All of the below is guesswork, so apologies in advance.

Reading the replaygain spec, the original talks about pink noise @ -20dBFS corresponding to a replay level of 83dB SPL. Since 89dB has been adopted now, we can assume that corresponds to a -14dBFS digital input signal. Curiously the largest Replaygain value you ever get is around -14.00dB.

So if Replaygain trackgain figures were not calculated with weighted noise curves, and you used the 95% level figure, it would be safe to boost Replaygain figures by approximate 13.3dB without clipping.

Certainly it stands to reason that it is safe to boost any replaygain figure up to a net gain of 0dB (you can't clip without any net gain). But if a track already has a large positive figure (curiously total silence is set at approx +64dB), is it ever safe to apply any boost and guarantee you won't clip ? I would guess you are safe up to about 12dB of boost (it falls under the theoretical 13.3dB), but is there a way to tell how loud the loudest peak (eg single 16-bit word) is and apply a boost based on that ?

I ask because I want a better level out of a portable. I am using a line driver to boost the signal, but I want the maximum potential out of the digital domain, before boosting the analogue signal (and the noise!)

hope someone clever can put this all to rights

Clunes

If the ReplayGain adjustment is 0dB (or you change it to 0dB) then this means no adjustment, so the track won't clip any more than it did already.

If the ReplayGain adjustment was, say, -8dB, and you applied +13dB gain to this, the overall gain is -8+13=+5dB relative to the original. You could do this by changing the ReplayGain adjustment to 5dB, if you wanted. Unless the original signal stayed 5dB below peak all the way through, then this will introduce clipping.

I know many people are insensitive to clipping, but 5dB of clipping is rather bad!

Cheers,
David.

Thanks for your answer - do you happen to know if there is anyway that you can calculate at what replaygain setting a track will clip ?


cheers

Clunes

If you plan to change RG-values on a per-track basis, then why are you using replaygain in the first place? Replaygain isnt about preventing clipping - its about equal loudness. The prevention of clipping is just a safety-feature which is implemented by some players. If you just want to prevent clipping, then you could as well use a normal peak-normalizer - no need for RG.

Someway, i really get the impression that RG should have been called "loudness-equalization" - many people seem to dont get its purpose.

- Lyx