Re: [Sursound] Bass Problem in crosstalk cancellation
Le Wed, 1 Jun 2011 19:38:10 +, Fons Adriaensen a écrit : > Ambiophonics makes a mess of it. Unless you use separate widely > spaced speakers for LF (driven by intensity-based stereo), as > some researchers have already recommended. I jut tried using 2 satellites at +-7.5 degrees (with the new XTC filter from DW) and 2 subs at +-30 degrees with no treatment (direct stereo), crossed at ~130hz. This configuration is very enjoyable; the bass seems more defined and localizable. -- Marc ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
Re: [Sursound] New filter
Le Wed, 1 Jun 2011 09:27:27 +, Fons Adriaensen a écrit : > On Tue, May 31, 2011 at 11:16:34PM -0400, Marc Lavallée wrote: > > > In the meanwhile, here's my jconvolver config > > (I hope it is valid; it works well for me): > > > > /convolver/new 2 2 1024 4096 > > /impulse/read 1 1 1 0 0 0 1 David_Wareing3_Left.wav > > /impulse/read 1 2 1 0 0 0 2 David_Wareing3_Left.wav > > /impulse/read 2 1 1 0 0 0 1 David_Wareing3_Right.wav > > /impulse/read 2 2 1 0 0 0 2 David_Wareing3_Right.wav > > Should work, but due to L-R symmetry you symplify it to: > > > /impulse/read 1 1 1 0 0 0 1 David_Wareing3_Left.wav > > /impulse/read 1 2 1 0 0 0 2 David_Wareing3_Left.wav > > /impulse/copy 2 1 1 2 > > /impulse/copy 2 2 1 1 > > Ciao, Right; I can see that the files are stereo and symetric. But I had to reduce the gain to avoid clipping; the IR file have a peak of -1.3dB. So now my config is : /impulse/read 1 1 0.5 0 0 0 1 David_Wareing3_Left.wav /impulse/read 1 2 0.5 0 0 0 2 David_Wareing3_Left.wav /impulse/copy 2 1 1 2 /impulse/copy 2 2 1 1 -- Marc ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
Re: [Sursound] Bass Problem in crosstalk cancellation
On 01/06/2011 22:01, Fons Adriaensen wrote: On Wed, Jun 01, 2011 at 08:59:30PM +0100, dw wrote: Computer say " i can give you itd=450us @ 20Hz" off the shelf. That is very near field, so should be unaffected by the room. I don't know whether it is true.. It depends on what you start from. If you have an intensity (panned) stereo LF source at e.g. 30 degrees left (so only in the L channel, assuming a normal stereo triangle), then reproducing this using two speakers at +/- 7.5 degrees is easy. It only requires a difference channel (L-R) gain 13 dB or so higher than the sum (L+R) gain, and independent of frequency. Some power is wasted but it will work. Things are different when starting with a binaural signal for for the same source. This will require a very high difference channel gain (proportional to 1/f, and complex), much more than could ever be used in practice. The whole concept of Ambiophonics is based on the idea of delivering the L,R signals in a stereo recording to the respective ears without crosstalk to the other. The basic assumption is that the signals are binaural. But for that kind of signal it can't work for low frequencies because it would require much more L-R gain than is possible in practice. If it works, and it does in many cases, it is because most recordings are *not* binaural to start with, but a mix of intensity based stereo (at LF) with *some* ITD at mid and high frequencies. For example the popular ORTF technique will deliver exactly that. For a pure binaural recording, Ambiophonics fails at LF. For a pure intensity (panned) stereo one, it will fail at mid and high frequencies because there is no ITD in the recorded signal (normally spaced speakers would create the ITD at playback). It Ambiophonics works well for some recordings it it just because these are intensity based at LF (requiring only moderate L-R gain), while still having some ITD at mid and high frequencies. Presenting it as a system that delivers binaural signals to the respective ears without crosstalk is misleading, that is *not* why it works in practice. Ciao, Nobody knows exactly how the ears work! I suggest you get a copy of Auberge - Chris Rea, some fine, small studio monitors, follow the instructions I provided and report back. ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
Re: [Sursound] Bass Problem in crosstalk cancellation
On Wed, Jun 01, 2011 at 08:59:30PM +0100, dw wrote: > Computer say " i can give you itd=450us @ 20Hz" off the shelf. That is > very near field, so should be unaffected by the room. I don't know > whether it is true.. It depends on what you start from. If you have an intensity (panned) stereo LF source at e.g. 30 degrees left (so only in the L channel, assuming a normal stereo triangle), then reproducing this using two speakers at +/- 7.5 degrees is easy. It only requires a difference channel (L-R) gain 13 dB or so higher than the sum (L+R) gain, and independent of frequency. Some power is wasted but it will work. Things are different when starting with a binaural signal for for the same source. This will require a very high difference channel gain (proportional to 1/f, and complex), much more than could ever be used in practice. The whole concept of Ambiophonics is based on the idea of delivering the L,R signals in a stereo recording to the respective ears without crosstalk to the other. The basic assumption is that the signals are binaural. But for that kind of signal it can't work for low frequencies because it would require much more L-R gain than is possible in practice. If it works, and it does in many cases, it is because most recordings are *not* binaural to start with, but a mix of intensity based stereo (at LF) with *some* ITD at mid and high frequencies. For example the popular ORTF technique will deliver exactly that. For a pure binaural recording, Ambiophonics fails at LF. For a pure intensity (panned) stereo one, it will fail at mid and high frequencies because there is no ITD in the recorded signal (normally spaced speakers would create the ITD at playback). It Ambiophonics works well for some recordings it it just because these are intensity based at LF (requiring only moderate L-R gain), while still having some ITD at mid and high frequencies. Presenting it as a system that delivers binaural signals to the respective ears without crosstalk is misleading, that is *not* why it works in practice. Ciao, -- FA ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
Re: [Sursound] Bass Problem in crosstalk cancellation
http://dl.dropbox.com/u/20268768/RACE/ITD.png (temporary link) This is for the existing filter without the to-mono crossover and with a 1 sample delay adjustment. ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
Re: [Sursound] Bass Problem in crosstalk cancellation
On 01/06/2011 20:38, Fons Adriaensen wrote: On Wed, Jun 01, 2011 at 09:42:42AM -0700, Ralph Glasgal wrote: But, the ear is not sensitive to crosstalk below say 100Hz so Where do you get this from ? Are you seriously saying that a low frequency signal delivered to one ear sounds natural ? Just try it. What this means is that one is not doing much cancellation as the frequency gets down to say 90Hz which is okay since one does not localize well or at all at low bass frequencies anyway. Now *that* is a fallacy if there ever was one. Agreed, if you play a 40 Hz sine in a small room without any acoustic treatment you won't be able to tell where it comes from. And if you put a piece of sandblasted glass on your computer screen then you won't be able to read this text. Things change if you allow them to. Apparently you have never heard a surround system that does reproduce low frequencies as they should be. Just plain intensity (panned) stereo gets close if the room doesn't destroy it. Ambisonic reproduction - even first order - gets this exactly right (under the same conditions). Ambiophonics makes a mess of it. Unless you use separate widely spaced speakers for LF (driven by intensity-based stereo), as some researchers have already recommended. To be clear: I don't want to denigrate Ambiophonics - it has is merits. But it would be better advocated using less pseudo science and by acknowledging its limits rather than by presenting it as something perfect invented by the gods and blessed by Alan Blumlein. In fact it's probably the most 'unnatural' way to reproduce sound - it's ill-conditioned by definition - even if sometimes it does work. Ciao, Computer say " i can give you itd=450us @ 20Hz" off the shelf. That is very near field, so should be unaffected by the room. I don't know whether it is true.. ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
Re: [Sursound] Bass Problem in crosstalk cancellation
On Wed, Jun 01, 2011 at 09:42:42AM -0700, Ralph Glasgal wrote: > But, the ear is not sensitive to crosstalk below say 100Hz so Where do you get this from ? Are you seriously saying that a low frequency signal delivered to one ear sounds natural ? Just try it. > What this means is that one is not doing much cancellation as > the frequency gets down to say 90Hz which is okay since one > does not localize well or at all at low bass frequencies anyway. Now *that* is a fallacy if there ever was one. Agreed, if you play a 40 Hz sine in a small room without any acoustic treatment you won't be able to tell where it comes from. And if you put a piece of sandblasted glass on your computer screen then you won't be able to read this text. Things change if you allow them to. Apparently you have never heard a surround system that does reproduce low frequencies as they should be. Just plain intensity (panned) stereo gets close if the room doesn't destroy it. Ambisonic reproduction - even first order - gets this exactly right (under the same conditions). Ambiophonics makes a mess of it. Unless you use separate widely spaced speakers for LF (driven by intensity-based stereo), as some researchers have already recommended. To be clear: I don't want to denigrate Ambiophonics - it has is merits. But it would be better advocated using less pseudo science and by acknowledging its limits rather than by presenting it as something perfect invented by the gods and blessed by Alan Blumlein. In fact it's probably the most 'unnatural' way to reproduce sound - it's ill-conditioned by definition - even if sometimes it does work. Ciao, -- FA ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
Re: [Sursound] Bass Problem in crosstalk cancellation
On 01/06/2011 17:42, Ralph Glasgal wrote: In most of the papers on crosstalk cancellation the point is made that at lower freauencies the power required to cancel the bass becomes prohibitive. I will make a feeble attempt to explain why this is a fallacy that derives from a common propensity to rely soly on mathematics when using the HRTF functions available. Yes, if you assume that one needs to cancel crosstalk at 20Hz then the usual HRTFs say the attenuation between the ears is negligible and so a high level signal is needed to do any cancellation and then more energy is needed to cancel this cancellation signal and the thing blows up. But, the ear is not sensitive to crosstalk below say 100Hz so one needs to take this into consideration. On can simply bypass bass frequencies around the HRTF bsased canceller but it turns out this is not really necessary in HRTF-less algorithms. One basic premise of RACE is that no HRTF functions need be used. But let us just concentrate on the bass region. RACE assumes a constant attenuation for a signal reaching the wrong ear. As the frequency declines this assumption becomes more and more inaccurate. But so what? What this means is that one is not doing much cancellation as the frequency gets down to say 90Hz which is okay since one does not localize well or at all at low bass frequencies anyway. In other words the amount of cancellation automatically declines with frequency so the overhead or power requirement does not change with frequency either. The head room needed is the same at all the normal real localization frequencies. (Very high frequencies are a different problem) You can see this in a brief note by Angelo Farina comparing RACE with other XTC methods. www.ambiophonics.org/papers/CrosstalkFilters.html Ralph Glasgal In fact I learned this from RACE. Instead of assuming cancellation at low frequencies, and then re-introducing large ITDs, either in the filter or counting on them from binaural recordings, I abandoned cancellation from 1kHz down and went for ITD directly. However there are only so many ways to squeeze a balloon without it bulging elsewhere.. BTW RACE has significant ITD, although smaller than mine :-) at lower frequencies. Perhaps this might explain why it works better than one might expect. ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
[Sursound] Bass Problem in crosstalk cancellation
In most of the papers on crosstalk cancellation the point is made that at lower freauencies the power required to cancel the bass becomes prohibitive. I will make a feeble attempt to explain why this is a fallacy that derives from a common propensity to rely soly on mathematics when using the HRTF functions available. Yes, if you assume that one needs to cancel crosstalk at 20Hz then the usual HRTFs say the attenuation between the ears is negligible and so a high level signal is needed to do any cancellation and then more energy is needed to cancel this cancellation signal and the thing blows up. But, the ear is not sensitive to crosstalk below say 100Hz so one needs to take this into consideration. On can simply bypass bass frequencies around the HRTF bsased canceller but it turns out this is not really necessary in HRTF-less algorithms. One basic premise of RACE is that no HRTF functions need be used. But let us just concentrate on the bass region. RACE assumes a constant attenuation for a signal reaching the wrong ear. As the frequency declines this assumption becomes more and more inaccurate. But so what? What this means is that one is not doing much cancellation as the frequency gets down to say 90Hz which is okay since one does not localize well or at all at low bass frequencies anyway. In other words the amount of cancellation automatically declines with frequency so the overhead or power requirement does not change with frequency either. The head room needed is the same at all the normal real localization frequencies. (Very high frequencies are a different problem) You can see this in a brief note by Angelo Farina comparing RACE with other XTC methods. www.ambiophonics.org/papers/CrosstalkFilters.html Ralph Glasgal -- next part -- An HTML attachment was scrubbed... URL: <https://mail.music.vt.edu/mailman/private/sursound/attachments/20110601/31e55b85/attachment.html> ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
Re: [Sursound] New filter
On Tue, May 31, 2011 at 11:16:34PM -0400, Marc Lavallée wrote: > In the meanwhile, here's my jconvolver config > (I hope it is valid; it works well for me): > > /convolver/new 2 2 1024 4096 > /impulse/read 1 1 1 0 0 0 1 David_Wareing3_Left.wav > /impulse/read 1 2 1 0 0 0 2 David_Wareing3_Left.wav > /impulse/read 2 1 1 0 0 0 1 David_Wareing3_Right.wav > /impulse/read 2 2 1 0 0 0 2 David_Wareing3_Right.wav Should work, but due to L-R symmetry you symplify it to: > /impulse/read 1 1 1 0 0 0 1 David_Wareing3_Left.wav > /impulse/read 1 2 1 0 0 0 2 David_Wareing3_Left.wav > /impulse/copy 2 1 1 2 > /impulse/copy 2 2 1 1 Ciao, -- FA ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
Re: [Sursound] Sound Externalization Headphone
ent head tracking? >>> >>> Hope this helps >>> Filippo >>> >>> >From: Archontis Politis >>> >Subject: Re: [Sursound] Sound Externalization Headphone >>> >To: sursound@music.vt.edu >>> >Message-ID: <4ddbca43.20...@gmail.com> >>> >Content-Type: text/plain; charset=ISO-8859-1; format=flowed >>> > >>> >Hi, >>> > >>> >I would add to J?rn's comments that apart from head-tracking, which is >>> >crucial, you probably have to apply some decorrelation to your synthetic >>> >binaural signals, and mix them with the normal ones. From anechoic hrtfs >>> >only, especially if they are generic ones it is easy to get the in-head >>> >effect. You can add decorrelation by some room simulation algorithm, >>> >artificial reverberation or simpler by convolving your signals with >>> >bursts of noise, passing them through allpass filters, applying varying >>> >delays, whatever you can do that will scramble the phase but not the >>> >magnitude of the sounds. >>> > >>> >I have heard demonstrations with room simulated binaural responses that >>> >were well externalised without head-tracking, adding head-tracking >>> >should be very effective. You can check the literature for audio >>> >decorrelation techniques or artificial reverberation. Have a look also >>> >on the following master thesis: >>> > >>> >Headphone Sound Externalization - TKK Acoustics >>> >www.acoustics.hut.fi/publications/files/theses/liitola_mst.pdf >>> > >>> >Regards, >>> >Archontis >>> >>> ___ >>> >>> Sursound mailing list >>> Sursound@music.vt.edu >>> https://mail.music.vt.edu/mailman/listinfo/sursound >>> >> >> >> >> >> > -- next part -- An HTML attachment was scrubbed... URL: <https://mail.music.vt.edu/mailman/private/sursound/attachments/20110601/5f226aae/attachment.html> ___ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound