IPB

Welcome Guest ( Log In | Register )

2 Pages V  < 1 2  
Reply to this topicStart new topic
A secret "known by signal processing experts" or nonsense?
Ethan Winer
post Jan 12 2013, 19:32
Post #26





Group: Members
Posts: 248
Joined: 12-May 09
From: New Milford, CT
Member No.: 69730



QUOTE (jmvalin @ Jan 11 2013, 16:26) *
Actually, I did a bit of research on that topic, for the purpose of decorrelating channels to help stereo acoustic echo cancellation. What I found was that you can make phase shift inaudible even when channels aren't matched and while varying the amount of phase shift. What's important is only matching the channels below about 2 kHz, where the ear is sensitive to interaural phase difference (IPD).

Interesting, you've done more types of testing than I have. Also, when I said "the phase shift is changing" I should have been clearer. The type of phase shift effects I've tested vary the center frequency of the phase shift, not the amount of shift at a fixed frequency. So by "changing" I mean sweeping the frequency. Though I imagine changing the amount of shift at a fixed frequency would also be audible at midrange frequencies.

--Ethan


--------------------
I believe in Truth, Justice, and the Scientific Method
Go to the top of the page
+Quote Post
Woodinville
post Jan 12 2013, 20:14
Post #27





Group: Members
Posts: 1402
Joined: 9-January 05
From: JJ's office.
Member No.: 18957



QUOTE (adamdea @ Jan 10 2013, 08:16) *
http://www.wickeddigital.com.au/index.php/...eally-necessary
Sets out what is described as a White Paper by Marco Manunta of M2Tech. It contains the following passage (my emphasis)

"But there is something more. Itís known by signal processing experts, and absolutely not popularized amongst music lovers, that converting an analog signal into a discrete-time one (as it happens when converting from analog to digital) destroys the phase information in the two top octaves of the resulting spectrum. In a CD-standard digital recording, all phase information are lost from 5.5kHz up to 22kHz,"

Can any signal processing experts comment on whether this is true or not, and if so why. As a mere layman I am not able to understand why it even might seem be true.



It is false. It is utterly false, absurd, and insane. Your cell phone, MP3 player, and cable modem could not work, for instance, if there was any truth to this utterly preposterious foolishness.


--------------------
-----
J. D. (jj) Johnston
Go to the top of the page
+Quote Post
Woodinville
post Jan 12 2013, 20:17
Post #28





Group: Members
Posts: 1402
Joined: 9-January 05
From: JJ's office.
Member No.: 18957



QUOTE (Arnold B. Krueger @ Jan 11 2013, 07:58) *
Above some frequency in what we call midrange, the ears lose their ability to discern phase.


Allow me to be more specific.

The ability to discern signal phase starts to degrade around 500Hz, and is almost completely gone by 2kHz, and completely gone by 4kHz.

HOWEVER, the ability to discern time delays in a non-stationary signal envelope at 15kHz is very good. So, time cues matter above 2kHz. A lot.


--------------------
-----
J. D. (jj) Johnston
Go to the top of the page
+Quote Post
Woodinville
post Jan 12 2013, 20:21
Post #29





Group: Members
Posts: 1402
Joined: 9-January 05
From: JJ's office.
Member No.: 18957



QUOTE (Ethan Winer @ Jan 12 2013, 10:32) *
QUOTE (jmvalin @ Jan 11 2013, 16:26) *
Actually, I did a bit of research on that topic, for the purpose of decorrelating channels to help stereo acoustic echo cancellation. What I found was that you can make phase shift inaudible even when channels aren't matched and while varying the amount of phase shift. What's important is only matching the channels below about 2 kHz, where the ear is sensitive to interaural phase difference (IPD).

Interesting, you've done more types of testing than I have. Also, when I said "the phase shift is changing" I should have been clearer. The type of phase shift effects I've tested vary the center frequency of the phase shift, not the amount of shift at a fixed frequency. So by "changing" I mean sweeping the frequency. Though I imagine changing the amount of shift at a fixed frequency would also be audible at midrange frequencies.

--Ethan


So, try delaying a 10kHz gaussian impulse with a 4khz bandwidth by half a sample in one channel. Tell me you don't hear that in headphones, eh?


--------------------
-----
J. D. (jj) Johnston
Go to the top of the page
+Quote Post
jmvalin
post Jan 13 2013, 08:46
Post #30


Xiph.org Speex developer


Group: Developer
Posts: 480
Joined: 21-August 02
Member No.: 3134



QUOTE (Ethan Winer @ Jan 12 2013, 13:32) *
Interesting, you've done more types of testing than I have. Also, when I said "the phase shift is changing" I should have been clearer. The type of phase shift effects I've tested vary the center frequency of the phase shift, not the amount of shift at a fixed frequency. So by "changing" I mean sweeping the frequency. Though I imagine changing the amount of shift at a fixed frequency would also be audible at midrange frequencies.


Actually, I was changing what you call the "center frequency" as well. If you remove the part of my filter that was designed to leave the lower frequencies alone, you have a "comb allpass" filter:

A(z) = (a + z^-N)/(1 - a*z^-N)

I was actually changing not only a, but also the order N to avoid the "nulls" in the phase change.
Go to the top of the page
+Quote Post
Ethan Winer
post Jan 13 2013, 20:03
Post #31





Group: Members
Posts: 248
Joined: 12-May 09
From: New Milford, CT
Member No.: 69730



QUOTE (Woodinville @ Jan 12 2013, 14:21) *
So, try delaying a 10kHz gaussian impulse with a 4khz bandwidth by half a sample in one channel. Tell me you don't hear that in headphones, eh?

If only I understood what that meant, or how to delay something by half a sample! biggrin.gif

I agree that left-right phase differences are audible. I haven't tested this enough to be more specific, but I'm glad you and others have.

--Ethan


--------------------
I believe in Truth, Justice, and the Scientific Method
Go to the top of the page
+Quote Post
greynol
post Jan 13 2013, 20:09
Post #32





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



Delay by a half sample: up-convert to a higher samplerate, perform the delay and then down-convert.

Pretty simple!


--------------------
I should publish a list of forum idiots.
Go to the top of the page
+Quote Post
jmvalin
post Jan 13 2013, 20:17
Post #33


Xiph.org Speex developer


Group: Developer
Posts: 480
Joined: 21-August 02
Member No.: 3134



QUOTE (Ethan Winer @ Jan 13 2013, 14:03) *
QUOTE (Woodinville @ Jan 12 2013, 14:21) *
So, try delaying a 10kHz gaussian impulse with a 4khz bandwidth by half a sample in one channel. Tell me you don't hear that in headphones, eh?

If only I understood what that meant, or how to delay something by half a sample! biggrin.gif

I agree that left-right phase differences are audible. I haven't tested this enough to be more specific, but I'm glad you and others have.

--Ethan


Actually, I've no idea what a "10kHz gaussian impulse with a 4khz bandwidth" would be, but delaying something by half a sample (or any fractional number of samples) is very easy to do, e.g. using a sinc filter. One thing I should have been more precise about in my earlier comment... you can get away with modifying the phase as long as it's above ~2 kHz and as long as the group delay (derivative of the phase response wrt frequency) is small enough. But as I said, as long as you're careful with how you do it, it's amazing how much you can get away with when playing with the phase. The real difficulty when the phase is involved is really how messing it up can cause other artefacts. e.g. changes in the actual power spectrum.
Go to the top of the page
+Quote Post
Woodinville
post Jan 14 2013, 03:16
Post #34





Group: Members
Posts: 1402
Joined: 9-January 05
From: JJ's office.
Member No.: 18957



QUOTE (greynol @ Jan 13 2013, 11:09) *
Delay by a half sample: up-convert to a higher samplerate, perform the delay and then down-convert.

Pretty simple!


Even easier, it's a gaussian impulse, it has no energy outside Fs/2 to speak of, so you can shift it analytically.

it's of the form

e^(((t-t0)^2))/c)*sin(2*pi*f*(t-t0))

So you just change t0.
Reason for edit: formula


--------------------
-----
J. D. (jj) Johnston
Go to the top of the page
+Quote Post
Ethan Winer
post Jan 15 2013, 21:42
Post #35





Group: Members
Posts: 248
Joined: 12-May 09
From: New Milford, CT
Member No.: 69730



Oh, you math guys...

Though I do understand up-sample, delay, down-sample.

--Ethan


--------------------
I believe in Truth, Justice, and the Scientific Method
Go to the top of the page
+Quote Post

2 Pages V  < 1 2
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 19th September 2014 - 16:25