Yes, of course the signal out of the DAC is purely analog. The signal is referenced to 0dBFS on the digital side and also something like dBu or dBv on the analog side (although it varies from part to part). I should have been clearer in stating this. :)
But the main point is that there are signals beyond all 1s in the data sent to the DAC if clipping exists. This is confusing to many because the digital math as commonly taught typically only refers to 'inside the box'. I leave this exception out of any workshop or uni course because of the confusion, but it should be known more widely.
Sounds like the zexy [limiter~] is better than many commercial plugins that don't upsample! As we know, the hard clip can create some very nasty aliasing artifacts and when done in the box there is no way to undo that damage.
Best practice is still to keep the peaks away from full scale even if using a very good DAC that can del with clipping well.
On Wed, Jan 1, 2014 at 2:03 PM, IOhannes m zmölnig zmoelnig@iem.at wrote:
On 2014-01-01 19:50, Chris Clepper wrote:
Nope, the DAC can freely construct intersample peaks as it sees fit and those can easily exceed 0 dBFS. It has been common practice in the industry for more than a decade to reconstruct clipped samples well
above 0
dBFS - partially to make up for shitty mixing and mastering prevalent in music, and also because it's the right way to do it.
+1
nevertheless you cannot send digital values to the DAC that exceed 0dBFS. the intersample peaks are *purely* analog. "FS" stands for "full scale" and refers to the full range of the digianl fix-point values. thus - by definition - 0dBFS only refers to digital values, and can never be exceeded. however on the analog side, the nominal 0dB can easily be exceeded in the reconstruction.
btw, zexy's [limiter~] tries to take intersample peaks into account by upsampling the signal prior to limiting,... so if you use it to limit between -1..+1, then the reconstructed analog signal should not exceed the nominal analog output range. (for practical reasons, upsampling is limited, so in some borderline cases you could still construct a signal that exceeds the 0dB analog)
gfmadsr IOhannes
Pd-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list