Mastering – Why Not to
“Mastering degrades the timbres, and is nothing but ‘Loud-Garbage-Out’.”
A quick tip:
Before I start with the nitty-gritty of what digital sound, sample rates, bit depth, headroom, and mastering work truly involves, let me point you to an article about mastering that I did a few years back. I explained what plugins you may want to use, and why use them, when mastering a track. Now add to that information, this tip:
“If you render out a sound file with ample headspace, and you import it into a new project in your DAW, then you can layer the file any number of times to raise the volume of the final product, apply effects like reverb to the master chain or any other chain, or in general, master individual chains by filtering out and working on very specific frequency ranges.”
Now let me add this information to the above tip, that I have myself tried this method on more than one occasion, and achieved exactly the same results in sound as any big commercial release, yet I have not released even a single song with that sound. I just didn’t like that sound! Call it a musicians’ ego if you want to, but there were good reasons (details below).
Getting a head around headspace:
Mastering needs 3-6 decibels headspace for the engineers to work their magic. But what really is this headspace; just lowering the volume of the mix (or individual tracks) down? At a more basic level; what is ‘Digital Sound’ really?
1. What is Sound: Any sound, from any source or instrument, is a mixture of audible frequencies, with bits scattered all over the spectrum, from the lower levels towards the higher levels. A few peak frequencies give a source its’ distinct voice. Try a low pass filter on hitats, and see how far below can you drop the shelf before you stop hearing any sound at all. Try it on a piano or a pad! You will know how each sound is a mix of many wavelengths and frequencies, each adding their own ‘timbre’ to the overall ‘timbre’ of the sound. Now consider a sound mix, like a song, where many sounds have been mixed together to create one sound. Do you think, with all their frequency ranges intertwined, you can really separate the instruments completely using any filters, equalizers, or the like?
2. Analog Sound: An analog sound recording is a continuous record of all the wavelengths and frequencies that make up a sound, along with their amplitudes, much like a line-graph. When recorded on a medium like a tape, it is played back in a flowing continuity.
3. Digital Sound: Digital sound, for example the industry standard PCM wave, on the contrary, is a disjointed record of the sound; that is, it is made up of chunks of data. From afar it appears like a line graph, but a closer inspection reveals its dotted make; the quantization (see below).
4. The chunks of digital data: Consider for example a ‘one second long sound sample’, made up of 16 different wavelengths. This sound’s analog record with include all the values that those 16 wavelengths will assume, as a part of their rising and falling amplitudes, over the course of one second. However, digital data is recorded in bits and bytes, which are not the same as a continuous line drawn on a piece of paper. Digital sound is recorded by breaking that continuous one second sound data, into many smaller chunks; for example, 44,100 chunks of a single second (44.1KHz), or 48,000 chunks of that second (48KHz). These chunks are called samples, and hence ‘sample rate’ defines how many samples per second a wave file contains.
Then of course, these chunks themselves have chunks of data; for our current example, all 16 wavelengths will be present as 16 small pieces, each with a range of data, albeit a much smaller range. Say if you chop a piece of pipe into smaller pieces, each piece would still have two ends, and water will still take time to pass from one end to the other. But a ‘Digital Sound’ can only be recorded using a set number of data points, with each point stating a precise value; for example, each sample can be described by using 16 (our current example), or 24, or 32 markers, with each marker having a single precise value (and not a range of values) for that particular sample. The number of markers so used to describe the details of a sound sample, they represent the depth of detail; the bit-depth.
Thus a one second sound chopped into 44,100 pieces, will be represented by that many blocks of data, with each block’s data encoded using 16, 24, or 32 markers. But what does each one of these markers really represent?
5. The data markers: In our example, let us assume that each marker is associated with one out of the 16 waves. Now each marker will represent the average amplitude of that wave’s piece in that sample. Thus the range between highest and lowest values would be lost in favour of one single average value. What does this means? This means, the analog wave that looked like a continuous line, gets converted into a dotted line in digital world, which would look continuous from a distance, but dotted up close. This is what is described as quantization error, as intervening values are lost in favour of single average value. In practical world, 50-60 KHz sampling rates yield results that can simulate an analog sound for the purpose of human hearing abilities. Anything beyond is non-detectable overkill. And that is why 48 KHz (more continuity in sound) is a good sample rate to render, with a 24 bit depth (more headspace).
But what has all this got to do with headroom, clipping, and mastering? Or wait; what did I just explain up there?
1. The working of bit depth: Suppose you have 16 one litre bowls to fill (16 bit information per sample), and a bucket full of water (a song made of many sounds). You pour water in each one of them to different levels (this is what your 16 bit wave sample would look like). Imagine each of those bowls represent a bit that is attached to a section of the audible frequency range, say bowl one to 20-100 Hz, bowl two to 100-300 Hz, and so on so forth. When you are creating a song, each instrument will add its’ sound (or water in our example) to more than one bowl; (because each sound is a mixture of many wavelengths). The bowl that would overflow, would lose water (that is, the amplitude for a particular frequency/wavelength section, that goes beyond what the system is capable of mapping, generally 0 db tops, that amplitude’s true value would be lost, and the resulting average information recorded wouldn’t be accurate, and would sound as a distortion, called clipping, and besides, the average value for extreme variations is generally inaccurate information). Remember how when you are producing a song, and all sounds are peaking well below the 0db level, and then you add one sound that just goes way over the charts and the master chain starts recording clipping; yeah, that is one bowl clearly full beyond full, and that full bowl is what the master mix is recording as error.
2. Dealing with near full bits: So when you send a song, with all the bowls full, or nearly full, there is nothing left for the engineers to add to that sound (of course, what they add is what I despise, as described below). So when a mastering engineer, say for example, needs to jack up the midrange, and that range’s bowl is already full, they have no choice but to empty some of it, to create space for the jacking up (layering and filtering tricks can help here). But of course, as you already know now; every frequency segment contains frequencies from more than one sound source. So you end up with one sound getting improved (allegedly), while many others take the hit. And this is why it is called GIGO!
On the contrary, if you give them plenty of room in the bits, to work with, the mastering engineers can really jack up the levels that need to be jacked up, without destroying anything else (allegedly, once more).
Putting timbres on timbre:
Of course, in a perfect world mastering engineers would be able to add or remove only the frequencies that need to be so, but unfortunately, in real world these frequencies, as already noted above, are intertwined. This is the very reason why the mastering engineers also don’t want you to add any effects to the song you dispatch to them. All the effects, like reverb, flanging, chorus, delays etc, they all add newer frequencies to the sound, covering an even bigger range, and complicating the maze. Now remember; you are sending a mastering engineer a single complex sound, a song, and not individual tracks that they could work on, and then mix together to create a gorgeous product for you to claim as your own artistic achievement. This is what makes mastering a good intentioned job done to create a poor end product.
1. The amazing sound timbres you create: All the gorgeous sounds that you create or mix, while creating a composition, they all have some very particular dominant frequencies. It is the sound of these frequencies, and how they mix with the rest of their own non-dominant frequencies, that makes them appealing to you. Each sound may have one or two peak frequencies, but the rest are not just absent, but are rather pushed out of prominence, in a prominent yet gradual way; consider a wave that stays low for a long time, then rises up sharply, and falls down quickly, and finally fades out slowly again. This is what all sounds are; some just have a smaller foot print in one or the other direction of their peak.
2. The timbres that mastering adds: When a mastering engineer is working with a mix of sounds, they may limit their impact to a particular frequency range, with an intention to improve a sound peaking in that range. Unfortunately however, whatever they do with that range, will also impact other sounds that might as such be peaking in other ranges, but have some bits of their limbs in the impacted range. Suddenly the overall character of those impacted sounds changes. Suddenly the ‘hihats’ may start sounding trashier even if you had tuned them to sound soft and rounded. The ‘kick’ may start sounding duller, even if flatter. The ‘bells’ may become rough, and ‘whistles’ and ‘pads’ might become heavier or shriller. The list is endless, and now you know why I haven’t released any of my songs with a commercially mastered sound in spite of my understanding of the subject matter.
This is what my dear friend, I call LGO. You lose the charm of the sounds that you create, only to get a louder product. So, do you really need mastering for your compositions? Well, this question is better answered by asking these questions:
a) Can you increase the volume of your song enough by layering some elements of the song (experience will teach you which ones)?
b) Are there really any sounds that matter so much to you that you would rather not master the track?
c) Do you rather want a commercial sound than pamper your artistic ego?
Answer these for yourself and you can figure out whether you need mastering in a particular instance, or not.
Fatal Urge Carefree Kiss