View Single Post
  #30  
Old 09-02-2018, 09:26 PM
DupleMeter DupleMeter is offline
Registered User
 
Join Date: Dec 2009
Location: Connecticut
Posts: 1,751
Default

Quote:
Originally Posted by rick-slo View Post
DupleMeter,

I would suggest you read "Digital Audio Explained" by Nika Aldrich and some of the myths including the one that sample rate recordings higher than the Nyquist frequency (about 44.1 kilohertz) inherently sound better.

Oversampling is useful during dynamic processing (say compression) of a waveform to prevent aliasing artifacts into the frequency range of human hearing. Depending on what processes Soundcloud goes through to get to the final 128 bit rate mp3 oversampling may not be relevant. I mentioned it because you mentioned it.

Also you are buying into “Mastered for iTunes” marketing. I would suggest a health skepticism. For example read: https://9to5mac.com/2012/02/28/maste...ser-to-the-cd/
So the 9to5 article is concerning itself with sounding closer to a CD release. That's not what we're talking about. He's simply null testing against a CD release. Anyone with even the slightest audio engineering background knows you will never null a PCM file against a lossy compression version. Heck, I can make a CD file fail to null against a copy of itself by adding even the slightest spectral change to one of them. That's just smoke & mirrors. And adding more confusion to this whole thing.

I have read Nika's work. I don't think you fully understand his stance about sampling rates. He does say there are times that lower sampling rates can sound better. But they are specific cases, with specific criteria. He also avoids getting into the math, which leaves out some important facts in exchange for being more accessible. I recommend reading Bob Katz's "Mastering Audio". Which goes into a lot more detail. it will fill in a lot of the holes.

I feel like I want to clarify what I'm saying with a quick visual. So here are 2 mp3s from the same file. The top is from the a 48k version of the file (original recorded sampling rate). The bottom is a conversion to mp3 after resampling to 44.1. You can see there is less info in the 44.1 file. including some areas of frequency totally missing (meaning the mp3 decoder has to interpolate that each time...with varying degrees of success). Now do the same with 96k (or 192k) vs 44.1 and imagine how much more info the mp3 decoder has to work with and not have to magically pull out of algorithm world? Also, keep in mind that with a spectrogram, the brighter the orange the more info is there. So that's significant too...as you will see differences in the 2 mp3s even in that. And this isn't even zoomed in.



I'm really not interested in arguing this. I don't want to get into a thing. It's just not worth it...I'm sure you have other things to do...I know I do. This would probably be best as a conversation over a few beers. My treat if you're ever in the NYC/CT area.
__________________
-Steve

1927 Martin 00-21
1986 Fender Strat
1987 Ibanez RG560
1988 Fender Fretless J Bass
1991 Washburn HB-35s
1995 Taylor 812ce
1996 Taylor 510c (custom)
1996 Taylor 422-R (Limited Edition)
1997 Taylor 810-WMB (Limited Edition)
1998 Taylor 912c (Custom)
2019 Fender Tele
Reply With Quote