital Hard-Disk-Based Recording
Recent developments in Internet technology may result in an even
more active method of collaboration: real-time interactive performance
over the network. It is now possible to reduce the time lag associated
with digital audio systems connected to the Internet to a delay short
enough to allow musicians around the world to play together. Our re­
search groups at Stanford and McGill have worked on particular algo­
rithms for reducing broadband latency as much as possible, and members
of the Stanford and McGill jazz community have conducted several live,
"low latency" internet jam sessions to demonstrate the technology. We
are still at the beginning of such possibilities, and there are some inherent
limitations to the process, but digital audio promises to provide new ca­
pabilities to musicians and music lovers in the future that are difficult to
imagine .
The Internet and digital music also present a problem for the creators
of music due to the ease with which sound files can be exchanged. Current
digital delivery media like the compact disc contain no method of copy
prevention, so anyone can extract, or rip the digital audio files from a
commercial CD to their computer and send them over the Internet to
anyone else with a computer. Preventing unauthorized copying is one
reason for developing new digital media. It has also helped promote the
development of new digital techniques of representing music that both
provide improved sound quality and easier protection from copying. Super
audio compact discs (SACD) use different encoding of the audio data that
simplify playback circuitry while preventing computers from reading or
playing the discs. Future media for music distribution will likely be secure
from unauthorized copying. While this protects the commercial producers
of music, it also makes it more difficult for independent musicians to use
the technology.
'
While many of the signal-processing devices, synthesizers, and se­
quencers now available provide preset, instant sounds, it is important
to fight the tendency to overuse them! In most cases, the designers of
these presets intended them as a starting point for exploration, not an
end point. The danger of using presets is that music will become increas ­
ingly homogenized as everyone starts to use the same effects, reverbs, or
synth sounds. It also introduces the danger of music sounding dated:
as soon as new sounds come out, everyone rushes to be the first to use
them, and then after several years those sounds become associated with
a particular era. As always, the best way to make fresh, creative , and
high-quality recordings is to let your ears guide you. Whether it is a
digital reverberation device, a new string synthesis program, or a guitar
amplifier simulator, we recommend that you play around with and modify
3 . Summation for the Defense?
27
th e para mete rs t o your liking. Even small modifications will give you a
mor e individu al, cust omized sound.
T he popul arit y and accessibility of compute r-base d recordi ng syste ms
has increase d t he nu mbe r of people mak ing music recordin gs, grea tly ex­
pandin g th e ma rket for recording equipment. T he demand for inexp en­
sive microphon es, preamplifi ers, mixers and monit or speak ers has be en
met with a new generat ion of manuf acturers who produc e equipm ent in
places where labo r is cheap an d who sell devices that look much like t he
expe nsive equ ipm ent of th e tr aditi onal record ing st udio. Ma ny of th ese
look-alikes actu ally perform t heir tas ks well, however some simply re­
semble the original devic e and fail to sound anyth ing like th e gear after
which t hey ar e mo deled. Th e availabi lity of decent inexp ensive recordin g
equipme nt has significan tl y increased , but one must know how to evalu ­
at e and select from the plethor a of new devices (an d presets) introduced .
The increased access to sophisticated recording equipment ena bled by
the personal comput er has led t o a shift in t he way musician s th ink ab ou t
recordin g: wh ere once a commercial st ud io was th e only pl ace to make
recordin gs, many people now favor doing it them selves at home. Several
larg e and well-kn own studi os have ceased operati on in par t du e to the
dwindlin g busine ss caused by t he shift to per son al an d pr oj ect studi os.
Sin ce many of th e do-it- your self pra ctitio ners have limit ed know ledge and
exp erience, th e avera ge qu ality of the music recordin gs may have declined.
However , as mor e and mor e hobbyist recording engineers gain kn owledge
and expe rience, we may exp ect a posit ive effect on th e qu ali ty of music
recordings.
Whi le a considerable community still regards musi c record ing as a job
for an alog tape, th e computer -based recording syst em is here to st ay. The
flexibilit y, power , and economy t hat it pr ovides are too comp elling, an d
young recordi st s ado pt th e com puter natur ally. Alth ough early digit al
syst ems were op en to crit icism for inferior soun d qu ality when compared
to an alog record ers, significant improv ements hav e res ult ed in comparabl e
sound from compu ter-base d syste ms. Th e convenience of t he edit ing and
mixin g capabilit ies of th ese syste ms are a final impetu s to join th e move
t o ad opt th is recordin g te chnology for ma ny. As we b ecome comfort able
with the new te chni cal requ irements of digital computer-bas ed studios ,
they will seem as familiar as th e old an alog st udio.
Alth ou gh tec hnology will cont inue to change rap idly, fund amental
prin cipl es do not : the pur pose of mu sic recording has always been t o
convey th e musi c as exp ressively as pos sible. Exc ept for thos e rare cases
when t he purpo se is to dem onstrate gear in a hi-fi store, a high qualit y
record ing of a song no one likes will never ta ke the place of a low qu al­
28
How Recordings Are Made II: Digital Hard-Disk-Based Recording
ity recordin g of a gre at song . When excellenc e in both comes t oget her,
the results are long-lastin g. Learnin g t o make good recording s, even with
high -quality and easy-t o-use compute rized equipment, take s time like any­
thin g else. Car e, att ent ion t o det ail , and a good ear are as imp ort an t as
t hey ever were, and th ose ar e thin gs th at ar e not likely to chan ge in th e
next five or even fifty years .