The headroom question is interesting. Ideally a 0 dB FS peak would be properly played without distortion due to nonlinearity. I have heard it suggested that since about -20 dB FS on a soundtrack is supposed to result in 85 dB (C) at reference position from a screen speaker, we should be able to play 0 dB FS and get 105 dB (C) there. However, SMPTE pink noise has a crest factor of 12 dB, so it is only possible to go up 8 dB before it is clipped by the digital recording. So, do we really need an extra 20 dB? Is SMPTE pink noise a good representation of soundtracks? Also, there will always be some distortion due to nonlinearity. How much is acceptable? I have been impressed with total distortion measurements using multitone (also called "spectral contamination"). The tone amplitudes can be weighted to represent typical soundtrack spectral distribution. Tone frequencies are chosen to avoid both harmonic and intermodulation frequencies so both harmonic and intermodulation distortion are detected.
Harold
Harold
Comment