[LAD] New stuff: zita-dpl

[Jörn Nettingsmeier]

On 12/05/2011 02:50 PM, Fons Adriaensen wrote:
> On Mon, Dec 05, 2011 at 12:20:24PM +0100, Jörn Nettingsmeier wrote:

>> * i'm assuming you are looking at the maximum level of all channels, and
>> then apply the same amount of gain reduction to each of them.
>
> Yes.
>
>> certainly
>> the way to go in speaker-based mixes. but as i already mentioned over
>> that coffee at ICSA, do you think it could be useful to add an ambisonic
>> mode which would apply the gain reduction only to the component that's
>> actually over? my hope is that the result is more subtle, because only
>> the source sharpness would change slightly, and with b-format, there is
>> no danger of irritating jumps of the source...
>
> Actually that is not true - sources would move. Imagine a simple WXY
> system. You have a source at 45 degrees and one at 90. The one at
> 90 (say some percussion) makes Y go into limiting. This means the
> one at 45 will move forward.

right, now i see the problem..

> So any gain change would have to affect equally at least all components
> of the same degree. And even that can only be done for a short time,
> as modifying the gain of one degree makes a complete mess of the decoding
> - rE will drop sharply and rV can take on any value, even go 'negative'.
> After at most a few tens of milliseconds the other components would
> have to follow.
>
> So that would require separating transient gain changes from longer
> ones. This would be possible in e.g. a compressor, but for a peak
> limiter (which has to ensure that no samples are over 0dB whatever
> happens, and operates in feed-forward mode) it can become quite
> difficult.
>
> I tried something similar: to make transient gain changes affect only
> the mid and high frequency part of the signal, but I had to abandon
> that idea (at least for now) - it really gets very hairy.
>
> Assuming it could be done, the limiter would have to know which
> channels belong to the same degree. No problem if you only have
> complete sets, but I want to support horizontal only and mixed
> order sets as well. The plugin system in AmbMixer can handle
> this. Others could mayby do it with a few more extensions, but
> those would always imply explicit AMB support by the host.

thanks for elaborating. i agree, it's too hairy to be worth it, and i 
dislike tools that try to be too clever - even if it could be done, i'm 
sure it will confuse some users in corner cases, and the final stage of 
a production is not a good place for interesting surprises...

best,


jörn