Author: steve albini
In Reply To: quibbles and bits (posted by scobie)
Subject: RE: quibbles and bits
First time caller, long time listener.
Thank you all for listening so closely to Silkworm's Italian Platinum. Those of you who are complaining about how not-loud it is may find a discussion of mastering to be of interest.
There are (basically) two mastering paradigms: Representational (intended for the listener at home to have the experience the band have listening to their music in the studio) and Stylized (intended to give the listener an "improved" or "exaggerated" version of the master.
It is impossible to simply "copy the master flat" onto a consumer medium (LP or CD) for several reasons, which I will address individually.
The biggest problem is dynamic range. CDs have an absolute dynamic range of about 90dB, but because resolution (sound reproduction accuracy) falls off by half (one bit) with every 4dB of level drop, the usable dynamic range of a CD is actually only about 30dB. Sounds recorded much lower than this are distorted due to the coarseness with which the waveforms are reproduced with fewer bits.
LPs have a much wider useable dynamic range, about 60dB or so, though dirty, poorly pressed and scratched records have a worse noise floor, limiting their dynamic range to about that of a CD or a bit better.
If a master were simply reproduced dynamically "flat" (no change in dynamic range) on CD, then the quiet bits would sound weird. For LPs, this is much less of a problem, so only in cases of very long or loud program material is limiting or compression required, but for CDs, something must be done to keep the signal in the "useable" dynamic range.
Dynamic range is generally limited by one of three means: Limiting, compression or multiband compression.
Limiting is a quick, automatic suppression of transient peaks (tiny fractions of a second where individual waveform excursions would be too large to be sampled accurately -- called "clipping"). This allows the program level to be raised somewhat, keeping more of the quieter moments within the useable range, while avoiding disasterous clipping of the transient peaks.
Compression is an overall reduction of dynamic range by "automatic volume control" means -- as the signal gets gradually louder, the compressor lowers the gain in the circuit, making the loud parts less loud (and, in effect, making the quiet parts less quiet).
Multiband compression is splitting the signal into separate frequency bands (bass, midrange and treble, for example) and compressing each before recombining them into a composite signal. This has the effect of making tonal changes in the music less noticeable. Taken to its extreme (an infinite number of bands compressing infinitely), this would result in any input signal turning into white noise at the output.
Since songs are generally mixed one at a time, there are naturally level and tonal differences that become apparent only when they are sequenced together into a final album.
If a master is reproduced tonally "flat," then a song that sounds loud and aggressive when heard in isolation may not have the same impact if it follows another song with an even louder coda. A song may sound too dull or too bright when listened to in the context of an album. These differences are adjusted in the mastering stage to make the album as a whole sound as intended, using an equalizer and level controls.
There are other, more arcane matters addressed in mastering (groove depth, land spacing, elliptical dynamics, etc. in vinyl, glitch suppression, DC offsets and assembly in CD mastering), but that's the lion's share of it.
Mastering is always a play of these matters against one another. Limiting and compression usually rob the music of some of its transients ("brightness" or "energy"), which can be simulated by adding high-end or midrange EQ. Strong bass signals require a lot of electrical energy, and may make a signal less "apparently loud," which will require a compensation in actual level, etc.
In the Stylized paradigm, the mastering engineer will try to present the "best" version of the album he can imagine, regardless of how different it is from the master tape. "Best," in this context, usually means "loudest" and "most aggressively jump-out-of-the-speakers in your face."
In the Representational paradigm, there is a presumption that the master tape is pretty much what the artist intends, with contextual problems solved, and maybe a few specific adjustments (decided-on in advance after repeated listening).
No piece of pie awarded for guessing which paradigm Silkworm has been in for its entire career.
Making a master "louder" cannot be a simple matter of turning up the volume, as the signal would overload. Simulating extra "loudness" is done by making the sound thinner (our ears are most sensitive to sounds in the midrange -- around 1kHz, much less so in the bass region), compressing it (so the quietest bits are brought up in level to be as loud as the loud bits), or adding distortion (called "exciting" the signal, it is actually adding distortion products which are harmonically related to the music to replace treble content lost through compression).
If you could sit in the mastering room and hear the effect these processes have on music, you would be aghast that musicians who care about their art would allow such damage to be done to their records.
The least destructive mastering allows a small improvement in apparent volume in exchange for a relatively un-fucked-up sound quality.
I know for a fact that Italian Platinum gained 4dB of actual level and probably another couple "apparent" dB in tone during the mastering process. Tim heard the effect of 6dB of level boost (and accompanying limiting), and the sound quality suffered, so he backed off.
Honestly, your Silkworm record may not be as "loud" as Tool or a friggin' Afghan Whigs CD, but if it was, it would sound worse, and you wouldn't want to listen to it as much.
It's not a trick, it's part of the unforgiving and difficult process of making a record.
best,
-steve albini