Question: "Compression Theory" (w/screenshots)

  • Thread starter Thread starter amra
  • Start date Start date
amra

amra

Well-known member
Screenshot - my target waveform and volume level for the recording I am working on

Screenshot - My current, unprocessed mixdown (note the overall lack of volume and the spikey transient signals that I can't kill)

Screenshot - The other extreme - How do they get this flat waveform and corresponding loud volume level without clipping?


Ok, so I think I have a general working knowledge about how compression works and how to use it during tracking. But I am trying to do a little more advanced type compression. I am having some trouble killing some quick 'transients' in my recording that are keeping my from getting the volume I am looking for. No matter how fast I set the attack it does not catch these little spikes, and then I can't boost the average db to where I want without causing these spikes to clip. So here are my questions. Feel free to answer them all, or just anwer one, any releveant information is welcome.


1. To get the overall volume of my recording up close to the levels in screenshot 1, I would think that a setting of: Threshold around -8 or -10 so, hard knee gain reduction, ratio at 7:1, attack at .5 ms, release at 4 ms, output gain set to 0db, would give me a nice tight waveform with peaks around -6, catching all those quick spikes and taking them down to at least -6db as well. But this is not the case. It IS tightning up the waveform some, though not to the degree in screenshot 1, but it is barely reducing those big spikes at all. If those spikes were gone, I could get another 3 or 4 dbs I am thinking. What settings will get me where I want to be? lower threshold, higher? harder/softer knee? quick/slower attack? what am I not understanding here.

2.Are the recordings in the first and 3rd screenshot using multiple passes through a compressor? or are they squashing it hard in one pass?

Thanks for taking the time to read this,
amra
 
Hard to say - Every mix is different, and no one setting will do the same thing to every mix. You'd be surprised to find out how much of that is clipped... The thing with clipping is that once you clip, you just turn it down a little and Presto... "Technically" it's not clipped anymore (as it's not full scale anymore).

It's still square, it's still smashed, it still sounds bad, but it's not all "1's."

HOWEVER -

Please, please, don't actually shoot for crap like that last one... How do they do it? Easy - Ram it through (anything) and then take the volume down. Or just use a digital limiter. IMHO, it normally sounds like crap. But it's reasonably well-known that I have issues with digital limters for the most part.

Don't use your eyes - Use your ears. But if you want to get it loud, ram it into a digital limiter. IF the mix holds up, great. It probably won't, or it would already be loud. As stated in my sig, it's FAR more important what you do BEFORE you hit the record button - That goes for volume as well. The potential for sheer volume in a finished mix is determined long before the mix is finished - or even started, for that matter.

[EDIT] Oh, Disturbed... Yeah, that's something to shoot for... I can't listen to 10 minutes of that album before my ears start trying to jump off my head... But keep in mind how separated the sounds are on that disc... How nothing really crosses over anything else too much. Thin, sparse, VERY heavily limited at the track level - That tune has a lot of open space. No dynamics to speak of, but a lot of emptiness between sounds. If it was much different, it wouldn't be nearly as loud before completely falling apart. [/EDIT]
 
Last edited:
You might want to try going in and editing the peaks through the waveform editing function of whatever program you're using. Zoom in on the peaks, highlight them, and manually gain reduce them so they're in line with the rest of the track. You should be able to squash till you heart's content after that...not that I would recommend doing that.
 
amra said:
I am having some trouble killing some quick 'transients' in my recording that are keeping my from getting the volume I am looking for. No matter how fast I set the attack it does not catch these little spikes, and then I can't boost the average db to where I want without causing these spikes to clip. So here are my questions. Feel free to answer them all, or just anwer one, any releveant information is welcome.
Click to expand...
Kaz is exactly right. This is where using a waveform editor as a waveform editor instead of just a host for plug-ins is called for.

Manually highlight the major offender peaks one at a time. You should do this at a high enough zoom level to ensure two things: that you are highlighting *only* the offendng transient, and that you are getting the whole transient peak from beginning to end. If the peak has zero crossings, start and end your highlights there, but often there are no zero crossings because the peak is riding on top of other sounds. Then just decrease the size of the transient peak by a couple of dB or so; enough to put it in line with the rest of the "normal" peaks in the mix. This will make compression and even (if you absolutely insist) normalization much easier to work with and much more effective in outcome.

Something to keep in mind: a compressor - whether plug-in or outboard iron - is noting more than an automation that applies changes to the shape of the waveform according to a defined set of characteristics; it's algorithm. Even a Manley tube compressor is just an old school version of a "wizard" that follows a certain mathematical "program" for adjusting the sound.

As with any automated algorithm (more familiar types might be noise reduction programs or beat/tempo matchers), there is a limit to what they can do and how they do it in the real world. The "programs" are just dumb programs that can't think or adapt to special or complicated situations. Sometimes the algorithm doesn't solve the problem, and the human mind and hands have to manually step in fix some things.

This is why God created NLEs. :D Just as manual editing techniques often work better or faster for noise reduction and audio restoration than a canned NR plugin does, sometimes in cases like this, you want/need to do some "spot compression" manually to the wave before throwing the fancy pre-programmed boxes and plugs at it.

As was pointed out in another thread also; this is one of those rare cases that seems to go against the "mix with your ears and not with your eyes" paradigm (to which I otherwise wholeheartedly agree.) But really, it's still mixing with your ears; it's your ears that told you that the compressor just was not giving you what you wanted. But in this case it took you eyes to tell you *why* that was. ;)

Why not just use brick wall limiting? Well, that workes perfectly well most of the time also. But again, it's an automated process that can do two things; it first limits all transients to the exact same volume, which could sometimes (though not necessarily) have a less "organic" feel to it, and secondly, it's circuitry/program could add coloration to the sound. Maybe that coloration is desired, in which case, go for it. But if you want absolutely zero coloration, manual editing is the way to go.

G.
 
You seem to want to achieve a level that has been pushed beyond its limits. You have noticed that a nationally released recording has a very high level. Most folks here will tell you that you don't want to go that high, that it is a bad thing. But, the damn song is making millions! What gives here? Let me explain.

Record Producers always want more volume. They figure that, in order to get everyone's attention when their newly released song plays on the radio, it needs to be just a little louder than all the other songs on the radio. That is a good strategy but, the Mastering Engineers know that if they push the volume level too high, (using compression,) the quality of the recording suffers. Dynamic range gets reduced to the point of sounding extremely unnatural. Volume peaks get sliced off at the top causing distortion. Subtle overtones and effects tend to get lost or buried in the mix. Things just start sounding uglier overall. Furthermore, when radio stations broadcast, they are forced to use heavy compression to accommodate the bandwidth limitations of FM carrier waves. The volume wars seem to have reached the ceiling but somehow they continue.

The Engineer is caught in the middle of this level tug-o-war and is sometimes forced to do things that he knows is wrong. If you want to copy-cat those "Bad Habits" you will likely end up with a crappy sounding recording.

Pushing the volume higher is usually a good thing as long as you understand how high is Too High. You can get away with squashing those peaks a little bit but, try to go easy on it. Better yet, let the Mastering Engineers make that decision.

RawDepth
 
1st question of mine is about the scale being used in your software (as shown in your screen shots). Is that a percentage .... the 100 being 0dBf?

If so, then you should convert it to a dBf scale as that is the scale you are using to set the compressor.

That could be a big reason why you are not getting the desired results you are hoping to achieve. Setting the compressor to a threshold of -10dbf means what to the scale of your waveform????

That being said.....Basically its a lot like sharpening a digital image ..... a couple/few passes at modest or conservative settings almost always produces a better result than one pass at extreme settings. I'm assuming you are using a plug-in that is not designed to be a "character" compressor that adds a lot of its own sound to the tracks and are only trying to tame levels and not impart a different "sound".


So, you could 1st put the track through a hard knee medium compression ratio with fast attack and release (in general ... use your ears) with the threashold set just to tame those highest peaks (about -40 on your screenshot #2, or what ever compressor settings match that scale).

Then run it through a limiter or 2nd hard knee compressor with really high ratio, fast attack and release set to a little HIGHER ratio (like -45).

THEN, if you want to raise the overall volume try running it through a 3rd compressor set to RMS mode (as opposed to peak) and a low ratio of about 1.2:1 to 1.5:1 and a really low threshold (maybe like the equivilent of -20 or -25) and medium attack and long release .... again use your ears and avoid any pumping... and then crank up the makeup gain to taste.

Avoid turning the makeup gain to the point of clipping ...... like that horrible mess made out of that disturbed song.

All of these setting are much better made if you can real-time preview in your editor.

The nice thing about doing this digitaly is that you dont have signal loss or noise added by chaining multiple compressors likie you can with analog chains.... although the noise floor will still go up and you may introduce latency on individual tracks.

As suggested by the others you could also manually edit the waveform but I have a little less patience than them ... lol.

-mike
 
Your 'target' mix has was is presumably the verses 6dB quieter than the chorus (50 on that scale corresponds to -6dB--you can switch to a dB scale in Wavelab).

Your unprocessed mix is simply really quiet. Only a few peaks go over -6dB. I dunno what dB level 40 corresponds to (-8dB?), but I'd be looking to slap down peaks to that level, then you could probably get 8 dB of gain or so. Your release time might be a little quick. Zoom in on the peaks and see exactly how long they are.

I agree with Massive--that third waveform is disgusting. It clips all over the place. Even so, I've seen much worse than that--that's actually tasteful for modern major label releases.

PS Weird Al does a funny polka version of that tune :)
 
Take a look at song 1 from korn's "the Untouchables".

Its so pegged....it sounds retarded.

Dynamicless....its is not music.
 
You must log in or register to reply here.
Back
Top