Originally Posted by Audiop
Loudnes war again? Or do you mean the material peaks should top at -3dBFS to 0dBFS?
Many DAC's have poor performacne close to 0dBFS so staying at -10dBFS would make sense. OTOH an occsional peak that hits close to 0dBFS will not cause problem for a typical listener.
Yes, the peaks should hit the highest levels are possible. I'm well aware of the diffuse "don't overload the AD/DA's analogue circuitry" arguments floating around the web since years. IMHO, this is not a black or white discussion.
You can't deny the simple fact that *not using* the full scale also means you are not using the potential of your AD or DA. In case a ADA manufacturer was stupid enough to ignore potential overloads of the analogue circuitry in his design, don't use his products! Such "near full scale" problems are maybe relevant to old and/or very very cheap DAs. I never saw a professional DA having problems with the conversion a full scale signal. It's their job, and trivial for manufacturers to give the final amp a dozen dB more headroom.
Not saying it's a good idea to record "hot" like in the old 16bit days, there are good reasons to leave generous headroom during recording. Especially since everyone uses 24bit for recording. But a -10dBFS recording will definitely be captured with greater fidelity than -20dBFS. This is clearly audible, measurable and provable too.
Just try to create 4-6 successive copies (playback, recording,playback, recording,...) of you favourite CD. Do it -1dB below the full scale and compare to the same thing with generous -20dB. The quality difference becomes very apparent with the second a third recording cycle. The fifth or sixth really sounds like an 8bit recording vs a proper recording.
I'm not really sure how far all this relates to loudness war. Isn't that more about brutish clipper/limiter abuse?