48000 to 44100 conversion - what is happening?




I’ve recently been converting 48000 FLAC files to 44100 WAVE files for CD burning.

I’ve noticed one particular 48000 FLAC file when opened in Gold Wave has a “final peak” of :

L -0.09db
R -1.91db

When I use DBPower Amp to convert it from 48000 to 44100 it’s “final peak” changes to :

L -0.02
R -1.93
(the same happens when burning the 48000 FLAC to CD using Nero).

I noticed if I merge a few of the 48000FLAC files and convert them at the same time in Gold Wave from 48000 to 44100 and split the files back into their individual files most files keep the same “final peak” as the master FLACS but the particular file mentioned above changes to :

L -0.16
R -2.00

Does anybody have any idea why this is giving a different “final peak” reading to the DBPowerAmp conversion from 48000 to 44100 (which is the same result burning the original 48000 FLAC file to CD using Nero gives).

I know my ear isn’t really going to detect a 0.14db difference but I’m trying to understand what is technically causing this difference ?

Should it remain the same ?



To me it looks like an artifact from resampling. Look at it this way, the FLAC is cut into 48,000 pieces but the WAV is only looking at 44,100 of these. To further complicate matters, the 44,100 samples are samples of a reconstructed waveform. There is no reason the peaks have to be the same. Now if you had a 44,100 sample FLAC and a WAV made from it had a difference, that would seem unusual.