mannyaudio85 wrote:So, creating a filter that manipulates the audio, in whatever way it needs to, to lower the crest factor to a value closer to 0 while not modifying the RMS, is an impossible task?
The only frequency in this world that can have a crest factor of 0 is 0Hz aka DC. A crest factor close the 0 means your signal has very little dynamics, even the worst mastered albums have a DR of 5-6, so your aim is to make a bad mix or maybe a distortion/fsu plug?
Just to reiterate: Crest = Peak - RMS;
Phase rotators, as used in radio for vocals as JCJR pointed out, only works because the glottal pulses of the human voice is asymmetrical to begin with, think of it as a pulse wave with 20% duty cycle, any symmetrical waveform with have little or no effect or even the opposite effect(increased crest).
As an increased dynamic example, take any mastered song(a whole song), put it in your DAW and put a phase rotator on it, a mastered song will obviously be slightly less than 0dBFS, now process it with a phase rotator, you'll see(with an accurate meter) that some signals will be greater than 0dBFS, aka clipping your converters.
On a side note about phase rotators used in radio, they are used because they had no choice -> it was a compromise decision, AM/FM transmission needs a high RMS level to work, it's a low fidelity technology, any $20 CD player will have better quality than a AM/FM radio broadcast.
Another thing about phase rotators, they also smear transients(make them mushy), a single one won't do much but add a couple hundred in series and you will get laser zap type sounds, this arrangement is exploited on some spring-reverb emulations.
A very cool plugin that doesn't do what you want(it increases phase correlation aka Peaks/DR), but works with phases non the less is this:
If you want to change the levels of individual instruments after they have been mixed, Melodyne might work for you, but it still won't exactly do what you want, in arithmetic terms, you want: 1 + 1 = 1.