|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
50.164.204.218
In Reply to: RE: Article Itself Has Some Fallacies...... posted by Todd Krieger on October 07, 2015 at 23:51:47
Only in the context that it's a continuous (or steady-state) sine wave, and a classic "sinc" filter is capable of reconstructing it "perfectly"..... (A time-resolute filter would still have some of the "jagged" artifacts.) Note that the closer the sine wave signal approaches half the sample frequency, the "longer" the filter must be in order to fully reconstruct it.But with transients, the same "sinc" filter introduces "ringing" to the signal. So it really isn't perfect.
This is not true if the original signal is band limited to less than fc/2. Ringing shows up in the impulse response of a reconstruction filter because the impulse response is not a band limited signal. If there is no content in the signal at fc/2, the reconstruction filter will not add any ringing.
If transients in the original music signal have content above the cutoff frequency of the anti-aliasing filter in the ADC, the anti-aliasing filter will produce ringing. But if the digital signal is band limited to below the cutoff frequency of the reconstruction filter, the reconstruction filter does not add any additional ringing. That condition is hard to meet with Redbook because the transition bands of both filters are crammed in between 20 and 22 KHz. But it's easy to achieve at higher sample rates.
Once again, this is only true with continuous (or periodic) signals..... It is NOT true with transients often found in the original analog music signal.Wrong. There is no requirement for periodicity, only band-limiting. If the transients are band limited, they can be properly captured by sampling.
Edits: 10/08/15Follow Ups:
"If the transients are band limited"
If the transients have been ban limited then can they still be considered true transients?
Hasn't the band limiting removed most/some of what defined them as "transients" in the first place?
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
The bandwidth of the signal is inversely proportional to the maximum value of the slope / fastest rise time in the signal, regardless of whether the signal transient or periodic. For example, a muted trumpet has a periodic waveform that looks a bit like an inverted impulse train, with steep spikes containing a lot of high frequency content. It's bandwidth is greater than a lot of transients like the striking of a piano note or drum. But not as much as a cymbal.
Here you go.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Chuck Mangione or even Wynton Marsalis spectra look much, much whiter than that.
nt
Hey, I play trumpet, and pretty darn well. And I'm here to tell ya, my sound isn't nearly that ragged.
LOL
"This is not true if the original signal is band limited to less than fc/2."
True.... The problem is the original live music event is **never** band-limited.
"Ringing shows up in the impulse response of a reconstruction filter because the impulse response is not a band limited signal."
Wouldn't make any difference here..... Although the signal would "already" have ringing if band-limited with a (analog) brickwall filter.
"If there is no content in the signal at fc/2, the reconstruction filter will not add any ringing."
There shouldn't be any content right at Fs/2.... But there *is* ringing for content just below Fs/2. (The video link below bears this out.)
"If transients in the original music signal have content above the cutoff frequency of the anti-aliasing filter in the ADC, the anti-aliasing filter will produce ringing."
But if transients were band-limited, the ringing would already be present.... There is no free lunch here.
"But if the digital signal is band limited to below the cutoff frequency of the reconstruction filter, the reconstruction filter does not add any additional ringing."
Stress the word "additional"..... The signal under this circumstance would already have ringing prior to reconstruction.
"That condition is hard to meet with Redbook because the transition bands of both filters are crammed in between 20 and 22 KHz. But it's easy to achieve at higher sample rates."
Very true..... This is why I think 44.1 kHz sample rate isn't sufficient in regard to not being able to differentiate between the original non-band limited signal and the digitized one.
" Once again, this is only true with continuous (or periodic) signals..... It is NOT true with transients often found in the original analog music signal.
Wrong. There is no requirement for periodicity, only band-limiting. If the transients are band limited, they can be properly captured by sampling."
Once again, while true, I'm talking about the original analog signal, prior to any limiting being applied to it.
While the ear would not be able to differentiate an analog vs. digitized sine wave, the fact people have strong preferences for certain types of filtering (or no filtering) for 16/44 signals is evidence alone that the ear *can* differentiate variants of band-limiting from the transients in the music signal.
I can't help but notice that you've criticized Monty Montgomery's claims multiple times on this site. On the other hand, I don't think Monty is aware of any of your criticisms.
Could you post your concerns on Monty's site so that he is able to reply to them? I'd love to see these issues hashed out in a two sided discussion.
You might need to contact Monty directly to tell him that you've posted:
monty@xiph.org
JE
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: