And the sampling frequency determines where the antialiasing filter has to be.
Otherwise you'd be saying that it is perfectly fine to use a 48k sample rate with a 35k antialiasing filter.
As it is a 48k SR rate requires a filter no higher than 24k so logically SR determines the possible filter frequency and thus the timing resolution/frequency response.
I would certainly edit that statement if it were in my power to do so. It's correct but incomplete. Perhaps somewhere else on that page is a statement about how noise at the LSB level decorrelates quantization distortion?
No, it does nothing to the peaks (unless something is wrong). The "dynamic range" thing is just like anything else. If you record (with properly set levels, etc) a performance with 115dB dynamic range to a cassette recorder, noise will be maybe 50dB down, and you'll have a hard time hearing any part of the performance that's more than 50dB below the maximum level.
The "magic" part of ADCs and DACs is you can make the audible noise LESS than the usual statistic of number-of-bits-times-6dB to get the noise floor (as in 16 bits x 6dB = the "standard" CD noise floor of 96dB below max).
There's two things that happen:
1. Dither eliminates (okay, technically it decorrelates) distortion of low-level signals in digital systems by adding white noise.
2. Noise shaping is the same as dither except that the noise is filtered so that it's in the least audible band (generally above 10kHz) and thus much harder to hear on playback.
I've always enjoyed reading this - it's not strictly about audio, but I feel it explains the dither thing really well:
http://www.ti.com.cn/cn/lit/an/snoa232/snoa232.pdf
There are also Wikipedia articles on these things:
https://en.wikipedia.org/wiki/Dither
https://en.wikipedia.org/wiki/Noise_shaping
Signed 16-bit numbers use two's complement arithmetic (there's surely a Wikipedia article on this). The range is from -32768 to +32767, so it really is 65,536 discrete levels.
It's not a locked page, so it's within your power. You [as in anyone who can access Wikipedia] can click "edit" and make it say anything you want, but vandalism and such tends to get fixed fast - there are even bots that detect and fix such things.
But Wikipedia is a bit of a strange beast. You can add something that's correct, and it could be taken out by someone else with the explanation "this other article explains that." I usually stick to fixing links and undoing silly vandalism and such.
But maybe here's "this other article that explains that:"
https://en.wikipedia.org/wiki/Delta-sigma_modulation
Last edited:
First of all I love that article. The arguments about why the higher sampling rates are worse are particularly interesting... inaudible high frequency components causing audible distortion artifacts.
If what you mean by "21 bits accuracy" is "21 bits dynamic range", that rings true. It may even be higher. In extremely quiet environments (ie. anechoic chambers) good human ears can detect sounds under 0dB. The upper limit is a bit more vague, since the ear starts distorting / compressing once you get up over 100dB. Plus y'know, hearing damage. On the other hand, good points have been made that we don't have microphones etc. good enough to capture this extreme range, nor do we have environments which can benefit from this dynamic range of sound playback, outside of research institutions.
There are so many better things to worry about than sample rate and bit depth. I mean come on, we are still (predominantly) using 2 speakers to create a soundfield. This is basically insane from any kind of "realism" perspective, which is usually where the "high-res formats" arguments come from.
16 bit audio isn't the bottleneck in audio record/playback systems - the 2 channel system is! That, and transducer technology (speakers and microphones).
Finally, I've yet to see demonstrated that sounds above 22kHz are perceptible in any way.
If what you mean by "21 bits accuracy" is "21 bits dynamic range", that rings true. It may even be higher. In extremely quiet environments (ie. anechoic chambers) good human ears can detect sounds under 0dB. The upper limit is a bit more vague, since the ear starts distorting / compressing once you get up over 100dB. Plus y'know, hearing damage. On the other hand, good points have been made that we don't have microphones etc. good enough to capture this extreme range, nor do we have environments which can benefit from this dynamic range of sound playback, outside of research institutions.
There are so many better things to worry about than sample rate and bit depth. I mean come on, we are still (predominantly) using 2 speakers to create a soundfield. This is basically insane from any kind of "realism" perspective, which is usually where the "high-res formats" arguments come from.
16 bit audio isn't the bottleneck in audio record/playback systems - the 2 channel system is! That, and transducer technology (speakers and microphones).
Finally, I've yet to see demonstrated that sounds above 22kHz are perceptible in any way.
The events I go to have a somewhat older audience. Flashing boobs is NOT what you'd want to see. A few years ago, my buddy Raymond Koonce and I went to a private show by Shawn Phillips (who's been playing with top acts for more than half a century- the guy played on Sgt. Pepper!). After the show, we helped him load his equipment onto his tour bus and I jokingly asked, "Where are the groupies?" Shawn sighed and said, "At my age, what you get are droopies!"
- Status
- Not open for further replies.