mastering audio by Bob Katz

[Audio~Geek]

don't mind http://planetoftunes.com either

shield your eyes before going to that one!

[kylen]

Isn't Katz book pretty much totally outdated if your recording digitally? Correct me if I'm wrong, but his K system and talk about headroom doesn't really apply to digital recordings. Don't clip in digital (stay below 0) and your fine.

K-system is for digital and/or analog as it eventually reflects reference SPLdb in a fully 'K' calibrated system as well as relative loudness in dbm while in the box (or digital hardware). Staying below 0dbfs in a rendered digital wave file is also good advise that's true unless you want your converters to add color (I have a few commercial releases where they did this... ).

The Bob Katz book is great and explains all this and more. It's the only audio book I brought out on the road with me besides a dsp book but that doesn't get used much as Kingston and a few others are here to expound.

A chapter a night!

[xander]

Isn't Katz book pretty much totally outdated...

I thought this thread was a little outdated...

...if your recording digitally? Correct me if I'm wrong, but his K system and talk about headroom doesn't really apply to digital recordings. Don't clip in digital (stay below 0) and your fine.

Hmm... I understand what you're saying but I think the larger part of BK's work is relevant and in fact sort of provides a solid basis/foundation for digital recording/mixing/mastering as well, IMHO.

[maxhodges1]

Isn't Katz book pretty much totally outdated if your recording digitally? Correct me if I'm wrong, but his K system and talk about headroom doesn't really apply to digital recordings. Don't clip in digital (stay below 0) and your fine.

The K system and headroom is *every bit* as important with digital systems. In fact, the K system was really designed with digital in mind, seeing as it's always tricky using lower levels on tape because of noise concerns.

The old myths, such as trying to track close to 0dBFS to ensure you get hot levels, or "if I turn stuff down I lose resolution and aren't I degrading my audio?" are outdated, incorrect and bad practice.

Record at 24-bit, track and mix at medium levels with plenty of headroom, and leave the limit-to-0dBFS process until you are mastering your tracks, and your music will sound the better for it.

And even the "don't clip in digital" is incorrect. You should never clip or even approach 0dBFS on your main output, which will distort, either directly or on the reconstruction waveform at your convertors. But most decent DAWs you can go way over 0dB in the mixer without problems as they have *lots* of headroom (typically around 1500dB or headroom in a optimum 32bit floating point mixer.

The one caveat to that is some plugins don't work in 32f, and instead use maybe 24bit fixed, so in thes you do need to watch your input levels, as their fixed point implementation means you can clip the inputs of those and cause unwanted distortion.

my understanding was that clipping distortion adds warmth in the analog world, but clipping is BAD digitally, but if you have compressors / limiters, etc, that keep you from clipping in a digital path, you'd go for the max and get as hot a signal to disc as possible

But I see what you're saying about the difference in track levels and master output. interesting. Didn't know about this... I'll experiment! Thanks

[Compyfox]

Isn't Katz book pretty much totally outdated if your recording digitally?

Actually he're more on the spot than ever. The K-System applies to both the analog and digital world, though with analog you can abuse the oversaturation effect for sounddesign.

The K-System is not only important in terms of productions (radio, CD, etc) but also for movies (especially surround work, cinema stuff). So it's not outdated at all. To understand the problems that accour nowadays, you have to understand the basics from back in the day. That seems a bit outdated. BUt trust me, the K-System isn't.

The problem with the system is however, that barely anybody adapts to it, because the clients and "so called engineers" (with setup minds) say "you have to make it louder to push through the crap that's out there" - which isn't true. The K-System in short is a guideline to percieve dynamics, but have a certain loudness level. Since I adapted the system, my tracks are obviously more quiet, indeed. Then again you can use up "certain limits" and still stick to the rules of the K-System to get a full dynamic, but still hot sound.

And this is what it's all about and why I adapted it! If I want a half orchestra half pop production, I don't want it squashed to bits, I want it loud, but fully dynamic. Try that with the "compress to max to get heard" rule.

im about half way through Mastering Audio by Bob Katz and i gotta tell ya... its a good read. lots of history and fundamentally important stuff to help you make your mixes better.

If you understood the book (especially the part how to mix in the K-System), adapt it and get better results, you're one step closer to a better world in terms of "sound". And we "engineers" got one recruit more on our side to spread infos, not to mention the "goods and the bads" about this whole loudness race.


Now if you'll excuse me, I wanna go hug my Redbook Audio CDs from the mid 90ies now.

[beej]

my understanding was that clipping distortion adds warmth in the analog world, but clipping is BAD digitally, but if you have compressors / limiters, etc, that keep you from clipping in a digital path, you'd go for the max and get as hot a signal to disc as possible

You just need to understand where in the digital path it is possible to clip, and there are actually surprisingly few places where you *can* clip in a well designed DAW.

You can clip your preamp/audio interface by setting too high an input gain. Once clipped your audio is permenently clipped in your recording and there's nothing you can do about it.

You *cannot* clip individual tracks, as when recording, the maximum digital signal level coming back from your audio interface that will be recorded into an audio file is 0dBS - ie, you are clipping the audio interface with too hot a level. The audio interface cannot send a sample value above 0dBFS to the DAW, therefore the DAW cannot clip, although many of them will report a warning if they detected levels hitting 0dBFS (either once, or more likely, a few 0dBFS sample values in a row) to alert you that your levels are too high.

You *cannot* clip individual tracks on playback, as the DAW will have sufficient headroom to raise the gain on individual tracks *way* past the 0dBFS point (think about it - if this wasn't possible, you wouldn't be able to mix 32 tracks together, each playing back at close to 0dBFS, as the combined mix (signals *add* together) would be way way way clipped and basically the whole system would be unusable.)

You *cannot* clip busses, auxes and other internal DAW routing channels, for the same reasons - plenty of headroom.

You *can* clip when using plugins on channels that have a fixed point implementation, and it's not always obvious or documented in particular plugin manuals whether this is the case. In this case, because you're moving from a high-headroom floating point implementation to a fixed-point implementation, we are back to having a single 0dBFS absolute maximum level, and therefore it's possible to clip it.

You *can* clip output channels, as once again, we are going from a floating point implementation back to a fixed point signal for finally outputing back through your audio interface or for printing into a file.

Basically, anywhere you are going from floats to fixed point you have the possibility of clipping to watch out for.

You *can* also clip output channels *even if* your signal *does not reach 0dBFS*, say, you've chucked on a limiter and set it to limit and normalise to -0.3dBFS.

This is because even though the individual sample values are less than 0dBFS, the reconstructed waveform made from those sample values *can* go over the 0dBFS point - while you won't generally hear this as distortion, it isn't desirable and may well be the cause why some people feel their mixes start to sound muddy or dead or cluttered when approaching the 0dBFS point, which doesn't happen when mixing with plenty of headroom. It will also cause your mixes to be rejected from any reputable mastering house if you sent your stuff out to be mastered.

Lower levels and headroom is a *good thing* in digital. You can track lower, meaning you don't have to worry about clipping your convertors or recordings, your resultant mixes will be easier to control, you won't run into distortion issues when using lots of plugins, and you can deliver natural sounding mixes, which can be limited to death to near 0dBFS later if required to "compete" with all the other bad sounding stuff out there.

And with low levels, you never need to worry about clipping *anything* - unless you want to do it creatively, of course...

[xander]

Man -- nice post beej!

[druid]

Thank you beej, I didn't know a couple of points there either.

You *can* also clip output channels *even if* your signal *does not reach 0dBFS*, say, you've chucked on a limiter and set it to limit and normalise to -0.3dBFS.

This is because even though the individual sample values are less than 0dBFS, the reconstructed waveform made from those sample values *can* go over the 0dBFS point - while you won't generally hear this as distortion, it isn't desirable and may well be the cause why some people feel their mixes start to sound muddy or dead or cluttered when approaching the 0dBFS point, which doesn't happen when mixing with plenty of headroom. It will also cause your mixes to be rejected from any reputable mastering house if you sent your stuff out to be mastered.

How much does the reconstructed waveform often go over the individual samples, or rather ... is there a way to calculate it or find out if it does? Let's say, in Audition 2.0, I can open the statistics window and it gives me the peak amplitude.. but that's the highest individual sample, right? So what am I looking for to find that out?

[SARcazm]

The Womb, with Slipperman, Mixerman and a bunch of other pros always has good mixing stuff in it. The distorted guitars from hell thread has great tips on comp & EQ (tho is a very long read!)
http://womb.mixerman.net/forumdisplay.p ... 77f76f&f=1

[afreshcupofjoe]

Here's a quote which jumped out and hit me in the eye :
"Toslink/fiber optic cable doesn't sound as good as AES/EBU or BNC"

yeah it's really insane stuff, hard to believe. i dont know if any human ear can hear these differences. Vestman is clearly superhuman he might've been crossbreeded with a dog to get such sensitive hearing.

Your statement implies that there are actual differences, which there are not.

Vestman is a complete kook. His statements are the worst kind of deception because he starts out talking about clock jitter, which is a real problem, and them somehow uses it as a leading statement to make many other claims which have no basis in science or reality at all. They would seem legitimate to someone who has little understanding of DSP, but I don't see how it's possible that someone who has some knowledge of digital technology could mislead himself and others so blatantly. Computers aren't some mystery magical boxes that we only have a limited understanding. The IS NO CONTROVERSY about how computers and digital technolgy work. We build them, and we understand the technology thoroughly. I just can't stand this shit. It's the same argument as digital summing differences. To claim that an issue like this is controverial is like claiming that a little green goblin lives inside your computer and spins cd rom drive is something up for debate.

I guess if nearly half(?) of americans can believe that a man woke from the dead, did a little dance, flew up to a paradise in the sky, and is coming back to earth to rescue all of the believers from being poked in the ass by a flaming ptchfork for all of eternity, then goblins living inside of their computers isn't a stretch at all.

[beej]

How much does the reconstructed waveform often go over the individual samples, or rather ... is there a way to calculate it or find out if it does?

There is lots of good info in the Sony Oxford limiter manual about this if you really want the hardcore details and the maths and stuff. Yoiu can download it from the Sony plugins site.

Paul Frindle, the creator of the Oxford plugins and an acknowledged expert in digital audio explains this much better than I can, and posts on ProSoundWeb - there have been plenty of threads on this stuff, and I got into it heavily on the Apple Logic forums recently.

Basically, most software meters do not have the facilities to display the reconstructed wave levels, you are correct in that that they mostly all show individual sample levels. The Sony Oxford limitor is one that does have accurate reconstruction levels metering.

It doesn't mean the final wave *will* clip the output convertors, but it's a possibility and it really depends on the waveform. Apparently, it's possible to clip the output convertors with the reconstruction wave with sample values as low as 70% of the full scale.

The point being, that without knowing about this stuff, it's yet one more reason why it makes sense to mix to lower levels - the K system makes sense for these kinds of reasons (and some other reasons to).

[The Chase]

I suggest checking out John Vestman's site so you can pick up great advice like mastering to a firewire drive because it sounds warmer than an internal drive.

Well Bob Katz said that the digital audio coming from a CD is different than that of a streaming hard drive...Just shows that people talented in a certain feild can sometimes be a bit clueless in the technical aspects of such a feild.

[IIRs]

Well Bob Katz said that the digital audio coming from a CD is different than that of a streaming hard drive...Just shows that people talented in a certain feild can sometimes be a bit clueless in the technical aspects of such a feild.

Where does he say that?

[tetsuneko]

I really recommend
John Watkinson's book
The Art Of Digital Audio (3rd ed.)

For all you 0dBfs digi-audio-phreax out there

[afreshcupofjoe]

I suggest checking out John Vestman's site so you can pick up great advice like mastering to a firewire drive because it sounds warmer than an internal drive.

Well Bob Katz said that the digital audio coming from a CD is different than that of a streaming hard drive...Just shows that people talented in a certain feild can sometimes be a bit clueless in the technical aspects of such a feild.

Actually, Bob Katz may be correct here. If he is speaking in terms of a commercial cd player vs. hard drive streaming, then yes. Even completely ignoring the DAC side of things, an audio cd player will probably have far inferior error correction compared to a hard drive in a computer. The important thing to know is that if you transfer the data from the cd to the hard drive, the data will be exactly the same as it is on the cd or it would be considered an error. Devices make mistakes when reading/writing data, but there are protections in place to catch these mistakes so that the data will remain exactly the same when transfered from one media to another.