- Joined
- Feb 23, 2011
- Posts
- 16,414
- Likes
- 3,088
Hypothesis: If HD music sampling rates (> 48 kHz) make a difference in audio quality, then downsampling HD music with signals above maximum frequency obtained from the Nyquist frequency defined by the Red Book standard (22.05 kHz) would degrade the audio quality because the natural ultrasonic frequencies that only HD sampling rates could capture would be eliminated.
The problem with that theory is that tests by the AES indicate that frequencies outside the range of 16/44.1 recording add absolutely nothing to the perception of sound quality of music. If we can't hear them, they don't matter to anyone but bats.
However a high sampling rate can wreak havoc in some home audio components that aren't designed to deal with super audible frequencies, creating artifacting and distortion down in the audible spectrum. In these cases, having the super audible frequencies in the recording makes it poorer quality, not better.
So basically, it's a no win situation. If the inaudible frequencies are there, it can only hurt the sound quality. It can't improve it.
Yes, but the AES papers don't analyze the music in a spectrogram before testing, or if they did, none of the papers I have seen do so. For all we know, there could be zero ultrasonic frequencies to begin with, so higher sampling rates would be absolutely useless with regard to my hypothesis. I specifically did a test on that song exactly because it has natural ultrasonic frequencies from cymbal hits. Some percussion music, such as gamelan, are very well-known to naturally produce ultrasonic frequencies. There are a few published papers about this, and likewise for ultrasonic content being able to be detected by humans. I don't believe any of the AES papers specifically target this music, where higher sampling rates would capture those signals. Instead, most of the papers I've seen just use a DVD-A, or the SACD, and that adds an additional source of error, as in the transport, not to mention the unknown variable of how the audio was mastered for those formats.
As for playback, yes you are correct about home audio components not being designed to handle supersonic frequencies.
This paper suggests recording in HD formats produces sound that is more like the live recording, but playback on loudspeakers with ultrasonic frequencies is harmful to that.
http://www.aes.org/e-lib/browse.cfm?elib=13967
The papers published from AES, although they're scientific papers, they don't list their impact factor. Thus, I often don't take these papers very seriously since I don't know how useful it is to the scientific community.
On the other hand, that paper used speakers with supertweeters, which are not used in headphone assemblies (to my knowledge). The HiFiMAN HE-560, although a useless value without standard deviations, is rated to produce sounds up to 50 kHz. I'm assuming any decent headphone amplifier can handle ultrasonic frequencies with linearity as Tyll measured from Innerfidelity. Keep in mind that the maximum signal from a 96 kHz sampling rate is 48 kHz. Many amplifiers measured there showed a linear frequency response from that frequency area.