Forums dot lukpac.org

I started the same thread at sh.tv, but I could use your experience as well.

I tried to compare the song "Peg" from my German unremastered Aja CD against a file sent by forum member DrJ from his US unremastered CD. I was always convinced that they were completely identical, save for some different leading lengths. However, there was this minimal EAC peak level difference: 88.7 versus 88.9.

Previously, I had just tested samples of 30 seconds of length or 1 minute. Once I had trimmed the leading lengths, inverted one file and pasted it over the other, I had always gotten an optical zero line, which means that they cancelled out completely. Or so I thought.

This time, DrJ sent me the complete song Peg. As usual, after trimming the beginnings, inverting one file and pasting, the resulting file looked as if it were totally silent. However, that was not the case. The peak level of that resulting "difference" file is 0.7%, and the music is still audible, however at a very low level.

If these two were from different masterings, they should not stay in synch, nor should the resulting "difference" file be of souch low volume.

If these two were from the same mastering, they should cancel out completely, not approximately.

I did the same attempt after normalizing the two files, and the result was similar: A "difference" file with very low volume music, peaking at 0.7%.

By the way, here is the "difference file":
http://download.yousendit.com/FA11DE187F0F6856

Two things:

1) I've seen this before. Same digital source, slightly different volume. IIRC, this is the case with the US and Canadian MBBB CDs. How does it happen? No idea.

2) Is it possible he has normalization set in EAC? EAC can automatically adjust the level of rips, which could explain this.

Even more strange is the case when two CDs don't stay in digital sync, yet the drift between them is constant, and they *will* cancel at a point when properly lined up. As if there was a time stretch of some sort. I'm thinking of the original and BMG copies of Who's Next.

Must be an MCA thing.

Could also be some slight futzing at the pressing plant level.

lukpac wrote:Two things:

2) Is it possible he has normalization set in EAC? EAC can automatically adjust the level of rips, which could explain this.

I tested this without and with normalization. The result was similar. In other words, it is not just a volume difference. His file was not normalized, it had the characteristic "88.7%" peak level, mine had "88.9%".

Let's say, someone futzed with the digital master, i.e. added some digital eq...wouldn't that always create a much bigger difference? If not for eq, what else can be done to a digital master which could produce those minimal differences?

The difference is not random noise.

You're right, it's not just a level difference. If it was, your null file (I just listened) would sound normal, just at a very low level.

Can't say I know offhand how/why that would have happened. Generally, differences in EQ would be much larger, yes.

Do a spectral profile comparison. If it's simply a level difference, the result will be a flat line above or below zero. If it's likely the same (digital?) master with different EQ, there will be a smooth curve.

comparison is done like so
1) do 'frequency analysis ' in cool edit/audition for version 1
2) click 'save to clipbaord' (this saves the xy values, not the image)
3) paste result to excel
4) repeat for version 2
5) in another set of columns subtract each track of 2 from the equivalent track of 1
6) graph result

krabapple,

I don't have either cool edit or audition.

But if it were simply a level difference, the two would cancel out completely if they were normalized before the null test, right? They did not.

They are definitely from the same digital transfer since they stay perfectly in synch.

Could you please check the two samples out? I already edited the beginnings so that they are in perfect synch. No normalization or inversion applied yet.
http://download.yousendit.com/6CCA2EE304F5D516
http://download.yousendit.com/0837EDE653CAD089

To within one decimal place, the average levels are the same for both snippets at all frequencies checked (from 20-20kHz, sampled approx every 1.5 Hz). In other words, the difference reads 0.0 dB at all frequencies.

The overall peak and average levels are also the same to one decimal place. However, the right channal peak (german) is -2.57 dB while it's -2.58 dB in the US version. But no one's going to hear a 0.01 dB level difference.

When I invert/mixpaste the two snippets, there are four or five tiny 'glitches' in the flat line, but I can't hear anything.

krabapple wrote:When I invert/mixpaste the two snippets, there are four or five tiny 'glitches' in the flat line, but I can't hear anything.

Thank you for checking it out.

If you turn up the volume of the null inversion result (i.e. normalize), you should hear something. It only looks like a flat line because the volume is so low (EAC says "Peak level 0.7%"). There is still music to be heard.

What is your educated guess as to what happened here?

Yes, if I zoom the 'flat' line way up a waveform appears.
I can faintly hear the music if I jack the volume way up. THe 'glitches' are kick drum hits. They 'peak' at around -46 dB, which gives you an idea of how faint this trace is.

No idea how this came about other than to say it;s a tiny level difference that you'd only 'hear' by a difference extraction, but I haven't given it terribly much thought either.

Thanks anyway. I will continue to call both of the CDs "Steve's mastering", even if they are not 100% identical.