i have try "bit compare 2 tracks" in foobar 0.7b9

i have create wav files with and without dsp

the files are the same with croosfader (only one track) , skip silence, volume control (0 db) BUT NOT with advance limiter.

In theory advance limiter not change signal if it is not clipped
then i have reduce volume to - 10 db (even if replaygain is used) but they are differences

note: with soft clipping at - 10 db no differences!

please, can you confirmed this problem?

Edit: same with 0.667

I can confirm it (EAC compare wav tool + Mix Paste on Cool Edit).

well I'm innocent, Peter made the last changes on the monkee limiter, I'll have a look at it and see what's fuxed.

Is normal that advanced limiter does not always prevent clipping ?
I noticed that (at least) one sample is still clipping even if limited....
Is this a bug ?

Here is how to reproduce it:
1) Take liberate.wav sample (from ff123's site)
2) Encode it with --alt-preset 128 --scale 1 (for example)
3) Decode the file with f2k (only advanced limiter DSP activated)
4) Cool Edit stats reveal 208 samples at maximum value.

I've previously posted this issue here but received no response.

samples at max value don't necessarily mean clipping, you need more than one in a row to qualify.

Or you can disk write to float wav with DSPs enabled, and write something to detect |x|>1.0 in the data.

...something like scanning the float WAV for ReplayGain track gain and examining the track peak value (should be 1.0000 or less).

Anyhow, advanced limiter is there to spot clipping in the original sample then create distortion that sounds better than the original clipping. Leaving a tiny bit of clipping (e.g. a couple of samples at 1.0500) may sound better than forcing even more distortion on the rest of the wave. I wouldn't worry too much even if it is retaining some minimal clipping distortion.

the advanced limiter is made to prevent all clipping occurances by successfully detecting clip or near-clip situations and limiting at such events.
Anyway, I'll tweak around with it today and you'd see a better version as soon as next beta is out. That would be in the changelog since this is my last post on these forums.


edit: plural: forums (HA)

Good idea !
With this procedure is easier (no need to write a program) to verify the eventual exixtence of clipping in floating point PCM files. I've noticed that the RG peak is 1 ,000000 so there is no clipping at all. If i understand correctlty, now we can conclude that the 16 bit integer file has some samples at maximum level but this is only due to rounding and not to clipping. Thank you all for clarifying this.

I certainly agree with you. We should always find a compromise, if clipping is not audible there is no good reason to limit the signal more. I've read here a lot of differents opinions about the use of limiting or the use RG peak value to prevent clipping and i think that the result is function of tastes and of music types too.

I'll just make this short since I'm awfully tired. It's a nice idea, but it's not possible to do, this for many reasons, the most fundamenatal being that Replaygain simply isn't accurate enough. The complexity is not satisfactory for such usage you propose.



btw. I can't duplicate these things you claim the limiter does, from what I've seen it performs ideally to what it should
peace out

are you leaving, why?

I don't understand what you are referring to but only ReplayGain talk I see is about peak values. Those are perfectly accurate and anything over 1.0 means clipping.

Why are you leaving the forums?

- Agent 86