SoylentNews is people
SoylentNews
from the some-art,-some-science,-and-a-whole-lotta-black-magic dept.
According to many Metallica devotees, the official version of the band's 2008 record Death Magnetic is not the one worth listening to. Upon the album's release, fan forums exploded in disgust, choked with complaints that the songs sounded shrill, distorted, ear-splitting. These listeners liked the music and the songwriting, but everything was so loud they couldn't really hear anything. There was no nuance. Their ears hurt. And these are Metallica fans—people ostensibly undeterred by extremity. But this was too much.
The consensus seemed to be that Death Magnetic was a good record that sounded like shit. That the whole thing was drastically over-compressed, eliminating any sort of dynamic range. That it had been ruined in mastering. Eventually, more than 12,000 fans signed a petition in protest of the "unlistenable" product, and a mass mail-back-a-thon of CDs commenced. The whole episode provoked a series of questions, not just about what had gone wrong with Death Magnetic but about the craft in question: What is mastering, exactly? How does it work? Beyond the engineers themselves, almost no one seems to know.
An article on sound engineering, but the real question is, people listened to Metallica after 2000?
(Score: 3, Interesting) by Alfred on Wednesday May 25 2016, @02:04PM
Nyquist is about being able to detect a signal at all. Nyquist says you need double the sample rate of the frequency of the signal to be detected. Everyone knows that.
What is usually skipped over:
*The phase difference between the signal and the sampling will directly affect the amplitude of the measurement. An exactly out of phase sampling will give you 0 amplitude or read as no signal when there could be a very large signal.
*For reproduction, a nyquist sampling of a signal can playback the same if the original signal was a square, sine or triangle input. This loses the musical nature of the signal because those sure as hell sound different in real life.
*Nyquist is F*2, you will also have signals at F+1, F+2, F+3...those signals will still show up and be manifest in the measurements. The beats phenomenon will be present but you are able to determine the presence of a frequency beyond what Nyquist says is possible. F-1, F-2, F-3... signals which are within the Nyquist rate will also be detectable but also with the same beats problems as the F+1 series. Though Nyqusit says those frequencies are detectable they are not detectable consistently well. The best signal capture is having a sample rate several times higher than the frequency to be measured, then you can have an idea of the shape of the wave too. Not always important but it is impotent to music.
*Nyquist is theoretically correct for detecting signals but it is also theoretically horrible for music signals. If all you have is one clean predictable signal then Nyquist is useful for knowing how crappy your gear can be and still kinda work. Nyquist is not so useful for many signals near the same frequency, or when there are overtones above the fundamental, or just many different frequencies at the same time.
Other comments:
*Internal processing of a signal in a mixer or DAW or other should always be several bits greater than the AD or DA conversion depth. Early digital reverb effects units suffered from not having enough internal resolution when the signal faded out or was otherwise quieter. The reverb tails would sound like they had a noise gate on them.
*Yes HD audio is a scam. A recording will benefit more from good mic placement upfront than with bit depth added later. Mic placement should not be a problem in a pro studio though.
*"Safely filter out"?? Mybe you can restate but those aren't signal I consider unsafe. I don't consider filtering the signals an unsafe act either. I think just about all AD converters have an analog filter stage before the conversion.
(Score: 3, Informative) by Scruffy Beard 2 on Wednesday May 25 2016, @03:24PM
Nyquist's theorem assumes infinite samples and an accurate clock. As long as the signal is not at exactly at 1/2 of the sampling rate, the phase will drift.
Every wave can be broken down into frequency components (with a Fourrier Transform) that are pure sine waves. A square wave is a sine wave with odd harmonics, for example. Nyquist's theorem requires that the input signal be bandwidth limited. That means that for example, any square wave will have rounded corners as the higher frequency harmonics are filtered out.
Yes, you are supposed to filter those out with a low-pass filter. That may be what you are getting at in your overall comment. No (physical) filter is perfect, so there is some benefit to a higher sample rate: just to make sure any unfiltered higher frequency components don't cause the problems you mention.
Maybe I misunderstand the design of "HD audio" equipment. If you can reproduce frequencies up to 96Hz, you can almost fit an FM stereo broadcast in the pass-band. A wider pass-band means that you are processing information that can't be heard anyway. If ultrasonics get into the signal, it may cause problems with equipment not designed to handle it. It can also be used for "traitor tracing" by embedding extra information during playback.
If "HD audio" is still using a pass-band of 20Hz-20kHz, then I suppose to only real problem is the waste of bandwidth (which may actually be a feature if you want to discourage lossless copying).
BTW, I agree extra bit depth is beneficial for processing.