I have found an interesting source, quoting previous research, that suggest I might not be "clearly wrong". I am currently reading the thesis and find it a good synthesis of localization.
http://sound.media.mit.edu/Papers/gardner_thesis.pdf
"Another approach is to perform localization experiments using binaural stimuli
synthesized with conflicting ITD and ILD cues, Using this approach, Wightman and Kistler (1992) have shown that low-frequency ITD cues dominate ILD cues; in other words, when the stimuli contain low frequencies, the position of the auditory image is determined by the ITD cue regardless of the ILD cues."
"Interaural time delay (ITD) cues are effective at localizing low-frequency sounds; they principally operate at frequencies below about 1500 Hz. The question of which ear has leading phase can be unambiguously determined for frequencies below about 700 Hz."
bzfcocon - I don't see where those quotes imply anything in regards to the relative importance above and below a certain frequency, only the relative importance of ITD versus ILD. ITD dominates at LFs and ILD at HFs (well known) - it does not imply which has the better localization ability.
I get everything but the first paragraph. If the system has constant group delay, like propagation through air, then doesn't it have to be correctable with a pure delay? I am not sure I follow your point.
You might want to re-read the quotes (or the paper).
The paper plainly states that in the presence of both - and conflicting - (low frequency) ITD and (high frequency) ILD the low-freq ITD always "wins" - that is, it dominates the localization ability. This is inline with Griesinger's paper suggesting to correct the stereo image collapse towards the middle of the auditory scene at low frequency (to Keep ITD and ILD in-synch).
To be fair, I've found a newer paper from 2010 (Localization of sound in rooms. V. Binaural coherence and human sensitivity to interaural time differences in noise) that somewhat relativize the above and show that:
- ITD localization threshold depends of the binaural coherence: ITD stops working if binaural coherence is lost (through reflections for instance)
- more over, this effect is stronger at lower frequencies, meaning the coherence must be significantly higher than in the midband for ITD to work.
However, the second paper uses 1/3 octave noise as stimuli, while the first uses broadband Signal. The authors of the second paper do actually acknowledge that ITD cues are better preserved with broadband signal (such as speech/music):
"First, the experiments by Shinn-Cunningham et al. were broad band (200–3000 Hz), which favors sharp cross-correlation functions and well-preserved ITDs. For instance, Hartmann et al. (2005) found that with a very wide band, KEMAR recordings showed excellent preservation of accurate ITDs, monotonic with azimuth, even for distant sources in a reverberation room."
I don't think that the quotes say this, although the paper might, I don't know. This seems to go against everything that I read about localization from Blauert. That is pretty much the only source that I use for this kind of thing. I knew Hartmann personally I wouldn't trust his work.
Have you read Blauerts Spatial Hearing? Do you think he's wrong? Can you get support for your position in that text?
For what they are the measurements look good, but they are not detailed enough IMO. A detailed polar map like I do would show all the flaws. which could be hidden in those coarse plots.
JBL likes lots of early room reflections, I don't. It reflects Floyds and my difference of opinion on this topic.
But, for what it is worth, those plots don't look markedly different than mine, which is to the point that I have made before - all really good loudspeaker designs will measure and sound pretty much the same. It will all converge on one set of similar measurements. There isn't a wide range of what is good. Its not "whatever sounds good to you."
They go for 50 Hz, I don't. If you are going to have multiple subs anyways then why ask the woofer to do all that unnecessary excursion?
There will always be subtle difference like 110 x 120 versus 90 x 90 - I'd rather have axisymmetric - but that's not patentable of course.
We're not reading the bible here searching for hidden meanings. The quote says:
"Wightman and Kistler (1992) have shown that low-frequency ITD cues dominate ILD cues; in other words, when the stimuli contain low frequencies, the position of the auditory image is determined by the ITD cue regardless of the ILD cues.""
I could not find the quoted paper online, but the abstract says:
"In the experiments reported here, the interaural phase relations in the processing algorithms are manipulated in order to produce stimuli in which the interaural time difference cues signal one direction and interaural intensity and pinna cues signal another direction. The apparent directions of these conflicting cue stimuli almost always follow the interaural time cue, as long as the wideband stimuli include low frequencies."
Well, only the second paper is from Hartmann (the one that is actually suggesting the conclusions of the first are to be taken with a grain of salt). However, they use 1/3 octave noise, which is rather unrealistic, I would think, while the original paper uses wideband noise.
Still, you're dismissing someone's work (published in a peer-reviewed (?) J Acoust Soc Am. ) because you don't like them personally ?
No, I have not, I'll try to get a copy. But AFAIK from the papers above, Blauert was mainly concerned with free field/anechoic, while we are more talking about in-room, where things like binaural coherence play a role.
Sorry, I wasn't too clear there.
Imagine an extreme case like a response with 90 degrees phase response, but flat magnitude response, everywhere except near 0 Hz. The slope of phase vs. frequency is zero, so group delay is constant. But the impulse response isn't an impulse (more like a doublet) and no pure delay will get all frequencies back to 0 degrees, because a quarter cycle at LFs is a lot more time seconds delay than at HFs.
My point was that flat phase or group delay alone isn't enough for waveform fidelity (at least on an oscilloscope). The response would have to also have a flat phase delay, with the constant slope pointing toward zero or 180 degrees at f=0.
Last edited:
So to you the room makes all the difference and that free field localization is one thing and in-room is another. The papers that you reference are not done with in-room test signals.
Let me present a perceptual-psychology way to look at this... many people will not know what to do with this,
Perceptions (such as the location of an oboe in an orchestra) are made up of a variety of cues and these tend to be softish (unlike a celluloid camera but quite like a processing modern digital camera) and sometimes conflicting (like when there are sound reflections, eh).
Your brain makes the best guess about how to use these various inputs in creating a solid perception which we "see" as solid, stable, etc.
So the issue for a recording producer (I always like to take that point of view because it is in some ways the creative mirror image of what your brain does) is to set up cues of time and loudness (AKA pan knob to locate the single-mic'ed oboe).
So there are general rules to recognize what cues the brain typically relies on - and these may surprise you. "Ecological validity" as we say in the trade.
Does that help?
Ben
Perceptions (such as the location of an oboe in an orchestra) are made up of a variety of cues and these tend to be softish (unlike a celluloid camera but quite like a processing modern digital camera) and sometimes conflicting (like when there are sound reflections, eh).
Your brain makes the best guess about how to use these various inputs in creating a solid perception which we "see" as solid, stable, etc.
So the issue for a recording producer (I always like to take that point of view because it is in some ways the creative mirror image of what your brain does) is to set up cues of time and loudness (AKA pan knob to locate the single-mic'ed oboe).
So there are general rules to recognize what cues the brain typically relies on - and these may surprise you. "Ecological validity" as we say in the trade.
Does that help?
Ben
wouldn't that say that the upper frequencies do not matter?We're not reading the bible here searching for hidden meanings. The quote says:
"Wightman and Kistler (1992) have shown that low-frequency ITD cues dominate ILD cues; in other words, when the stimuli contain low frequencies, the position of the auditory image is determined by the ITD cue regardless of the ILD cues.""
I could not find the quoted paper online, but the abstract says:
"In the experiments reported here, the interaural phase relations in the processing algorithms are manipulated in order to produce stimuli in which the interaural time difference cues signal one direction and interaural intensity and pinna cues signal another direction. The apparent directions of these conflicting cue stimuli almost always follow the interaural time cue, as long as the wideband stimuli include low frequencies."
Well, only the second paper is from Hartmann (the one that is actually suggesting the conclusions of the first are to be taken with a grain of salt). However, they use 1/3 octave noise, which is rather unrealistic, I would think, while the original paper uses wideband noise.
Still, you're dismissing someone's work (published in a peer-reviewed (?) J Acoust Soc Am. ) because you don't like them personally ?
No, I have not, I'll try to get a copy. But AFAIK from the papers above, Blauert was mainly concerned with free field/anechoic, while we are more talking about in-room, where things like binaural coherence play a role.
Doesn't that say that only the woofer determines directionality?
It that is the case then why bother with he upper frequencies at all since the woofer determines the directionality?
No, it does not mean that. If no ITD cues or no low frequency content is present, then ears fall back to ILD, hence you'd better get high frequencies right as well.
Myhrrhleine - I agree. I am having trouble with the implications of what is being said. It goes against my understanding and it goes against experience. If I did not have so many other irons-in-the-fire I would search out this issue. But right now I will just have to let it lie as an unresolved question - one of many I suppose.
I do not mean to say that localization should be fundamentally different in free fielad vs. in -room, just that the room does change the game via reflections - something that the Hartmann paper puts in Terms of "binaural coerence" altered by reflection. Griesinger is also pointing it out when he talks what the room is doing to localization.
It's true that the first paper does an anechoic conditions analysis (headphones). The second paper from Hartmann does not use in-room Stimuli as such, but ones that simulate lack of coherence that should be similar to the one created by room reflections.
I do not know exactly how Blauert's vision on this looks like and certainly understand your point here and the fact that the cited papers go against that spirit. Maybe there is a common point somewhere. But definitely a interesting area.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.