Mastering and editing a recording project is analogous to assembling an art exhibit. Every piece may be great on its own; however, the pieces will have a much greater impact as a collection once they are cleaned up, properly framed, and displayed in a way that shows each in the best possible light. This kind of attention to detail makes the whole artistic experience greater than the sum of its parts.
The goal of mastering is to make the music sound good on a variety of systems and delivery media and to unify the program from beginning to end through careful quality control. Mastering a project from one studio and one set of sessions is straightforward because the masters will be on one format and the sound will be consistent from one piece to the next. Assembling compilations, on the other hand, makes interesting demands on the mastering engineer.
Compilations run the gamut, from tribute albums to holiday records to ethnic music collections, and they range in size from a single CD to multiple-disc box sets (see Fig. 1). Because each song usually comes from a different artist and a different set of recording circumstances (including location and studio recordings), it is likely that the engineer will receive a wide variety of masters on various analog and digital media.
The mastering engineer's challenge is to achieve a consistent, unified sound throughout the compilation. This is the primary focus as the engineer masters each piece and makes improvements to the mixes by using EQ and compression, changing the stereo image, and adding reverb.
PREPARATIONAs with any phase of studio recording, the amount of preparation you do can make or break a session. This is especially true with multiple-source mastering jobs. Media compatibility is a common problem, and dealing with it can be particularly time-consuming. I have received masters on almost every imaginable format, from old reel-to-reels and cassettes to assorted digital media, and I've learned the hard way that it is essential to know ahead of time the format of each mix I will receive. That way I can have the proper machines on hand.
Before the session begins, make sure you have received documentation and noise-reduction specifications for all masters used in the project. Cue any cassettes involved ahead of time. If there will be an outside producer/ coordinator, don't assume that they know what formats the mixes are on; a little detective work may be required. Beware of the rare client who doesn't understand that he or she has to mix the multitrack masters before bringing them in for mastering. Asking the right questions well in advance of your session will avert potential disasters.
Mental and physical preparations are equally important for a demanding mastering session. There's no room for the "fix it in the mix" approach at this stage, and the quality of the finished product depends entirely on your alertness, attention to small details, and ability to make critical listening decisions for hours on end. Make sure you're rested, start your session with fresh ears and a full belly, and forget about using recreational drugs or alcohol until the computer is turned off and the final CD is burned and safely nestled in its jewel box.
Of course, your monitoring environment must be suited to mastering. It's unwise to master on speakers that don't offer basically flat frequency response, or to master in a room that hasn't been treated or tuned to eliminate major resonant modes. You aren't doing anyone a favor by mastering in an unreliable monitoring environment, and you might do irreparable damage not only to the recordings but to your reputation as well. (For more on mastering, see "Mastering on a Budget" in the October 1998 EM.)
TRANSFER LOGISTICSWhen you have your room, your attitude, the necessary equipment, and the source tapes and other materials together, it's time to get to work. On a single-source mastering project (for instance, with a band that is recording an album), I usually plan on spending about six hours (approximately 10 to 12 cuts, 50 minutes total time). An hour-long compilation, done to high standards, can easily take ten hours or more.
The first half of any mastering session is generally spent listening to the source tapes, discussing various processing options with the client or producer, and making sure that he or she understands what can and cannot be done. This is the stage at which you make most of your crucial listening decisions. You should check each piece for inconsistencies and other problems; adjust it to an optimal gain setting; equalize it; convert it to 16-bit, 44.1 kHz digital format, if necessary; and import it into a 2-track editing program.
When assembling a compilation CD, you can reasonably expect to receive material in more than one media format, including a mixture of digital sources recorded at 44.1 and 48 kHz. To simplify the task of sample-rate conversion, presort your source material and transfer all the recordings with a similar format together (for example, all the 48 kHz DATs) before moving to the next format. This will not only save time but will also reduce the potential for human error, such as transferring a DAT at the wrong sampling rate.
Although any mixdown media can be used for mastering, by far the most common for compilations are digital audio tape, CD-R, and cassette. As I load tracks from analog masters into the computer, I run them through a 2-channel parametric equalizer, and from there I send them into a digital multi-effects processor for additional processing and analog-to-digital conversion.
For digital sources that are recorded at 44.1 kHz (DAT, CD, MDM, MiniDisc, or digital audio workstation), I use the 5-band digital parametric EQ and other processing functions of a TC Electronics M2000. For DAT sources recorded at 48 kHz, I either make an analog transfer through a parametric EQ or simply connect the analog outputs of a professional-quality DAT machine to the analog inputs of the M2000 using high-quality balanced cables. I have found both of these methods to be sufficient for maintaining the resolution of the original digital signal while making a sample-rate conversion in real time from 48 kHz to 44.1 kHz. Many digital editing programs can do sample-rate conversion, although this can be somewhat time-consuming on older computers.
As you enter the source material into your computer, keep a detailed log that includes source-tape information, EQ and other transfer processing, start and end times for each track, markers for editing, playlists, and any other information that might be helpful in the future to you or another engineer.
CORRECTIVE EQEqualization is a major part of most mastering sessions. This is a process in which an inexperienced engineer can do more harm than good. To avoid making any big mistakes at this stage, take the time to familiarize yourself with the way that professionally mastered records sound on your system. Always A/B any EQ changes for yourself and your client before you commit to them-and be conservative, especially when boosting levels. A single cut or boost of 2 dB can have a major impact on any mix, and with good equipment subtle adjustments of 0.5 dB or less are usually audible.
For the sake of continuity on a compilation, some pieces may need major sonic surgery, and this may require multiband cuts and boosts of 3 to 4 dB or more once everything is in the computer. Bear in mind that most people listen to recorded music on relatively small, inexpensive speakers (car stereos, boom boxes, bookshelf systems, and so on) at moderate volume levels, so make sure the tracks have sufficient definition at a soft listening level. Pay special attention to the extreme highest and lowest frequencies, which often get lost on consumer-grade monitors. Keep your ears fresh throughout this part of the process by limiting loud monitoring time and taking breaks as needed.
I like to deal with corrective EQ in real time during the transfer stage. I listen to a few passes of each individual piece at a moderate volume level and formulate an EQ curve to compensate for deficiencies in the recording medium, the mixing environment, or the acoustical space where the music was recorded.
For example, stereo-miked live recordings often have at least one band of pronounced, boxy room tone between 250 and 400 Hz, which can be easily detected by boosting the gain and sweeping the frequency of a medium-bandwidth filter within this range. When the EQ boost centers on this acoustical resonance mode, you will notice an audible increase of mushy midrange reverberation; gently attenuate the resonance while adjusting the frequency bandwidth as needed.
You can determine the proper amount of attenuation by comparing the EQ in its active and bypassed positions. You can also fine-tune the EQ parameters so that an acceptable increase in midrange clarity occurs with only a minimal decrease of overall warmth, instrumental tone, and other desirable low-end qualities. A skillful attenuation of the room's resonant frequencies often reduces the natural reverberation characteristics of the acoustical space, making it possible to further enhance a live recording by adding reverb while mastering. Begin with a conservative cut of 1 to 2 dB, increasing the attenuation as necessary.
You can deal with high-frequency harshness (typical of digital multitrack recordings, particularly on the original ADATs and early DAT recorders) by using a similar process, this time concentrating on the frequency range between 2 and 6 kHz. Be careful not to take too big a bite out of the highs in this range, and do a thorough A/B comparison to make sure that you aren't dulling the high frequencies or losing important detail and definition.
Broadband EQ boosting is another effective mastering tool. When applied skillfully, it can enhance the fullness and overall impact of a recording. Typical scenarios for this type of EQ processing include high-shelving boosts above 10 or 11 kHz to add shimmer and upper harmonics; low-shelving boosts below 80 to 100 Hz to enhance the power of the low drums and bass fundamentals; upper-bass boosts between 200 and 400 Hz for more punch and roundness (especially on digital multitrack recordings); and upper mid or treble boosts between 1 and 8 kHz to increase overall brightness and to aid the projection of vocals, guitars, drums, and solo instruments. EQ increases in the 400 Hz to 1 kHz range are rarely needed due to the fact that most microphones and monitors exhibit fairly even response in this region. However, increases in this range, when used cautiously, can bring out chordal instruments, warm up vocals, and make thin-sounding mixes bigger.
You may choose to do corrective EQ in the computer rather than during the transfer. Certainly the precision, affordability, and high quality of many equalization plug-ins (not to mention the allure of multiple undo levels) make this an attractive option. If you plan to EQ in the computer, keep your input levels a little lower during the transfer to allow for potential signal boosting and the subsequent increase in gain it may produce. Any overall gain changes, normalization, or compression should be done after all EQ decisions have been made. If you need to equalize a normalized region, avoid digital overloads by dropping the overall gain of the track by 2 or 3 dB before adding EQ.
FURTHER ENHANCEMENTSDuring the transfer stage there are additional enhancement options, plus a few critical quality-control checks, that you need to make whenever you work on master tapes from various sources.
Occasionally I add digital reverb to an entire mix, either at the client's request or on my own. I generally choose a short room or ambience setting with a considerable low-end cut and a decay time of one second or less. Then I listen closely to how the added reverb affects the drums, bass, vocals, and lead instruments. A small amount of reverb (7 to 15 percent) is usually sufficient to lend some extra air, sustain, and spaciousness to the mix without making it sound artificial.
Some mastering engineers apply compression during the transfer stage. Since most mixes I receive have some degree of compression or limiting already, I usually address compression later in the mastering process. But when it is warranted, as on live recordings and uncompressed mixes with infrequent amplitude peaks, I use a high-quality compressor in the analog domain or employ the digital compressor in the M2000.
Most semiprofessional, analog stereo compressors add significant coloration to the signal chain, which is one reason I don't recommend them for mastering. If your compressor has a hard-wired bypass, click it in and out to see how the circuitry affects a stereo mix before you use it on a mastering session. For units without a true bypass function, feed a signal into your computer with and without the compressor in the chain and compare the differences.
MAXIMIZING LEVELSWhether or not you decide to compress at this stage, level control is an essential part of the transferring process. If you have analog sources or analog processing, be sure to maintain proper gain staging throughout the signal chain to avoid adding noise with signals that are too low or introducing clipping distortion from signals that are too hot.
For digital transfers, level control is generally not available on DAT machines and CD players. Digital multi-effects units provide level control for digital transfers, ranging from +6 to -16 dB, with 16-bit dithering on the output. Note that every time you chain digital units together in this fashion, it's vital that the digital source (DAT, CD, and so on) is functioning as the master clock for all downstream processors (including the computer) to avoid jitter or resampling errors.
Make sure that each piece is transferred into the computer as hot as possible so that amplitude peak values are within 1 or 2 dB of the maximum digital ceiling. It's also important to check the left/right balance, particularly on analog sources, to ascertain that the channel gains are matched and that any center-imaged information (usually the kick drum, snare drum, lead vocals, and solos on studio recordings) actually appears midway between the speakers. Don't rely solely on the level meters; they are not always calibrated perfectly. Use your ears, and monitor constantly for distortion or imbalance in the stereo signal before it goes into the computer.
Vague center imaging may indicate out-of-phase information in the stereo spectrum, ranging from slight phase cancellation to complete 180-degree phase reversal of a single channel. The easiest way to test for this is to sum the stereo pair to mono by panning the left and right channels to the center. If some instruments or frequency ranges change timbre or disappear completely, there is a phase-coherency problem with the track (see Fig. 2). Fixing it is not as easy as finding it, and unless one side is truly 180 degrees out of phase, this is a job for a professional. A track that is completely out of phase can be fixed in the analog domain by using a phase-reversing, balanced XLR or TRS connector during the transfer to digital.
A pair of tracks that are 180 degrees out of phase can also be spotted visually and remedied in computer software (see Fig. 3); in this case the left and right channels will appear as mirror images, and the side that displays its transients as a negative voltage can be corrected using the DSP Invert function.
DROPOUTSIt is also important that you be able to recognize digital dropouts and problems that can occur because of high block-error rates (BLERs) during the transfer to the computer. You may already be familiar with the unmistakably "fritzy" sound of DAT dropouts, which produce brief periods of unpleasant fuzz or stuttering silence within a mix.
High error rates also can cause minor dropouts and slight "CD skipping" artifacts or subtle graininess that is most audible in the highs. Although BLERs can be monitored on some DAT machines, there is no way to fix dropout problems short of editing around damaged sections, mastering from a safety copy or an alternate mix, or playing the tape in different machines. (Occasionally, DATs recorded in one type of equipment will have compatibility problems when played in other machines.)
DIGITAL EDITINGOnce everything is loaded into the computer, the next step is to edit the tracks and assemble them in the proper order. When editing the beginning of a piece, start the region as close as possible to the first significant sound, without trimming it so tightly that an important aspect of the sound is removed. The start of the piece may sound artificial without a vocalist's breath, a guitarist's initial pick stroke, or a drummer's hi-hat tick. Begin your region a little to the left of such sounds (see Fig. 4).
Likewise, be careful not to cut off any significant room sounds or reverb at the tail of the piece. If you're drawing a fade-out over the end, try a longer fade region first, then shorten it by small increments if necessary. It's easy to make a fade-out that sounds too abrupt, so err on the conservative side, and check your fades on speakers and headphones. Sources with significant background noise or tape hiss may require longer fade-outs. Quick fade-ins of 500 milliseconds or less will reduce the shock of a noisy beginning coming out of digital silence.
For tops, tails, edits, and DSP functions, always start and end your regions on zero crossings, where the waveform crosses the horizontal "zero" line on both the left and right channel displays (see Fig. 5). This will greatly reduce those annoying little clicks, pops, and waveform distortions that plague even the best editors from time to time. If you're using a playlist-based editing program with crossfade capabilities, learn what the various crossfade options are-and don't be afraid to use a crossfade of 10 to 100 ms to get rid of a clicky edit.
ON THE SPOTYou can use EQ and gain "spot" changes-whether on an intro, a single note or phrase, or an entire section of a piece-to correct mixing or performance flaws, remove clicks and hum, clean up quiet sections, and even smooth over a troublesome edit. There is often a lot of this detail work involved in cleaning up semiprofessional and live recordings; how much you want to do depends on your time and workload, processing power, skill, and, most of all, patience. At this stage, you should also do the DSP Invert phase correction mentioned earlier, along with any other specialized operations such as pitch shifting, reverb, and time compression/expansion.
Before moving on to the final mastering stages, you should have each of the songs trimmed, edited, tweaked, and sounding the way you want, at a reasonable gain level. Note that it is not necessary to normalize every piece prior to the final consistency check; in fact, doing so may be a duplication of effort, especially if you're planning to use mastering-compression software.
ASSEMBLYThe final sequencing of the project is done by selecting the individual items from a menu and placing them in their proper order. For pop CDs, I start by inserting two to three seconds of silence between pieces. Punk and alternative bands tend to favor shorter gaps than this, while jazz discs and projects with lengthy pieces often work better with four to eight seconds of silence between cuts.
Audition these silences and judge the impact of the gaps between the pieces. This will also give you the opportunity to double-check the tops and tails on each piece and get a sense of their relative levels and EQ. To accurately gauge how much of a gap is needed between pieces, it's important to audition at least the final 45 seconds of each track and listen through the gap into the beginning of the next cut. This may seem like a long time, but a shorter listening period will often fool you into thinking that a smaller gap is needed, resulting in a program that seems to rush the listener along from one selection to the next.
One more crucial task awaits you: the final consistency check, where your collection of masterpieces is touched up and finally put on display. If you've done your EQ correctly, each piece will sound full on its own and compare favorably with every other cut on the CD. The disc as a whole should match the tonal characteristics and bass-to-treble balance of other CDs in that genre.
Chances are, however, that one or more pieces won't make the grade, even with their gain bumped up to match their neighbors'. As you listen, make detailed notes of these deficiencies before returning to your editing program to make corrections. If you're feeling particularly ambitious, you can return to the source material, access your detailed notes, and transfer the piece again with an improved EQ curve.
MASTERING COMPRESSIONMost mixes have a few volume peaks (usually on vocals, solos, or drums) that are at least 2 to 3 dB above the average level. Normalization brings these occasional peaks up to 0 dB on the digital scale, which should make a piece loud enough for your purposes. But if normalizing doesn't do the job, only mastering compression can effectively tame these peaks while bringing the overall level up by 2 to 3 dB.
For continuity of levels, it's important to compare the beginning, middle, and ending of every cut to determine an average volume or core level. During this process, a single track (or group of tracks) with a hotter-than-average core level serves as a benchmark for the CD as a whole.
Load the tracks, in sequence, into a CD-R burning application (I use Digidesign's MasterList CD) to determine the amount of processing you will use in the mastering compression program. Raise the gain of the quieter pieces in 1 dB increments until the core level of all pieces is consistent. Be prepared to lower the gain on pieces that seem unnaturally loud in the context of the compilation; these are often mixes without drums, or quiet and drony pieces whose lack of dynamic range allowed them to be transferred into the computer at a core level close to 0 dB.
These gain changes in the CD-R program are not final; they are merely a handy mock-up. By using mastering-compression software to implement the gain boosts (typically 1 to 4 dB on normalized material) rather than boosting in the CD-burning program, I avoid getting digital overloads when I burn my CD-R. Return to the digital editor to use the mastering compressor and make any last-minute changes based on your findings during the continuity check.
READY TO BURNAt this point there is one more important decision to make regarding the overall compression level for your CD. If the level of the "hot" pieces is satisfactory, you can just implement the calculated gain changes for the quieter pieces using your mastering compressor, and then you're ready to burn your CD master.
But if the core-level continuity you've achieved for this project is still noticeably below that of comparable CDs, you may decide to raise the gain of all the hottest pieces by a set amount (for example, 3 dB) to bring it up to a more competitive level. If you choose to go this route, it means that you have to add 3 dB of gain to all of the changes calculated in the playlist mock-up, as well. The resulting cumulative boost of 4 to 6 dB that some of these pieces will now receive should prompt you to check again for overcompression before going back to the CD-burning software.
In most instances, I use a mastering compressor (I prefer Waves' L1 Ultramaximizer) on all of the pieces (see Fig. 6). Even when pieces are at unity gain and no level change or compression is needed, I like to take advantage of L1's powerful dithering and increased digital resolution (IDR) functions. Dithering to 16 bits is a standard last step for any files, regions, or playlists that have undergone gain changes, equalization, compression, fades, or other digital signal processing that generates requantized (say, greater than 16-bit) digital word lengths.
It may not be the most exciting reading, but it's worth your while to read up on the technical uses of dithering and explore the options that your software or converter provides.
FINAL PASSNow sit back and listen through the entire project. If you're exhausted (as I often am at the end of a complex multisource job), burn a reference copy on CD or DAT and check your work after you've recuperated.
If you're dealing with a client or outside producer, make sure he or she gets a reference copy (clearly marked as "not for production"), and be prepared to back up the project or leave it on your hard drive for a few days until you get approval from all interested parties. When everyone's happy with the final version, burn a master CD to send to the replicator, and make a duplicate disc in case the master gets lost or damaged. If there is any doubt in your mind that you haven't heard the last of this project, do a complete data backup; save all playlist and editing information, if possible.
Remember that mastering is a highly skilled vocation, and there are no software shortcuts or hardware substitutes for the finishing touches that an experienced mastering engineer can add to a CD compilation. But there's also no harm in practicing the tricks of the trade in your own studio. Take your time, listen closely to every change you make, and learn all you can from your mistakes. The time and effort you spend will pay off every time your finished project is heard, whether it's on a boom box or over the ritziest audiophile stereo system.
Myles Boisen is a guitarist, producer, composer, and head engineer/instructor at Guerrilla Recording and the Headless Buddha Mastering Lab in Oakland, California. He can be reached via e-mail at firstname.lastname@example.org.