I agree completely.
but having 99% of the problem right is better than 90%.
some of the points of people advocating hires simply reveal lack of understanding. I've seen the argument of how a 19kHz waveform at the output of a NOS filterless DAC looks like being used a few times.
at the same time I have my own - I dare to say more subtle - doubts.
though there is a "je ne sais quoi" which makes me refrain from going into the detail here and now. maybe it's the fact that I'll likely be opposed by a minority of the scientist's camp that says "Shannon/Nyquist are right" without first trying to make sure all the conditions for using it are met.
since I'm not exactly a math guy I'm not sure a have a good enough grasp on this either.
Well, lookee here.
I have NEVER heard a "redbook" playback that was anywhere near as good as a live feed.
So, if the argument is if something is better than redbook or indestinguishable from it, the best question to be asking is if it is equal to the LIVE FEED or not!
Isn't that a big part of the "quest"? To equal live?
Just throwing this chum into the shark infested waters...
_-_-bear
I have NEVER heard a "redbook" playback that was anywhere near as good as a live feed.
So, if the argument is if something is better than redbook or indestinguishable from it, the best question to be asking is if it is equal to the LIVE FEED or not!
Isn't that a big part of the "quest"? To equal live?
Just throwing this chum into the shark infested waters...
_-_-bear
Problem is, by the time you get to hear something made to Redbook spec, the music has gone through the engineering/mastering process, and you can count on massive dynamic range compromises, brutal EQ, and distortion boxes. Someone on this forum (can't remember who) made a terrifically insightful comment: The CD medium is great, it's the CDs that suck.
Gee,
And here I though he really understood the localization issue.
But if you prefer Manfred Schroeder at Bell Labs did a series of experiment with computer generated waveforms to see how much phase shift could be detected. Using waveforms with the same absolute amplitude the result seemed to be 5 degrees at 20 Khz.
Rupert Neve has mentioned while designing a mixer one channel seemed a slight bit different than the others. On investigation it turned out there was about a 5 degree phase difference at 20 Khz. Fixing that solved the issue.
So lets assume then that we need the timing to allow for less than 5 degrees of phase shift at 20 Khz.
ES
Then you need identical anti-aliasing filters in each stereo channel, with less than 5 degrees difference in phase shift at 20kHz. And a sampling rate above 40kHz.
Thats the same as moving an ear (1/20 ft * 5/360) about 1/1200 of an inch. Do you listen to music with your head in a clamp?
Show me some music with those kinds of rise times.
Dan Larvy (from Larvy Engineering) published in 2004 an article that concludes that 192kHz (as opposed to 88.1 or 96kHz) is useless. All the above considerations (including the impulse response) are covered.
I agree with his conclusion... not that matters anyway 🙂
I agree with his conclusion... not that matters anyway 🙂
Last edited:
..........
Yes, this was talked about in the Meyer/Moran paper. That's one reason they added in an extra D/A -> A/D step.
Read very carefully: 5 degrees of phase shift of a 20 KHz signal (or 5 microseconds of delay) will be accurately recorded and reproduced through a competently designed 44.1 KHz A-D-A chain. Feel free to disagree, but understand that your argument is with Nyquist, not with me.
Quote:
>So lets assume then that we need the timing to allow for less than 5 degrees of phase shift at 20 Khz.
>>>Thats the same as moving an ear (1/20 ft * 5/360) about 1/1200 of an inch. Do you listen to music with your head in a clamp?
Earphones are a problem if there's phase shift. Your head is in a perceptual clamp. Small phase shifts at high frequencies can affect perception of timbre of instruments with high energy, high frequency harmonics such as trumpets and cymbals, even listening with speakers.
Quote:
>... The human ear can register rises and falls at a rate of 5.8 microseconds per dB. ...
>>>Show me some music with those kinds of rise times.
Cymbal crash rise time is @ 1millisecond. That's a pretty good rate of rise/dB
This is interesting They found @ 20 microsecs - I'm sure there are other similar, later papers out there. Kunchur, I believe, saw about 7 in his work but there's some objection cuz he's an astronomer or something like that, but his bibliography is good.
http://www.google.com/url?url=http:...sg=AFQjCNGAizFYy6MLJaIUykxfKYkpcrVJeg&cad=rja
>So lets assume then that we need the timing to allow for less than 5 degrees of phase shift at 20 Khz.
>>>Thats the same as moving an ear (1/20 ft * 5/360) about 1/1200 of an inch. Do you listen to music with your head in a clamp?
Earphones are a problem if there's phase shift. Your head is in a perceptual clamp. Small phase shifts at high frequencies can affect perception of timbre of instruments with high energy, high frequency harmonics such as trumpets and cymbals, even listening with speakers.
Quote:
>... The human ear can register rises and falls at a rate of 5.8 microseconds per dB. ...
>>>Show me some music with those kinds of rise times.
Cymbal crash rise time is @ 1millisecond. That's a pretty good rate of rise/dB
This is interesting They found @ 20 microsecs - I'm sure there are other similar, later papers out there. Kunchur, I believe, saw about 7 in his work but there's some objection cuz he's an astronomer or something like that, but his bibliography is good.
http://www.google.com/url?url=http:...sg=AFQjCNGAizFYy6MLJaIUykxfKYkpcrVJeg&cad=rja
Rarely are the conditions required satisfied in practice, Nyquist?Shannon sampling theorem - Wikipedia, the free encyclopedia
When i read the Sony CD white paper in 1980 i happened to be taking a grad level sampling course. My 1st comment was that they would need to increase sampling by 8 times before it would be indisguishable from analog. This was further reinforced by a question i had wrt digital scopes. The scope jockies said, to paraphrase, that they wouldn't trust a scope that didn't have sampling 5-10x higher than strictly necessary.
dave
- Status
- Not open for further replies.