The Song Fight Bulletin Board

Here's a confession: I like loud. But, as everyone knows by now, <a href="http://www.youtube.com/watch?v=3Gmex_4hreQ">loud can sound pretty damn terrible</a>. When mastering my own songs the penultimate stage (dithering comes last) is always volume maximization, which I achieve through a combination of Brickwall Limiting and Clipping. Just a quick refresher: Brickwall Limiting is like a compressor with an infinite ratio, and when mastering rock songs it increases the song's volume by softening the loudest elements of the mix (typically the kick and snare hits); Clipping is like your headphones breaking, and it increases the song's volume by distorting the loudest parts of the mix (again, typically the kick and snare hits). In breakfast terms, aggressive limiting makes the song sound "mushy", while aggressive clipping makes the song sound "crunchy". Notice that both processes make the song sound <em>worse</em>, NOT better, so I try to minimize the damage I do when applying either.

Recently I've been noticing that limiting is sounding worse and worse to me, while clipping is sounding more transparent. I've been finding that a well-mixed song can handle being clipped a lot more gracefully than it can handle being limited. So, I did this little experiment with one of my Nur Ein songs, <a href="http://www.forkbomb.ca/songs/demos/puce_bp.mp3">Brownie Points</a>. In the picture below, the red waveform in the background is the un-maximized version of the song: notice that it has a nice, natural looking ebb-and-flow to it, with reasonable dynamics and obvious drum peaks. I took this mix and applied about 4db of peak reduction, producing two "maximized" versions using limiting for the first one and clipping for the second (in the image the "maximized" track has been volume-matched and superimposed in green over the original). Notice that all the drum peaks have been truncated: from 1:38 onward the song looks literally square-wave.

<center><a href='http://www.forkbomb.ca/wordpress/wp-con ... veform.gif' title='Brownie Points Clipped (but not from the Betty Crocker catalogue)'><img src='http://www.forkbomb.ca/wordpress/wp-con ... veform.gif' width="483" alt='Brownie Points Clipped (but not from the Betty Crocker catalogue)'></a></center>

Here are the resulting tracks (at a high bitrate). If you want to compare them you'd probably be best off skipping to the end of the song where things start to go a little crazy.
<blockquote><li><a href='http://www.forkbomb.ca/wordpress/wp-con ... cessed.mp3' title='Brownie Points Unmastered'>Original, Untreated Mix</a></li><li><a href='http://www.forkbomb.ca/wordpress/wp-con ... atched.mp3' title='Brownie Points Clipped'>With ~4db of Clipping</a></li><li><a href='http://www.forkbomb.ca/wordpress/wp-con ... atched.mp3' title='Brownie Points Limited'>With ~4db of Limiting</a></li></blockquote>
I then did several <a href="http://en.wikipedia.org/wiki/ABX_test">blind ABX tests</a>, and was surprised to find that I could hardly tell the difference between the clipped version and the original. The main difference I noticed was that the kick drum sounded a little more constrained in the clipped version. However, I had no problem distinguishing the limited version, as the softening of the kick and snare during the song's loud sections makes it sound noticeably wimpier than the original. Although I won't post them here, I also did a few mixdowns with a blend of limiting <strong>and</strong> clipping, but found that nothing sounded as transparent as just straight clipping. Here is the mastered track without the volume matching (WARNING - this is a lot louder than the other tracks): <a href='http://www.forkbomb.ca/wordpress/wp-con ... lipped.mp3' title='Brownie Points Mastered'>Brownie Points Loudified</a>.

So, what have I learned from all this? From now on when I'm mastering my own material I'll avoid using a limiter if I don't have to.


Taken from my blog.

Footnote: To apply clipping on these tracks all I did was turn up my master fader in Cubase, and then do a mixdown (Cubase applies hard clipping to prevent overs). You could also use GClip on its default settings; they both use the same algorithm.

I for one do appreciate your lessons. I don't always apply them to my mixing. I find that different styles and different mic configurations lead to different applications. However, reading all this helps my production all 'round. It gives me a better handle on what I should do over all.

Don't stop. Ever.

I don't want to turn this into a new thread about the pros/cons of this type of processing of a song. I do want to say that I like your songs but I have a hard time listening to several of them in a row because of this. I know it's rock music, and that's the way rock is produced these days, but personally I don't like how it sounds. I'd love to have the pre-squashing versions of your tunes, for easier bulk listening.

The clipped version has more balls, so I like that. The other two sounded fine, but too soft around the edges.
Nice tech writeup and examples, Puce. This stuff is gold to the entry level studio guys like myself, and not so entry level guys too. I know it when I hear it, but it's nice to put a term or a visual to the process.

I've been playing with these techniques myself of late. I don't want my final song to be out of control with peaks and clipping, but at the same time, I don't want it mushy and pussafied.
My latest song is "So Weird" for this weeks title. I've been experimenting with new styles to my old way of thinking.
In my bridge, I have a progression from mild cross stick snare that builds to aggressive crashes. The guitar builds with it and on the finale, if you will, it gets very aggressive and distorted to convey the emotions of the subject matter.
I intensionally hard clipped the second guitar to create the "over the top" sound, but in the mix down, I EQ'd and compressed with no clipping. Then in the mastering I maximized the levels pretty high. In hindsight, the max level I went with was my only iffy choice after listening a few times. My breaks and punch ins weakened a bit, but that might be a good thing for the average listener.
Here's a sneak peak to So Weird. You can just skip to the bridge to hear what I'm referring to. Feedback of the "production" would be great. I'm afraid this falls into the "not good for bulk listening" category that Lunk mentioned.

Lunkhead wrote:I do want to say that I like your songs but I have a hard time listening to several of them in a row because of this. I know it's rock music, and that's the way rock is produced these days, but personally I don't like how it sounds.

I noticed the same thing, which is what triggered this experiment. I've always limited songs when mastering because I figured, hey, that's what you're "supposed" to do, but re-listening to old songs (or even recent ones) I've found them quite fatiguing. I'm going to try to adopt the Hippocratic oath when it comes to mastering my own stuff. That said, I'd be interested to know if the clipped (and level-matched) version sounds worse to you than the untouched mixdown, or even different at all. In a blind test I couldn't consistently identify which one I preferred.

Billy's Little Trip wrote:Here's a sneak peak to So Weird. You can just skip to the bridge to hear what I'm referring to. Feedback of the "production" would be great. I'm afraid this falls into the "not good for bulk listening" category that Lunk mentioned.

Cool song Billy. Drums and bass sound great, and that snare is nice and snappy. I would prefer the song with more pronounced changes in volume, but that's probably more of an arrangement preference than a mastering one.

For what little it's worth, these days I always have a brick-wall limiter as the penultimate effect on my final mixes. This may seem trivial and it's probably not something Puce wouldn't have already tried, but a limiter is not just an on-off thing. You can back off on your master volume (or the limiter's pre-gain; doesn't matter, just lower the input to the limiter) quite a bit-- it's a way to get a good deal of sensitivity without losing much of the percieved loudness that you want.

Think of it as the difference between lightly tapping on a wall, and banging your head against it. You end up in the same place either way. And while the softer parts of your song don't get quite the boost, it still seems to make a big difference.

Another So weird preview! This track has the limiter's input padded by 2.5dB, which isn't much, but it seemed like enough at the time. Double that isn't uncommon for me. It's not as loud as you'd expect a Puce mix to be, but it's significantly louder (and clearer overall) than it would be limitless. EQ plays a large part in that, too, but that's another story entirely.

I don't have much practical experience with clipping, though it sounds like what you want from that is more of a tape-style saturation. There ought to be some magic ratio in which you could combine the two for maximum effect.

It's hard for me to comment objectively about ear fatigue, because people are generally tired of listening to Charcoal long before the effects of overcompression would set in. It doesn't bother me to loop whatever track I'm working on for a couple of hours at a time, so maybe that's a positive.


Finally, for those who know what "penultimate" means, my last processor is a very subtle reverb. Half the time I can't tell if it's there or not but I like to think it also helps to smooth things out a little, so the limiting doesn't feel so... limiting. It's probably just psychological, though.

Puce, I'm not sure what you mean by pronounce changes in volume. But I think you mean playing parts softer or quieter in, say, the first verse, then get louder in the second verse? Or could it be the way I pushed the volume in the mastering which seems to even the highs and lows out.

I guess I should have posted the original mix down and the one with the maximized volume, so here they are side by side. Keep in mind, both have the same EQing and compression. I'm not "exactly" sure what changes take place after converting to MP3, but both are at 192 bps.

The original mix down

The volume maximized to 2.0

By the way Charcoal, AWESOME! I'd love to use that in one of my surf videos someday. Someday meaning, when I actually quit being lazy and edit my vids into something watchable.

Billy's Little Trip wrote:Puce, I'm not sure what you mean by pronounce changes in volume. But I think you mean playing parts softer or quieter in, say, the first verse, then get louder in the second verse? Or could it be the way I pushed the volume in the mastering which seems to even the highs and lows out.

I mean that the chorus is rockin', so it would be nice if it got a bit louder than the verse, to help increase excitement and energy and add variety to the song. But the bass, drums, and guitars all stay pretty much the same volume (give or take ~1 db) throughout the song's nearly 5 minute duration. The bridge is a nice change, but--like always--I want more.

Puce wrote:

Billy's Little Trip wrote:Puce, I'm not sure what you mean by pronounce changes in volume. But I think you mean playing parts softer or quieter in, say, the first verse, then get louder in the second verse? Or could it be the way I pushed the volume in the mastering which seems to even the highs and lows out.

I mean that the chorus is rockin', so it would be nice if it got a bit louder than the verse, to help increase excitement and energy and add variety to the song. But the bass, drums, and guitars all stay pretty much the same volume (give or take ~1 db) throughout the song's nearly 5 minute duration. The bridge is a nice change, but--like always--I want more.

I agree with all said here. Time limits are a bizatch, as you well know.
But yes, I felt the chorus was too subtle too. I also had some cool additions for the bridge and a few other places, but didn't get them done in time.

BLT-- glad you enjoyed it, though I think you'll be severely disappointed by the version with lyrics that I actually submitted. My off-key yodeling is enough to put anyone off breakfast.

Having thought about this some more, I've decided that EQ is part of this same story, after all. For starters, this is a Har-Bal graph of Puce's unprocessed mix. It is basically a graph of how much energy there is across the audible frequency spectrum. The lighter yellow line is a standard (?) reference that comes with the program and is supposed to represent some ideal of "rock" music, but I don't know offhand what the source is. It was a little surprising to me that the three versions of his song all looked more or less identical, but you have to remember that Har-Bal is a holistic look at your song and transient things like "slightly duller kick drum" probably wouldn't register much. At any rate, neither clipping nor limiting are having any major effect on the spectrum of the song. (I expected the clipping version to have more upper-mid to high frequency energy; say, 3-8k. Not enough that you'd necessarily hear the difference, but enough that you'd see it here. I'm often wrong about things.)

Anyway, the graph looks like you'd expect a slick, well-produced rock track to look. It seems like a slight rolloff at the high end might help. You've got a bit much in the 8-12k range, and while the 5-8k matches up better with the reference standard, that is also where a lot of harshness (and loudness! sadly) comes from.

For comparison, this is my "So weird" track given above. There's clearly too much going on at 200Hz and below; this may add some warmth but it's mostly just muddiness. It's also a big part of why you can't hear a lot of the bass guitar fundamental notes (they get as low as the 41Hz E in spots) and why the kick is not very snappy.

The gap between reference and song jumps dramatically somewhere between 4 and 5K and I think this is significant. The song is not as shimmery or airy because of that, the cymbals aren't as gloriously cymbalic as they could be, and it doesn't sound as loud, but it isn't annoying (for me, anyway) to listen to on repeat for an hour or two at a time.

So, all that crazy talk and crazy pictures aside, I guess what I'm sayin' is to try rolling off the highs a bit. It's another duh revelation but I've typed all this crap and uploaded all these screenshots and I'm not gonna delete it all now. You may find a golden point where you can smooth out the harshness without smoothing out too much of the perceived volume, and it might not even matter what kind of limiting you do after (or before? I'd try both but after seems more promising) that.

Puce wrote:I mean that the chorus is rockin', so it would be nice if it got a bit louder than the verse, to help increase excitement and energy and add variety to the song. But the bass, drums, and guitars all stay pretty much the same volume (give or take ~1 db) throughout the song's nearly 5 minute duration. The bridge is a nice change, but--like always--I want more.

I tend to think of this as more of an arrangement problem than a mix problem. Again my So weird: it's mostly just two guitars and a sporadic organ. In the parts that you might call a chorus there is a third guitar with a lot of fuzz, the drums emphasize the cymbals more (important!), and the organ has a bit more of the forte to it. That third guitar is way below the first two -- maybe 8-10dB -- but you can hear it. The overall volume of the song doesn't really change at all, but it feels like there's more energy, and that's what matters. It's a better way to go about solving the problem than turning up all the faders when the chorus kicks in, I think. Just some off-topic food for thought which I'm rambling out mostly because I still have enough humanity left to enjoy talking about myself.

BLT-- glad you enjoyed it, though I think you'll be severely disappointed by the version with lyrics that I actually submitted

When I add music to my vids, I quite often look for music without vocals. What I like about that song is that it has an up to date progressive Dick Dale feel to it, which is perfect for action vids.

bzl wrote:For starters, this is a Har-Bal graph of Puce's unprocessed mix. It is basically a graph of how much energy there is across the audible frequency spectrum... You've got a bit much in the 8-12k range

<center>[ Insert Generic Har-Bal / Eyes vs. Ears Rant Here ]</center>
If you're going to use Har-Bal, one thing you should know about it is that you are looking at the Peak graph, as opposed to the RMS graph (that's the green one). Peak is useful for showing where certain frequency ranges have more or less compression; according to the graph you posted, my bass is over-compressed (I agree / intentional) and the 12khz range is under-compressed (or, less compressed than Har-Bal recommends). If you look at it in RMS mode--which, unlike Peak mode, analyzes sound the same way the human ear does--you'll notice that the 12khz bump is much less pronounced.

But, most importantly, does it sound to you like it needs a high rolloff?

I may have officially taken this thread off topic.

Puce wrote:But, most importantly, does it sound to you like it needs a high rolloff?

Yes, but your ears are better than mine (which is why I have to use my eyes as supplemental help), so take that with a bucket full of salt. As a quick A/B I tried a 6db shelve at 6k (on the unprocessed version) and liked it better on than off. I suppose you'd probably disagree, but, hey, it's only opinions. That's a pretty drastic chop, of course, but it didn't seem to make the song feel any less loud. Anyway, part of my liking the "dulled-down" version better is the feeling that I could listen to it a lot longer without it becoming grating.

I know that any advice/ideas I have are necessarily on the simple side. Think of it as aimed more towards other folks that will read this and aren't as experienced as you. Not that I don't want to help you, too, but I'm not so confident that I actually can.


I hope I don't come off sounding snide or anything, by the way. I certainly don't mean to.

bzl wrote:I hope I don't come off sounding snide or anything, by the way.

Not at all; I realize that I may have come off as sounding defensive. I more just meant it as "OK, now I know what your eyes tell you, but what do your ears tell you?". You bring up some good points about eq and perceived volume, which is something that's probably worthy of a thread of its own.

Nothing personal charc, but I don't get how eyes can produce music. I 100% have to rely on my ears. If I have to give them a rest, I do.
I like the graphs and all the tech stuff, but after being entertained by it, I go back to mixing the way I hear it. I honestly don't know the number settings of my last song. I can go back and look, but how is that going to help someone mixing in a completely different style than me?
For instance, I like how you mixed you're song. You kept the guitar, which is the moving force, high in the mix. Perfect for an action movie. Sure, I'd like to hear some trem bends, but that doesn't hurt the song.

Billy's Little Trip wrote:Nothing personal charc, but I don't get how eyes can produce music. I 100% have to rely on my ears.

I look at frequency power graphs to give me ideas on how to improve the sound. If I relied 100% on my ears, I'd spend a lot more time hunting and pecking for the most effective processing. And given that none of my equipment is great, I doubt my ears get the least biased information.

The "optimal" line on that chart is fascinating. I've never thought to compare my final mix to some standard distribution. I usually check for "flatness" and see where the holes are. I usually don't have enough high end, and it's nice to know at what frequency the roll-of starts.

I've had very good luck using graphs. I'm curious how often the graph and the ears say different things.

Mostess wrote:

Billy's Little Trip wrote:Nothing personal charc, but I don't get how eyes can produce music. I 100% have to rely on my ears.

I look at frequency power graphs to give me ideas on how to improve the sound. If I relied 100% on my ears, I'd spend a lot more time hunting and pecking for the most effective processing.

Maybe this is why I seem to take so long in production, because I do a ton of hunting and tweaking before I get what I want.