Loudness normalization with limiting

Does Audacity have an effect that combines Loudness Normalization with Limiting? The Loudness Normalization effect allows normalizing loudness (RMS or LUFS) TO a target level but doesn’t include a Limiter feature applicable to peaks. The Limiter effect allows both increasing loudness [with input gain] BY a user-specified db and limiting peak db to a user-specified level, but this could be risky in a macro when you don’t know in advance the loudness of the raw track to which it will be applied.

Would defining a macro with steps in turn Loudness Normalization (for specifying target loudness db) and Limiter (for specifying max peak db) do the job? (Would this risk amplifying peak db into clipping before Limiter limits it to below clipping?)

How about defining macro with steps in turn Loudness Normalization (setting loudness to a moderate target level) and then Limiter (input gain amplifying loudness BY an amount compatible with the preceding loudness normalization, and limiting peak db to a target below clipping)?


– In Limiter, can Input Gain drive peak db into clipping (exceeding 0db) before “Limit to(db)” is applied?

– Please clarify the difference between Hard Limit and Hard Clip, and between Soft Limit and Soft Clip:
– if peak db is BELOW 0db before “Limit to (db)” is applied. E.g., are the “Clip” settings relevant here?
– If peak db EXCEEDS 0db before “Limit to (db)” is applied.
These distinctions are unclear in Audacity’s manual.

Limiting will always introduce some amount of distortion.

“Hard clipping” acts only on the part of the waveform above the threshold, pushing it down to the threshold level (in effect it is chopping off the peaks) and produces the most distortion.

The other forms of limiting start to act before the threshold is reached, so they operate more progressively, thus producing less distortion.

For Audacity’s limiter, the types of limiting in order of “how much distortion” (from least to most):

  1. Soft Limit
  2. Hard Limit
  3. Soft Clipping
  4. Hard Clipping

However, the Soft Limit algorithm only works for audio below 0 dB. If the peak level is above 0 dB, then it drops down to the “Hard Limit” algorithm for those parts.

Also note that the amount of distortion increases with the amount of limiting. 3dB limiting with “Soft Limiter” has very low distortion, but increasing the amount of limiting effect will greatly increase the amount of distortion. You have to decide how much distortion is acceptable for you.

For best sound quality, normalize to a level that does not exceed 0 dB peak.

To normalize to a higher level while minimising distortion: Normalize or amplify to 0 dB Peak first. Then apply “Soft Limit”, and then Normalize again.

Yes you “could” just Normalize and then limit (as the all-in-one tools do), but that will compromise the sound quality, which may or may not be acceptable for your needs - it is certainly easier.

(Full disclosure: I wrote Audacity’s Limiter effect)

The Audiobook Mastering procedure/macro does that. It adjusts to the target level, then limits. (Oddly, the ACX spec says the peaks can’t go over -3dB.)

No. It can push the levels in to potential clipping but Audacity uses floating-point internally so it essentially has no upper or lower limits and it won’t clip… You can amplify , or boost the bass, etc., and it might “show red” for potential clipping but it’s not clipped yet.

You can clip your DAC if you play it at “full digital volume”, or if you export to a regular (integer) WAV it will be clipped. But as long as you lower the volume before exporting everything is OK.

I’m not 100% sure, but in my limited experiments (pun unintentional) both hard & soft limiting seem to use look-ahead so they don’t clip or otherwise distort the waveform.

This topic was automatically closed after 30 days. New replies are no longer allowed.