Mastering: You Can Do It Yourself (Part Two)
In the last post we covered a description of mastering and a couple of its basic operations, level and frequency balance. This time we’ll cover some equally important but often overlooked elements, namely editing, spreads, and exports. Everyone but mastering engineers thinks that making a track sound better is the only job of mastering, but these other operations are mission-critical for a great product. Let’s take a look.
Also available in this series:
- Mastering: You Can Do It Yourself (With a Little Caution)
- Mastering: You Can Do It Yourself (Part Two)
Editing during mastering has gone through a complete metamorphosis in just a few short years. Until the mid-80’s when mastering entered the digital age, most editing was still done by hand using a razor blade and splicing magnetic tape on an analog two track recorder. But as the demand for CDs began to rise, razor blade editing no longer worked in a digital world and quickly gave way to electronic editing in the digital domain using a Sony DAE-3000, which was basically a modified video editor, and two BVU-800 (and later DMR 4000) 3/4” video decks which carried the digital audio (see figures 1 and 1A).
Today, unless you’re in some sort of alternate reality that’s behind the one we’re in by 20 years, all editing is done on a Digital Audio Workstation, which as you know is a hardware/software package using a personal computer as the engine.
While the editing speed and capability varies from unit to unit, the main operations required by the mastering engineer remain the same. The mastering engineer must supply fades (both fade-ins and fade-outs), basic additions/subtractions to the song via cut and paste techniques, and sometimes spreads (the time between songs). As with most mastering operations, what may seem easy can actually be quite difficult without the proper knowledge of how to apply the proper tools.
Just about anyone with a workstation knows how to apply fades, but does that mean that they’re the right fades? Another one of the main elements of professional mastering is making sure that the outro fade (if there is one) sounds smooth. As a result, the mastering engineer is frequently called upon to either add the fade himself, or help out one that’s not quite the best it could be.
Even in these days of automated mixing and fades that can be drawn directly on the waveform in the workstation, many mix engineers still actually leave the master fade-out completely up to the mastering engineer. It’s not unusual for a mastering engineer to get a note from a mixer that says something like, “Please add a 4 second fade to the end.”
There are two schools of thought on the fade-ins or headfades; one uses a sharp “butt cut” and the other a more gradual algorithmic fade. Regardless of which type of fade is chosen, the principle is to get rid of count-offs, coughs, and noise left on the recording before the song begins. Although this seems to be an easy procedure, care must be used in order to maintain the naturalness of the downbeat. This is another task that should have been completed during mixing, but it’s surprising just how much it’s overlooked.
The type of fade selection used can make a big difference in the sound, as you’ll see. The temptation is to use a linear curve to make a fade as in Figure 2. However, an exponential curve (Figure 3) is sometimes smoother and much more realistic sounding.
Even when a fade is made during the mix, it sometimes needs some help due to some inconsistencies, especially if it was done manually with a master fader. “Following the Fade,” means drawing a curve that approximates the one on the mix, only smoother.
There are a lot of potential edits to be done during mastering. The first is shortening or altering the arrangement of the song. For example, after living with a mix for a while, it’s not uncommon to come to the conclusion that the outro could be longer, or the bridge shortened, or the intro half as long. No need to remix, just edit during mastering. Perhaps a music bed is to be used for television cues, but you need cuts that are 29.5 seconds, 9 seconds, and a 3 second bumper - time to get creative with your editing and use a little bit of Elastic Time or other time-stretch apps to get it to where you need.
Whatever the reason, it usually (but not always) easier to do this kind of editing during mastering. Just remember, if it sounds clipped or edited, it’s not acceptable and you’ll probably have to mix it again and do your editing there.
Different versions of a song are sometimes required, like in the case of offensive lyrics. The mastering engineer can accomplish this edit in two ways. The best sounding way, but the one that requires the most work, is to have two versions of the same song - one with the original lyrics, and a TV track. A TV track is a separate mix that has everything but the lead vocal so that a singer or rapper can perform live or on television just using the backing track (I know it’s cheating, but it’s also a big part of the business). The section of the TV track is substituted for the section in the song that has the offensive lyric.
Another way to cover up offensive lyrics in mastering is to cut in a sound effect or test tone in the spot where the offensive lyric occurs. This can work OK as long as the sections of the song are short, but if they’re over a second or occur frequently, it can definitely disturb the flow of the song. That’s why it’s better to have a TV track (we’ll cover the various mix versions that are commonly delivered in another post.)
Mix fixes are yet another editing job that frequently occurs during mastering. If there’s a brief digital clip, glitch or noise, it has to be fixed. Sometimes it can be fixed by zooming in and lowering the level for a few milliseconds, but if you can hear the fix, then the song has to be mixed again. A noise elimination plug-in can sometimes work as well.
The spread is the time between each song, when mastering for CD or vinyl distribution where all the songs will be played back in a sequential order. While the spread might seem to be quite arbitrary in many cases, the savvy mastering engineer usually times the spread to correspond with the tempo of the previous song. In other words, if the tempo of the first song was at 123 beats per minute, the mastering engineer times the very last beat of the first song to stay in tempo with the downbeat of the next. The number of beats in between depends upon the flow of the album.
Please note that a timed spread might not be appropriate in all cases since each project is unique, but it’s a good place to start. Many times a smooth flow between songs is not desirable and a longer space is far more appropriate. The spread in that case is replaced with a 2 or 3 second (or longer) area in between songs to keep them disconnected. Occasionally a cross-fade is used between songs so there’s no real spread, but that’s still a decision usually left for mastering as well.
Export (sometimes called Delivery)
Once upon a time (about 5 years ago), exporting was a bigger deal than it is today. Since CDs were the primary music format, it was important to get a master to the replicator with as few errors in the digital medium as possible. In order to accomplish this, a format call DDP (Disc Description Protocol) was used and printed to an Exabyte tape (common for computer backup at the time), which was then sent to the replicator.
Today, most replicators prefer to get the DDP file (which contains all the fades and PQ information) via FTP from the pro mastering engineer. For everyone else, a CD-R master sent to the replicator works just fine, but it’s best to send a second one as a backup, just in case your original master contains an area so corrupted that it requires a replacement.
Generally speaking, every major project will have a number of masters cut, depending upon the marketing plans and policy of the label. This usually breaks down as follows:
The CD Master – This is the master that the glass master at the replicator will be cut from, which in turn will ultimately make the replicated CDs. If an artist is to have a world-wide release, a separate CD master for each region of the world is made.
The Vinyl Master – If a vinyl record is desired, then a separate master from the CD master is required since the song sequence is usually different from the CD due to the split sides of the vinyl format. This master is sometimes just a CD with 30 seconds of dead space to indicate a side switch, but sometimes it’s a separate master for each side. Ask which one the replicator prefers.
The On-Line Master – Since the on-line portion of sales is now such a large part of the overall sales picture, a separate MP3 and/or AAC master is made. This master is specially tweaked to provide the best fidelity with the least amount of bandwidth.
Backup Masters – Most labels will ask for a backup master that they will store in the company vault. The mastering engineer should also make a “house” backup as well to save time should a new master be required at a later time.
Before a master is sent to the replicator, a pro mastering facility verifies it’s integrity in several different ways. If the master is a CD-R, the disc may be tested using a StageTech EC2 error detector (See figure 4). If the error rate is too high, the disc is rejected and another is made.
Most major mastering facilities will also employ some type of audio verification as well, where a production engineer will listen to the contents of the master (sometimes with headphones) to ensure that it’s free from pops, clicks or glitches. Although the StageTech error analyzer isn’t practical unless you’re a major facility, it’s easy enough to listen to the entire disc in detail just to make sure that your master is acceptable. Don’t cheap out on this. It’s really easy to take the easy way out and cut the master, then send it out without checking it. Make sure you listen first just to be sure.
Encoding an MP3 of your mix may seem easy, but to make it sound great it requires a bit of thought, some knowledge and some experimentation. The idea is to encode the smallest file with the highest quality, which is, of course, the tricky part. Here are some tips to get you started in the right direction so you won’t have to try every possible parameter combination. Remember though, that the settings that might work on one particular song or type of music might not work on another.
The Source File
Lossy coding like MP3 (which actually throws away some of the digital data to make the file smaller) makes the quality of the master mix MORE of an issue because high quality audio will be damaged much less by this type of encoding than low quality audio will. Therefore, it’s vitally important that you start with the best audio quality (highest sample rate and most bits) possible. That means it’s better to start with the 24 bit mix master or make the MP3 while you’re exporting your mix, than using something like a 16 bit CD master as the source for your MP3 encodes.
It’s also important to listen to your encode and perhaps even try a number of different parameter settings before settling on the final product. Listen to the encode, A/B it to the original, and make any additional changes you feel necessary. Sometimes a big, thick wall of sound encodes terribly and you need to ease back on the compression and limiting of the source track master. Other times, heavy compression can make it through the encoder better than with a mix with more dynamics. There are a few predictions one can make after doing it for a while, but you can never be certain, so listening and adjusting is the only sure way.
Here are some things to consider if your mix is intended for MP3 encoding:
MP3 ENCODING TIPS
- Start with the highest quality audio file possible.
- Filter out the top end at whatever frequency works best (judge by ear). MP3 has the most difficulty with high frequencies - cutting them out liberates a lot of processing for encoding the lower and mid frequencies. You trade some top end for better quality in the rest of the spectrum.
- A real busy mix can lose punch after encoding. Sparse mixes, like acoustic jazz trios, seem to retain more of the original audio punch.
- Make sure your level is reasonably hot. Use the “Tips For Hot Level” in the previous Tut or even normalize if you must, but it's far better to record at a good level in the first place.
- Try different bandwidth settings. Your encode might actually sound a better at 32 kHz than at 44.1 kHz since the encoding algorithm can concentrate on the more critical midrange.
- Don't totally squash your mix with a compressor/limiter. Leave some dynamic range so the encoding algorithm has something to look at.
- Use multi-band compression or other dynamic spectral effects very sparingly. They just confuse the encoding algorithm.
- Set your encoder for "maximum quality", which allows it to process for best results. The encoding time is negligible anyway.
- Remember, Mp3 encoding almost always results in the post-encoded material being slightly hotter than the original material. Limit the output of the material intended for MP3 to -1.1dB, instead of the commonly used -.1 or -.2dB, so you don’t get digital overs.
Unfortunately, all MP3 encoders are not created equally, and therefore don’t provide the same quality output, so using a good encoder is the biggest advantage you can give yourself.
An MP3 encoder to consider is LAME, which is an open source application. LAME is a acronym for LAME Ain't an MP3 Encoder, although the current version really is a stand-alone encoder. The consensus seems to be that LAME and the Fraunhofer encoder (who co-invented the MP3 format) produce the highest quality MP3 files for average bit rates of 128kbit/s and higher. Another good MP3 encoder is the one found in iTunes.
Regardless of the encoder, there’s really only one parameter that matters most in determining the quality of the encode and that’s bit rate, which is the number of bits of encoded data that are used to represent each second of audio. Lossy encoders like MP3 provide a number of different options for its bit rate. Typically the rates chosen are between 128 and 320 kilobits per second. By contrast, uncompressed audio as stored on a compact disc has a bit rate of about 1400kbit/s.
MP3 files encoded with a lower bit rate will result with a smaller file and therefore download faster, but they generally play back at a lower quality. With a bit rate too low, "compression artifacts" (i.e. sounds that were not present in the original recording) may appear in the reproduction. A good demonstration of compression artifacts is provided by the sound of applause, which is hard to data compress because it’s random. As a result, the failings of an encoder are more obvious and become audible as a slight ringing.
Conversely, a high bit rate encode will almost always produce a better sounding file, but also results in a larger file, which may take an unacceptable amount of storage space or time to download.
Bit Rate Settings
For average signals with good encoders, many listeners once considered a bit rate of 128kibit/s (providing a compression ratio of approximately 11:1) to be near enough to compact disc quality. However, listening tests show that with a bit of practice, many listeners can reliably distinguish 128kbit/s MP3s from CD originals. When that happens many times they reconsider and then deem the 128kbs MP3 audio to be of unacceptably low quality. Yet other listeners, and the same listeners in other environments (such as in a noisy moving vehicle or at a party), will consider the quality quite acceptable. That being said, 160kbs has mostly become the norm for acceptable quality MP3s.
- 128kbs – lowest acceptable bit rate, but may have marginal quality depending upon the encoder. Results in some artifacts but small file size.
- 160kbs – lowest bit rate considered usable for a high quality file.
- 320kbs – the highest quality, large file size but may be indistinguishable from CD.
Constant vs. Average vs. Variable Bit Rate
There are 3 modes that are coupled to bit rate that have a bearing on the final sound quality of the encode.
- Variable Bit Rate Mode (VBR) - maintains a constant quality while raising and lowering the bit rate depending upon how complex the program. Size is less predictable than with ABR (see below), but the quality is usually better.
- Average Bit Rate Mode (ABR) - varies the bit rate around a specified target bit rate.
- Constant Bit rate Mode (CBR) – maintains a steady bit rate regardless of the complexity of the program. CBR mode usually provides the lowest quality encode, but the file size is very predictable.
At a given bit rate range, VBR will provide higher quality than ABR, which will provide higher quality than CBR. The exception to this is when you choose the highest possible bit rate of 320kbps where, depending upon the encoder, the mode may have little bearing on the final sound quality.
There are some additional parameter settings that can have a huge influence on the quality of the final encode. These include:
- Mid-Side Joint Stereo (sometimes called MS Joint Stereo) - encodes all of the common audio on one channel and the difference audio (stereo minus the mono information) on the other channel. This is intended for low bit rate material to retain surround information from a surround mix source and is not needed or desired for stereo source files. Do not select under normal circumstances.
- Intensity Joint Stereo – again intended for lower bit rates, Intensity Joint Stereo combines the left and right channels by saving some frequencies as mono and placing them in the stereo field based on the intensity of the sound. This should not be used if the stereo audio contains surround-encoded material.
- Stereo Narrowing – Again intended for lower bit rates, allows narrowing of the stereo signal to increase overall sound quality.
It’s better not to check any of the above parameters when encoding stereo files that originate at 16 bit or above. With these disabled, the encoding will remain in true stereo, with all of the information from the original left channel going to the left side and same for the right channel.
For Best MP3 Encodes:
- Don’t Hypercompress the source master
- Cut some of the high frequencies
- Use Variable Bit Rate
- Turn OFF Mid-side Joint Stereo, Intensity Joint Stereo, and Stereo Narrowing
- Try not to use a Bit Rate below 160kbs (higher is better)
- Set the Output to –1.1dB since encodes export hot
Exporting for iTunes
iTunes uses the AAC (which stands for Advanced Audio Coding) format for it’s store and, contrary to popular believe, it’s not a proprietary format owned by Apple. In fact, it’s part of the MP4 specification and generally delivers excellent quality files that are about 30% smaller than a standard MP3 of the same data rate. All new music destined for the iTunes store must be encoded at 256kbs at a onstant bit rate with a 44.1kHz sample rate. The iTunes store stopped selling 128kbs songs in April 2008.
Here’s some info on some of the parameters of the AAC encoder.
Stereo Bit Rate: allows you to select the bit rate. The standard setting is now 256Kbs and the highest-quality setting for this format is 320Kbps.
Sample Rate: enables you to select the sample rate. Never use a higher sample rate than the rate used for the source. You making the file larger and not gaining anything.
Use Variable Bit Rate Encoding (VBR): This option helps keep file size down, but quality might be affected. VBR varies the number of bits used to store the music depending on the complexity of the sound. If you select the Highest setting from the Quality pop-up menu for VBR, iTunes encodes up to the maximum bit rate of 320 Kbps in sections of songs where the sound is complex enough to require a high bit rate. Meanwhile, iTunes keeps the rest of the song at a lower bit rate to save file space. The lower limit is set by the rate that you select in the Stereo Bit Rate pop-up menu.
Channels: This pop-up menu enables you to choose how you want the music to play through speakers — in stereo or mono. Select the Auto setting to have iTunes use the appropriate setting for the music.
Optimize for Voice: This option is meant for podcasters and filters the sound to favor the human voice, which is obviously not something you’d want for music.
It’s best to select the highest bit rate in the Stereo Bit Rate pop-up menu and leave the other two pop-up menus set to Auto.
Now you can see that there’s a little more than just affecting the level and tonal balance to do a complete mastering job. But if you follow all the tips in this post, with a little bit of practice your mastering job will sound truly professional.