Ok. So following the advice from CliveB and Radish, I selected my test track:
My God by Jethro Tull from Aqualung. Track peak is 1.0 but the album as a whole needs a positive replaygain value - always a good sign for me :-) This track appears to be well mastered, sounds good and has a very good dynamic range. Looking in Audacity you can see where it hits 0 - not very often, and only on some snare drum beats. Most of the time the track level is well below this. I used the demo version of the standard Declip tool with default settings and then shoved the original and declipped version into Audio DiffMaker. The Declip demo limits you to 29 seconds so I'd previously extracted a 20-sec wav using Audacity to ensure I got a few of those peaks... Audio Diffmaker comparison was interesting, since it eliminate the 6dB level reduction that Declip introduced. A/B'ing before and after I could hear no difference at all - good! The ADM correlation was > -80dB ( so down to the lowest two bits broadly speaking ). This passes my personal transparency test so I'd be happy to let declip process everything (and declipping as it needed to). I'm not worried about the 6dB (1 bit) insertion loss of the process, especially as Declip can actually convert to 24-bit on output (another story). Anyway, onto Radish's idea. I then used Audacity to increase the level of the original track by 6dB, creating clipping. Note this is brutal hard-limiting at 0dB, not compression!. Then I repeated the above exercise. It's early days for me with the software but this is what I found: 1) I couldn't easily hear the clipping (that I knew was there - I could see it in Audacity) and a/b'ing wasn't dramatic 2) The ADM correlation depth was -45dB which is quite low and indicates that audible changes had been made to the file 3) The difference track (with some mild boost) was obvious - the snare drum beats were now being accompanied by a nasty harsh "crackle" - what you'd expect digital clipping to sound like. My conclusions: 1) It seems to work so far. The declipped version did not have any clipping in it, either visually or audibly (note - I haven't diffed the boosted(clipped)+ then declipped version against the unboosted original yet - that's my next test 2) It's as transparent as I need on non-clipped tracks. 3) you can make it compute internally in 32-bit FP and write its output in 24-bit (like the SB) which means that the level changes are handled in a similar way to SB volume (i.e. very well) 4) Clipping of this sort is not easily audible inside the actual track, because it is masked - in this case by the core snare drum sound. So actually what you are hearing is a snare beat with some odd-order distortion underneath it. I think that's why a quick A/B doesn't often reveal the problem. But... over time your brain must get tired of masking out that almost-subliminal crackle that it knows shouldn't really be there. You might be surprised at just how much clipping of this sort you have to get into before its immediately obvious that the track is distorted in a "fluff on the needle" sort of way... I suggest you have a try. It took me about 30 mins altogether to do this test. See what you think. Phil -- Phil Leigh You want to see the signal path BEFORE it gets onto a CD/vinyl...it ain't what you'd call minimal... SB3 (wired) - TACT 2.2X (Linear PSU) + Good Vibrations S/W - MF Triplethreat(Audiocom full mods) - Linn 5103 - Aktiv 5.1 system (6x LK140's, ESPEK/TRIKAN/KATAN/SEIZMIK 10.5), Townsend Supertweeters, Blue Jeans Digital,Kimber Speaker & Chord Interconnect cables Outdoors: Boom ------------------------------------------------------------------------ Phil Leigh's Profile: http://forums.slimdevices.com/member.php?userid=85 View this thread: http://forums.slimdevices.com/showthread.php?t=57872 _______________________________________________ audiophiles mailing list [email protected] http://lists.slimdevices.com/mailman/listinfo/audiophiles
