View Single Post
Old February 14th, 2006, 03:39 PM   #20 (permalink)
chadsxe
Super Rad Member
 
chadsxe's Avatar
 
Join Date: Dec 2005
Posts: 1,136
COMPRESSION

I'm "reprinting" this so it is easier to find. By popular request I typed up a quick and dirty tutorial on better use of compression.

Here's the big secret of compression:

You should *barely* hear it working except as increasing your overall volume within the parameters you need. The average person may not even hear it working much. THAT is how the pro's set 75% of their compression, the other 25% is super squish city reserved for things like submixing drums in stereo and mixing it back in at low levels to beef stuff up.

Jim's rules for compression:

First let's define what a compressor does--which is to affect the amplitude of a signal by selectively reducing it. Compressors tend to have the following controls:

Compression ratio: this determines how 'hard' the compressor is supressing the signal. Usually described as a ratio such as 2:1, 4:1 and so forth. What this means is, after you cross the threshold setting, how many db's you have to go over to effect 1db of volume change. Thus a 4:1 ratio means that once you go over the threshold for every 4db over you will only get 1 db of amplitude change.

Threshold: this sets the decibel level that the compressor starts to work. Signal underneath the threshold will be unaffected--signal above it will be hit by the compression amount determined by the ratio. Needless to say, setting the threshold above the peaks of the signal will NOT do jack shit to the signal. You gotta set it in the path of the signal, so to speak. This is always expressed in negative db, thus a -24 threshold will compress any audio above -24db, and leave everything below it alone. (*Note, soft knee compressors start to work a bit before the threshold!)

Attack time: how long, in milliseconds, it takes for the compressor to kick in. This keeps your transient peaks unaffected and is the trick for getting a "punchy" kick or snare (the front end crack will be uncompressed and thus louder than the following signal).

Release time: once the signal falls below the threshold how long, in milliseconds, it takes for the compressor to 'let go' of the signal. For vocals and other similar instruments you want this to be fairly long like 200-250ms. For drums 75-125ms is great.

Special note on soft-knee compressors: some compressors have a soft knee function. What this does is start compressing the signal lightly as it approaches the threshold, and as you get closer to the threshold it will compress harder and harder until you reach the threshold and the full compression ratio will be utilized. This provides for fairly transparent compression and is great on vocals. Personally it sucks for drums unless you are squishing a stereo submix of drums.

Another note on stereo compressors: you should *always* link stereo sides of compressors when processing stereo signals. Once a side reaches the threshold BOTH sides get the compression. Failure to do this can lead to, for example, drums that leap in volume on one side but not the other... very assy unless that's what you really wanted. (Why god, why???)

Moving right along.....

Here are some guidelines off the top of my head:

2:1 ratio--overheads, distorted guitar, soft vocals, most synths
3:1 ratio--clamping down on overheads, acoustic guitar, most singers
4:1 ratio--bass, snare, kick drums, toms, crap singers
8:1 ratio--bad bassists, screaming vocalists, squishing the life out of stuff
12:1 ratio--out of control peaks or when you want to sound like limiting but still keep some life to it

Compression ratio and threshold are intertwined, so set both accordingly!

If you need dynamic range--LOWERthe ratio
If you need more regularity in levels--RAISE the ratio
If you just need to shave off some peaks--RAISE the threshold
If you want to affect a lot of the signal--LOWER the threshold

Here's the tricky parts that require hard decisions:

If you want more smooth sounds--LOWER attack time (under 6ms)
If you want more punch--RAISE attack time (between 7-50ms)
If you need "more" compression--LOWER attack time more
If you need "less" compression--RAISE attack time

If you need 'invisible/natural' compression--RAISE release time
If you need 'audible/percussive' compression--LOWER release time

Now pull out yer ears:

If it pumps and breaths--RAISE release time (unless you want that)
If the compression seems to disappear--LOWER release time

Finally the number one rule for compression:

ALWAYS match relative volume levels (by ear) before and after compression using makeup gain--meaning that they should be peaking about the same. If you record using my "-15 to -12dbfs with peaks no greater than -6dbfs" rule this is easy; if you tend to record sloppy and "hot" you may need compression to keep you out of the red. Don't do this to yourself.

The idea for this is that LOUDER often equates to sounding better to us, fooling us into setting duff and mookish compression settings. When dialing in compression make sure that the before and after levels are identical so you can hear the compression and not the jump in volume.

Here are some guidelines on setting makeup gain:

The lower the threshold the more makeup gain you need.
The higher the threshold the less makeup gain you need.
The higher the ratio the more makeup gain you need.
The lower the ratio the less makeup gain you need.

Further modified by:

The faster the attack the more makeup gain you can get away with.
The slower the attack time the less makeup gain you can use.
The faster the release time the less makeup gain you can use.
The slower the release time the more makeup gain you can use.

You can also calculate the amount of makeup gain you need by looking at the signals peak levels or RMS--figure this out, and then:

COMPRESSION THRESHOLD = T
SIGNAL DB PEAK = P
COMPRESSION RATIO = R

(T-P)/R = A

P-A = MAKEUP GAIN

Thus if your threshold was -24db and your signal is peaking at -12db the amount of gain being compressed is 12db total; i.e. (T-P). Divide this amount, 12db, by the ratio of 4:1--making A = 3db reduction of peaks.

Then take the signal db peak and subtract the peak reduction--in this case -12db and subtract 3db... meaning you require 9db of makeup gain to approximate the original signal level.
chadsxe is offline   Reply With Quote