Author |
Message |
bachus
Joined: Feb 29, 2004 Posts: 2922 Location: Up in that tree over there.
Audio files: 5
|
|
Back to top
|
|
|
bachus
Joined: Feb 29, 2004 Posts: 2922 Location: Up in that tree over there.
Audio files: 5
|
Posted: Wed Apr 23, 2008 7:12 pm Post subject:
|
|
|
Playing with this and looking at it I realize that though it is time correct it is not amplitude correct. When an acoustic source is moved to the "far" left the amplitude at the right microphone dose not drop to zero or anything like it. Thus the HRTFs must have a dominant role in in localization. _________________ The question is not whether they can talk or reason, but whether they can suffer. -- Jeremy Bentham |
|
Back to top
|
|
|
bachus
Joined: Feb 29, 2004 Posts: 2922 Location: Up in that tree over there.
Audio files: 5
|
Posted: Sat Apr 26, 2008 5:30 am Post subject:
|
|
|
'nother file update
I think at this point I will abandon this and look at VST in c++ so I can play with adding HRTFs in the pan path.
What I want ultimately is something that models the ambiophonic recording method. If anyone can point to help with such it would be appreciated. _________________ The question is not whether they can talk or reason, but whether they can suffer. -- Jeremy Bentham |
|
Back to top
|
|
|
durwood
Joined: Apr 25, 2008 Posts: 30 Location: Chicagoland
|
|
Back to top
|
|
|
bachus
Joined: Feb 29, 2004 Posts: 2922 Location: Up in that tree over there.
Audio files: 5
|
Posted: Sun Apr 27, 2008 10:23 am Post subject:
|
|
|
Thanks for the link. These are interesting technologies. However the panner does seem to be for ambiosonics which is for multi-channel (>2) systems What I want is one that enhances both ambiophonic and (I think) therefore std. stereo playback. _________________ The question is not whether they can talk or reason, but whether they can suffer. -- Jeremy Bentham |
|
Back to top
|
|
|
bachus
Joined: Feb 29, 2004 Posts: 2922 Location: Up in that tree over there.
Audio files: 5
|
Posted: Sun Apr 27, 2008 7:10 pm Post subject:
|
|
|
More bugs found. I'm still getting the hang of debugging Reaktor files. I recommend that this not be used till I have a few weeks of futzing around with this. Very sorry _________________ The question is not whether they can talk or reason, but whether they can suffer. -- Jeremy Bentham |
|
Back to top
|
|
|
durwood
Joined: Apr 25, 2008 Posts: 30 Location: Chicagoland
|
Posted: Thu Dec 11, 2008 7:32 am Post subject:
|
|
|
I was discussing this with mosc and he suggested to post about it here. First a little background theory to help explain my idea, bear with me.
A little something I put together showing a single source of sound location around a head. This will should you how azimuth angle can change the time difference between the two ears. Notice how different head diameters can create different time differences.
Two important AES papers where I gather some of this info:
"The Stereosonic Recording and Reproducing System" - Clark, Dutton and Vanderlyn
"Phasor Analysis of Some Stereophonic Phenomena" - Bauer
The equation is taken from one of the papers I read.
D=diameter of the head
c=speed of sound
theta = azimuth/angle of source
Time difference (s) = D/c * sin (theta)
Note: negative time indicates switch over from L-R to R-L. The 5 and 10deg numbers are an important set of numbers for an ambiophonics playback system.
Code: |
Head 0.16/ 0.17 /0.18/ 0.19/ 0.20/ 0.21/ 0.22/
Diam (m)
________________________________________________________________
Source Time Difference Between L& R @ L ear (ms) for a single source
Angle
(deg)
________________________________________________________________
0 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
5 0.0410 0.0436 0.0461 0.0487 0.0513 0.0538 0.0564
10 0.0817 0.0868 0.0919 0.0970 0.1021 0.1073 0.1124
15 0.1218 0.1294 0.1370 0.1446 0.1522 0.1599 0.1675
20 0.1610 0.1710 0.1811 0.1911 0.2012 0.2112 0.2213
25 0.1989 0.2113 0.2237 0.2362 0.2486 0.2610 0.2735
30 0.2353 0.2500 0.2647 0.2794 0.2941 0.3088 0.3235
35 0.2699 0.2868 0.3037 0.3205 0.3374 0.3543 0.3711
40 0.3025 0.3214 0.3403 0.3592 0.3781 0.3970 0.4159
45 0.3328 0.3536 0.3744 0.3951 0.4159 0.4367 0.4575
50 0.3605 0.3830 0.4056 0.4281 0.4506 0.4731 0.4957
55 0.3855 0.4096 0.4337 0.4578 0.4819 0.5059 0.5300
60 0.4075 0.4330 0.4585 0.4840 0.5094 0.5349 0.5604
65 0.4265 0.4532 0.4798 0.5065 0.5331 0.5598 0.5864
70 0.4422 0.4698 0.4975 0.5251 0.5528 0.5804 0.6080
75 0.4546 0.4830 0.5114 0.5398 0.5682 0.5966 0.6250
80 0.4634 0.4924 0.5214 0.5503 0.5793 0.6083 0.6372
85 0.4688 0.4981 0.5274 0.5567 0.5860 0.6153 0.6446
90 0.4706 0.5000 0.5294 0.5588 0.5882 0.6176 0.6471
95 0.4688 0.4981 0.5274 0.5567 0.5860 0.6153 0.6446
100 0.4634 0.4924 0.5214 0.5503 0.5793 0.6083 0.6372
105 0.4546 0.4830 0.5114 0.5398 0.5682 0.5966 0.6250
110 0.4422 0.4698 0.4975 0.5251 0.5528 0.5804 0.6080
115 0.4265 0.4532 0.4798 0.5065 0.5331 0.5598 0.5864
120 0.4075 0.4330 0.4585 0.4840 0.5094 0.5349 0.5604
125 0.3855 0.4096 0.4337 0.4578 0.4819 0.5059 0.5300
130 0.3605 0.3830 0.4056 0.4281 0.4506 0.4731 0.4957
135 0.3328 0.3536 0.3744 0.3951 0.4159 0.4367 0.4575
140 0.3025 0.3214 0.3403 0.3592 0.3781 0.3970 0.4159
145 0.2699 0.2868 0.3037 0.3205 0.3374 0.3543 0.3711
150 0.2353 0.2500 0.2647 0.2794 0.2941 0.3088 0.3235
155 0.1989 0.2113 0.2237 0.2362 0.2486 0.2610 0.2735
160 0.1610 0.1710 0.1811 0.1911 0.2012 0.2112 0.2213
165 0.1218 0.1294 0.1370 0.1446 0.1522 0.1599 0.1675
170 0.0817 0.0868 0.0919 0.0970 0.1021 0.1073 0.1124
175 0.0410 0.0436 0.0461 0.0487 0.0513 0.0538 0.0564
180 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
185 -0.0410 -0.0436 -0.0461 -0.0487 -0.0513 -0.0538 -0.0564
190 -0.0817 -0.0868 -0.0919 -0.0970 -0.1021 -0.1073 -0.1124
195 -0.1218 -0.1294 -0.1370 -0.1446 -0.1522 -0.1599 -0.1675
200 -0.1610 -0.1710 -0.1811 -0.1911 -0.2012 -0.2112 -0.2213
205 -0.1989 -0.2113 -0.2237 -0.2362 -0.2486 -0.2610 -0.2735
210 -0.2353 -0.2500 -0.2647 -0.2794 -0.2941 -0.3088 -0.3235
215 -0.2699 -0.2868 -0.3037 -0.3205 -0.3374 -0.3543 -0.3711
220 -0.3025 -0.3214 -0.3403 -0.3592 -0.3781 -0.3970 -0.4159
225 -0.3328 -0.3536 -0.3744 -0.3951 -0.4159 -0.4367 -0.4575
230 -0.3605 -0.3830 -0.4056 -0.4281 -0.4506 -0.4731 -0.4957
235 -0.3855 -0.4096 -0.4337 -0.4578 -0.4819 -0.5059 -0.5300
240 -0.4075 -0.4330 -0.4585 -0.4840 -0.5094 -0.5349 -0.5604
245 -0.4265 -0.4532 -0.4798 -0.5065 -0.5331 -0.5598 -0.5864
250 -0.4422 -0.4698 -0.4975 -0.5251 -0.5528 -0.5804 -0.6080
255 -0.4546 -0.4830 -0.5114 -0.5398 -0.5682 -0.5966 -0.6250
260 -0.4634 -0.4924 -0.5214 -0.5503 -0.5793 -0.6083 -0.6372
265 -0.4688 -0.4981 -0.5274 -0.5567 -0.5860 -0.6153 -0.6446
270 -0.4706 -0.5000 -0.5294 -0.5588 -0.5882 -0.6176 -0.6471
275 -0.4688 -0.4981 -0.5274 -0.5567 -0.5860 -0.6153 -0.6446
280 -0.4634 -0.4924 -0.5214 -0.5503 -0.5793 -0.6083 -0.6372
285 -0.4546 -0.4830 -0.5114 -0.5398 -0.5682 -0.5966 -0.6250
290 -0.4422 -0.4698 -0.4975 -0.5251 -0.5528 -0.5804 -0.6080
295 -0.4265 -0.4532 -0.4798 -0.5065 -0.5331 -0.5598 -0.5864
300 -0.4075 -0.4330 -0.4585 -0.4840 -0.5094 -0.5349 -0.5604
305 -0.3855 -0.4096 -0.4337 -0.4578 -0.4819 -0.5059 -0.5300
310 -0.3605 -0.3830 -0.4056 -0.4281 -0.4506 -0.4731 -0.4957
315 -0.3328 -0.3536 -0.3744 -0.3951 -0.4159 -0.4367 -0.4575
320 -0.3025 -0.3214 -0.3403 -0.3592 -0.3781 -0.3970 -0.4159
325 -0.2699 -0.2868 -0.3037 -0.3205 -0.3374 -0.3543 -0.3711
330 -0.2353 -0.2500 -0.2647 -0.2794 -0.2941 -0.3088 -0.3235
335 -0.1989 -0.2113 -0.2237 -0.2362 -0.2486 -0.2610 -0.2735
340 -0.1610 -0.1710 -0.1811 -0.1911 -0.2012 -0.2112 -0.2213
345 -0.1218 -0.1294 -0.1370 -0.1446 -0.1522 -0.1599 -0.1675
350 -0.0817 -0.0868 -0.0919 -0.0970 -0.1021 -0.1073 -0.1124
355 -0.0410 -0.0436 -0.0461 -0.0487 -0.0513 -0.0538 -0.0564
360 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
|
|
|
Back to top
|
|
|
durwood
Joined: Apr 25, 2008 Posts: 30 Location: Chicagoland
|
Posted: Thu Dec 11, 2008 7:38 am Post subject:
|
|
|
Ok moving on, sorry for more long text to read.
Typical panning knobs in stereo are not totally correct since our hearing relies mostly on Time diffences <1.5Khz. When you attempt to use a typical amplitude panner for ambiophonics, it doesn't work right since width will be lost and you will only be albe to pan +/-15 deg.
The problem is that by using amplitude you kill off the cancelation that is allowing you to have width, so it locks the image to whatever side you pan it to. This is the same problem with pan-pot stereo as well IMO, but not as bad because the speakers are at least spaced out more so it is less noticeable. For closely spaced stereo dipole/ambiophonic setups, this is noticed even more. The image should be allowed to move beyond the speaker, otherwise you end up with opposite effect you should get by panning and end up with a mono source trapped at the speaker right? Not good IMO.
I suggested this in mosc's ambiophonic VST thread, and decided to try a quick experiment to see if this would work by adding another delay on the direct signal. If you have two sources + two ears, you have 4 paths. In order to truly pan by azimuth angle, you have to compensate for all 4 paths/delays. See my attached diagram.
Rough Examples:
Centered
Cross Delay L=R=60us
Direct Delay L=R=0s
Panned Left (ex. 20us change)
Cross Delay L=80us
Cross Delay R=40us
Direct Delay L=20us
Direct Delay R=-20us
Panned Right (ex 20us change)
Cross Delay L=40us
Cross Delay R=80us
Cross Delay L=-20us
Cross Delay R=20us
You can't have a negative delay, so you would have to add/pad additional delay to allow for maximum panning so add an additional 60-80us to all signals. I used a generic 20us value for my examples, but that is not totally correct. Maybe later I will try to work out the real math.
Also, there would have to be some type of gain/amplitude adjustment that gradually comes in towards the higher frequencies based on ITD to ILD switch over. Ideally, you would tie this to one knob that adjusted all parameters at once-I call it an "azimuth panning knob" as that makes the most sense to me.
I can test it by manually setting the values just like above and it seems to work, but I cannot test this variably. Does this make any sense? Is it possible to make a plugin that does this? There is an ambisonic VST that does something similar to this for surround that is a rotate/tumble/tilt that works kind of similar to this on the York website.
I'm attaching my crude idea minus amplitude/gain compensation. It's basically mosc's MAP + additional delay block to allow time correct panning. Can this be made into a VST of sorts?
Description: |
|
Filesize: |
14.33 KB |
Viewed: |
810 Time(s) |
This image has been reduced to fit the page. Click on it to enlarge. |
|
|
|
Back to top
|
|
|
bachus
Joined: Feb 29, 2004 Posts: 2922 Location: Up in that tree over there.
Audio files: 5
|
Posted: Thu Dec 11, 2008 8:42 am Post subject:
|
|
|
durwood wrote: | Typical panning knobs in stereo are not totally correct since our hearing relies mostly on Time diffences <1.5Khz. |
This is kind of tangential but the relative importance of time delay vs. HRTFs is not clear to me. Do you have any links to empirical or other evidence/studies for this.
thanks _________________ The question is not whether they can talk or reason, but whether they can suffer. -- Jeremy Bentham |
|
Back to top
|
|
|
durwood
Joined: Apr 25, 2008 Posts: 30 Location: Chicagoland
|
Posted: Thu Dec 11, 2008 10:19 am Post subject:
|
|
|
Are you asking if I have come across a study that indicates ITD is the only important factor? No, this I have not. ITD and ILD are both important parts of HRTF's so yes, HRTF is better way to describe it since it covers more bases. I'm suggesting for horizontal azimuth panning, ITD is important I didn't mean for my statement to be so general.
The time/intensity trading is a gradual thing but if we generalized it and said which are more dominant, for most of the information available according to Raleigh's duplex theory.
- Frequencies less than 700hz - ITD
- 700-1500 ITD+ILD (Although ITD is still more dominant)
- Frequencies greater than 1500hz starts to become ILD dominant
Of further note, my understanding of ILD for us is mainly used for elevation.
Is that what you are asking?
In ambiophonics where we have two closely spaced speakers, a panning device only based on level adjustment is going to cause hard left panned material to only reside at the left speaker (5-10deg), and hard panned right material will reside at the right speaker (5-10deg) because we won't have the cross-cancelation to "throw" the source out of bounds. My experimenting with my crude panning concept above seems to prove this to me unless I am doing something grossly wrong.
I'm trying to figure out how you can do two things
1) Rotate or pan a virtual image or source around a head/sphere using ambiophonics
or
2) "Shift" the entire stage left or right using ambiophonics. Last edited by durwood on Thu Dec 11, 2008 10:42 am; edited 1 time in total |
|
Back to top
|
|
|
durwood
Joined: Apr 25, 2008 Posts: 30 Location: Chicagoland
|
Posted: Thu Dec 11, 2008 10:39 am Post subject:
|
|
|
Also, we are completely neglecting the effects of the room itself which have a huge impact on how successful any of this is. With ambiophonics the speakers are more centrally located and farther away from side walls that can ruin stage width. In stereo the speakers are farther apart, and if proper steps are not taken to minimize side wall reflections or at least delay them enough, then the width will suffer. |
|
Back to top
|
|
|
bachus
Joined: Feb 29, 2004 Posts: 2922 Location: Up in that tree over there.
Audio files: 5
|
Posted: Thu Dec 11, 2008 12:14 pm Post subject:
|
|
|
durwood wrote: | Is that what you are asking?
. |
Partly but more sleazily I was hoping for some thing that was sufficiently detailed that I could implement full HRTF algorithms from it. _________________ The question is not whether they can talk or reason, but whether they can suffer. -- Jeremy Bentham |
|
Back to top
|
|
|
durwood
Joined: Apr 25, 2008 Posts: 30 Location: Chicagoland
|
Posted: Thu Dec 11, 2008 1:44 pm Post subject:
|
|
|
Hmmm, well I've seen curves, impulses, all sorts of things but they are all still approximations. I don't think an accurate easy model exsists. Google HRTF models and you will get lots of different methods.
I've heard of systems that utilize a laser tracker and switches impulses in real-time to location of the head, but outside of virtual reality and/or headphone users I'm not sure too many people would want to wear such a contraption.
HRTF models suffer from some of the same problems digital room correction do. The room interaction. |
|
Back to top
|
|
|
bachus
Joined: Feb 29, 2004 Posts: 2922 Location: Up in that tree over there.
Audio files: 5
|
Posted: Thu Dec 11, 2008 5:54 pm Post subject:
|
|
|
I had come under the (uncertain) impression that for panning HRTF FR curves were of secondary importance after amplitude and that the shape of an individual’s pinna determinative for those curves. And that the pinna shape could be broken down into something like 16 morphological approximations, one of which would almost certainly “fit” any given individual. And that this dependence of the HRTF FR curve on the individual’s pinna shape “class” was the major stumbling block for improved panning (ambiophonic or otherwise). And that inter aural time delay issues were tertiary. Is this not correct? _________________ The question is not whether they can talk or reason, but whether they can suffer. -- Jeremy Bentham |
|
Back to top
|
|
|
durwood
Joined: Apr 25, 2008 Posts: 30 Location: Chicagoland
|
Posted: Fri Dec 12, 2008 8:52 am Post subject:
|
|
|
Very interesting. I have to honestly say this is new info to me. I cannot say if I believe it is right or wrong. Do you have a source that discusses this? I would love to read about it.
Just to be clear you are saying the order of importance for panning a virtual source with more than one loudspeaker is
1) HRTF manipulation
2) Amplitude levels
3) Time
By altering the timing of certain signals, would you agree it can affect amplitude (for various frequencies depending on combfiltering), which could also affect HRTF's. The question is, does this effect create the right result? |
|
Back to top
|
|
|
durwood
Joined: Apr 25, 2008 Posts: 30 Location: Chicagoland
|
|
Back to top
|
|
|
bachus
Joined: Feb 29, 2004 Posts: 2922 Location: Up in that tree over there.
Audio files: 5
|
Posted: Mon Dec 15, 2008 3:11 pm Post subject:
|
|
|
I was under the uncertain impression the order was:
1) Amplitude levels
2) HRTF manipulation
3) Time
I'll need to read this several times. Looks like an excellent link, and many thanks for it. _________________ The question is not whether they can talk or reason, but whether they can suffer. -- Jeremy Bentham |
|
Back to top
|
|
|
|