You may need to re-view the video ... You are stillloking at graphs ...That may not change your point of view if you make of a it a belief. Oh Well ....
New Xiph.Org video: "Digital Show & Tell"
- Thread starter xiphmont
- Start date
-
- Tags
- digitalaudio objectivism xiph
You may need to re-view the video ... You are stillloking at graphs ...That may not change your point of view if you make of a it a belief. Oh Well ....
..But when the the signal is converted back to analog the gaps are filled in. The question is how acurrate is the resonstruction. Nyquist tells us the reconstruction is perfect when the sampling rate is at twice the badwidth.
So there are no gaps after the digital signal is converted back to analog.
Intuitvely we would think the greater the sampling rate the more accurate the reconstruction. We would have to prove Nyquist wrong.
So there are no gaps after the digital signal is converted back to analog.
Intuitvely we would think the greater the sampling rate the more accurate the reconstruction. We would have to prove Nyquist wrong.
There aren't any 'gaps' - that's an illusion. Its sampled data system - like the contents of a spreadsheet showing (say) unemployment data, month by month. To talk about 'gaps' in the sampled data for audio is like saying that there's a gap between January and February in the unemployment data.
Please correct me if I've gotten this wrong. There are no gaps because they are "filled". More samples however DOES improve the accuracy of the pdf in determining what goes in there (noise).
More samples means higher SR - more bandwidth but no better capture of the audio band. What's pdf?
It's not quite right to say they're filled-- discrete time is a different mathematical domain where time itself only exists in discrete steps. This is subtle, tricky, and requires some abstract thinking to grok. I certainly had trouble with it back when I was a student. (I have to fess up that one of the OpenCourseware classes I linked in the vid, 6.003, I took in person and I got a C at the time. It nearly sunk my chances of grad school ;-)
More samples only increases the channel bandwidth. It doesn't change representational accuracy at all. The only things that affect representational accuracy are bit depth (currently topping out at about 21 bits, any more does not help) and the quality of the ADC/DAC implementation.
Good luck with that.
As I've explained countless times, if the "gaps" were not reconstructed properly that would manifest as distortion. That fact that the distortion of many / most DACs is vanishingly soft proves that reconstruction works as intended.
--Ethan
Ethan I had this discussion with Amir long ago. All I know is digital is flawed in the execution. At least to my ears. There is such a thing as "lossy." Not my term, I assume that lossy does not meet the Nyquist standard. HD Tracks has 192 and 44,1 recordings both are superior to other commercial digital. For that matter Cheskys analog recordings are superior also.
Chesky has analog recordings? They used to license some older analog recordings to release in on CD, but I don't think they have made any analog recordings?
Mostly re-releases. I bought a considerable classical music collection a couple of years ago, and found about 15 Chesky releases in it.
http://www.discogs.com/search?q=chesky&sort=score&sort_order=desc&type=all&format_exact=Vinyl
http://www.discogs.com/search?q=chesky&sort=score&sort_order=desc&type=all&format_exact=Vinyl
I don't remember . but here is what you said:
Soundproof said:
That would imply some were original. I'll post any more inormation I can find. I am getting old I purchased them in the mid to late '80s.
Yeah and just to add as I do not disagree.
Possibly what some do not consider is that it is not bettering capture but higher sampling rates such as 192khz push the alias artifacts further out, and one reason a NOS DAC will look smoother at higher rates.
The sinewave "data" is all there for both 44.1khz and say 192khz, but higher sampling rate provides greater flexibility with regards to filtering-alias artifacts effects that create a jagged-step like look.
Both Paul Miller and John Atkinson has shown this.
It is worth noting the benefit of either higher sampling rate or greater bit depth is different, in general greater bit depth helps with quantisation (but then we can overcome many challenges with dither) and greater sampling rates help with alias artifacts (but then this can be overcome with the associated filter).
It is more complex than this but helps to emphasise the difference and that both have technological solutions, but a greater bit depth and higher sampling rate may help in some ways beyond data captured.
Cheers
Orb