This question was posed in almost the same manner in another thread (headroom/digital recording).
I answered the question with "try to use as much of the recorder's capacity as possible" and then bring the faders down to a reasonable level in the mix to try to start with a fully 24bit recording if possible. I suggested tracking with peaks no lower than -12dbfs but low enough to avoid clipping.
I was roundly insulted by Otitis Media for being stupid. Rather than spitting back "Yo Momma, D*ckweed!", I thought a little about my twenty years of digital recording, Sony F1, Sony 1630, Sony 3324, DAT, Pro Tools, MX 2424, etc to see if I have been doing something wrong. I don't think so. The point here and in the other thread IMHO is that I want every bit (sorry) of resolution to begin with so that I can use all of that processing power in the console and still end up with the highest resolution mix master. Otitis said that it was stupidity to record hot and then bring down the levels. However, when you bring down the levels in a 32 bit mixer, you are still preserving the full resolution of the 24 bit multitrack recording. If I track at -20dbfs on the front end, I am throwing away resolution. And yes I know that 20bit resolution exceeds the dynamic range of most playback systems, however I am intent on creating a 24bit recording, not a 20bit recording played back through a 32
bit mixer.
Now for most of my career I have used digital multitracks interfaced with analog SSL consoles. In that setup, I could hit the (16 bit) 3324 right up to -1dbfs peak and the SSL liked it fine. My gain staging approach has always been to generate all of my gain with the mic pre and if my equalizer drove the level too high, to bring down the mic pre, not the fader. On an SSL, I always track with my little fader at 0db. On an analog console I like to run the master fader at 0db attenuation as well. This is the most accurate taper on the fader and insures that I am not adding noise through additional gain or allowing distortion on individual channels and correcting for too much gain with final stage attenuation.
I recognize that 32 bit digital consoles do not adhere to these same parameters. However I still contend (even though I don't seem to be argueing with either Eric or Nika) that attempting to retain the highest quality from the source and on through the chain is the best way to end up with a solid product.
Best Regards,
Bill