How do you do this, Kim?

[Uchdryd]

Kim Lajoie's song, Horse Head (Show me your human) (from here) is so incredibly loud that I use it as a benchmark for my music to make sure it's at least in the same ballpark with regard to volume. But I can't seem to even get close!

When I load this song into an audio editor I notice some (to me, at least) shocking things:

1) There is practically no headroom at all--80% of the song is right up against 0 dB--and yet there's no audible clipping!

2) The RMS power is an unbelievable -12 and the DC offset is an equally unbelievable 0%.

Granted, the guy has equipment and expertise that I can only dream of having, but I would love to get some idea as to how this is even possible. I use compressors/limiters, manually remove transients, attempt to auto-correct DC offset problems (which are significant in my case), and make other feeble attempts to increase loudness, but can't seem to even get close. Am I missing something?


Uchdryd

[frequency_algorithm]

First off all you can use a fairly good limiter or look ahead compressor with a high 20/1 ratio to do this, but of course good ones will sound a lot better. If you also mix your music well it will be easier to boost it. Lastly, you can drive analog compressors/limiters much harder that you can with virtual/plug in compressors. The song appears loud because the frequencies that are boosted are the mid highs/highs (a lot of times called "presence") which are very eminent in the vocals and the piano/snare. The mid high/high range of the spectrum is what our human ears are most sensitive to and so it always seems even louder when boosted in this area. Louder isn't always better, cause i bet if I blasted this in my 1200 car audio system my tweeters would blast your ears out and you would be pissed. Everything is relative.

[afreshcupofjoe]

No doubt Kim is a great producer, but that song is way overcompressed even for my tastes. I wouldn't use that as a target, more like an example of loudness gone too far. No offense Kim, you are a much better producer than I am, but that shit hurts my ears.

Regardless Uchdryd, you're not going to get that kind of clarity and loudness without a great tonal balance and some very smart eq tweeks. It's all about giving everything its own space. If you try to do any heavy compression on a track where everything is fighting for space in the mix then killing the transients will suck the life out of everyhting and just leave a muddy distorted mess. But if you give everything it's own room to breathe, then you can squash the hell out of the song and the drums will still sound punchy, the bass will keep pumping, and the vocals will still cut, without it sounding muddy and distorted. There is lots of frequency distortion going on in Kim's mix as a result of all of the compression, but you just don't notice it because he did a kick ass job of mixing it.

I'm probably totally wrong, and Kim can give you some real advice, but the question seemed very appropriate to what I've been experiencing lately, so I thought I'd comment. Just my 2 cents, YMMV.

[frequency_algorithm]

You are prolly going to end up using a software virtual compressor so just put that over your mix and bring the threshold knob/slider down until you see the reduction meter peak at about 3 db. Use the auto release button if your compressor/limiter has one. Make sure your final gain is at 0.2 db. There, now you are a pro.

[bduffy]

Yeah, and just get the average RMS of your mix; if the average is -20dB, then you need to apply 10dB of gain reduction on your limiter to get it up to -10dB RMS, which is generally where today's nasty, over-compressed masters hit. Actually, they're usually louder than this now.

[Kim Lajoie]

Please don't let this descend into sarcasm and infighting.

For purists to say "Loud is as easy as clipping the master limiter" is just as useless as over-producers to say "Purism is as easy as plugging in your most expensive mic and hitting record".

Let's talk about techniques and artistic judgements.

-Kim.

[Kim Lajoie]

Kim Lajoie's song, Horse Head (Show me your human) (from here) is so incredibly loud that I use it as a benchmark for my music to make sure it's at least in the same ballpark with regard to volume. But I can't seem to even get close!

What music are you working on? It only makes sense to reference against tracks that are similar to your own. I mention this because unless your song is very similar to Horse Head, you won't get close no matter how hard you try (and you'll butcher your sound in the process).

Having said that, is it really a good idea to be using my work as a benchmark or reference? Horse Head is by no means an example of professional mastering. It really depends on what your music is, and what your goals are. I'm not yet convinced that I have enough influence in the music industry to make "Kim Sound" a winning formula.

When I load this song into an audio editor I notice some (to me, at least) shocking things:

1) There is practically no headroom at all--80% of the song is right up against 0 dB--and yet there's no audible clipping!

2) The RMS power is an unbelievable -12 and the DC offset is an equally unbelievable 0%.

As afreshcupofjoe wrote, the key is getting it right in the mix. There are two things to keep in the back of your mind when mixing for "loud" - Peak/Average ratio and saturation.

By getting the Peak/Average ratio of each instrument right at the mix, you won't need to apply excessive compression or limiting at the mastering stage. Horse Head is actually a good example because I screwed it up. I didn't treat the kick very well in the mix, so it ended up having a peak far in excess of its impact. In other words, the actual peak level of the track was too high, given its audible effect on the mix. This meant that to get the kick at the right level, the peaks were actually "poking out" of the mix.

When it came time to master, those peaks on every kick made processing very difficult - I almost went back to the mix to fix it (in retrospect, I probably should have). Because the peaks were "poking out", I couldn't bring the average level up of whole master without squashing those peaks. However, squashing them with a compressor had a ducking effect on the rest of the mix. I had to apply a careful combination of compression, limiting, and saturation to control those kicks. If you listen to the final master, you can still hear the kick ducking the rest of the mix (and I'm still not happy with it).

Saturation is another trick to increase the perceived loudness of a track. Unlike compression or limiting, which makes loud peaks quieter, saturation makes loud peaks noisier (and only quieter if you overdo it). Unfortunately, saturation is often very audible on a master, due to intermodulation distortion (where instruments will affect the sound of other instruments, in a largely uncontrollable and ugly manner). Saturating a mixdown with a high peak/average ratio (very dynamic and open) will make the mix sound clipped and "broken", wheras saturation a mixdown with a low peak/average ratio (squashed) will make the mix sound fuzzy and less defined. It will also make it louder.

To counteract the effects of master saturation (and intermodulation distortion), I apply saturation to almost every track during the mixdown. This was I can control the type and amount of saturation on a per-sound basis, and apply it with much more control. Cumulatively, this has a similar effect to master saturation in reducing the peak/average ratio, but allows me to do it in a way that's most appropriate to the song. For example, a snare drum can take a lot more saturation than the lead vocal (which I usually do not saturate at all). This is a stylistic choice I make, it might not be appropriate for your style of music. I happen to like the sound. When I get to the master, I don't need to apply as much saturation as I otherwise might, and the effects of the master saturation are masked by the track saturation - effectively I can push it more.

Granted, the guy has equipment and expertise that I can only dream of having, but I would love to get some idea as to how this is even possible. I use compressors/limiters, manually remove transients, attempt to auto-correct DC offset problems (which are significant in my case), and make other feeble attempts to increase loudness, but can't seem to even get close. Am I missing something?

You have all the tools you need. You're missing The Knowledge To Make It All Work. In increasing order of importance:

3) Read everything you can. Not just about mastering and mixing, but about audio in general. Physics, acoustics, electronics, DSP, everything. Not everything you read will be useful. Not everything you read will be true. Learn to tell the difference.

2) Learn to listen. A/B your work with tracks that sound great. A/B your work with tracks that sound terrible. Listen to music everywhere and anywhere you can. Get to know your speakers. Learn to tell the difference between enjoyment listening (for fun), critical listening (for composition and mixing), and scientific listening (for mixing and mastering). Learn to be able to switch between them at will. Learn to be able to listen in different modes simultaneously. Learn to hear the change before you make it.

1) Practice. Review your work after you practice to make sure you learn from it and improve for next time. Learn your tools so you can use them without thinking about them. Read the manual, then learn the tools well enough that you know more than the manual can teach. Ruthlessly rid yourself of tools you don't need. Compose more so you can have more mastering practice. Rearrange your life so you can spend more time on music. Stop watching TV. Post your work here on KVR for critical analysis. Try out people's suggestions, decide what will and won't work. Develop your own set of techniques. Develop your own style.

Hope that helps.

-Kim.

[Kim Lajoie]

By the way, I don't have "equipment and expertise that [you] can only dream of having". There's no esoteric hardware, no analogue compressors. I used two plugins - Electri-Q (posihfopit edition) and T-RackS. There are many other tools that do similar work. Some are free.

-Kim.

[Uchdryd]

Wow--I'm glad I asked! Thanks for explaining how you did it, Kim, and thanks to everyone else for your advice and opinions.

While everyone has his/her own opinion on the issue of audio dynamics, the simple fact that this song can be made that loud and still sound as good as it does is a testament to Kim's skill and craftsmanship. I personally think that the production techniques used in that particular song work well for it. It has a very sharp, abrasive quality that suits that style of music. Clearly it wouldn't be wise to do that kind of thing with all genres of music.

I wasn't suggesting that my goal is to make my music that loud, or that it should be, but I did want to have some idea about how to do it for the sake of understanding the principles of music production. And now I do.

The DC offset is what's causing most of my problems, I've determined. The minimum peaks are far, far greater that the maximums on the song I'm working on ATM, which I guess is due to poor-quality recording hardware.


Uchdryd

[afreshcupofjoe]

Kim, did you say (in another thread) that was your sister singing on that track? Damn, she has a great voice! What's up with her? Is she pursuing a music career, does she have an album out, and is she still writing music, recording? I'm not trying to hit on her or anything like that; I'm just really impressed by her vocals. I would gladly buy her album if she's got one. It would be a shame if she was wasting such talent. Is there any more collaborations between you two in the works?

[Kim Lajoie]

Uchdryd, the quickest way to kill a DC offset is a highpass filter (DC offset is equivalent to signal at zero hertz). I usually highpass every track in the mix except for kick and bass. If you've got DC offset problems, it might be feedback-based effects (such as flangers or filters), or (more likely) there's some rumble in your recordings that you're having trouble hearing. Low frequency rumble, like DC offset, will fight you all the way up the loudness hill. If you're going for a pop/produced/loud sound, you've got to give your highpass filters a workout. Again though, this won't be appropriate for some styles of music - you have to make an artistic judgement.

There are a few songs on Tori Amos' Boys For Pele where there's some beautiful noise from the piano's sustain pedal. It's not rhythmic foundation, but the song would be poorer had it been highpassed out.

afreshcupofjoe, that is indeed my sister singing on Horse Head (and Undone, and Sunlight). I think she has a great voice and performance style too. Unfortunately for her fans, she's not persuing a music career, and she doesn't have an album out. There are only these three recent songs, and a few we worked on in the distant past (which I'm not publishing, because they do us both a disservice). We'd both like to work on more songs, but we've both been busy with other work.

-Kim.

[Uchdryd]

By getting the Peak/Average ratio of each instrument right at the mix, you won't need to apply excessive compression or limiting at the mastering stage. Horse Head is actually a good example because I screwed it up. I didn't treat the kick very well in the mix, so it ended up having a peak far in excess of its impact. In other words, the actual peak level of the track was too high, given its audible effect on the mix. This meant that to get the kick at the right level, the peaks were actually "poking out" of the mix.

I think I can answer my own question here, but how exactly is it that the audible effect can exceed what is reflected graphically in an audio editor? Is it that the frequencies are too low, or too broad, and so your ear doesn't register what your equalizer does?

When it came time to master, those peaks on every kick made processing very difficult - I almost went back to the mix to fix it (in retrospect, I probably should have). Because the peaks were "poking out", I couldn't bring the average level up of whole master without squashing those peaks. However, squashing them with a compressor had a ducking effect on the rest of the mix. I had to apply a careful combination of compression, limiting, and saturation to control those kicks. If you listen to the final master, you can still hear the kick ducking the rest of the mix (and I'm still not happy with it).

Yes, I've encountered this problem, and when I do I know I have to go back a step or two and fix the problem at the individual track level so as to avoid that "ducking" problem.

Saturation is another trick to increase the perceived loudness of a track. Unlike compression or limiting, which makes loud peaks quieter, saturation makes loud peaks noisier (and only quieter if you overdo it). Unfortunately, saturation is often very audible on a master, due to intermodulation distortion (where instruments will affect the sound of other instruments, in a largely uncontrollable and ugly manner). Saturating a mixdown with a high peak/average ratio (very dynamic and open) will make the mix sound clipped and "broken", wheras saturation a mixdown with a low peak/average ratio (squashed) will make the mix sound fuzzy and less defined. It will also make it louder.

I plead ignorance here. How does one add or remove saturation? And why would I want to make my song sound more (as you say) "clipped and broken" or "fuzzy and less defined?"

You didn't mention anything about DC offset, or, more importantly, the problem of having minumum peaks that far exceed maximum peaks (or vice versa), which of course can occur on any track and on either the left or right channel, so it's a common problem (at least for me).


Uchdryd

[Kim Lajoie]

By getting the Peak/Average ratio of each instrument right at the mix, you won't need to apply excessive compression or limiting at the mastering stage. Horse Head is actually a good example because I screwed it up. I didn't treat the kick very well in the mix, so it ended up having a peak far in excess of its impact. In other words, the actual peak level of the track was too high, given its audible effect on the mix. This meant that to get the kick at the right level, the peaks were actually "poking out" of the mix.

I think I can answer my own question here, but how exactly is it that the audible effect can exceed what is reflected graphically in an audio editor? Is it that the frequencies are too low, or too broad, and so your ear doesn't register what your equalizer does?

First of all, I'm not sure I made myself clear - I was trying to say that the kick had too much "physical" level, and not enough "audible" level.

You have to understand that the digital waveform display and the level meters are not what you hear. They are a representation of the "physical" aspect of the sound - what is recorded on disk, what the plugins will process, how the speakers move to vibrate the air. It is not the same as what we hear - what we percieve. This is psychoacoustics. For example, a 50Hz tone and a 1kHz tone at the same "physical" level will not be heard at the same level (the 1kHz tone will be louder). Likewise, a 50ms noise burst and a 500ms noise burst at the same "physical" level will not be heard at the same level (the 500ms noise burst will be louder). You have to be able to translate between the two worlds of "physical" sound and percieved sound.

In the case of the Horse Head kick, there was a lot of "physical" level, whilst not being (audibly) loud. One thing I could have done in the mix was saturate the kick. This would have allowed me to greatly reduce the "physical" level of the attack (the first 50-100ms) of the sound while keeping the same audible impact. Driven further, it would have created some upper harmonics - increasing the audible loudness while keeping the same "physical" level.

Saturation is another trick to increase the perceived loudness of a track. Unlike compression or limiting, which makes loud peaks quieter, saturation makes loud peaks noisier (and only quieter if you overdo it). Unfortunately, saturation is often very audible on a master, due to intermodulation distortion (where instruments will affect the sound of other instruments, in a largely uncontrollable and ugly manner). Saturating a mixdown with a high peak/average ratio (very dynamic and open) will make the mix sound clipped and "broken", wheras saturation a mixdown with a low peak/average ratio (squashed) will make the mix sound fuzzy and less defined. It will also make it louder.

I plead ignorance here. How does one add or remove saturation? And why would I want to make my song sound more (as you say) "clipped and broken" or "fuzzy and less defined?"

Sometimes clipped and broken or fuzzy and less defined is a desirable part of a sound's character. That's an artistic judgement you need to make.

Other times, it's a (controllable) side effect that you allow in order to gain the other desirable effects - such as lower peak/average ratio and increased percieved loudness.

I find that track saturation can make a mix sound more "real". This is probably because a lot of music we grew up listening to had been through many stages of analogue processing, which can result in a more distant or "covered" sound, because we're hearing the original recording "through" layers of electronics. It can also add subtle complexity to an otherwise simple sound, as the analogue electronics were not perfectly linear or deterministic. This is in stark contrast to unprocessed digital recordings, which I find sound "naked" and bland. Again - one is not always better than another, but it's useful to know and understand it to make meaningful and informed artistic judgements.

And (for the purposes of this discussion) you can't remove saturation that's already been "baked in" to a sound.

You didn't mention anything about DC offset, or, more importantly, the problem of having minumum peaks that far exceed maximum peaks (or vice versa), which of course can occur on any track and on either the left or right channel, so it's a common problem (at least for me).

I did mention DC offsets in my last post.

Personally, I don't think about DC offsets much at all, and I don't ever think about asymmetric peaks. They're just not concerns of mine. DC offsets are easily dealt with by highpass filters (which I use anyway), and asymmetric peaks are either a result of a DC offset or a feedback effects like a flanger or filter. I've occasionally had a DC offset caused by a bitcrusher plugin, but it's rare and easy to fix.

Keep in mind that DC offsets and asymmetric are not super-evil. They're only a problem if you're going for a clean pop/produced/loud sound. Even then, they're probably going to be taken care of by fixing any bigger problems you've already got (like too much subbass).

-Kim.

[Uchdryd]

I feel like I've just been through Music Production 101. Or maybe 201. Now I'm eager to get back to making music again, armed with this knowledge.

But first, two clarifications, to make sure I understand you correctly, Kim:

1) A filter "squashes" a sound signal horizontally by simply eliminating a range of frequencies, whereas a compressor or limiter "squashes" a sound signal vertically by attenuating the entire frequency spectrum of the sound equally (when it reaches the threshhold), correct?

2) And by using a filter, you can reduce the physical level of a sound while not noticeably reducing the audible level of that sound, right?

3) Saturation is a purely analog device that's similar in effect to a limiter except that it introduces distortion at peak levels instead of attenuating the sound signal at those peaks, correct?


Uchdryd

[spuddle]

Watch:

Laugh.

Continue.