Outspoken would be an accurate term to characterize mastering engineer Bob Katz. He has strong opinions on many audio-related topics and is particularly adamant in his condemnation of the “loudness race” — the continual increase in average level, and the consequent reduction of dynamic range, that has resulted from the heavy compression of CDs.
Katz devotes an entire chapter of the recently released second edition of his book, Mastering Audio: The Art and the Science (Focal Press, 2007), to this very subject. He makes a convincing case, backed up by a slew of charts and graphs, for proving how overcompression has hurt the overall sound quality of contemporary music. But Katz's book offers a lot more than that, also giving a thorough explanation of the mastering process — from setting up the studio to sequencing the material to monitoring levels, among many other issues.
Mastering since the 1970s, Katz has a number of Grammy Award winners on his résumé. He has also patented a couple of mastering-related products, including the K-Stereo DD-2, which is a processor that gives the mastering engineer control over ambience and stereo image.
Katz works from his own studio, Digital Domain (see Fig. 1), which is located outside of Orlando, Florida. I had a chance to talk with him at length about a variety of mastering issues, including practical advice for recordists who are bringing their products in for mastering, techniques that mastering engineers use, and, of course, his feelings about the loudness race.
What kinds of improvements can a mastering engineer make to a recording?
In some cases, all that's needed is the most subtle polish that you can imagine; in other cases, mastering can be a major transformation. And it depends on whether the mix engineer or producer is expecting a large transformation because they were having trouble with the mix, or because they are looking for the best possible translation of their existing good mix.
Are frequency issues — such as an overly boomy bass — easier to compensate for than some other types of problems?
Well, yes and no. When it comes to frequency issues, the extremes of the spectrum are easier to handle than the middle. For example, if someone says the electric piano stands out too far, that really is a mix issue. If I start to deal with the frequencies of the electric piano, I'm probably going to affect the vocal, I'm probably going to affect the guitars and other instruments that are there in the midrange. But if somebody comes in with a boomy recording, we have to look to see whether the recording is fairly neutral to begin with. If the relationship of the bass drum to the bass is very good to start with, then some careful equalization on the mastering side might just keep that relationship intact, and we'll end up with a good result. But if the problem is caused by the bass drum, and the bass drum is too loud and the bass instrument is a little bit weak, then we'll have to work really hard to find a way to reduce the bass drum without affecting the bass.
What might you do in a case like that?
There are many tricks that we've learned over the years. Let's start with a bass drum that's going kuh-thump, kuh-thump, kuh-thump — and it's too loud, but the bass instrument is too weak. In the first (and, hopefully, the easiest) case, I might take a highpass filter and put it in at, who knows, 40 Hz, you never know, and see if I can get the bass drum to be softened enough so that it isn't perceived as being too loud. That's the easy solution.
Are you talking about the thumpy part of the bass drum or the clicky part?
That's part of the problem, because you can't distinguish the beater that well from the body of the bass drum. The beater has harmonics right on up to 1 or 2 kHz. So imagine if someone tried to apply the kind of equalization that you might apply to a bass drum [in the mix] in a mastering situation. When you're mixing the bass drum, you might add a significant amount at 1 kHz, 2 kHz, 900 Hz, 800 Hz, 700 Hz — that would certainly make a vocal sound pretty nasal, wouldn't it? So we have to resort to other tricks. If it's the clicky part of the bass drum, we might [be able to] ignore it. That's usually the best solution, because often in cases like this, the cure sounds worse than the disease. And often it's the mix engineer who brings it up and says, “Listen, the bass drum's really clicky.” And we say, “Well, you know, it didn't bother us too much.” Because when the mix engineer is really close to the music, he's still going to be hearing mix issues, even if he comes and attends the mastering session. And we have to say, “Sit back, because now we're trying to make your existing mix sound the best way possible, not remix it.”
But with the clicky part of the bass drum, it's possible to take a multiband compressor and adjust the center frequency fairly narrowly around where that area is, and adjust the attack time of it so that it just touches that part of the bass drum without touching anything else. And then you cross your fingers. But in many cases, if we end up doing more than 1 dB of that kind of correction anywhere, it starts to affect something else. But if it's the boomy part of the bass drum, certainly, getting in under 40 Hz (if the bass instrument doesn't touch those notes too much) may produce a very acceptable result.
So you'd roll off at 40 Hz?
I'm not saying this as a general rule, but for a bass-drum problem, it's a good place to start.
So in the technique that you're describing, you're using some sort of shelving filter below a certain frequency.
The steepness of that filter can be very important, actually. If it's the bottom of the bass drum that's offensive, and you want to stay away from the bass, [using] a very steep, linear-phase filter such as what can be done in the Weiss or the Algorithmix Red [see Fig. 2] might be the most transparent solution. Because if you do a gentle filter, you might get up into 50 or 60 or 70 Hz, right into the bass. So a very sharp filter might be the solution. But a sharp filter with a normal equalizer, known as a minimum-phase equalizer, tends to produce a phase shift several octaves above it. You can really hear that, so you want to use an equalizer that's as transparent as possible. And I nominate a linear-phase highpass. Now in the reverse situation, where the bass drum is too light, you can have some good luck centering a fairly narrow — say, half-an-octave wide — boost at around 60 Hz to get a little bit of that kuh-thung back into the bass drum. Another situation where the linear phase sounds best is dealing with “1-note bass,” as the linear-phase EQ does not produce phase shift when doing a narrow-band cut.
Do you come across bass-frequency problems a lot?
Yes, bass-instrument problems are the most prevalent of all the kinds of problems that we find in mastering. That's because more and more mixes are being done in tiny control rooms with tiny loudspeakers. That's why we have to have a very neutral room ourselves, so that we know that the problem is not being caused in our own room. And I've resorted sometimes to producing an equalization that looks like the Grand Tetons. Finding out the key [of a] song is sometimes the easiest and quickest way to deal with it. As soon as you know the key of the song, and [if] you have good relative pitch, you know when the bass is playing the tonic or the dominant or the second or whatever, and I can look it up on our frequency chart faster than I can sweep an equalizer — especially if it's a walking bass. How can you possibly sweep an equalizer for a note that goes by so quickly?
So you figure out what the note is, then you figure out its frequency, and then you apply EQ at that point?
Exactly. It can be much, much, much faster than any of the traditional techniques, [such as] sweeping the equalizer. Just think about it as a musician, and you can see the logic in it [see Fig. 3].
Talk about how you use M-S (middle-side) technique to manipulate individual elements within a 2-track mix.
M-S manipulation is something that we mastering engineers turn to when and if the mix has a problem, and they couldn't remix. It's a remix tool. M-S, when used properly by someone with good ears and a set of monitors that has good stereo separation, can be an amazing tool to minimize artifacts. For example, if the vocal is sticking out a little bit too far on the peaks but the band has great dynamics, if you use a stereo compressor on that, you're going to squash the band as well as cheat the vocal down. But [it can work] if you use a compressor that only compresses the M channel, and you limit its bandwidth so that it stays away from the bass (which is also usually in the center), and you make the attack long enough that the snare doesn't lose its snap, and you only have about 1 dB of gain reduction. If you try to stick within the vocal frequencies, you can polish that mix and take it from an A to an A+. If the mix was so bad that you need more than 1 dB of gain reduction in the M channel to try and cheat that vocal down, it should have been remixed in the first place.
So you would be treating only what's in the very center of the stereo image.
In that case, yes. So in that case, M-S manipulation can be done so transparently and so beautifully that it becomes a cure with no perceptible disease.
As a mastering engineer, is there a huge difference in approach that you take between, say, an acoustic project and an electronica dance-music project?
Every project is unique. But definitely, you have to have a different approach for each one, and if you go in with a preconceived notion, it's also just as bad. One of my specialties is that my ears were trained classically from day one, much like Bob Ludwig. He was a classically trained musician before he became an engineer. I know what live music sounds like produced by live musicians. And that helps in every kind of music that I've dealt with, including dance. Sure, you have to know the genre. You have to know it really well. But one of the secret weapons that I'll pull out, if I'm working with dance music that might have come in a little overcompressed, is to try to enhance the dynamics. What moves your body when you're on the dance floor? The rhythm. I know what a good, real, live bass drum sounds like. And if the samples they used are wimpy and not too involving, and they can't remix it (because I'll tell you that the best solution is to go back and remix), I will apply my sense of what a live rhythm section sounds like to this electronic form. I may use an expander on the bass frequencies, up to a decibel, to try to get the impact back that they lost using cheap samples and maybe overcompressing them.
Speaking of overcompression, let's talk about the loudness wars. I know you feel strongly about how the dynamic range in pop music is being reduced to unacceptable levels in an effort to make the final product sound louder and louder.
In the second edition of my book, I was privileged to get a diagram of ideal dynamic range for different venues [listening environments] from my good friends at TC Electronic. And that, in color — right there on the front cover of the book, there on the left-hand side — you see what their research found [see Fig. 4]. You could put the dance floor, which isn't in the diagram, around the fourth or fifth bar. [By overcompressing the music,] producers are currently making dance and metal and hard rock fall into the eighth bar, where in-flight entertainment is. Now you know what the in-flight entertainment sounds like — you've put the earphones on.
So the green bars in the chart represent average level?
And the orange is peak level, and the blue is the noise floor (or what we would consider the unacceptable low level). And that really tells a story. I can tell you that a good piece of dance music — disco or whatever from the '70s or '80s — will have the dynamic range right around the third bar. Put on Donna Summer or you name it, and it's going to really kick. But notice how these are all normalized on this to an average level of 0 dB. If I were to put on the Donna Summer from the '70s disco in a current disco, she would sound way too low. That's because her average level on the disc is relatively low, and the disc jockey would have to turn up his fader to play it. And when he did, she [Donna Summer] would then have all the impact and “shake-your-body-and-move-it,” and would totally blow away any electronica (except some, where my clients are enlightened) that's being made today. The point is that the increase in CD levels has gone too far.
Is there ever a situation in which compressing a song that heavily actually helps it?
Oh sure, but usually because it's not a good mix. Our expectations have changed as well. Take the Black Sabbath song “War Pigs,” from the '70s. That was considered to be radical [in terms of level], and you put it on today and you just turn it up — it's incredible. That thing really kicks. That was called “metal” back then, and today we call it “hard rock.” Today's metal in general, Limp Bizkit or whatever you want there in that genre, really, really sucks — there's no dynamics left.
So with more-current heavy bands, the peaks are all getting squashed down so that the overall level can be raised.
If you compare it to the Led Zeppelin or the Black Sabbath of 1978, Limp Bizkit sounds — limp. Does that help? This is the first time that any book has ever talked about this in as clear a manner. I think we've crossed new boundaries, because while it may seem like this is pretty evident stuff, I really didn't quantify it until seven years after my first edition. And Thomas Lund of TC Electronic, whose research resulted in that graph [represented in Fig. 4], and I had a simultaneous revelation. We've been in this industry for 20 or 30 years, and it hit both of us at once. So this is really groundbreaking material. I'm sorry for sounding repetitive, but we need people to understand that the ability to peak-normalize is the root cause of why the digital loudness race has resulted in such extreme compression.
But doesn't peak-normalizing simply raise the level of the program material relative to the highest peak? Isn't that basically harmless?
If you took your album that you had for mastering, and at the end of the day the highest peak was -1 dB from below full scale, and [then] you raised the entire album by 1 dB using your peak-normalization algorithm (which is the same as measuring it and saying, “Okay, I've got 1 dB headroom, I can add 1 dB”), that's relatively harmless in the following sense: all of your careful aesthetics between the songs remain the same, and your album gets 1 dB hotter. But if you tried to peak-normalize on an individual song basis, you would screw up the whole relationship of the feeling from song to song. If a ballad was 4 dB below the top, and the hottest song was only 1 dB below, then the ballad would be raised too high, relatively speaking.
Okay, but how does peak normalization contribute to the loudness race?
It encourages people to compress.
Because a song with a compressed dynamic range, and therefore lower peaks, can get much louder than an uncompressed song when peak normalization is applied?
Yes, with extreme average levels that could never have been produced in the days of analog.
Do you find a lot of sonic issues with projects that come out of people's home studios?
Yeah. There's a lot of inexperience out there. A lot of people are saying, “Well, gee, I can save $1,000 by not hiring a mix engineer. I can do it myself.” Or “I worked with mix engineers for a few years. I can mix this one myself.” I'm doing a project now for a jazz singer, and I've worked with the producer on several of his albums before, where he worked with pros. But now he's mixing it himself, and the mixes came in clipped, overloaded, distorted. And I told him, and he looked back and he said, “Oh yeah, my levels are really hot, aren't they?” So the most fundamental things get lost.
Other than being organized with their files, what do you recommend clients do when preparing to bring their material in for mastering?
One of the best things that people can do is to communicate. If you're in the middle of a mix and you know that you're going to be done tomorrow, it's kind of late for you to be consulting the mastering engineer to find out if you did it right. The time to do it is right after your first mix, the one you think you're really happy with. Send it to the mastering engineer that you think you're going to be working with. Let him or her listen to it and [tell you] if you might need stems, if your bass is in good shape, if you're overcompressing, or if you have good dynamics. Many mastering engineers will do this on a consultation basis for no extra charge. At no charge for maybe one song. If a client asks me to listen to their whole album, I'd have to charge them.
You're saying that if they consult in advance, then you won't have to go through as much trying to fix things.
The less we have to do, the better it will probably sound.
When people are mixing, should they go light on the master bus compression, leaving that for the mastering session?
I'm not a fan of it. Instead of saying to be light with it, maybe we can get to that through a different philosophical approach. Let's talk about old-school mixing, and this is for the mixing engineer: if you're not sweating bullets when you're mixing, you're probably overcompressing. A good mix engineer will probably be riding faders a lot in order to get the balances to work.
These days, it may be riding using DAW automation.
That's right. But I'll tell you, the mouse encourages overcompression, because you tend to get lazy the tenth time you have to move that little rubber band up and down in Pro Tools. And you tend to say, “Let's just turn the compressor up a little bit more.” Are you telling yourself to turn that compressor up a little bit more because you don't want to go in there and move that little automation line, or are you doing it because, as an experienced listener, you think that it may need a little bit more compression? So if you're sweating bullets and you're actually using a controller in a DAW — moving faders up and down and doing a lot of passes — then you're probably not overcompressing. But if all your faders are in a straight line and the mix mixes itself, then I think the answer is self-evident. So in answer to your question about being light on the bus compression: it depends on what your musical philosophy is. If you're looking for a particular sound, I'm not going to tell you what to do. But if you're trying to get a great open sound that has impact, you should reserve [using] the bus compressor, if at all, for after you've gotten the best balance possible or you're well along into the mix.
But also, aren't you concerned that if people compress the master bus a lot, it will leave you with less room to work with dynamically?
Right, but I prefer not to express it in those words, because it gives the impression that the mastering engineer wants to have control over things. No, what we're trying to say is if you're looking for a recording that sounds open and impacting, that the bus compressor actually works against that. As things go up in your mix, it takes them down. Isn't that kind of counterintuitive?
Mike Levine is EM's executive editor and senior media producer and the host of the twice-monthly Podcast “EM Cast” (emusician.com/podcasts).
Bob Katz: A Selected Discography
Steve Dyer, Native Art (Sony/BMG, 2007). World music. Afro-influenced rock/pop/jazz.
Marley's Ghost, Spooked (Sage Arts Records, 2006). Folk/rock/reggae/country, with incredible harmonies. Produced by Van Dyke Parks.
Shrift, Lost in a Moment (Six Degrees Records, 2006). A journey through real-world sounds, music, and dreamlike textures.
Afro-Cuban All Stars, Step Forward: The Next Generation (Globe Star Recording, 2005). The best Cuban salsa musicians around. Produced by the Cuban genius Juan de Marcos González.
Midival PunditZ, Midival Times (Six Degrees Records, 2005). Hot electronica, mixed with traditional East Indian vocal.
McGill, Manring, Stevens, Controlled by Radar (Free Electric Sound, 2002). Hot fusion trio.
Remastered and Restored
Willie Colón and Rubén Blades, Siembra (Fania Records, 2006). The biggest-selling salsa album of all time.
Joe Cuba, Bang! Bang! Push, Push, Push (Fania Records, 2006). Classic boogaloo. Great studio sound from 1966.
Recorded Direct to 2-Track and Mastered
Rebecca Pidgeon, The Raven (Chesky Records, 1999). Folk/pop artist, audiophile.
Paquito D'Rivera, Portraits of Cuba (Chesky Records, 1996). Big band, audiophile. Grammy Award winner for best Latin jazz performance.
Digital Domain studios