A neat trick I learned...
 

Feed both channels the same mic signal, except add a -20dB pad to one of them. When you edit in post, you can fade between the two signals as necessary to get something like 280+dBs of range...good for scenes when the actor yells in the middle of a whisper or something.

Probably old hat for most of you, but I thought it was rather clever.

This is great for 16bit, but not really needed in 24bit, as you can afford to leave tons of headroom.

True, although those of us who record audio to the camera hardly ever run into 24bit.

...oops, duplicate post

Wouldn't it be more prudent to do a quick run through, and use that run through to set one channel for the normal/whispered dialogue, and one for the yelling? Rather than just arbitrarily apply the -20 pad? I would use an arbitrarily lower setting for one channel in an uncontrolled type of situation. . .documentary, ENG, live, etc. where you don't know what's going to happen beforehand.

I record drag racing, its not possible to do a run through. This can be an invaluable tip for that type of recording. I run a single shotgun mic, I want voices as well as up close shots of the starting line. and its common for an engine to start up unexpectedly. there is no way to predict the volume levels, and I can't always adjust them in time. and of course, the automatic levels dont sound right

That's what I was saying. . .in those types of situations, the OPs idea is a good one. But the OP mentioned actors specifically, so I assumed he was referring to narrative/controlled shoots.

This is an excellent idea whether you're recording 16 bit or 24 bit, not so much to expand the dynamic range of the final mix per se - in fact, compression to reduce the dynamic range is often done in the final mastering stages - but because it's very easy for an unexpected loud noise to drive a track into clipping and distortion. By setting up one primary mono track at the nominal recording level and the secondary at a signifigantly lower level, an accidental over can be salvaged in post from the unclipped insurance track. I like the idea of a pad since there are a number of consumer DV cameras that don't have independent recording level controls for each channel. Not sure if I'd put the insurance channel 20 dB down though, 10 dB seems better to me.

When I split the track I usually do 6 dB or 8 dB. I think that's standard. I try to avoid it altogether though, as post often screws it up.