Re: Fwd: [Ardour-Cvs] Ardour source branch meterbridge updated. 3.2-153-gf885622

On 07/06/2013 08:34 PM, Robin Gareus wrote:
> On 07/06/2013 06:27 PM, Jörn Nettingsmeier wrote:
>> hi robin, i'm not currently tracking the meterbridge branch, so i may be
>> talking nonsense, but here goes:
>> peaking should be defined as a configurable number of samples at full
>> scale (or equal to or greater than 1.0f, in the case of floats) , not
>> some arbitary level.
>
>
> Hi Jörn,
>
> Do you have a reference for this *should be* defined as? I don't think
> that makes much sense.
>
> Actually it's rather the other way 'round. Since the digital signal is
> band-limited there can be peaks in the audio even if the digital samples
> are not exceeding 0dbFS [1].

maybe i misunderstood the purpose of this code snippet (like i said, i 
don't know the context). if this is about a peak hold feature, ignore 
me. if it's about peaking in the sense of "clipping", bear with me.

within a digital audio system based on floats, i don't have to care 
about peaks at all. only at the last step before going back to 
fixed-point or analog do i have to adjust.

so all i want to see from a peak indicator is "has my data been clipped 
at some earlier point in the chain?". i don't really care about 
inter-sample peaks at this stage.

fwiw, the current implementation in ardour is not foolproof either:
when i overload my micstasies, the peak indicator will show 0.0dB but 
not turn red, so i've learned to look for the numbers rather than the 
color. of course, it could also be that the preamp does not ever deliver 
fullscale for some reason...

> Either way this is not relevant yet. The meterbridge branch (which will
> be merged back to master soonish) is currently mostly concerned with GUI
> - not the actual calculations.

i see.

> NTL, the level-metering in ardour will see a major rework in the not to
> distant future - and doing that needs a rethink.
>
> Your suggested approach has one major downside though: it cannot be
> efficiently accelerated with SSE2 (thanks to Sampo, the current
> max-peak-hold is accelerated).
>
> Other level-meters standards/algorithms will have the same issue and
> because every port in ardour has a peak-meter and you can have thousands
> of ports this can become a major DSP-load bottleneck.

thousands of ports? never :-D

> Anyway - this needs careful planning.
>
> Thanks for kicking off this discussion..

looking forward to this new code!

btw, i should have written to you about it, but i've been using the 
video timeline for a little project last week (usual workflow: video guy 
decides edits and delivers the final video with rudimentary guide track 
from built-in mics). worked like a charm, thanks a lot!

best,


jörn



-- 
Jörn Nettingsmeier
Lortzingstr. 11, 45128 Essen, Tel. +49 177 7937487

Meister für Veranstaltungstechnik (Bühne/Studio)
Tonmeister VDT

http://stackingdwarves.net