Interesting Post on the clipping = distortion concept.
 

A guy over on the Ableton forum made a youtube video that demonstrated that you can push your tracks into the red (over 0dB) in a 32 bit float daw and not incur any distortion into your audio signal.

This was kind of eye opening as I always thought that anything above 0dB in the digital world was bad.

http://www.youtube.com/watch?v=oipad34MYn4

note, make sure your volume is turned down when he starts the test tone.

oh noes...

I think the reason for this to work is because 32 floating point decimal means that the values are constantly calculated (that´s why it´s floating) so when he lowers the volume so that it peaks under 0 dB the shape is calculated back to a pure sine . It will be shite if you export to 16 or 24 bit wave file...

Some nerdy techniocal stuff about floating point: http://en.wikipedia.org/wiki/Floating_point

maybe I am missing something, but I thought he actually did export the +8dB audio file to a wav file and then reimported it back into ableton live? Then when he zoomed in on the wav file, it showed that it wasn't distorted.

He exported to a 32 bit floating point wave file that he re-imported. It was first when he turned down the volume that it wasn´t distorted (he doesn´t show the waveform before he has turned down the volume).

From video: He also points out that it will only work with 32 bit floating point format:

So mixdowns to CD quality is a big NO NO...

ah, ok. thanks for clarifying that onkel.

so then my thought that I should keep each track's signal below 0dB still the proper way to work?

Well as long as your audio tracks are recorded in 32 bit floating point and you keep the the overall volume under 0 dB when doing mixdowns to less than 32 bit you should be on the safe side (i think). But to be on the safe side i would keep below 0 dB in each track...