
Originally Posted by
Simon Barden
I'd immediately advise you to swap to 24 bit recording. With 16-bit, you need to get the recording level as hot as possible (without overloading and getting digital clipping) to maximise the theoretically available 96dB (in practice a bit less) of dynamic range and headroom above the sysytem noise floor. With 24 bit, you can run at much lower input levels e.g. -18dBFS (which is still 21 bits of data) and get a much better theoretical dynamic range and a lower relative noise floor. All the time you are recording at say -1dB or -0.5dB, even though you may not clip digitally, inter-sample peaks mean that when the D/A converter will clip the sound produced.
You should have somewhere in your software (or find on the web) a true peak meter, which calculates the levels of inter-sample peaks (ISPs). Use it as the last insert on your master bus, and this will tell you if any of your recordings will distort. ISPs are worse on waveforms with sharp, near-vertical, leading edges (such as true square waves) and percussive transients, where they can be almost 3dB above the maximum recorded digital value. ISPs exceeding 0dBFS are one reason why some recordings sound really nasty when converted to MP3s etc.
So always 24 bits, and take things easy on the recording levels.