[sdiy] Screwing with Square Waves

[rburnett at richieburnett.co.uk]

> Of course a
> Hilbert wave has a much higher peak-to-peak value than a square wave
> of the same volume.

Definitely.  That's why a technique called "phase rotation" is used in 
broadcast processing ('Optimod') to increase loudness.  It consists of a 
chain of all-pass filters that aggressively rotate the phases of 
low-frequency components so as to minimise the peak to average ratio of 
the signal.  You can then turn up the volume some more before clipping 
so your station gains perceived loudness against the competition in a 
competitive market.

To be honest, the phase rotation is one of the most benign things that 
goes on in broadcast processing! ...but it has an interesting 
interaction with CDs that are mastered with clipping present.  The phase 
rotation tends to move the clipped portions of the audio waveform away 
from the peaks and troughs and towards the zero crossings.  Then the 
audio gets clipped again at the peaks and troughs during the stations 
own normal "loudness processing" leaving a severely butchered waveform 
with little actual detail remaining at all!

The lesson from this is don't master your recordings with clipping to 
make them sound loud, because it fights against what the broadcasters 
already do to make it sound loud, and ultimately sounds awful.

-Richie,