I’m curious as to what level you record at? Do you try to get the signal as hot as possible without peaking either on your software, in my case Cubase SX3 input level or your interface, or do you leave some headroom on both? If so, how much do you leave. Let’s assume that we’re recording at 32 bit.
I’m pretty sure we’ve covered this already in replies to previous questions of yours, but I’ll go over some things again.
First off, it’s a total waste to record in 32 bit mode since there are no such things as 32 bit A/D converters. You are just wasting disk space and performance, and those extra 8 bits just get filled with zeros. Even the best 24 bit converters have, at best, 20 to 22 bits of real data, as the last few bits are below the noise floor of the analog side of the circuits themselves. The software does all the audio processing in the mix engine, and for all inserts and other effects at 32 bits with its internal audio engine, so all the critical math is being done in 32 bit floating point mode anyway, regardless of what bit depth you record at. So, set your recording to match the resolution of your converters, which is 24 bits for most modern converters and soundcards (some old ones might only do 16 bits). Even for my final mix down file, I still use 24 bits.
As far as the level goes when recording, it used to be recommended that you record the signal as hot as possible, without clipping, when recording digital. BUT, this was really for the first generation of digital multitrack machines that recorded in 16 bits and had fairly bad sounding converters by today’s standards. In those days, it was believed that it sounded better if you tried to keep the levels as close to clipping as possible with the 16 bit machines with crappy A/D converters.
Now days, though, 24 bits gives you MUCH more of a dynamic range than 16 bits does, and, as mentioned above, even the best converters in the world can’t give you a full 24 bits of real data. So, we’ve got some headroom to work with. Also, it’s because of the analog side of the gear which is meant to operate at 0Vu as the optimum operating range of the analog components. 0Vu is typically set up to be around -18dBFS in the digital world. So, you would want to try to set your levels so that the AVERAGE RMS level is around -18 on your digital meters, with the loudest peaks not going above about -6. This gives you some extra headroom for the typical situation where the talent plays MUCH louder once you start recording than during your level checks, and also keeps you in the optimum range for the analog side of your A/D components.
This is especially important for the cheaper soundcards or A/D converters. The more expensive A/D converters can probably handle being pushed to the limit better than the cheaper ones, but it’s still best to try to stay within the optimum range.
So, that’s what I try to do these days. I record in 24 bit mode, since my converters are 24 bit, and I try to keep my peaks no higher than -6 dBFS. This also gives me more room for tweaking during mixing without worrying about overdriving plug-ins or having to turn the master fader down because all my individual channels have been recorded too hot.
Hope this helps!