The Big Mastering, CD, and Digital FAQ

A digital audio signal is defined by these two parameters – the sample rate is essentially how many times per second a digital converter takes a “snapshot” of your audio – a 44.1khz signal means that 44100 times in a second, a section of the incoming or outgoing signal is converted.  The bit depth is the detail of that snapshot, represented by a binary value of the size specified by the bit depth: a 16-bit audio sample is represented by a number between -32,768 and +32,768.  A 24-bit signal is +-8,388,608.

Okay, great.  So what does that mean, really?

The sampling rate defines the frequency that can be reproduced – the “Nyquist frequency” is one-half the sampling rate, so the maximum frequency a CD can reproduce is 22khz, or about 2khz higher than humans can hear (at the best of times, most of us top off at around 17khz).  Bit depth is the resolution, and subsequently the “detail.”  Higher bit depths mean higher possible dynamic ranges, lower noise floors, etc.

The main reason is portability and translatability.  A mastering engineer’s job is, mostly, to ensure that your final product sounds good everywhere.  Along the way that means punching a few things up, evening out EQ, adjusting loudness, managing dynamic range, etc.  It also means prepping the music for the final delivery medium – CD, Tape, Vinyl, etc.  It’s the final polish, and the final prep involved in getting your music where it needs to go.

Well, you can.  It’s just not usually a great idea.  Mastering Engineers usually sink time and money into specialized gear, resources, and knowledge specific to the process.  Plus they’re a second set of fresh ears.  But that said, a lot of people do it themselves, for better or worse.

Since time immemorial, producers and artists have been pushing mastering engineers to increase the loudness of audio.  The nonstop march of technology has been making it possible to do this.  But why?  Well, to get your attention.  Listen to the radio, and the song that grabs you out of your drive-time zombie haze is the loudest one.  Watching TV, the loud commercials certainly get your attention (So much so there are now international standards to combat that).  On a 3-minute pop song, it’s fine.  When it’s all you hear?  Fatiguing in the extreme.

This really depends on your engineer.  The common formats are all digital, but if you’re a guy who records to 2-in tape, there’re engineers out there who work with it.  The general rule is, though, you want to give your engineer the highest-possible quality mxidown you can provide, in whatever format the engineer will accept.  So don’t send a maxed-out MP3 file.  Usually a 24bit, 44.1khz audio file in a standard, uncompressed format (wav, aiff, SDII) is sufficient.  Again, discuss this with your engineer, as each may have different requirements

Yes.  It doesn’t have to be much – preferences vary from engineer to engineer – but usually anywhere from 1 to 6 db seems to be common.  You don’t want to be slamming your meter at just-below-clipping 0dbfs, because it gives the mastering engineer a little less breathing room (and it also means there’s a higher chance of “intersample clipping” – see below).  In the old days, you didn’t want too much, because everyone recorded at 16bit, and too much headroom meant a much lower signal-to-noise ratio, but if you’re recording at 24 bit you’ve got something like 48db of headroom before you’re risking signal degradation.

This is a complicated question.  A lot of very talented engineers use a compressor on their master bus.  There are a lot of software products that an in-the-box mixer can use to emulate the same high-end gear these engineers have.  It’s very popular to compress the master bus.  HOWEVER.  It’s usually done pretty gently, to “glue” stuff together, and Eric “Mixerman” Safran likes to point out it prevents surprises when sending a mix off for mastering.  This is all true.  Where it becomes a problem is when it’s done to excess – at some point 2bus compression is going to stop being a subtle glue and start becoming part of the finalizing process that the mastering engineer usually handles, and whatever you do as that process can’t be undone or adjusted by the masterer.

There are a number of schools of thought on this one.  Steve Jobs, Neil Young, and George Massenburg all have said at some point in their careers that they envision a future where all audio is delivered as 24-bit, 96khz audio.  And yet some big digital audio guys say this is ridiculous and a waste of time and storage.  The primary argument for using it is that when doing signal processing “in the box” on a computer, the noise and distortion caused by digital signal processing will be at a frequency far above the range of human ears, resulting in a cleaner audible signal.  This may be true, but it’s slightly debatable – a lot of ITB software either already does what’s known as “oversampling” to handle this problem at lower sampling rates, and some software doesn’t behave entirely predictably at high sample rates (while a lot of it does, you’d be best knowing what will and won’t work correctly).  Additionally, some hardware may or may not function as well at these high sample rates.  And finally, if you deliver this stuff to someone who doesn’t use these higher rates, you’ll be forced to downsample, and not every sample rate converter is created equal.

So it’s kind of a provisional “maybe” – if you’ve got the disk space, and you’ve got gear that you know works well at these sampling rates, and you can convert without problem, and you’re convinced you can hear the difference and it works for you, then…well, go for it.

When you convert from a high bit depth, like 24 bit, to a lower one, like CD-standard 16, there’s going to be a loss of quality and an addition of distortion.  It’s not a lot, but it can be some.  What dithering does is it adds a special noise of its own, usually at frequencies above the range of human hearing, that (approximately) compensates for the distortion added by changing bit depth.  Darkroom mastering has been kind enough to write up an excellent detailed explanation.  Generally, dithering is not something you have to worry about when mixing, and only the mastering engineer needs to care.

Digital audio is by nature, a quantized format.  When a converter reconstructs old-fashioned sound from a digital signal, it basically is doing a lot of math to figure out what shape of a wave is defined by these digital numbers.  Sometimes, even if the numbers themselves don’t exceed a certain maximum, the mathematical curve the define could loop up past it.  This is known as “intersample clipping” – namely, clipping that happens to your signal between the individual sampling points.  It’s generally not a big deal, but it can add some audible harshness to your signal.  The easiest way to avoid it is to not peg your meter at 0db – these intersample peaks are rarely much higher than the samples around them, so adding a bit of headroom basically keeps this from happening.

All of the above mentioned are standards for measuring and calibrating loudness.  Loudness in an audio signal is different than volume – volume may be the actual scale of signal power, but loudness is how loud an audio signal is perceived, which is a combination of volume, dynamic range, and such measurements over time.  Because modern signal processing technology has made it possible to completely blast the loudness of a signal, certain international broadcast standards like EBU R128 and ATSC have been enacted to prevent huge variations in loudness between certain types of broadcast signals (primarily so commercials don’t shatter your eardrums when you’re watching a quiet TV program with the volume up).  The K-system is a system developed by legendary mastering engineer Bob Katz (who quite literally wrote the book on mastering) that requires an engineer to first calibrate their playback system to a standardized baseline (c-weighted 85db per channel for a calibrated pink noise signal).

For more info, see http://www.tcelectronic.com/loudness/loudness-explained/

Redbook is the 1980-defined standard fro Audio CDs.  It defines CD audio as 44.1khz, 16-bit linear PCM-encoded audio, with a number of parallel tracks called P, Q, R, S, T, U, V and W which encode various bits of data useful to the CD.  P and Q codes in particular handle things like “track start” and “track end.”

There’s also “orange book” and “yellow book” and a few other colors used for defining other CD standards, like CD-ROM, CD+G, etc.

Well.  Most CD authoring programs will get you close. iTunes, Winamp, Nero, etc will burn a CD that adheres to some of the standard, and will make CDs that play in most players – mostly because most CD players are pretty forgiving.  However, these CDs are not guaranteed to work everywhere, and if you’re trying to make a master to send to a replication plant, you want something that will be full-on compliant, or there’s a good chance they’ll reject it.  This can range from something simple like Toast Platinum to something expensive like Sequoia.

As far as most CD players are concerned; not much.  Accurately, though, a “replicated” CD is one that’s been pressed by a plant – it’s a thin layer of aluminum embedded in plastic, manufactured to have a reasonably long lifespan.  A duplicated CD is essentially a CD-R, a burned copy of your master disc.  CD-Rs are thought to have a shorter lifespan than a full replicated CD, but are also significantly cheaper to produce and are able to be produced in smaller quantities.  The shortest run for a replicated CD is usually 300, and it’s not generally financially advantageous until quantities of 500-1000 or above, whereas duplicated CDs can be done in much shorter lots.

Is there an advantage to one over the other?  It’s really a personal and financial choice.  It can be argued that a replicated CD looks more professional than a duplicated CD, and for large runs is a cheaper option.  For short runs, a replicated CD can be comparatively expensive to produce, and if you don’t think you can move 1000 units of that CD, you’ll be stuck with a closet full of CDs you paid a significant amount for.

DDP is a “disc description protocol” file – basically a specific type of image for an audio CD. CD replication plants often take these files in lieu of physical CDs, which can be quite nice since there’s nothing to get scratched, damaged, lost in the mail, etc.  It’s usually only accepted for replication, though, and not short-run duplication.

PQ codes are the parallel data tracks that define certain aspects of CD metadata – most importantly track start, track end, pauses, and some metadata codes.  Generally you don’t have to deal with these directly, unless you’re trying to do something clever like put hidden tracks before track one or whatever.  CD creation software handles this stuff for you and usually hides it behind friendlier “track start” and “track end” interfaces.

CDs are, of course, limited to 44.1khz, 16-bit audio.  There’s a maximum of 99 tracks on a CD, a maximum length of 79 minutes (most top out at 74), and the minimum track length is 4 seconds (including a 2-second track pause).  The pause before the first track must be 2 seconds, although subsequent tracks may not need to have any pauses at all.

Digital formats like mp3, AAC (iTunes) etc employ “lossy” compression, meaning they achieve smaller file sizes by removing and approximating some audio data.  While modern encoders are generally pretty good, there are still some issues with frequency response, headroom, and intersample clipping that need to be addressed.  While you can just rip directly off a CD to mp3, you’ll get a better quality final product if your material is mastered specifically for digital.  This includes testing against encoder types (how does it sound as an AAC?  As an mp3?  At various bit rates?), verifying headroom, checking for intersample clipping, etc.

There are a number of ways. If you’re being distributed by a label or a distributor like IODA, TheOrchard, or some others, this stuff is often handled for you.  If you’re an indie, you can use a service like TuneCore.com, CDBaby, DistroKid, or Reverbnation (among others) to package your content and put it on a variety of digital services around the world.

If you wanna get even more indie, there’s bandcamp.com, which doesn’t license to other services but also has a different pricing model and is very artist-driven.

“Mastered for iTunes” is a certification program used by Apple to hold recordings to a certain set of standards.  Essentially, albums that bear the “Mastered for iTunes” label are mastered with sufficient headroom and encoded using a specific set of tools, supposedly ensuring maximum quality.  Most indie artists are not able to supply their own encoded files or participate directly in the MFI program, but the tools used are available from Apple, and even non-certified music might benefit from at least having the MFI process applied to verify that it will transition well to a digital format.

That has gotten somewhat complicated.  Pandora used to take all comers, but since they have gotten reasonably popular they now highly curate their selections based on criteria only they know.

The process begins by filling out a form at submit.pandora.com.  From there, you will be asked to upload a sample track.

This sample track will be the basis for whether they accept or reject your album.  If it is rejected, they will not give you any further information as to why.  If you are accepted you will be provided with further instructions regarding submitting the remainder of your album.

iTunes, Windows Media Player, and…well, basically every reputable music player out there uses an internet service called Gracenote to identify CDs. (if you’re old-school, you might remember when it was called “CDDB”).  In order to make this work, you need to stick your brand-new CD in a compatible player, type in your data, and then use the player to submit the data to the Gracenote database.  Gracenote provides complete instructions, including their rules for track naming, how to handle Classical releases, etc, on their FAQ site.

Why, the mark of the beast, of course!  HA, I KID.  The UPC/EAN is a numerical code, usually associated with a barcode, that uniquely identifies your product for sale.  You need to buy one, usually, for a nominal fee.  Sometimes your label has ’em.  Sometimes you can get them from your distributor (Tunecore and ReverbNation will allow you to get a UPC for your release).  If you want to sell your music in stores, physical or virtual, you need a UPC.  The UPC code also can (and should) be embedded in the CD metadata itself.

ISRC is the International Standard Recording Code – it’s a unique identifier for every recording you make.  It’s unique to an artist, and the country of origin, and the year released.  You don’t need them, per se, but if you’ve got them, great.  They’re useful for tracking songs for copyright and statistical reasons.  Say your recording appears on 5 compilations – the ISRC code would stay the same across all of them, amkign it possible to track sales of that one particular song.  This is particularly useful in Europe, where this is tracked well (less so in the US).  If you want an ISRC code, you can usually get them assigned to you by a label or a distributor, but if you want to assign your own, there’s a $80 one-time fee (in the US) from http://www.usisrc.org/.  If you want to assign them to other artists, you need to pay an annual fee.  This varies from country to country, so if you live in Andorra, I’m not sure who you need to go through.

So…if you’ve got an mp3 or AAC, and you want to have info like the song title, artist, genre, bpm, key, etc show up in your player, you would fill in that info in the file’s ID3 tags.  It’s simply put a standard for digital music metadata.  When you rip music from a CD, most encoders, if using the aforementioned Gracenote information, will convert the CD’s data to ID3 tags.

CDTEXT is a standard for encoding certain text fields onto a CD, including track title, performer, composer, and a short message.  While a supported standard, very few consumer CD players actually bother to implement this (it seems to be most notably used in car stereos), and only a few computer-based CD playing programs pay much attention to it (one or two will sue CDTEXT, if present, only if Gracenote info cannot be found).  It’s not a bad idea to have it on your CD, and it certainly won’t hurt anything, but it’s not something to lose sleep over, either.