Quote:
Originally Posted by MattGray  Well there are other reasons why I'm reluctant to adjust them all destructively. There is 3 different mixes per converter file & some converters alter the levels differently from one piece of music to the next even though they were all run directly through the converter as one continuous audio file. So for example one sample might have the same peak level as the original digital file but be hotter by 0.1dB RMS but then in a different piece of music the same converter file might might be down -0.2dB in it's peak level but is still only 0.1dB RMS louder. Then in the 3rd piece of music it might be down -0.1dB in Peak levels but up +0.2dB in RMS levels. Should I ignore peak values & just change RMS values per piece of music?
Also if I lower levels should I be adding 24bit TPDF dither? Wouldn't adding it or not adding it slightly change the sound?
You can see what I mean when you open this can of worms... what is the best way to match levels in a case like this? Obviously for the file that's 2.1dB in difference it needs to be done but with small differences of 0.1 should we bother or are we going to do more harm than good?
Just saying... |
I understand; and and not changing the different songs' respective levels makes much sense indeed.
Also, my suggestion was with regards to high level differences. When we're talking about 0.1 or 0.2 dB, that's obviously very close anyway, so I guess I've been actually raising a non-issue. Sorry for that. I admit I was reacting to the figure of 2.1dB.
For academic / geek points only, I'd like to reply to some things you mentioned:
I'd say that one way or another, whether samples are matched by peak or rms measurement, by ear, or even not matched at all (beyond the unit's initial calibration), it's an active decision made, an influence imposed on the test. I.e. I don't think the can of worms is - or can be - sidestepped at all.
The reason I suggested matched files (as the default download) is that I believe them to produce a better comparison setup than random / non-matched samples. I.e. the benefit of matched levels will significantly outweigh the (imo purely academic) degradation through a simple level change (in 24 bit, be it dithered or not). Or put another way: I believe it's fair to say that unmatched levels (if there are significant differences) will very likely affect the outcome of the comparison considerably, whereas a properly done gain change will very likely not. Quote:
Originally Posted by MattGray Ok here is what I've done... I've used iZotope Ozone with 24bit TPDF & gain only (no other modules active) to raise the level of 'Converter File F' by 2.1dB destructively.
All the others need the gain adjusted as follows if you look at RMS values only... |
Awesome, thanks! Can't wait to be done with work and have a listen
Quote:
Originally Posted by macc removing another argument for the inevitable onslaught of 'this test is flawed' posts  |
Hope my posts weren't understood to mean that
