Ah, my bad. You are right, sampling at exactly twice does "kinda" violate nyquist.
But sampling just above produces amplitude modulation. No violation but inaccurate reconstruction.
Jn
When I used "complicated" I meant to boil down a common thrust in way too many of these arguments: music is constantly changing; static measurements or sampling or some other evil force can't capture it. Etcetera, etcetera, as Yul B put it.
The exceptions to sampling seem to always involve a violation of Nyquist, but they go on year after year. Folks with a real interest would be more fruitful if they were attacking issues of bandlimiting, IMO of course. Nyquist is settled law.
All good fortune,
Chris
It is unworthy of you, sir.
tip: Explain to me how you can sample half a period with sampling at twice its frequency. With a little drawing, for those who, like me, have, um, difficulty understanding ?
An please, be honest for a while, don't draw your sampling point in the middle.
Last edited:
No one.. not me. is suggesting violating Nyquist. A higher sampling rate is needed and for good reason.
THx-RNMarsh
We're talking about the limits of the theory so you mean 22.049999kHz is OK? It's a continuum, as you approach the Nyquist limit it takes forever to reconstruct the signal. This is not of much practical interest in audio IMO. There is little or no information at frequencies near or beyond 22.05kHz and there are several controlled studies that show the brickwall filtering is not audible.
Let's scale it up to Red Book in order to use existing commonly available data. Harmonic distortion at full scale 20kHz or nearabouts - very small distortion numbers - and the same for 19/20 IMD.
As strange and unbelievable as it sounds, perfect (ideally, of course) sampling and reconstruction of a bandlimited and dithered signal is possible, as long as Nyquist isn't violated. I didn't believe it myself until I'd drawn it all out with pencil and paper.
Always the best,
Chris
As strange and unbelievable as it sounds, perfect (ideally, of course) sampling and reconstruction of a bandlimited and dithered signal is possible, as long as Nyquist isn't violated. I didn't believe it myself until I'd drawn it all out with pencil and paper.
Always the best,
Chris
I just did please try to understand the concept of limits and numbers that are expressed as the sum of an infinite series, etc. or values that are expressed as integrals from plus to minus infinity.
I guess we just take your word?
Heck, I could do it with an excel sheet and provide proof.
Jn
SW and JN: I guess I'm just not following you here (what a surprise!). Are you really saying that something bad is happening that doesn't show on a distortion analyzer?
Much thanks, as always,
Chris
Much thanks, as always,
Chris
You overlook the fact that the ear does not work like a microphone followed by a brick wall filter. And I thank you very much for examining the waveform of the impact of a drummer stick on the bell of a cymbal.
There are a few reasons why the music industry is increasingly moving towards higher sampling frequencies. And it's not just because it looks prettier on a datasheet.
Examining the arguments of each other with the alder of his sympathies is not a very scientific approach. Refer to a theory while neglecting the context in which it is called either.
Last edited:
You go ahead and try, perhaps that exercise will show where your confusion lies. You may find you need an infinite time to build the spreadsheet.
Think of soundstage imaging. How does one guarantee accurate inter channel temporal information at the 2 to 5 microsecond level when sampling at 22 microsecond intervals.
Jn
Nothing to do with distortion. There is enough compute power available now to take an entire 3min song and FFT filter it at 22.05kHz. The transform is completely reversible and nothing is lost.
You provided no proof, no graphs, no visuals.
When I claimed a shorting ring lowered inductance and increased resistive losses, I lathed a brass, aluminum, and steel ring, measured the impact on a voice coil, then slit the ring to measure bulk eddy losses.
Actual measurements, actual build of hardware, proof. That is how I roll.
I would hope for the same from you.
I am not confused by the theory...just by your statements.
Jn
What is “inter channel temporal information”, who says it needs to be between 2uS and 5uS, and how is this related to the sampling period.
Are there commercial products out there that reconstruct a song using a 3 minute window?
Whoa, I guess I've been hibernating.
Jn
Humans are capable of 2 to 5 uSec interchannel discernment for localization..i.e., soundstage.
Actually, 1.2 uSec tested, but that was dithered.
Jn
Ps.. I've pointed that out on this thread for 15 Effin years now... Really?
- Status
- Not open for further replies.