Multi band/frequency aware (dynamic-) compressor

[Runoratsu]

Heya everybody!

One feature I still miss on every player I used to this date, and haven't found in rockbox (or as a suggestion in this forum, for that matter), is the ability to use a frequency dependent compressor.

With most players, the louder you turn up the volume and/or the higher a volume peak, the more unbearable the treble/presence gets, sometimes even the whole frequency response gets bent out of shape.
It would be cool to be able to counteract this with a multiband compressor to tame the overzealous frequencies, or at least a compressor with a user definable attack frequency and a nice rolloff to lower frequencies…


I don't know if something like this is even possible with the current players' hardware capabilities, and I sure don't have the ability to code it… but I'd be quite happy if someone else would try ^.^

[saratoga]

With most players, the louder you turn up the volume and/or the higher a volume peak, the more unbearable the treble/presence gets, sometimes even the whole frequency response gets bent out of shape.

This is just distortion from either digital clipping (if your player's software is badly designed) or distortion in your headphones.  I don't think a DSP effect trying to work around it makes much sense, nor would it work very well.

[Runoratsu]

If nothing else, it'd be a nice tool to counteract the volume dependent weighted frequency perception in human hearing… and I can only say for myself that I find treble and upper mids often unbearable even at volumes where my stereo DEFINITELY doesn't clip or distort yet, that's why I usually use a multi band compressor on the computer that's serving it, and limit their volume that way.

I got a 215W/channel stereo Yamaha power amp driving old huge Cantons that can take >180W RMS, that's SO far above levels I find comfortable I doubt they'll EVER clip when I'm listening to them.

On headphones, where one is even more trapped in with the sound it only gets worse IMHO.

Btw, I listen to the rockboxed iPod via Beyerdynamic headphones and a powerful headphone amp (FiiO E11) that's not even halfway to full volume, so I doubt they clip either. My files are also level adjusted so that they should not clip, and the iPod is working at about 60-70% volume.

Of course that does not mean and should not be read as me demanding the suggestion I made to be implemented! I just wanted to explain my rationale behind it again. =) I'm pretty happy with all the functionality Rockbox delivers as is!

[Llorean]

Are you sure you don't want the multi-band equalizer to compensate for the weighted frequency compression?

I would imagine between the compressor and the multi-band equalizer you'll be able to address what you're describing.

[Runoratsu]

Are you sure you don't want the multi-band equalizer to compensate for the weighted frequency compression?

I would imagine between the compressor and the multi-band equalizer you'll be able to address what you're describing.

I use the paramEQ already, and I love it!

Yet, this can only correct the frequency response statically, but (as you probably know anyway) the human hearing gets "flatter" with rising SPL, so that the sound signature of many player/headphone combos sounds awesome at lower levels, to which they are adjusted, but at higher levels, the boosted treble is louder than would be necessary for the now flatter frequency perception curve.
Bass, too, but as bass is generally much easier on the ears, that's less of a problem. And I think many amps compress the bass "by design", because they just can't deliver the necessary power to sustain it at higher levels.

[Chronon]

I usually listen to my music at pretty close to the same level.  Can't you set the EQ levels for that volume?  Anyway, it doesn't seem like frequency dependent compressor is what you want but rather volume dependent EQ.

[Runoratsu]

Nonono, I also have a preferred level I listen & EQ to! The problem isn't differen listening level, but rather the songs' dynamics. If you have rather dynamically compressed songs it's no problem, but with very dynamic songs you get level peaks all over the spectrum, and with your hearing being more sensitive to treble and bass at higher volumes relative to the sensitivity at the "usual level" of the song, these peaks are hurtful in the treble range, while bass and mids are less of a problem.
So I'd like to compress those treble peaks, while still allowing the whole dynamic range in mids and bass.

[saratoga]

I think you're over thinking this problem.  Peaks far below max generally aren't heard since they're masked.  Furthermore, the shape of the equal loudness contours changes only a little with intensity at typical levels.   I suspect that if you took the time to actually implement the processing you're proposing, you'd find that it made very little difference because the change in per frequency gain would be nearly identical over the entire dynamic range of almost any track you have.