Two versions of same track. Do they measure and sound the same?

[PeterSt]

OK, I'm afraid you have to count me out because I seem to have an issue with the sampling rate I am using (8 x 96 in this case) - and not much time to hunt down the issue.

 

So instead I have this hint (for self and others):

One of the files reaches a maximum decimal value of +32768 when mapped onto 16 bits.

I saw this yesterday and already typed in a post "fishy !".

 

 

6 hours ago, fas42 said:

only thing I saw was that Audacity considered the second, tiny bit louder, version to be clipping at a single point

 

 

[danadam]

8 hours ago, PeterSt said:

One of the files reaches a maximum decimal value of +32768 when mapped onto 16 bits.

I saw this yesterday and already typed in a post "fishy !".

Huh? There is only one sample that reaches full scale and it is the 16'889'889 sample in the left channel of the second file:

]$ hexdump --skip $((44+16889889*6)) --length 6 --format '6/1 "%02x " "\n"' "Melle - Running out of Time2.wav"
00 00 80 18 82 d9

0x80 00 00 is -8'388'608 and it is a perfectly valid value. Mapping to 16 bits gives 0x80 00, which is -32'768, which is also a perfectly valid value.

Obviously it is a peak sample in that file. A corresponding sample in the first file is also a peak sample in that file and it is:

]$ hexdump --skip $((44+16889889*6)) --length 6 --format '6/1 "%02x " "\n"' "Melle - Running out of Time1.wav"
ac 01 80 9a 82 d9

0x80 01 ac is -8'388'180.

[PeterSt]

3 hours ago, danadam said:

Huh?

 

Huh, as in: this is surely no coincidence with the maximum value in the other file at 32766. So it has to be about that - and possibly nothing else.

Impossible to be audible.

[danadam]

49 minutes ago, PeterSt said:

Huh, as in:

My "huh" was mostly about the "+" in your "+32768". First, unless we are talking about some exotic encodings, the limit for positive values in signed 16 bits is +32767. Second, the sample in the file (after converting to 16 bits) is "-32768", which is the limit for negative values.

[Rexp]

FYI I did nothing special to these tracks, just used Audacity 'normalize' to increase the volume in one and 'amplify' in the other. 

[March Audio]

23 minutes ago, Rexp said:

FYI I did nothing special to these tracks, just used Audacity 'normalize' to increase the volume in one and 'amplify' in the other. 

Thanks for the info.

 

If you are normalising tracks I would recommend only going up to about -3dB.  This will avoid the risk of any intersample overs.

[fas42]

47 minutes ago, March Audio said:

Thanks for the info.

 

If you are normalising tracks I would recommend only going up to about -3dB.  This will avoid the risk of any intersample overs.

 

Yes. The other trick I do is to oversample, say by 4x, the track rate - this shows where the bad buggers are ... so, if you want to squeeze max volume out of the track, this will give you a good number to amplify by.

 

1 hour ago, Rexp said:

FYI I did nothing special to these tracks, just used Audacity 'normalize' to increase the volume in one and 'amplify' in the other. 

 

Did you hear any difference?

[Rexp]

24 minutes ago, fas42 said:

 

Yes. The other trick I do is to oversample, say by 4x, the track rate - this shows where the bad buggers are ... so, if you want to squeeze max volume out of the track, this will give you a good number to amplify by.

 

 

Did you hear any difference?

I did, no one else did, the recordings not great though. 

[pkane2001]

On 6/4/2021 at 9:52 PM, Rexp said:

@pkane2001 @March Audio any thoughts? Anyone else care to offer an opinion? 

 

No audible differences. PK Metric applies perceptual weighting to the difference between the two files, measures at -167dBr (rms):

 

Unweighted RMS difference is at -129dBFS.

 

Phase difference:

 

Spectra difference:

 

Spectrum of the difference between the two files:

 

 

Oh, and the files are bit-perfect up to 20 bits, so all the differences are below 20 bits.

 

[idiot_savant]

On 6/6/2021 at 4:29 PM, idiot_savant said:

I reckon one has had dither added possibly

 

do I win a prize?

 

 

 

your friendly neighbourhood idiot

[danadam]

8 hours ago, Rexp said:

FYI I did nothing special to these tracks, just used Audacity 'normalize' to increase the volume in one and 'amplify' in the other. 

So neither of these is the original recording? What was the max level in the original recording then? (or in another words, by how much did you increase the volume?)

[Rexp]

11 hours ago, danadam said:

So neither of these is the original recording? What was the max level in the original recording then? (or in another words, by how much did you increase the volume?)

Original  https://www.dropbox.com/s/quy078zf1dyd8a5/Melle - Running out of Time.flac?dl=0