|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
64.252.138.132
In Reply to: RE: Early reflections: 5 ms time window for sound source localisation? posted by KlausR. on July 05, 2010 at 22:20:52
5 milliseconds is the point at which fusion between the original source and its reflection no longer occurs in a significant number of listeners:FUSION ECHO THRESHOLDS
Freyman et al. 1991 clicks 5–9 ms lag heard on 50% of trials
Yang and Grantham 1997a clicks 5–10 ms lag clearly audible on 75% of trials
Litovsky et al. 1999 clicks 5–10 ms lag clearly audible on 75% of trialsDISCRIMINATION CRITICAL THRESHOLDS
Freyman et al. 1991 clicks 5–9 ms d851
Yang and Grantham 1997 clicks 5–10 ms discrimination 75% correct
Litovsky et al. 1999 clicks 5–10 ms discrimination 75% correctLOCALIZATION CRITICAL THRESHOLDS
Litovsky et al. 1997 clicks 11.4 ms lead location chosen on 75% of trials
Litovsky et al. 1997 clicks 8 ms lead location chosen on 75% of trialshttp://www.waisman.wisc.edu/~litovsky/papers/1999-3.pdf
As you can see, 5 msecs is a fairly optimistic figure. Better results will be achieved if reflections are suppressed for a longer period.
Edits: 07/13/10 07/13/10Follow Ups:
> 5 milliseconds is the point at which fusion between the original source and its reflection no longer occurs in a significant number of listeners. <
On their German site Audio Physic specifically refer to the time window for sound source localisation, not echo thresholds for clicks: "only the first 5 ms are used for source localisation, everything after that is suppressed". If by that they actually mean echo threshold, then why for clicks, why not for noise, or speech, or music?
> As you can see, 5 msecs is a fairly optimistic figure. Better results will be achieved if reflections are suppressed for a longer period. <
Figures often mentioned are 20 ms/20 dB, or 15 ms/10 dB. However, apart from some indications there is no thorough research that I’m aware of showing that in the 2-channel stereo case, using music as signal, first reflections are disturbing as a matter of principle.
Klaus
Clicks are a popular test signal I think because the ear is maximally sensitive to them, so they have utility in mapping the perceptual limits of the ear/brain system. But the fusion threshold is known to depend on the nature of the program material, see forex Fig. 16:http://www.harman.com/EN-US/OurCompany/Technologyleadership/Documents/Scientific%20Publications/13686.pdf
Despite what Toole implies in his paper, early first reflections have deleterious effects:
"The requirement to place speakers at least 1m away from reflecting surfaces (measured from the tweeter) had been found empirically for box and panel type loudspeakers. All reflections are then delayed at least
6 ms, except for the floor reflection."http://www.linkwitzlab.com/publications.htm
Linkwitz has made some attempts to codify what engineers, reviewers, and listeners who have played with dipole radiators have known for many years, that side wall first reflections should be suppressed or delayed (in my experience, the more the better) and that rear wall reflections should also be delayed as far beyond 5 msecs as practical and are then instrumental in the reproduction of depth. Also that polar response, both frequency and phase, must be uniform.
But as you point out, it seems that little systematic research seems to have been done on the issue. It remains for the most part a matter of practical art rather than scientific understanding.
Edits: 07/20/10
> Despite what Toole implies in his paper, early first reflections have deleterious effects:
"The requirement to place speakers at least 1m away from reflecting surfaces (measured from the tweeter) had been found empirically for box and panel type loudspeakers. All reflections are then delayed at least 6 ms, except for the floor reflection." <
I had a look at all publications Toole refers to w.r.t. early reflections, and some more, and apart from some indications there is no evidence for the deleterious effects of first reflections. These indications are based on
a. experiments with speakers with variable directivity index (Klein+Hummel O85a)
b. experiments where 4 (identical) loudspeakers were arranged in pairs of 2 such that in one pair the speakers were arranged vertically, in the other pair horizontally
c. Linkwitz’ monopole and dipole (AES paper 7162)
The requirement of placing speakers at least 1m from reflecting surfaces looks as if it was based on the Allison effect, ensuring that boundary reinforcement occurs for frequencies below 34 Hz, which is sufficient to avoid bass boom for the great majority of loudspeakers.
> Linkwitz has made some attempts to codify what engineers, reviewers, and listeners who have played with dipole radiators have known for many years, that side wall first reflections should be suppressed or delayed (in my experience, the more the better) and that rear wall reflections should also be delayed as far beyond 5 msecs as practical and are then instrumental in the reproduction of depth. <
Toole doesn’t exclude the possibility that early reflections may have nasty effects, e.g. in the case of loudspeakers with bad off-axis behaviour (http://www.aes.org/e-lib/browse.cfm?elib=5430). In the case of dipoles, side wall reflection may indeed be beneficial.
> But as you point out, it seems that little systematic research seems to have been done on the issue. It remains for the most part a matter of practical art rather than scientific understanding. <
Naqvi’s active listening room looks like very promising approach, but so far no results for 2-channel stereo: http://www.aes.org/e-lib/browse.cfm?elib=13418
Klaus
> I had a look at all publications Toole refers to w.r.t. early reflections, and some more, and apart from some indications there is no evidence for the deleterious effects of first reflection. <
If you're fond of 40 dB nulls in the midrange. There really is a science here, whether Toole has discovered it or not.
> The requirement of placing speakers at least 1m from reflecting surfaces looks as if it was based on the Allison effect, ensuring that boundary reinforcement occurs for frequencies below 34 Hz, which is sufficient to avoid bass boom for the great majority of loudspeakers. <
Early arrival first reflections impair midrange clarity and imaging. And in general, the further you move loudspeakers from the wall, the greater their ability to reproduce depth.
Why does Toole say that side wall reflections are beneficial in the case of dipoles? One of the main advantages of dipoles is that they minimize side wall reflections without the need for diffusion and absorption at the first reflection points. I've found that orienting a dipole so that the first sidewall reflection is nulled at the listener's position significantly improves midrange clarity. Floor and ceiling reflections, though, are beneficial in the case of a low frequency line source, since they approximate an infinite line source.
> Naqvi’s active listening room looks like very promising approach, but so far no results for 2-channel stereo <
Their description of the results are similar enough to what one observes with two channel stereo. It will be nice to have more information about audibility thresholds, though.
> > I had a look at all publications Toole refers to w.r.t. early reflections, and some more, and apart from some indications there is no evidence for the deleterious effects of first reflection. < <
> If you're fond of 40 dB nulls in the midrange. There really is a science here, whether Toole has discovered it or not. <
There is science, and there are thresholds:
http://repository.tudelft.nl/assets/uuid:7f0331e3-bc1a-4d7f-8d2a-eb5d6cc04fbf/as_salomons_19951220.PDF
So yes, reflections may generate audible coloration, but not always. In this respect Brüggen, “Coloration and binaural decoloration in natural environments”, Acustica/acta acustica 2001, vol. 87, p.400 presents interesting results.
> Early arrival first reflections impair midrange clarity and imaging. <
The simple experiments I made to test this didn’t show any such effects. There is no scientific evidence to support that statement.
> Why does Toole say that side wall reflections are beneficial in the case of dipoles? <
My fault, I intended to say that reflection treatment may be beneficial.
> One of the main advantages of dipoles is that they minimize side wall reflections without the need for diffusion and absorption at the first reflection points. <
Minimize yes, but to levels below perception thresholds? I was perhaps a bit hasty when saying that dipoles may benefit from reflction treatment. Loudspeakers with odd-looking off-axis response, and that doesn’t exclude dipoles, may benefit.
> > Naqvi’s active listening room looks like very promising approach, but so far no results for 2-channel stereo < <
> Their description of the results are similar enough to what one observes with two channel stereo. It will be nice to have more information about audibility thresholds, though. <
These pilot experiments merely confirm audibility of early reflections in Naqvi’s setup. The cues used for detection are similar or identical to those used by Olive&Toole, Bech, Schubert, Seraphim etc. Obviously, in order to detect the presence of a reflection, there must be a change in timbre, imaging or loudness, this is known since long so the Naqvi study doesn’t present anything really new. If that is interpreted as the reflection having deleterious effects, so be it. You will have noted that there is no agreement among listeners as to what cue is being used for detection (85% of the subjects, most listeners, some listeners).
Naqvi’s paper: how were reflection levels determined, do they represent real-life levels? The panels don’t work below 500 Hz, has this an effect? What about the radiation behaviour of the NXT? Did Naqvi take the absorption characteristics of real walls, floor and ceiling into account for the spectrum of the generated reflections? Are these possible flaws of Naqvi’s study? The paper has passed peer-review, were these (and other) issues discussed?
Klaus
> > Early arrival first reflections impair midrange clarity and imaging. < <
> The simple experiments I made to test this didn’t show any such effects. There is no scientific evidence to support that statement. <
First reflections are typically within the fusion range and their amplitudes are within the range that has been shown to affect tonal character and alter imaging. Besides which, referring to the figure:
http://www.avguide.com/forums/first-reflections
The remarkable thing is that the brain can so effectively minimize combing of that sort (more effectively with both ears than one, as the Salomons paper points out). I'm not sure why your experiments didn't produce an audible result.
> Minimize yes, but to levels below perception threshold?
I have no way of knowing, but I don't see why not, if precisely on the dipole null and for wavelengths that are small with respect to the side wall. In practice, other considerations make it unlikely that the dipole null will be at exactly the right angle, so acoustical treatment could potentially produce an audible improvement. The general rule of thumb seems to be "maybe, but not as necessary as with omnidirectional speakers." I doubt that science could add much here, since rooms and speaker placement vary so widely.
[Naqvi]
This is a good illustration of the disjunction between basic research and practical engineering. It's frequently unclear as to how and to what degree idealized tests of perceptual thresholds apply in real-world circumstances. Had Naqvi attempted for example to emulate the acoustic impedance of various common construction materials, the experiment would have grown unwieldly, and it would still have been of questionable utility, since the acoustic impedance of materials varies -- sheetrock, plaster, brick, glass, concrete, wood, not to mention the furnishings and objects that are found in the typical listening room (and which are sometimes used as a rough-and-ready form of acoustical treatment).
Thus, to a large extent, our scientific understanding of these phenomena lags behind our practical understanding and there's still a fair amount of practical art involved. Which, however, didn't prevent Stradivari from making a better violin than we can today, or, arguably, the designers of 18th- and 19th-century concert halls from making acoustically superior halls.
Avery Fischer notwithstanding, what usually happens I think is that our scientific understanding gradually advances and supplants practical art as it becomes available, with usually superior results. But as far as I can tell, we're still very far from having sufficient understanding of these phenomena to rely entirely on theory here. In fact, some pretty important phenomena still don't have a solid explanation, such as the superior depth rendition of dipoles. There are theories, yes, but to the best of my knowledge, little by way of solid research.
> First reflections are typically within the fusion range and their amplitudes are within the range that has been shown to affect tonal character and alter imaging. Besides which, referring to the figure:
http://www.avguide.com/forums/first-reflections <
I don’t know of any research that shows that first reflections affect tonal character and alter imaging, other than corresponding cues being used for determining perception thresholds for the single reflection scenario.
Of course, treatment makes a huge improvement, on Ethan’s graph! As long as you don’t correlate such measurements with confirmed perception thresholds, measurements alone tell nothing about the audibility of the measured effects. Also, what is worrisome is that Ethan uses, again, the recording engineer/recording mike argument to make his case. A mike will record comb filters as audible coloration, because it doesn’t have the binaural decoloration mechanism humans have. Ethan should know that.
> The remarkable thing is that the brain can so effectively minimize combing of that sort (more effectively with both ears than one, as the Salomons paper points out). I'm not sure why your experiments didn't produce an audible result. <
The experiments were simple A/B and without any controls, but if the effects were as pronounced as it is always claimed I should have heard something. Maybe the off-axis behaviour of my Genelec and Klein+Hummel is beyond reproach:-) Maybe any image change was masked by localisation blur.
Klaus
> I don’t know of any research that shows that first reflections affect tonal character and alter imaging, other than corresponding cues being used for determining perception thresholds for the single reflection scenario. <
I believe that Haas demonstrated this in 1949.
> The experiments were simple A/B and without any controls, but if the effects were as pronounced as it is always claimed I should have heard something. Maybe the off-axis behaviour of my Genelec and Klein+Hummel is beyond reproach:-) Maybe any image change was masked by localisation blur. <
I've noticed three main phenomena as speakers were moved away from room boundaries: changes in the bass and midbass response, clearer sound (correlating to intelligibility), and better (or sometimes just different) imaging (correlating in part to differences in image spread and location). Pretty much what I'd expect on the basis of theory, except perhaps in the case of the effect on depth perception, which AFAIK hasn't been adequately explained in the scientific literature.
> > I don’t know of any research that shows that first reflections affect tonal character and alter imaging, other than corresponding cues being used for determining perception thresholds for the single reflection scenario. < <
> I believe that Haas demonstrated this in 1949. <
I don’t have the Haas paper with, so I can’t comment.
> I've noticed three main phenomena as speakers were moved away from room boundaries: changes in the bass and midbass response, clearer sound (correlating to intelligibility), and better (or sometimes just different) imaging (correlating in part to differences in image spread and location). <
When moving the speakers away from room boundaries you are changing several parameters simultaneously (i.e. boundary reinforcement, room mode coupling, delay and possibly spectrum of early reflections, probably also the interaural cross-correlation), so you don’t know for sure that it’s the early reflections that are responsible for the perceived effect. In my experiments I added absorption at lateral reflection locations (Klein+Hummel) and I placed the Genelecs on a table in the middle of a large room, and placed room doors against that table to generate lateral reflections. In the first case reverberation time might have changed slightly due to the absorbers, in the second case the reflections were generated geometrically, hence without any changes to the acoustics of the setup.
Klaus
> > I've noticed three main phenomena as speakers were moved away from room boundaries: changes in the bass and midbass response, clearer sound (correlating to intelligibility), and better (or sometimes just different) imaging (correlating in part to differences in image spread and location). < <
> When moving the speakers away from room boundaries you are changing several parameters simultaneously (i.e. boundary reinforcement, room mode coupling, delay and possibly spectrum of early reflections, probably also the interaural cross-correlation), so you don’t know for sure that it’s the early reflections that are responsible for the perceived effect. <
Not all of the are due to early reflections. The bass changes are a consequence of room modes. The midbass variations are a consequence of the Allison effect. Interaural cross-correlation of the direct sound isn't an issue, since the angle subtended by the speakers is constant. That leaves early and late reflections. By rotating a pair of dipoles, it's possible to tune out the first sidewall reflections. The effect is immediately apparent. Also -- and I'm not sure why you didn't hear this -- the effect of adding diffusion or absorption at the first reflection points. I'm particularly puzzled as to why you didn't hear the presence of the doors, assuming they were at the reflection point from the perspective of your listening position. I've heard the effects of proximate boundaries many times. For example, console bounce is a common problem when setting up mini monitors.
> Also -- and I'm not sure why you didn't hear this -- the effect of adding diffusion or absorption at the first reflection points. <
As I said, these experiments were without any controls. My speakers are on the long room wall, so maybe the reflection level is already close to audibility threshold so that any additional measure won’t have audible effects.
> I'm particularly puzzled as to why you didn't hear the presence of the doors, assuming they were at the reflection point from the perspective of your listening position. <
I checked the reflection points with a mirror. As far as not hearing the presence of the doors is concerned, any image shift could have been “masked” by localisation blur, the generated comb filters could have been below detection threshold. I’m not generalizing here, but in my case reflections don’t seem to have a detrimental effect.
> I've heard the effects of proximate boundaries many times. For example, console bounce is a common problem when setting up mini monitors. <
If the desktop reflection is the only reflection, which probably is the case for nearfield listening with desktop monitors, one could, in view of the corresponding statements in Salomons’ thesis, expect an audible problem. Genelec probably has taken that particular boundary into account (http://www.aes.org/e-lib/browse.cfm?elib=13683).
> > I don’t know of any research that shows that first reflections affect tonal character and alter imaging, other than corresponding cues being used for determining perception thresholds for the single reflection scenario. < <
> I believe that Haas demonstrated this in 1949. <
I’ve looked at the original paper and the setup Haas was using was two loudspeakers at 45º, half to the left and half to the right, one as direct sound source, the other as source for the single reflection. Both loudspeakers played at the same SPL. Once again, the single source/single reflection scenario. What happens in 2-channel stereo (phantom source, see also http://www.aes.org/e-lib/browse.cfm?elib=12718) with multiple natural reflections, has not yet been thoroughly investigated, to the best of my knowledge that is.
Klaus
> I’ve looked at the original paper and the setup Haas was using was two loudspeakers at 45º, half to the left and half to the right, one as direct sound source, the other as source for the single reflection. Both loudspeakers played at the same SPL. Once again, the single source/single reflection scenario. What happens in 2-channel stereo (phantom source, see also http://www.aes.org/e-lib/browse.cfm?elib=12718) with multiple natural reflections, has not yet been thoroughly investigated, to the best of my knowledge that is. <Mine either. As far as I know, all we have are the fusion experiments and the practical experience and lore, which suggests
- Side reflections -- all reflections -- add a sense of spaciousness to a stereophonic image, but alter timbre and broaden the image when within the 10-30 msec fusion range
- Rear reflections are important to the recreation of depth and distance cues (this has been established by informal experiment)
- Linkwitz's assertion that room reflections should be spectrally similar to the original so that the brain interprets them as reverberation, a formalization I think of some very old observations regarding the spatial fidelity of dipoles and omnidirectional loudspeakers but also I think an excellent result. Again, there have been informal experiments here, e.g., the observation that adding a rear-firing tweeter to a loudspeaker with limited high frequency dispersion improves spatial rendition.
Linkwitz did do an interesting experiment and make some measurements, but his papers on the subject are more in the manner of throwing out some hypotheses and a simple experiment than a systematic investigation. The design and implementation of the sort of experiment that would significantly expand our knowledge in the manner of Haas's fusion and intelligibility experiments would I think be fairly challenging.
Edits: 08/21/10
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: