Topic: 24/48 to 16/44 conversion introducing clipping...

[breakonthru]

Sample Rate Conversion (SRC) can introduce peak levels beyond 0 dbfs due to interpolation. So the order of operations is the key.

Gordon

there was a huge debate about this
“Intersample overs” back in the day when benchmark introduced the DAC 2. While I generally find the technical discussions of John Siau to be impenetrable, the “community” as it were just poo-poo’d the idea as if it wasn’t a thing, or at least not as significant as benchmark made it out to be

https://benchmarkmedia.com/blogs/application_notes/intersample-overs-in-cd-recordings

[tim in jersey]

I use Sound Forge and normalize to 98.14% (not sure what the dB value of that is), and then dither with iZotope's mbit process.  I never noticed any clipping after dithering.

Looks like 98.14% is about -0.16dB.

I was just coming to post that!

I'm curious as to the math behind y'all coming up with that number. Care to enlighten me?

[nulldogmas]

I'm curious as to the math behind y'all coming up with that number. Care to enlighten me?

[Chilly Brioschi]

FWIW, the sharpest mastering engineer that I know says -3db, and why not -6db for 24 bit?
Let those peaks breathe!

Pop music be damned

[Gutbucket]

Goldilocks middle way.  Not too much, not too little.  No need to overcompensate when normalizing by trying to get it as close to 0dBFS as possible.  Just need enough extra space so the complete waveform doesn't get chopped off at the top.

If peaks are way down then normalize to a peak level closer to 0dBFS. If unsure exactly how high that peak actually goes because it might happen to crest in between two sample points (an intersample over) choose a slightly lower target peak value.  That's it.  One or two dB down from 0 is plenty and equates to a more or less insignificant difference in the position of the listeners playback volume knob. 

-3dBFS is not unreasonable. ..and that or even -6dB is likely what a mastering engineer would prefer to receive from a client.  But a target of -6dBFS as a target for final release gains nothing and will require additional playback gain to compensate, as it will most likely sound significantly quieter than whatever is played before or after.  The additional 3dB of empty space gains nothing. -6dBFS is totally fine as a "released as it was recorded peak level" that isn't so low it really needs normalization to be listenable, yet is an overly low target for final normalization of a finished recording where levels are being actively adjusted in post production.

[rocksuitcase]

I do not normalize, I add gain up to -0.2dB to my final wav 24/48 files.
I *may* convert down to 16/44.1 for some odd reason, but gave that up as a workflow habit many years ago.

[nulldogmas]

I do not normalize, I add gain up to -0.2dB to my final wav 24/48 files.

Isn't "adding gain up to X dB" exactly what normalizing is?

[Gutbucket]

it is.  Was just typing the reply below..

I do not normalize, I add gain up to -0.2dB to my final wav 24/48 files.
[..snip]

That is normalizing.  What you describe is the process you use to achieve the desired normalization of level.

There are various ways to do it, which including tools that automatically calculate how much gain to add to reach the desired target level.  Complicating things is the often unstated "level of what?".  We're talking about peak levels not being overly low nor exceeding 0db Full Scale.  But level normalization could also be referring to loudness normalization, which is a perceptual measure that relates more to RMS value normalization or some measure other than "peak".

[rocksuitcase]

I do not normalize, I add gain up to -0.2dB to my final wav 24/48 files.

Isn't "adding gain up to X dB" exactly what normalizing is?

YES and NO.  Not "exactly".
Depends on which type of normalization.
https://www.izotope.com/en/learn/audio-normalization.html
 
Peak vs. Loudness Normalization:

    Peak Normalization: Focuses on adjusting the gain to bring the loudest peak of the audio to a target level.

Loudness Normalization: Adjusts the gain to bring the average loudness of the audio to a target level, often measured in LUFS (Loudness Units relative to Full Scale).

So given all that, I prefer to do it "manually" so to speak.