What is the ideal directivity pattern for stereo speakers?

[LineArray]

Some summarized stuff on direction perception:

http://www9.dw-world.de/rtc/infotheque/sound_perception/sound_perception.pdf

Graph on left vs. right ear level vs. frequency with source 30 degrees on the left
in Section "THE BASICS OF SPATIAL PERCEPTION"

HeadWize - Technical Paper: The Elements of Musical Perception by HeadWize

A level difference of approx. 4dB to 10dB between left and right ear in the range of 400Hz and 5Khz is not
really a crosstalk cancellation. It may be called a "crosstalk attenuation" ...

Due to the small crosstalk attenuation stereo interchannel level differences have to be
much higher than ILD whould have to be in a natural (single sound source) hearing situation,
to perceive the image at the same angle as a stereo phantom.

http://www.sengpielaudio.com/InterchannelLevelDifferenceTimeDifference2.pdf

 

[Elias]

When a mono signal is playing on two properly set up stereo speakers in a properly set up room, both ears will receive the same signal with the exact same amplitude, phase, and time delay.

...

At mid frequencies up to ~800Hz phase is compared to look for left-right positioning, because both ears are receiving the same signal in the same phase, this localizes the sound along the centre line.

...

(If only one speaker was playing there would be a phase difference between left and right ear, but with both speakers playing the phase of left and right speakers sum together at each ear to result in the signal at each ear being in phase)

...

Again amplitude will be identical in both ears at high frequencies, which strongly localizes the sound to the centre line.

...


In a 60 degree stereo triangle there may still be some pinnae response shift, but there is no amplitude imbalance between ears - both ears are receiving equal amplitude high frequencies with the same frequency response contour, so one major left right localization cue is eliminated, forming an image in the centre.

...

It's true that the precise frequency response contour will be different than an actual directly ahead single sound source, something that's been discussed in this (?) and other threads in relation to errors in tonal balance of stereo, but without the amplitude imbalance between left and right ear to signal an off centre sound source, how can the brain distinguish this reliably from only the pinnae's modification to the frequency response ? I think the answer is that it can't, at least not at narrow angles.

...

So I'm calling BS on your

Sorry DBMandrake but what you say is BS.

All of your claims are based on false assumption that HRTF, or better yet ATF, would be indentically symmetric. It is not.

An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.

http://www.embracingsound.com/docs/SingleBipolarLoudspeakersystemforstereoreproductionb0.91.pdf


- Elias

 

[Pano]

Speaking of phantom images, I got a really weird and surprising one yesterday in MONO (sort of).

Sitting backstage running video. I had a little monitor speaker about 4 feet in front straight of me to hear the presenter speaking. I was wearing intercom with the right ear fully covered. The sound wasn't coming out of the little monitor speaker, but about 3 feet to its left. Striking effect.

I say "sort of" mono, because there was a lot of indirect sound coming thu the backdrop behind the monitor speaker. So it was a mono point source with a very big, defused background channel - in a manner of speaking. Still, if I removed the intercom headphone from my right ear, the sound would go straight to the little monitor speaker (with the reverb of the hall behind it.) Cover the right ear again and the image was off to the left of the monitor speaker. A phantom mono image.

Don't think I've ever noticed that effect before.

 

[markus76]

All of your claims are based on false assumption that HRTF, or better yet ATF, would be indentically symmetric. It is not.

That's doesn't really render anything Simon said false because our ears don't change. We learn to listen with whatever shape. Pinna cues for 30° might be physically different for the left and right ear but perceptionally they're the same.

P.S. Please stop name calling. Makes you both look bad.

 

[Pano]

Agree.

From the linked paper.
"The unsymmetrical ILD is the reference for this unique listener and is therefore correctly resolved."

 

[erjee]

To claim as fact that it's 50% of the general population without any research or carefully controlled listening tests with a large sample size is preposterous though.

So is advocating the abandonment of stereo as a useful medium on the basis of a few people who don't perceive it the same way as the majority. Should we not bother making colour TV's because a few people are colour bind and can't get the full benefit from them ?

+1

 

[LineArray]

The quasi stochastic nature of most music material may IMO
contribute largely in interference being not as detrimental for
phantom localization than would be expected with sinusoidal signals.

Crosstalk "cancellation" below 5Khz solely can IMO hardly explain
that.

But again: Who observes <= 10 cm lateral shift from median
plane when listening to speakers separated about 3m for listening
distances up to 3m ?

Most speakers are (by most customers most of the time) used
outside the stereo listening area.

 

[Elias]

That's doesn't really render anything Simon said false because our ears don't change. We learn to listen with whatever shape. Pinna cues for 30° might be physically different for the left and right ear but perceptionally they're the same.

P.S. Please stop name calling. Makes you both look bad.

Agree.

From the linked paper.
"The unsymmetrical ILD is the reference for this unique listener and is therefore correctly resolved."

But that's hardly the point.

Some are provided more of the advance asymmetry of HRTF / ATF than others. With better tools the job is more easily done.

I remain in my earlier quoted claim:
"it has been shown that physiological differences may result in some listeners performing much better than others on auditory localization tasks."


- Elias

 

[erjee]

I know its been pointed out before, but please don't underestimate the degree of left-right cross talk cancellation above 2Khz afforded by the head and pinnae.

Simple sound wave interference apps like the one you've linked to assume 100% cross-talk between left and right speakers at all frequencies (eg, single omni microphone) which is not how we hear.

We have a head. We have an ear on each side of our head. The ear on one side of our head is occluded from sounds coming from the other side, which leads to a high level of attenuation at high frequencies for sounds from the wrong hemisphere.

Provided that the angular separation of the speakers is great enough, there is enough cross-talk reduction above 2Khz to minimize comb filtering. Because relative phase shift between left and right ear is ignored above 2Khz, only the amplitude variations at each ear caused by comb filtering are important.

If there's very little treble from the left speaker reaching the right ear canal to sum with the signal from the right speaker, there will be very little comb filtering and thus very little amplitude change with lateral head movement.

...

...which is nicely illustrated with this simple simulation:

 

[speaker dave]

Sorry DBMandrake but what you say is BS.

All of your claims are based on false assumption that HRTF, or better yet ATF, would be indentically symmetric. It is not.

- Elias

"The unsymmetrical ILD is the reference for this unique listener and is therfore correctly resolved." (Sentence directly above the figure.)

HRTFs are known to vary greatly in detail, while being generally similar for all. Also note the expanded scale.

and people wearing glasses hear better stereo!

David S.

(I see Pano spotted the same line.)

 

[Pano]

Yeah, I thought was great.

 

[DBMandrake]

That's doesn't really render anything Simon said false because our ears don't change. We learn to listen with whatever shape. Pinna cues for 30° might be physically different for the left and right ear but perceptionally they're the same.

That's kind of what I thought, although I didn't mention it explicitly.

In an equidistant stereo set up a mono signal will experience significant channel cross-talk at low frequencies, and significantly reduced cross-talk at high frequencies, but the end result is still that the two pinnae will receive the same signal with same amplitude, phase, and time delay, from the same (but opposite) incident angles, eg +/- 30 degrees or so.

Whether the specific HRTF of each ear is a bit different for the same azimuth doesn't matter because our brain has learnt the intimate details of our HRTF over many years. We know our own ears, and the perceptual cue says ~30 degrees.

The interesting bit (and I'm not sure if this is answered in research or not) is why a simultaneous high frequency HRTF cue of 30 degrees left in the left ear and 30 degrees right in the right ear can resolve to straight ahead.

Does the brain just average the estimated angle from each individual ear, obtaining a straight ahead direction ? (Normally the average angle derived from each ear of a single sound source would actually point at the source, with the near ear probably providing a more accurate estimation of azimuth)

Or is left-right amplitude asymmetry at high frequencies an essential cue for azimuth detection ?

In other words, if we eliminate the amplitude asymmetry, presenting each pinnae with the same signal, does the brain decide that additional contradictory azimuth cues from the shape of the high frequency response "don't add up" and just ignore them ?

 

[markus76]

But that's hardly the point.

Some are provided more of the advance asymmetry of HRTF / ATF than others. With better tools the job is more easily done.

I remain in my earlier quoted claim:
"it has been shown that physiological differences may result in some listeners performing much better than others on auditory localization tasks."

- Elias

Yes, Blauert mentioned some people performed better with other people's HRTF in lateralisation experiments but this doesn't say anything about the cause. You believe it's symmetry, I'd say it's unproven.

Btw, highlighting your beliefs doesn't make them come true.

 

[DBMandrake]

...which is nicely illustrated with this simple simulation:

It is ?

 

[markus76]

In other words, if we eliminate the amplitude asymmetry, presenting each pinnae with the same signal, does the brain decide that additional azimuth cues from the shape of the high frequency response "don't add up" and just ignore them ?

Moulton believes that our brain interprets it as 2 reflections with the direct sound missing. I believe we don't have any real explanation.

 

[erjee]

It is ?

Ripple Tank Simulation

This java applet is a simulation of a ripple tank. It demonstrates waves in two dimensions, including such wave phenomena as interference, diffraction (single slit, double slit, etc.), refraction, resonance, phased arrays, and the Doppler effect.

Be sure that the 'fixed edges' checkbox is off for acoustic simulations.

The grey blob is the head. There are no walls and the source is in the upper right corner. One can easily observe the shading and even some interference/comb filtering due to reflecting waves off the head.

 

[markus76]

It is ?

Ripple tank applet. It shows how the waves (top right corner) bend around an object (circle lower middle).

 

[DBMandrake]

Ripple Tank Simulation



Be sure that the 'fixed edges' checkbox is off for acoustic simulations.

The grey blob is the head. There are no walls and the source is in the upper right corner. One can easily observe the shading and even some interference/comb filtering due to reflecting waves off the head.

Ahhh, gotcha.

I like the iPhone / iPad version too, very cool.

 

[erjee]

Ahhh, gotcha.

Nice eh Like 'a picture says more...'

 

[speaker dave]

Ripple tank applet. It shows how the waves (top right corner) bend around an object (circle lower middle).

Curved wavefronts!