Because the drivers are not coincident and the larger the source the further the nearfield effects will be present.
You can fix one single point in space, but the correction will be significant which will mean it is extremely position dependent and even the distance between your ears will be a problem.
Exactly, if the defect is LTI then it can be completely corrected and the further the defect moves away from that the less successfully it can be corrected. Trying to correct large sources widely separated in the nearfield is not on the successful end of correction.
No, to represent a 20khz sine, you can do this within a 44.1 or 48khz bandwidth, using 96khz does not represent the 20khz sine any more accurately. Not ... one ... iota.
It is time for you to start using the question mark (i.e. the "?"). The conversation will be much better if you indicate when you are sure about something and in fact are doing a statement or when you are guessing.
Your "questionstatement" posting style is cumbersome as I see it
//
Looks more accurate to me. Something else must be going on, on the digital side of this...?
According to what I just dug up "Therefore, we need to measure the wave at least two times per full cycle to accurately capture its frequency. ".... Capturing a generalized frequency and capturing an exact waveform are not he same thing. How does it know whats happening in between the samples?
Attachments
Last edited:
yep, you are misunderstanding the way sampling and the sampling theorem itself works, but dont worry, it's a very popular misconception . Let me search for a decent youtube or reference to do a better job of explaining than I have the time, or ability to do. The Nyquist sampling theorem will get you on your way, but there was a very good link posted some years ago that did a very good job of explaining the subject.
a sine is perfectly represented with just 2 points, one at the top and one at the bottom. we need JUST more than double the sample rate vs the max FR we want to sample in order to get a PERFECT result, remember we are talking about digital representation here. The discrete samples are interpolated between (over time), (sort of like a Spline in 2D and 3D modelling) rather than a straight line being drawn between them.
. Let me search for a decent youtube or reference to do a better job of explaining than I have the time, or ability to do. The Nyquist sampling theorem will get you on your way, but there was a very good link posted some years ago that did a very good job of explaining the subject.a sine is perfectly represented with just 2 points, one at the top and one at the bottom. we need JUST more than double the sample rate vs the max FR we want to sample in order to get a PERFECT result, remember we are talking about digital representation here. The discrete samples are interpolated between (over time), (sort of like a Spline in 2D and 3D modelling) rather than a straight line being drawn between them.
The classic myth, explained in Sound on sound article
Reconstruction filter is able to convert the "stairstep" visualization of digital data back to "analog line"!
Reconstruction filter is able to convert the "stairstep" visualization of digital data back to "analog line"!
Last edited:
Yep, the way its drawn by some, with the lines in between (which are utterly meaningless and misleading) went a long way to causing this broad misunderstanding, even by people interested in audio. I would have expected it to have been cleared up a bit by now, almost a decade after that youtube vid was made though ...
the part where he describes how only a single frequency will pass through all of the points is important and relates to what I was trying to tell you. the points/samples of a 20khz wave, will only superimpose on a 20khz sinewave.
the part where he describes how only a single frequency will pass through all of the points is important and relates to what I was trying to tell you. the points/samples of a 20khz wave, will only superimpose on a 20khz sinewave.
Last edited:
No, it is not. Nyquist theorem inequality is strict and this requirement is not optional.
yes, OK, an infinitesimal amount higher than 2. or any amount higher than 2. so 2.00000000001 would be ok. my apologies for the oversimplification and I agree its an important distinction, I just didnt want to get into that before the basics were understood.
hmm, still not sure i'm describing that correctly. More easily, to represent 10000Hz you cant do it with 20000hz sampling, it would need to be 20001hz. Does that work?
hmm, still not sure i'm describing that correctly. More easily, to represent 10000Hz you cant do it with 20000hz sampling, it would need to be 20001hz. Does that work?
Last edited:
If we are talking about stereo recording of live music, then the sampling period of 48 kHz turns out to be longer than the ability of the human brain to detect the difference in the time of arrival of a sound wave between the right and left ears. Our hearing aids have surprisingly high spatial resolution in horizontal plane.
Not sure what that has to do with the current topic? or the quoted text?
Ah, it seems you are just providing a reason not to go to 96khz audio? wasnt really the topic at hand and I would probably agree with 48khz, but maybe not 44.1. 96 has become standard enough across the board that I would just leave it there and not argue we need to go back to 48
Ah, it seems you are just providing a reason not to go to 96khz audio? wasnt really the topic at hand and I would probably agree with 48khz, but maybe not 44.1. 96 has become standard enough across the board that I would just leave it there and not argue we need to go back to 48
Last edited:
no, a sine wave is not perfectly represented by two points.
One would end up something like a triangle wave, although it kind of depends at what moment in time the signal is being sampled.
In theory one could sample exactly in the nulls for example.
Although from a sampling point of view, music is pretty random in nature.
Nyquist only tells us the absolute bare minimum to work.
But basically only when the signals are in perfect sync
(something they don't always tell at school)
That being said, the audio information around 20kHz is extremely low.
Also the natural low pass roll of of the human ears will make everything above 8-10kHz basically a sine wave, since higher harmonics are being attenuated drastically.
Frequencies that high are very directional, moving your head, listening or loudspeaker position a few centimeters will already affect it significantly.
So in the end there are so many more issues at play at these higher frequencies that a lot of other issues are heavily masked.
In the end, like a lot of other psycho-acoustic effects and artifacts, are extremely hard or sometimes even totally impossible to "un-hear" for the very well trained listeners.
Embrace the evolution of your brain for that.
One would end up something like a triangle wave, although it kind of depends at what moment in time the signal is being sampled.
In theory one could sample exactly in the nulls for example.
Although from a sampling point of view, music is pretty random in nature.
Nyquist only tells us the absolute bare minimum to work.
But basically only when the signals are in perfect sync
(something they don't always tell at school)
That being said, the audio information around 20kHz is extremely low.
Also the natural low pass roll of of the human ears will make everything above 8-10kHz basically a sine wave, since higher harmonics are being attenuated drastically.
Frequencies that high are very directional, moving your head, listening or loudspeaker position a few centimeters will already affect it significantly.
So in the end there are so many more issues at play at these higher frequencies that a lot of other issues are heavily masked.
In the end, like a lot of other psycho-acoustic effects and artifacts, are extremely hard or sometimes even totally impossible to "un-hear" for the very well trained listeners.
Embrace the evolution of your brain for that.