compressions an art, not a science...
a simple example of when i first played with computer tracking might be helpful, i remember being initially confused about the compressor, especially because i had never used a real-life compressor before.
I was recording two short tracks, an
acoustic guitar track, then vocal track for the same guy.the guitar track was fine, but i had problems with the vocals. It was a great track, as far as it sounded good, he was in time, singing on pitch, etc. there were brief periods, however, when he was "bringing it home" lol where he would lean in and get louder. This made my mixing job much harder...what to do? I was introduced to a crash course on the compressor. Hard limiting was annihilating the track. A limiter, christ is this hard by WORDS. The person who told you to fiddle with it, and to note the results was right. Also great the guy who said to use too little, not too much. A compressor is a power tool that can work wonders, or ruin things, if used with inexperieice.
picture youre wav. I know you can "look" at a representation of the wav because you are talking about recording into the comuter. The waveform you view is doing much the job of an oscillioscope to the electronics technician. You are looking at a graph of voltage (height) versus time. When the waveform gets taller (in height) it also gets louder at that point when you are listening thru the speakers or headphones.
when you are recording into a computer, or transferring a track into a computer, same thing, this smooth, unbroken VARYING VOLTAGE is having little "snapshots" taken of it thousands of times a second. If enough snapshots are taken per second, we hear it as continuous. Much the same as a movie on the screen appears continuous motion, though it is really a series of individual frames or snapshots.
Whether you are recording from a microphone or amp or whatver, the analog varying voltage, smooth and unbroken, is going thru what is known as a analod to digital converter (A/D) in the sound card, and the resulting snapshots are stored as a WAV file. If, while recording onto the computer, this voltage EXCEEDS (in height...) a certain height, digital clipping is the result, and it is a nasty, unmistakable sound. In this case a real life compressor before it hits the computer will control this. In my case, there ws no digital clipping, but the vocal track exhibited TRANSIENTS...sudden spikes in voltage (height) that were not sounding good. I applied a software compressor to the wav file to "even out" the vocal performance.
when you play with the compressor settings, real life or software, the following terms become useful. If you picture a waveform of the voal snippet "I want to BE the one...", the "BE" was way too loud, and was represented by a jumping spike on the waveform view...while the rest of the vocal line swung the meter from -40 to -15, this "spike" was hitting like -2.
threshold = how high the spike has to go before the compressor will react. This could be a dial on a real life compressor, or a software setting expressed in dB...either way, its how high the wave (height, voltage) has to go to trigger the compressor.
ratio of compression = how much the spike will be squished. Some people call the compressor the squisher.
attack = time, in milliseconds, it takes the compressor to begin working once the transient crosses the threshold level.
release = time, in milliseconds, it takes the compressor to stop working.
sustain = time in milliseconds between the attack and the release.
1000 milliseconds = 1 second. IE, 100 milliseconds is 1/10 of a second
Every transient is different, as is the nature (shape) of every instruments waveforem, which is why no one can give you "best settings". Some would be under squished, and others over squished, newbies tend to use too much than too little.
a drum transient, for example, must be clamped down on very quickly, and so has a very short attack setting, and a fairly quick release, and sustain is not too much. the very nature of a drums sound or waveform is that is gets loud very quickly, and goes away very quickly. A long attack setting and you will "miss" the transient you meant to control.
A vocal performance, by comparison, the voice rises in voltage (height/loudness) a lot slower, and singers HOLD their notes longer than drums typically. For this reason, the attack is somewhat slower (more milliseconds), and the sustain is much longer (often half a second or more, or 500+ milliseconds (ms)
the settings can be crucial. if you clamp (attack) too slow, you might let the transient thru. Too fast, and you get a quite unnatural sound out. too little sustain, and you quit working on the long vocal note too soon, then the thing kicks on again, quite unnatural sounding.
there are other approaches to exhaust before reaching for compression. You could just turn the mic amp down, or the volume knob in the software down. You could move the microphone further away from the performer, or turn the amplifier down. any of these things will reduce the peaks of voltage more naturally sounding than reaching for the squisher.
in my case, however, even if i moved the mic away, the word "BE" was going to be way too loud in relationship to the surrounding words...so i compressed. In the same way one picked guitar note might be too exhuberant against softer sounds around it...the comp will "even out" the performance. By experimentation (lots...lol) and constant use of the "undo" feature in my software, i finally found out that 100ms of attack was about right...sustain on the order of 400ms or so, as he held "BE" quite a long time. The threshold and ratio has to be played with. did I want to turn in on at -14 or -15 dB, and use heavy (20:1) ratio?...no...that sounded quite "flat", and indeed the waveform picture started looking more like a plateau than a mountain it looked like originally before the squisher "lopped off" (made flat) the top of the waveform. the other words were going from -40 to -15, about a 25 db "swing" or jump in meter, so i used a 2:1 ratio, threshold (start point) of just above -40. Then, when it swung almost 40 db up to near top meter, the 2:1 ratio made it only swing about 20db, and the resulting -40 to -20 swing was perfect...but since the surrounding words also started at -40 and only swung up to about -15, they got ratio'd down to like a -40 to -25 swing...this wasnt right...every mountain was reduced in height by half, so the word "BE" was still twice as loud as its neighboring words...with MUCH EXPERIMENTATION, i finally found that a ratio of something like 7:1 or 8:1, thresholding further up like -20 or so was about right...finally, the word "BE" was in relation to its other words in the vocal line...but it took quite some time.
i would reccomend to see how this works (sounds) a plug in compressor, like xKompressor, to play with...feed it a short section of drums, a short section of vocals, etc etc...and turn the settings all the way up...all the way down...some up some down...experiment and listen to what it does for each instrument sound. You can fiddle with the real-life looking knobs until you think it sounds evened out, yet still natural...and always go with less, rather than more. Anytime you are reaching for the compressor constantly, with high settings (high ratio, los threshold, quick attack and release, long sustain...) on many tracks, and on final mixes, you are using it way too much. This is why:
beautiful sound SWINGS the meter. This is true in radio electronics, and it is true in mixing. The same sound swinging less is much less beautiful and much less interesting. This is why a solo piano piece can sound so soulful and beautiful and stir emotions...it is the only sound at the time in the mix, and the engineer can gleefully let the meter swing from all the way down to all the way up freely...the sound is quite beautiful, and without compare. Ditto for a short vocal solo temporarily acapella...it really stands out and sounds beautiful. the same exact piano, forced into an already over crowded pop mix, cant swing as much to fit in, and doesnt have nearly the impact. This is where a cold signal sounds worse...youre recorded wavefore only swings the meter 10db, and when you turn the volume of the track up, it swings still only 10db...just further up the meter. its louder (more quantity of) sound, but it isnt more beautiful (more quality of sound)(total swing=beauty).
it is the mark of learning how to mix, and i am only just beginning to learn, i assure you, that you have other options before compressing. Sure, i could compress both guitars to get them into the mix, but i can also pan one left and one right apart a little, and preserve their swings and not dull the sound with the compressor. If the notes are longer lasting, i can start one guitar several milliseconds earlier than the other, and then they both swing, but dont compete at the same time. A great mixer can get many instruments, and tracks, to sit in the mix by judiciously using l/r pans, good eq practice, proper tasteful use of certain delays and effects and such. This is the mark of the professional...the crude beginner simply compresses everything and turns them all up, than when it sounds dry and uninteresting and flat, starts spraying effects too heavily around to try to put life back in the soup. Its rather like over cooking a potentially fine filet of orange roughy to kill it, then trying to cover up the mistake by gobbing too many spices and peppers all over it in an attempt to resurrect the taste...it doesnt succeed by half. You can always go back and compress a little MORE later if you have to, but you cant undo the damage if you find you used too much and the final mix is now dull and lifeless.
it is perhaps not fruitful to compare your mix effort to a similar sounding CD as a reference...the commercial CD has been mastered and is louder, hell, to a lot of peoples taste, the commercial CD has maybe had all life squished out of it. get a good mix engineer to reccomend a good reference commercial cd that really swings the meter and produces some emotion and beaty...and go for that sound.