Windows vs Linux: round 2 (Off Topic)

by PeterSt. , Netherlands, Monday, August 13, 2007, 13:10 (6094 days ago) @ PeterSt.

Just some more comments towards Klaus :

When indeed you applied to 48K upsampled material the poorer the SRC program, the more aliasing you'll have. It is difficult to explain what happens without showing it on a screen, but this is "capable" of creating frequencies at volume levels of the exact same volume of the original (the base where these "harmonics" come from). So, a frequency of, say, 18K at digital level 20000, will incur for a 20Hz frequency with level of 20000. The 20Hz frequency is completely fake, and is audible at the same level obviously.
This would explain the sub woofer honk with no further normal bass audible !!
Btw, it would workout the other way just the same : an original frequency at 20Hz may incur for an 18KHz frequency of the same level.

Since these aliases come from squarish sound, and are squares themselves, the energy of them is already uplifted because of them being squares (the math upon sines versus squares).

The aliases will be all over the place, and individually they harm, but possibly inaudible. Where it goes wrong is when aliases meet. So, there are a few special (or prone to this) frequencies where all aliases (which create aliases by themselves !!) meet. This typically happens at the beginning and at the end of the "official" spectrum, with 22050 and 0 being mirrors. Remember, these are INSIDE mirrors, and not outside as the aliasing we normally talk about. Outside mirroring happens with an 18K frequency that occurs at 26100.

If I must guess this gives the lacking mid with the Django (which really would be over expressed bass and high aliases and which is how it sounded), and without the Django all shifted more to the outsides of the audible spectrum, that possibly being incurred by loosing the 2 bits of resolution in that setup (I just don't know what would be the result to aliasing and less resolution).

If you didn't apply upsampling, this all doesn't count as a relative difference (!!). It would be there anyway because of the nos-DAC principle but less, and probably more randomly organized (less chance that aliases meet).

Just trying to find explanations ...

Peter

Tags:
0