Feedback Compressor II As Mix Limiter?

[David Else]

I am using Voxengo Elephant to limit, but not sure i like the sound of its oversampling everything. Your compressor is bit perfect when its not compressing, so its a tempting option

Let us say i have a mix with just a few short transient drum peaks i want to chop off, some maybe as small as they can be, maybe just a sample in length. The old best practice would be to use a sample editor.

I understand it is not a true brickwall limiter, but for my limited example would it work?

I would set the TDR Feedback Compressor II to work on release peak only, max ratio, threshold acting only on peaks, hard knee,... how about attack/release time? At what attack speed would it catch everything with best results... what speed equates to one sample at 44.1, is it able to do this?

FANTASTIC compressor!!

[FabienTDR]

Interesting question.

Keep in mind that the true timing values on the FB Comp are ratio dependent. The indicated value only gives a rough idea of the true speed. In a FB configuration, the true speed varies by substantial amounts depending on the material, especially with a deep "knee" and depending on the active control path and the SC HP setting, too. (btw, the ratio itself is frequency and content dependent, too! This makes any attempt to estimate the true speed very difficult)

And to be exact, the timing value is defined/measured as the time needed to reach 80% of it's input value or recover 20% of the original value after a DC input. That is, the whole thing takes into account that Infinite Impulse Response filters (theoretically) never reach 0% or 100% respectively, they just approach it asymptotically. In other words, you never know with certainty when the timing filter will ever reach it's target value.

Finally, the differential oversampling approach isn't perfect either. While it nicely preserves the original bandwidth, it can also increase the level at Nyquist under certain conditions! These conditions happen when non-"musical" input (unnatural signals such as super high level high freq noise) meets extreme amounts of gain reduction. These problems practically never appear in the context of an audio compressor (because of the low ratio). But they make it practically impossible to deliver anything close to a brick-wall behaviour.

These facts make the FB Comp II a very bad overload protection device. No matter how fast the timing values are set. Of course, the compressor is very fast and can help to reduce overloads. But it really can't prevent them due to its design restrictions.



Limiters need extremely precise and careful design to do their job properly. The task is much more complicated than it seems and it's an even bigger challenge if "sound" matters.

Maybe we'll see a Tokyo Dawn Labs Limiter sooner or later?



I am not trying to demotivate you! Really sorry for the complicated and technical answer. Have you considered using a FB Comp II combined with Vladg's limiter 6 clipper module? Works great.

[David Else]

Thanks for the detailed reply!

I think i will have to stick to Elephant then. I hate digital limiting and loudness, but sometimes its a necessary evil.

[FabienTDR]

Elephant is a solid alternative IMHO.


I personally like to use a compressor with a "conservative" clipper in some cases (an oversampled one of course), especially when facing recordings of "real" instruments.

But it's still a difficult compromise.

[David Else]

Is there much advantage to oversampling when you are only limiting a very few peaks in the whole track?

I see the Vladg's limiter 6 clipper has a similar feature to yours in that it can only oversample the actual limited peaks and leave the uncompressed signal untouched - 'upsampling GR' mode. He does not mention bit perfect as you do, but i assume it is the same principle.

The manual then says I need the 'protection' mode too to prevent overs... not sure how they relate. I just want the cleanest clipping with the very least effect on the rest of the signal. I am wondering non oversampled Elephant at 44.1 or Limiter 6 in 'upsampling GR' mode? mmmm...

Is there a way to compare 2 bits of audio for being bit perfect and show you where the differences were if there were any? Maybe show an overall percentage of difference too... how cool would that be?!

[FabienTDR]

With regard to Limiter 6, Vlad still needs to confirm the assumption, but I'm pretty sure he implemented it in the differential way. Look here in case you're interested: http://www.google.de/patents/US6337999

(however, and as Vlad mentioned, this approach can't deliver true brick-wall behaviour without additional processing)

I definitely recommend an oversampled clipper. One that runs about 16x the audio rate and uses proper re-sampling filters. (Vlad's clipper is fine). Switch off any additional audio rate clipper, just give the clipper 2-3dBs of "space" to handle the clipping more or less musically.

If you're looking for the typical dub-step sound, you will most probably intentionally ask for a naive (and heavily aliasing) clipper instead. It sounds hard, metallic and loud. I don't like it. But for "simple" music that isn't very musical anyway, it truly doesn't matter! (without questioning these genres!)

About comparing files, keep in mind that "percent" isn't really practical in the case of audio signals. These differences usually happen below -100dB (which would be 0.001%). That's a reason why I find so called "null-tests" to be pretty confusing in practice. The "visual" difference is often too small to be "seen", but our ears act on a considerably more detailed scale than our sight does. The only perfect "Null-Test" would be a CRC check. But this is just a true/false thing. If you're open for true nerd activities, check this out: http://stackoverflow.com/questions/8385 ... o-binaries

[xtrax]

deleted