Tips & Tutorials | December 16, 2013
It's not easy being a mastering engineer these days. It sounds so simple — a little assembly editing, a little EQ, some compression, just tweaking clients' tapes so they sound really good over REALLY good monitors in a REALLY GOOD room. Sounds like a lotta fun to me.
But there are clouds in paradise. In my last piece on mastering I mentioned a friendly little game played by clients called The "Make My Record LOUDEST Of All!" Game." In this highly competitive blood sport, clients beg, threaten, cajole and extort the mastering engineer to get him/her to make their particular mixes sound louder than any other CD on the planet. They often bring in their favorite LOUD CDs and say, "Beat this!" Lotta fun, especially when you've got a stack of unpaid bills.
Now, this game is based on the assumption that the CD that is louder will sound better to consumers. We'll consider that idea below. It also assumes that such louderness will translate into better record sales. Such assumptions are central to the process, but first we've got to consider the physics of the problem.
In digital audio, the loudest level without distortion we can have is 0 dB Full Scale (FS). Because there are no other bits available above this level, we encounter clipping instantly, and often that clipping is brutally offensive and violent. If we take a 0 dBFS sine wave and amplify it by, say, 60 dB, it will become a 1 kHz. square wave, with a level of +3 dBFS. We cannot reproduce a sine wave above 0 dBFS. Period.
With digital audio, then, the game has changed. We no longer have a safe and sane analog output level specification that has any correlation to perceived loudness, but rather a digital clipping specification: 0 dBFS, which is detected by a peak meter that gives no indication of perceived loudness and which lacks correlation to any VU or RMS sort of amplitude measurement (which does correlate well with perceived loudness). Interestingly, digital console manufacturers have acknowledged and standardized this, so that engineering practice has evolved into pushing the peak levels right up to 0 dBFS and treating that level as a "nominal overload level." "Hot" mixes may cruise along at levels as loud as +22 dBu (10 volts RMS, 30 volts peak-to-peak)! To drive the point home even further, manufacturers have replaced the VU meter and its RMS-based brothers with a "peak" meter, which indicates ONLY the peak levels. Such levels are usually 6 to 20 dB hotter than VU levels, except for steady state tones, and there is no reasonable correlation between them, or with what we hear. In a nutshell, we're now obsessing about peak levels instead of good levels, and ignoring our perceived loudness levels into the bargain.
In a previous article, I wrote, "One of the primary goals of mastering, as driven by client preference, is to make the CD as LOUD as possible. . . . The louder of two otherwise identical recordings will be preferred by end-users. The engineer that can most effectively push the level right up near 0 dBFS wins!"
I got some irate mail about this paragraph, and, frankly, I think the criticism is somewhat justified. While my statements reflect industry practice in a general way, they do not take into account the basic nature of the problem, and they make some unfounded assumptions. So, let's go a little further with this subject.
Interestingly, I don't personally know of any study that rigorously shows that a louder version of a recording is preferred to a softer one, but nonetheless any subjective test that doesn't have precisely controlled levels will be regarded with suspicion as unreliable. I've casually assumed that someone did the research, probably years ago, and that that is why the practice is ingrained in all of us who do subjective testing.
Therefore, I suggest that it is safe for you to assume such a preference in listeners. I may actually test the assumption some day, especially if I can't find any supporting research and can get someone to hire me to do the study. At the same time, my studio experience has been that, when comparing two otherwise identical versions of a recording, louder is better, within a range of normal listening levels. There is considerable psychological and neurological justification for such a principle, including increased intelligibility, emotional intensity and the adrenaline stimulation that are all caused by increased sound levels. However, such a phenomenon exists only as a function of the DIFFERENCE between two amplitude levels. A recording that is constantly loud defeats this effect, and usually ends up being tiring, irritating and generally unpleasant.
When we're mixing, and compress a track and then compare between compressed and bypass, if the apparent loudness goes up when the compressor is engaged, we often get the mistaken impression the sound has "more energy, more punch," when really, it's simply, well, louder! As soon as we adjust the compressor's gain so it has the same subjective loudness as the bypassed signal, we are often seriously surprised to hear that the compressor isn't doing anything good to the sound. Worse, we find it may be doing some really bad things. I've been bagged by this little mistake at least a zillion times!
Similarly, mastering engineers may use compressors to help the sound of a "flabby" or weak recording. But when such compression derives its impact primarily from increased loudness, we lose. The quality of the sound, including such essential musical and technical attributes as clarity, transient response and stereo separation, often goes down when too much compression is applied.
How can we deal with this problem? This is where the whole thing enters the twilight zone. Except for one particular playback modality, the car, the whole problem shouldn't exist! It shouldn't matter whose recording is loudest (at least within limits), because the actual loudness of a recording in playback is under the direct control of the listener, and the limits of that loudness are not a function of 0 dBFS, but of the listener's loudspeaker/amplifier system and listening environment! If the music is a little soft, the listener can simply TURN IT UP! It's that simple!
The limits to this are (1) laziness, and (2) troubled environments, cars in particular. Laziness has been dealt with, sort of, with the general implementation of remote controls that have level adjustment on them. Cars are tougher, because of their extremely limited dynamic range. We'll discuss that in a moment.
This all came to a head recently in a posting to the ProAudio listserv on the Internet, by mastering engineer Bob Katz. For those of you who do not know, Bob is a well-established and highly regarded mastering engineer. Further, he is an extremely good RECORDING engineer, whose work for Chesky Records (among others) has earned him the reputation of being a recording engineer's recording engineer (AKA "pure envy!"). He also hosts an excellent website,www.digido.com. I recommend him to you most highly.
Here is Bob's ProAudio rant:
"A potential client . . . had come to me for mastering months ago and was very pleased with what I did on a country-music mastering job. So I thought I would call and [push for] doing some mastering for their hit stuff. [The client] did a little song and dance about their European people wanting to do their own mastering. And another song and dance about their wanting to build a mastering studio themselves. Then I said, 'If I can prove to you that we can do it better, then you wouldn't have to build your own mastering studio.'
"Then he said, 'You know, your ideas about overcompression and stuff, we want our songs to sound as loud as possible on the radio.' I said, 'They will. Do you know that processing on top of overprocessing actually ruins the sound of the music?' He said 'I've never done the testing...' I said, 'The kids will love the sound of the record I can make when it makes it on the radio. And if you feel you have to do it, make the album sound real good, but make a special single version, and squash that.'
"He said, 'Hmmm, those are interesting ideas,' but I sensed he wasn't convinced... Overcompressed sound makes me sick, nauseous to the stomach, if you want the truth. So, I have to prove to him it doesn't matter what you do when it gets to the radio, that good-sounding records sound great at home AND on the radio, and all the rest... [I feel like] I'm pushing against a mountain of ignorance... The sound of the CD is in jeopardy!!! Does anyone realize this... that because of today's completely-unjustified overprocessing, we made better-sounding LPs 25 years ago than the CDs we're making today?
"To prove it, I put on a Simon and Garfunkel LP and a Pink Floyd LP, both 25 to 35 years old. They sound great! Moderate low level upward compression at the bottom end was done to keep the signal above the record noise. And relatively little mastering compression, probably just limiting to keep the cutter from frying... Sounded G-R-E-A-T.
"Do you want to make C-R-A-P the rest of your mastering life? No? Then fight this overcompression thing, before it beats us. Current commercial pop records are so squashed they have incredible distortion and occasionally [as little as] 3 dB peak-to-average ratio. It's worse than tragic. This is a situation directly attributable to the new digital processors.... we never had to process that much in the days of analog and we never could even if we wanted to. What is needed desperately today are some ears, a lot of restraint, and a lot of education and some voices of sanity out of the wilderness. HELP!"
Bob makes a strong point. The loudness mythology is firmly in place, and mis-education is still a big problem across the recording business. The idea that someone wouldn't hire Bob because of a desire for overcompression is scary, bordering on utterly lunatic. Meanwhile, many contemporary CDs are ridiculously hot. I recently auditioned the new Gloria Estafan CD and found that its levels cruised between -3 and 0 dBFS, while a 3-year old Toni Braxton CD is equally whacked. In both cases, levels and EQ are really hard-edged, fatiguing and unpleasant. While there may be some argument that they are being mixed for Auratone-type playback venues, they still sound awful over a general array of decent loudspeakers (and keep in mind I REALLY LIKE the music in the Braxton CD! This is not a musical issue, but more akin to listening to great recordings through a boombox). Also, it's not a knock on the mastering engineers, either. We all have to do what we're told, if we wish to continue working in the field. If you don't believe me, re-read Bob's post more carefully.
Why does this happen? The reasoning goes something like this: radio broadcast is essential to record sales, and if it doesn't sound good on the radio it won't sell in the record stores. Meanwhile, most radio listening is done in cars. So there's a double whammy — it's gotta sound good on radio (which has serious limitations, particularly in the AM band) and it's gotta sound good in cars, which have VERY limited dynamic and spectral ranges, not to mention stereo limitations.
Radio in cars sounds like, well, radio. Heavily compressed, usually, with plenty of EQ thrown in. Sometimes reverb as well. Why? Because the chief engineer has a mandate from the boss to jazz up the transmitter sound both to get the level as hot as possible within the power limits imposed by the FCC, as well as to make the station's signal sound as attractive as possible for listeners in cars, who are the primary consumers of radio.
The silver lining to this is that the radio chief engineer is the one doing all the radio preparation for us, and so, at the mastering stage, we don't have to do it. In fact, we probably shouldn't, because then our mixes will be double-dipped in heavy compression! Not cute! Not pretty!!
So, if we leave radio audio up to the radio guys, then we can concentrate on making our CDs more suitable for other end-user climates. Which is to say, we can concentrate on making better-sounding CDs.
Unfortunately, all the cassette and CD manufacturers have been sticking cassette decks and CD players in cars, and cars are now a primary listening venue for music, maybe THE primary listening venue.
How to deal with this? The noise floor in a car is around 70 dBA SPL, and the maximum playback level is a little over 90 dBA SPL. Some sort of compression is called for (some of my classical CDs, for instance, are simply unlistenable in the Spyder at, er, 55 mph), and compression needs to be done for such venues, with care and thoughtfulness as well as some real sensitivity to the musical issues. Further, such compression should have general consistency.
What the CD or cassette doesn't have to be is LOUDEST! Instead of scanning the dial searching for loud and intelligible stations, the listener is going to be scanning CDs or cassettes. Anything in the audible ballpark will be fine, and the listener will simply tweak the overall level to bring it to a satisfactory level and leave it there until the CD is done, or until he/she slows down for a toll booth and gets a really strange look from the toll collector.
There's a lot of craft to this, and the best mastering engineers do it really well (I include Bob Katz on this list — I actually first discovered his recordings over the radio in a car, while I was doing the Berklee commute, and the stuff sounded good enough there that I said to myself, "Self, you gotta pick up that CD!"). The actual dynamic signal range we need for CDs, in both cars and other venues, is about 20 dB for acoustical music and 12 dB for pop/rock. Further, we can hear music 10 dB below the noise floor, and will put up with that without complaint, for brief periods anyway. If we can fit our clients' recordings into that range of levels, we can make recordings that will work, dynamically, in a wide range of venues.
But, the real advances here are going to come via the development of more active and high-quality car playback systems. Manufacturers have recognized that premium sound systems are BIG sales arguments, and that for many people, the car is the primary listening space. The problem is not a mastering problem, but rather a hostile and underdeveloped listening environment. I just returned from an AES Conference on Small Room and Automotive Acoustics, and the sense there was that we have a LONG way to go before we've tamed the car, but at the same time we have begun to work quite seriously at it
So what do you with all of this?
First, you don't hire a mastering engineer on the basis of whether or not he/she makes the loudest records. Instead, you hire the mastering engineer on the basis of his/her track record and perceived quality of work.
Second, you don't worry about radio. Forces are at work that will take care of that for you, whether you like it or not!
Third, you work at sane levels. I recommend (and use) -14 dBFS as a nominal signal level, and I calibrate all analog gear so that level is equal to +4 dBu. I suggest you do the same. RMS (VU) levels should get as high as +3 VU, while the peaks can hit +14. If you're confused about this, buy a pair of Dorrough mastering meters, which show both peak AND VU levels simultaneously. An education, all by themselves!!
Fourth, remember that we're up against some fairly brutal signal processing practices in mastering. It isn't just Bob. A couple of years back I interviewed a major mastering engineer, and he pointed out that some major-label clients often asked for so much level that he would just run the mix right up to + 3 dBFS, so they could hear the distortion. When they then complained he would back it down just until they stopped complaining. Then, he would advise them to cut a production safety master 10 dB lower, so that if and when it ever became time for a Greatest Hits compilation, they would have something in the vaults that wasn't quite so, well, dated-sounding, distorted and downright embarrassing!
So, you might want to try it Bob's way. My suspicion is you'll like it a whole lot better, particularly in the long run.
Happy dynamic ranges!