Welcome back to the second part of MusicTech.net’s complete guide to that mysterious process – mastering…
The Shadow Hills mastering compressor has both slow and fast response compressors (and, let’s face it, looks amazing).
Mix-buss compression is a tricky area – get this right and your mix can really come together. The instruments will all punch as one unit, and the gentle boost will bring up the quieter sounds, thickening the overall sound. However, get it wrong and you can suck all the life out of a mix. Mastering compression is very similar to mix-buss compression. It is a two-channel compression stage applied to the stereo mix, and so many mix engineers will prefer to leave compression off their mix, hoping that the mastering engineer either has better tools or more experience in this task. This is often a mistake. When used from the start of a mix, compression can influence the mix itself. Fader-rides will change; the very balance of the song itself will alter. Applying a compressor at the last stage of a mix, then, is going to make changes to your mix, and this technique is most often where mixes fall apart. In mastering there is no option to have the compressor on the mix from the start, so you have to be more careful to get the settings right.
The first decision to make when compressing is what compressor (or compressors) to use. Do you want something with a bit of its own colour, or do you want a cleaner signal? The safest bet is usually to go with a cleaner compressor, as colouration can be added in many other ways. However, take a look around most professional mastering rooms and you’ll usually see at least one analogue compressor famous for its unique character. The most common compressors in mastering are vari-mu or optical designs such as the Fairchild 670 (vari-mu) or Maselec MLA-2 (optical). Contrast this with the most common mix-buss compressors, such as the SSL G-Buss and API 2500, which are VCA-style compressors.
However, it is worth noting that this is not a strict rule and VCA designs do also make their way into mastering studios. Vari-mu and optical compressors tend to have slower responses than VCA compressors and some mastering engineers will even use two compressors together so they can have one fast compressor and one slower compressor. The Shadow Hills Mastering Compressor (pictured on this month’s front cover) combines both an optical and VCA compressor in one unit for exactly this reason
The Maselec MLA-2 is an optical compressor with a very subtle, open and transparent compression characteristic.
Slow compressors will allow transients to pass through untouched, working instead on the body of sounds and enabling you to gently bring up the level without reducing the impact of the drums. However, with slower attack and release times you might get audible pumping as sounds duck just after being played, before slowly coming back up. This is especially noticeable on longer, sustained notes.
Fast compressors can grab hold of transients and reduce them quickly, and with a fast release you shouldn’t get any audible pumping. However, this can reduce the impact of the sounds, and so a good blend of both fast and slow compression characteristics is often useful.
Remember, though, that a brickwall limiter applied at the end of the chain will act like a very fast compressor, so don’t compress too heavily with faster compressors at this stage. Some models of software compressor, such as the FabFilter ProC, provide powerful visual feedback of any gain reduction taking place, which can be incredibly useful for seeing how the compressor is responding to the transients in your audio. More commonly, however, you will simply see an LED or VU meter that displays the amount of gain reduction, but isn’t as easy to spot the subtle attack and release characteristics with these devices.
We usually start experimenting with a slow optical or vari-mu-style compressor at low ratios, perhaps 2:1 or even lower (sometimes as low as 1.2:1). Using fairly long attack and release times (around 50–100ms), adjust the threshold to a point at which the loudest part of the signal is being compressed by no more than 2–3dB. This is a good starting point, and while some tracks may require more, others may need less. Again, be sure to A/B the compressed mix with a non-compressed version and make level adjustments to ensure that you are comparing like with like. Not all mixes need additional compression at all – don’t get sucked into thinking that because you are mastering something you have to make changes, the best mixes will need no alterations. It can be just as rewarding to hand back a master with little to no changes made to it as it can having made drastic changes to a track.
FabFilter’s ProC offers great visual feedback – always useful for seeing how a compressor responds to transients in the audio you are mastering.
Another tool worth experimenting with is the compressor sidechain. By filtering out the low-frequency content from the compressor sidechain it is possible to let the high-energy, LF information through uncompressed and focus the attention of the sidechain on peaks in the mid-band and HF content. This can reduce any pumping effects that may otherwise reduce the effectiveness of a compressor, especially when using slow attack and release parameters. Remember that unlike a multiband compressor, the whole spectrum is still being compressed, including the low end. It is simply that the compressor threshold does not respond to the low-end information.
If we are going to use a second compressor we will usually put it in the chain immediately after the first compressor, using much faster attack and release settings, perhaps 10–20ms. This compressor will then be used to simply shave off the tops of transients before they hit the limiter. Again, we’re likely to use a low ratio, although perhaps not quite as low as the slow compressor – something around 2:1 or 4:1 will do the job, depending on how big the transients in the mix you are working on are. As we’ve explained, pulling down the transients can take the impact off the percussive elements in a mix, and the brickwall limiter will be doing more of this later on in the chain, so don’t go crazy with the fast compressor. Even as much as 1dB may be enough to remove the spikiness from an overly transient mix.
T-RackS’ Soft Clipper can be placed before the final brickwall limiter to yield another couple of decibels.
When using two compressors it is also important to think about how the characters of each will affect each other. We usually prefer to have a slow, characterful compressor paired with a transparent, fast compressor. So a tube vari-mu compressor (like the Thermionic Culture Phoenix) can be paired with a second software VCA-style compressor such as the UAD Precision Buss Compressor.
Sometimes a mix comes in where even multiple compressors in series and a whole host of EQs before and afterwards isn’t enough. For these situations multiband compression becomes a very useful tool. However, where single-band compression has the capacity to spoil a mix, leaving it lifeless or pumping, multiband compression can not only cause all the same problems but has the capacity to ruin the tonal balance, too. Careful adjustment and regular A/B’ing is vital throughout.
The hardest parameters to set on a multiband compressor are the crossover frequencies. Get this right and you are on the way to clean, transparent compression. Most multiband compressors will let you audition the band being compressed. Use this to hear what the compressor will be responding to and compressing. If there are any problem frequencies in your mix this is the perfect opportunity to hear and respond to them.
Take it to the Limit
While your full mastering chain may include any number of additional processors – from harmonic exciters to saturation effects and even reverbs – the final stage of mastering is almost always limiting. As a rule of thumb we will always place a brickwall limiter as the final element in our mastering chain, setting the ceiling to -0.3dBFS and using the threshold to bring up the level of the track. If more than 1–2dB of gain reduction is required at this stage we will usually precede the brickwall limiter with another form of limiting known as soft-clipping. Soft-clippers such as the T-RackS Soft Clipper emulate a trick some mastering engineers realised they could do by driving high-end converters.
Instead of hard-clipping the signal, the converters would soft-clip it, enabling mastering engineers to get a couple more dBs of level cleanly before going into the brickwall limiter. For extremely loud masters the trick of balancing the gain reduction across a range of different processors is incredibly important. While it is possible to throw a single brickwall limiter across your master buss and crank the threshold until you’re getting 7–8dB of gain reduction, the amount of resulting distortion will be huge compared to a well-mastered equally loud version placed through three or four well-configured processors.
The Loudness War
Even as early as the 1960s, record companies were trying to make sure their records were loud. The Motown record company adopted a strategy called Loud and Clear to ensure their records were as loud as they could be for the time, without compromising clarity. This strategy involved keeping records to less than three minutes – enabling wider grooves, boosting in the 8–10kHz region, and filtering out everything below 70Hz. To compensate for this, the second harmonic of the bass instruments was boosted, which gives the impression of a louder bass than there actually is thanks to a psychoacoustic property called the missing fundamental effect, in which the brain re-creates the missing fundamental pitch from the harmonic information without it having to be there.
Other techniques, including loudness envelopes that reduced the level in 0.5dB steps in order to create room for the level to go up later, were also used. This approach was especially important for 45RPM 7-inch singles, which would be played on jukeboxes in bars. The labels wanted to know their songs would be loud enough to play over the background noise of the venue. This wasn’t such a problem for 12-inch 33RPM LPs, which were usually played in homes, so less attention was paid to keeping these as loud and the grooves would be placed closer together to get more music on a side rather than more level.
The Loudness Wars have resulted in increasingly loud masters. The top track here is City’s Full, by Savages, released this year. Note how it is almost a straight line compared to the 1986 Let It Bleed CD release of the Rolling Stones’ Gimme Shelter, below
AM and FM radio stations further escalated the Loudness War, using their own compressor and limiter chains to increase the signal strength of their stations so that listeners would find them clearly when tuning through the dial. They also needed to know that, once tuned in, the listener would be able to hear the radio station over the noise of their car and so on. However, this processing was applied at the broadcast stage rather than the mastering stage by the radio stations themselves. Of course, it didn’t stop record companies demanding louder masters so that they could ensure their songs would be the loudest on the dial.
When CDs launched in 1982 the RMS level was kept around -18dBFS, with peaks rarely reaching up towards 0dBFS. By the late 80s this had increased such that 0dBFS was occasionally being reached, but RMS levels were still as low as -15dBFS. The first records to hit 0dBFS on nearly every beat didn’t occur until the 90s – Metallica’s Black Album being an early example. This enabled music to reach RMS levels around -12dBFS, with peaks hitting 0dBFS regularly. However, compared to today’s releases this is considerably dynamic, with RMS values closer to -6dBFS being common in much of today’s chart releases. Some releases have even been measured with an RMS of -4dBFS. At some point along this way to louder and louder masters the object of clarity got lost. Louder and louder tracks began to sound audibly distorted. This isn’t always a problem – electronic music and hard rock styles can afford to have some distortion – but even these can only go so far.
Some audio engineers and audiophile listeners have been complaining about the ever increasing loudness of masters and the accompanying reduction in dynamic range throughout this time, but only recently has the average listener and the major press begun to get involved. A huge backlash against Metallica’s Death Magnetic led mastering engineer Ted Jensen to distance himself from the project, claiming the mixes arrived brickwalled at his studio. Death Magnetic is the most commonly cited example as it was simultaneously released as both a heavily limited audio CD and a more dynamic Guitar Hero version, which gave fans a direct opportunity to compare the two. Tens of thousands of fans signed a petition asking for the album to be remastered from the Guitar Hero stems.
This backlash hasn’t quite put an end to the Loudness Wars, but it has led to more attention being given to the amount of limiting, and albums are now being praised for their open and dynamic sound when not over-compressed, with many big-name artists now insisting their albums are released without heavy limiting. The new Daft Punk album, Random Access Memories, for example, has been mastered with plenty of dynamics left in and this has been praised by both critics and fans alike.