What makes you think room reflections are necessary for the formation of a centre phantom channel in a stereo playback system ? Is this just a hypothesis of yours, or have you actually tested it ?
Headphones localize the image inside your head because the HRTF of the outer ear is bypassed. (and in the case of in-ear-canal headphones the pinnae is also bypassed)
If you haven't tried it before, set up your speakers outdoors at normal indoor listening distances and angles but well away from any walls or objects - you'll hear a perfectly good centre phantom channel, in fact it will usually be better than indoors.
The human brain is perfectly able to localize images without room reflections - can you figure out which direction a noise is coming from outdoors ? Of course you can.
In fact the opposite of what you say is true - the brain has to do a lot of extra processing to try to ignore or "window out" reflections so that the direction of the original sound (the first arrival) can be determined despite time delayed reflections coming from different directions also being present.
This is all part of the precedence effect, and probably the reason the precedence effect exists in the first place, as an adaption to help determine the true direction of a sound source in a confusing acoustic environment.
Last edited:
Hi Keyser,
brave but good attempt
I think people here will not come to a final agreement. History in forums shows that there does not seem to be a consensus. In the end it all boils down to preference and duty and probably there is no right or wrong.
My current simplified picture of the two extreme positions:
It seems that people, who are after enjoying the music as such, prefer speakers with wide dispersion and more reflections. People, who are doing a recording, mixing or whatever job will likely prefer pinpoint type of speakers with narrower dispersion because they need the feedback if the mics have been placed in the right spot for what they are trying to capture. The "music lover" does typically not care so much about that aspect and in turn prefers a representation that is closer to a live venue with a nice mixture of direct and indirect sound; also on the recording itself. Pinpointing does not play a major role as “there is no pinpointing in Carnegie Hall either”.
It is no secret that I am more in the camp of the music lovers. On the other hand it is my experience that dipoles and an omni like Pluto are able to combine the best of the two worlds, although I believe that there is no "one fits all" approach. That means, if implemented well, such speakers could cover the needs of a good part of a population if they have an appropriate room, which is no real challenge.
brave but good attempt
I think people here will not come to a final agreement. History in forums shows that there does not seem to be a consensus. In the end it all boils down to preference and duty and probably there is no right or wrong.
My current simplified picture of the two extreme positions:
It seems that people, who are after enjoying the music as such, prefer speakers with wide dispersion and more reflections. People, who are doing a recording, mixing or whatever job will likely prefer pinpoint type of speakers with narrower dispersion because they need the feedback if the mics have been placed in the right spot for what they are trying to capture. The "music lover" does typically not care so much about that aspect and in turn prefers a representation that is closer to a live venue with a nice mixture of direct and indirect sound; also on the recording itself. Pinpointing does not play a major role as “there is no pinpointing in Carnegie Hall either”.
It is no secret that I am more in the camp of the music lovers. On the other hand it is my experience that dipoles and an omni like Pluto are able to combine the best of the two worlds, although I believe that there is no "one fits all" approach. That means, if implemented well, such speakers could cover the needs of a good part of a population if they have an appropriate room, which is no real challenge.
The ear-brain mechanism is fully intriguing. Since the only axis 2 horizontally displaced ears can analyze for image location DIECTLY is left to right (X axis), it appears that we use reflection and frequency response cues to sense Y and Z axis. Say what?! I've been researching this for decades.
The pinea of the ear has a unique shape that is well understood by the ear-brain mechanism (from experience if not genetics). The brain knows that when sounds go up and down physically, there is a well known variation in relative frequency response in the approx. 8kHZ region that is created by the reflections in the pinea (outer ear). This works best when there is change, and less when the response shape is static.
To sense where on the Z axis a sound is, I don't see how else we could detect that with two horizontally displaced ears except by analyzing all related reflections, and placing it by relation to all other sounds. A comparative analysis of the reverb/reflections.
As far as the X axis (left to right), it appears that we sense image location more by amplitude comparisons in the upper frequencies, and more by timing comparisons in the lower frequencies (below about 1kHZ).
A nice demo on distance and reflections: Initial Time Delay Gap ITDG
FWIW, the most spacious stereo I've ever heard cam from narrow pattern loudspeakers crossfired midway between the listening position and the loudspeaker.
Dan
FWIW, the most spacious stereo I've ever heard cam from narrow pattern loudspeakers crossfired midway between the listening position and the loudspeaker.
Dan
There is still the reflection from the ground outside. If there are no reflections and only direct sound from source to ear, and of course you have to be equidistant and at the centerline of the two sources, the image will form in your head.
The reason, perhaps, that it images better outside is because the ground reflection is early enough to enhance the sense of localization. Also, in my limited experience, it is most certainly more tiring to listen to a speaker with narrow dispersion than one with a wide dispersion. The brain is working harder to localize the sound, but it is not getting enough cues from the reflected sound.
Only when reflections start coming in late that the sound becomes muddy and hard to follow.
And this is not only my hypothesis:
Linkwitz-Publications
Listen to "Talk with Q&A"... its pretty long, but worth it.
Last edited:
This book can help with understanding localization: http://human-factors.arc.nasa.gov/publibrary/Begault_2000_3d_Sound_Multimedia.pdf
Dan
Dan
There is perhaps a different way of looking at the sound inside the head thing.
With headphones the left ear receives only sound from the left headphone and visa versa. But with a pair of speakers, even if there are no reflections present the left ear still receives a lot of information from the right loudspeaker and the same is true for the right ear and the left loudspeaker.
Lets not forget that we also determine the position of sound sources by the difference in sound between the left and right of the head principally because the shape of the head and then the ear alters the frequency response of the incoming sound so that what the left and right ear hear are quite different. The brain then works its magic and we say - it's over there.
If there happened to be a single loudspeaker replaying sound in a reflection-less environment you'd still be able to determine where it is because of these differences. You don't need reflections to know that it is over there. If you then brought in a second loudspeaker, say in a typical triangle configuration, as would be used in stereo listening, you'd know that the left speaker is over there and the right one were over there. The sound would not be 'in your head' for each individual loudspeaker, so why would it suddenly jump inside your head if you turned both on at once?
With headphones the left ear receives only sound from the left headphone and visa versa. But with a pair of speakers, even if there are no reflections present the left ear still receives a lot of information from the right loudspeaker and the same is true for the right ear and the left loudspeaker.
Lets not forget that we also determine the position of sound sources by the difference in sound between the left and right of the head principally because the shape of the head and then the ear alters the frequency response of the incoming sound so that what the left and right ear hear are quite different. The brain then works its magic and we say - it's over there.
If there happened to be a single loudspeaker replaying sound in a reflection-less environment you'd still be able to determine where it is because of these differences. You don't need reflections to know that it is over there. If you then brought in a second loudspeaker, say in a typical triangle configuration, as would be used in stereo listening, you'd know that the left speaker is over there and the right one were over there. The sound would not be 'in your head' for each individual loudspeaker, so why would it suddenly jump inside your head if you turned both on at once?
You're looking at it only from the perspective of comb filtering here, and while that's a part of the issue it's not the whole thing, and possibly not even the most important aspect.
There's a couple of psychoacoustic effects I'm thinking of applicable to the high midrange low woofer approach (there may be others too) - one is the widely different time delay of the floor reflection of each driver and how it relates to the precedence effect, the other is apparent source localisation of sounds whose different frequency components are produced by drivers at different heights, and how both of these relate to possible crossover frequencies.
I think everyone agrees that for good quality midrange you can't have the midrange driver too close to a side-wall or floor boundary, but why ? I would say that the reason is to make sure that there is a certain minimum time delay for the first reflection, such that the precedence effect can work its magic.
With sufficient time delay we perceive the direct arrival above ~300Hz, and we don't really hear the bad effects of comb filtering. Too short a delay and it all fuses together, allowing us to hear the response irregularities of the comb filtering.
Another possible aspect of this (just a hypothesis though) is that any resonant behaviours of the reflective surface (such as a panel resonances on a wall) will be audible if the time delay is short enough to allow aural fusing, but largely ignored as a component of the delayed signal if the time delay is in the precedence window. (The same probably applies to the floor and wall behind the speaker too)
So although a steady state measurement might show show the 2nd, 3rd, 4th harmonic notches in the comb filtering from the midrange driver being present, we won't really hear them above the Schroeder frequency due to sufficient time delay of the reflection. (Also, with multi-path reflections any deep notches would tend to be eliminated at the listening position even in the steady state signal)
The implication from this is that the crossover frequency should be at or close to the Schroeder frequency - for higher frequencies where the brain can separate first arrivals from reflections, we want those produced by a driver with a sufficiently delayed 1st reflection, provided to us by sufficient height. On the other hand for frequencies below the Schroeder frequency where we tend to hear the steady state response we want those produced by a driver close to the floor, whose close proximity puts its comb filtering artefacts above it's passband.
The second factor is localisation of the sound source when you split bass and midrange into two drivers with considerable vertical separation. Low frequencies can't be localised, midrange frequencies obviously can, but what is the cut-off point ? If the source of bass can be any direction in the room, most people say 80Hz, but if it's much closer to the midrange driver we can go a lot higher. In fact if the woofer is vertically below the midrange driver, I think we can go up to about 300Hz without the localisation shifting from the midrange driver. (empirically derived)
So if the crossover frequency is too low (say 80Hz) audible holes in the steady state response below the Schroeder frequency will open up due to boundary cancellation, while if the crossover frequency is too high (say 500Hz) midrange clarity will be compromised due to a significant part of the midrange being produced by a driver close to the floor, as well as the localisation of the sound source starting to stretch vertically - the woofer will start to draw attention to itself, when ideally the sound should sound like it's coming from half way between midrange and tweeter with the woofer not being noticed as a separate sound source.
Another consideration I'd throw into the mix is that the proximity of the woofer to the floor has a lot more effect on the 2nd harmonic floor/ceiling mode (typically between 100-150Hz) than it does at the floor bounce differential frequency (typically ~300Hz for a high mid-woofer) and that a lot of the benefit of putting the woofer close to the floor applies in this 100-150Hz region rather than the calculated floor bounce frequency. I've typically found the calculated floor bounce notch disappears once you get more than a couple of metres away in a room back due to multi-path reflections from multiple boundaries, whereas the effects of the floor/ceiling mode cancellation tend to persist at most listening locations, so aren't as temporal.
One final thought (more questions than thoughts actually) is the effects on the directivity index / power response of a low mounted woofer, versus a high mounted one. If you design a speaker with a high mounted woofer (or a 2 way using a mid-bass) and apply a theoretical 6dB baffle step correction you'll end up with a certain frequency response in room and a certain shift in directivity index from bass to midrange.
If you then move the woofer so that it's close to the floor, the measured bass response at the listening position, particularly in the upper bass region will increase by many dB. To get a balanced overall result it's necessary to equalise this in the network somehow, for example by lifting the drive level of the midrange and tweeter.
What does this do to the power response and directivity index at low frequencies ? By putting the woofer closer to the floor we have less destructive interference and greater response at the listening position for the same power level, but have we increased the effective efficiency or have we really just increased the directivity index ?
How about if we also have the baffle width such that the baffle step half way frequency occurs at the crossover frequency, then what will we find in directivity index shift between bass and treble vs a speaker whose drivers are all a long way from the floor ?
Seems like a very successful approach would be - (1) crossover frequency at the Schroeder frequency, (2) baffle width to put the baffle step frequency also at the Schroeder frequency (3) woofer very close to the floor, (4) midrange high up, (5) network attenuation levels adjusted to give the midrange and tweeter a bit less than 6dB of baffle step correction to allow for the floor gain of the low mounted woofer.
The constraints of the different factors I've discussed would lead to the crossover frequency being somewhere between 200-300Hz.
Last edited:
Guys, You should make the distinction between lateralization and localization - two different things
You should also check this one out:
Sound reproduction: loudspeakers and ... - Google Books
let me quote:
bolds are mine
the reality of an anechoic chamber is just as ra7 describes it - a disaster, sound in the head
it is not just hypothesis, it was rather painfully tested, what a disappointment...
The illustration below from Spatial Hearing by Jens Blauert demonstrates the principle:
The picture tells us the ear has trouble distinguishing what spectral information is coming from the source and what is caused by the Pinna and ear canal.
Dan
but this is for the median plane only
OTOH there is hypothesis that ATF - anatomic transfer function - plays also an important part in localisation mechanism in the horizontal plane through "interaural spectral differences" cues as ITD/IPD and IID/ILD cues tend to be ambiguous/unreliable in normal reverberant environment
but I couldn't find any studies dedicated specifically to this question
but we can ask the question - Does the duplex theory still hold?
and in fact such question is exactly being asked even in case of "single sound source in free field":
http://dafx10.iem.at/proceedings/papers/Majdak_DAFx10_Tutorial4.pdf
and this is the easiest case for the theory! now how about multiple virtual sound sources in reverberant space?!!
Dan does the chart above assume one ear or two? I can easily see that one ear might be pretty bad at localisation, but two be fine. Kind of like how no matter how good a single eye is it cannot give you depth perception, but two mediocre eyes can.
If you get 'sound in the head' syndrome in an anechoic chamber, what then happens to the sounds that don't occur equally in both the left and the right loudspeaker? I am assuming here that some other process besides simple localisation takes over for the brain to think that the sound is 'in' the head. Ie localisation would tell us it's over there if the loudspeakers were played individually, but when two are combined the phantom image pops inside the head.
I do not have a problem with this, but what would happen for example if the sound was slowly panned from the left to the right. In a standard room with loudspeakers you'd expect the sound to move across the sound stage. But from what you're saying, in an anechoic chamber, it sounds like the sound would start out to the left then move over to the right whilst moving closer to you as the listener, then when the signal is equal in both the left and the right, it would pop into your head, then as it pans to the right the sound would leave your head and then move over to the right loudspeaker.
No doubt if that happens it would create some pretty funky sound stage, with some stuff in your head, and then other stuff off centre and in front of you. With headphones you get the 'in the head' experience, but any stuff beyond only exists to your left and to your right, not in front of you.
If you get 'sound in the head' syndrome in an anechoic chamber, what then happens to the sounds that don't occur equally in both the left and the right loudspeaker? I am assuming here that some other process besides simple localisation takes over for the brain to think that the sound is 'in' the head. Ie localisation would tell us it's over there if the loudspeakers were played individually, but when two are combined the phantom image pops inside the head.
I do not have a problem with this, but what would happen for example if the sound was slowly panned from the left to the right. In a standard room with loudspeakers you'd expect the sound to move across the sound stage. But from what you're saying, in an anechoic chamber, it sounds like the sound would start out to the left then move over to the right whilst moving closer to you as the listener, then when the signal is equal in both the left and the right, it would pop into your head, then as it pans to the right the sound would leave your head and then move over to the right loudspeaker.
No doubt if that happens it would create some pretty funky sound stage, with some stuff in your head, and then other stuff off centre and in front of you. With headphones you get the 'in the head' experience, but any stuff beyond only exists to your left and to your right, not in front of you.
You should also check this one out:
Sound reproduction: loudspeakers and ... - Google Books
let me quote:
bolds are mine
the reality of an anechoic chamber is just as ra7 describes it - a disaster, sound in the head
it is not just hypothesis, it was rather painfully tested, what a disappointment...
Thanks for that Graaf! I didn't think anybody would go through the pains of verifying this was true. Well, they were expecting something different anyway.
The point of all of this is that wide dispersion is good. But it must be smooth so as to maintain the tonal balance with the on-axis (direct) sound.
ra7:
Regardless of the effects exploited in spatial hearing, being it ITD, ILD,
interaural time shift of envelope, i cannot see reflections from
room walls or even ground being needed for lateralization
or estimation of elevation of a sound source.
Localisation works more precise, when the head is allowed to move,
which is normally the case. Moving the head in a sound field changes
ITD, ILD, different HRTFs are chosen by moving the head.
With headphones the "sound field" is moving with the head, no change
occurs due to motion of the head, which may contribute to
"in the head localisation".
I would like to see some evidence for that claimed
"in the head localization" occurring with center phantom sources
using loudspeakers in anechoic chambers (head not fixed).
I also interpret SL differently, like he is trying to find conditions that
support the ear/brain in attempting to put the listening room "beyond
the acoustic horizon". The room is not needed essentially for localization
in stereo listening or even natural listening.
There are conditions however which make a room's reverberant soundfield
less disturbing for localization.
That a concert hall contributes to fidelity in terms of "fusion" and "envelopment"
is without question, and to maintain both in the home environment is desirable.
But "support of localization" is not what the listening room provides or adds
to a given recording IMO.
Regardless of the effects exploited in spatial hearing, being it ITD, ILD,
interaural time shift of envelope, i cannot see reflections from
room walls or even ground being needed for lateralization
or estimation of elevation of a sound source.
Localisation works more precise, when the head is allowed to move,
which is normally the case. Moving the head in a sound field changes
ITD, ILD, different HRTFs are chosen by moving the head.
With headphones the "sound field" is moving with the head, no change
occurs due to motion of the head, which may contribute to
"in the head localisation".
I would like to see some evidence for that claimed
"in the head localization" occurring with center phantom sources
using loudspeakers in anechoic chambers (head not fixed).
I also interpret SL differently, like he is trying to find conditions that
support the ear/brain in attempting to put the listening room "beyond
the acoustic horizon". The room is not needed essentially for localization
in stereo listening or even natural listening.
There are conditions however which make a room's reverberant soundfield
less disturbing for localization.
That a concert hall contributes to fidelity in terms of "fusion" and "envelopment"
is without question, and to maintain both in the home environment is desirable.
But "support of localization" is not what the listening room provides or adds
to a given recording IMO.
Last edited:
'head not fixed' is the operative phrase here. I don't move my head constantly while listening. Why would you do that in an anechoic chamber? It was implied, but not stated, that the head is more or less fixed.
The point is that it is the speaker's dispersion that is more important than the room it is put in.
The point is that it is the speaker's dispersion that is more important than the room it is put in.
I guess slight movement of the head from time to time is quite
normal ...
Same for estimation of distance of sound sources, there are more
mechanisms than just the initial time delay gap, which may also be
captured by a microphone during a recording.
Types of sources known to the listener - e.g. "male speaker" - change
their characteristic spectrum of speech sounds depending on distance
even in free space. Every source has its specific frequency dependent
radiation pattern, which allows estimation of distance even without
reflecting boundaries. All those cues may very well be delivered with
the recording even if artificially introduced in a radio play e.g.
If the reverberant field in the listening room would be essential in judging
lateralisation and distance of phantom sources properly, control rooms
would have higher reverberation time and omni sources would be the
preferred monitors in studios ... which obviously is not the case.
normal ...
Same for estimation of distance of sound sources, there are more
mechanisms than just the initial time delay gap, which may also be
captured by a microphone during a recording.
Types of sources known to the listener - e.g. "male speaker" - change
their characteristic spectrum of speech sounds depending on distance
even in free space. Every source has its specific frequency dependent
radiation pattern, which allows estimation of distance even without
reflecting boundaries. All those cues may very well be delivered with
the recording even if artificially introduced in a radio play e.g.
If the reverberant field in the listening room would be essential in judging
lateralisation and distance of phantom sources properly, control rooms
would have higher reverberation time and omni sources would be the
preferred monitors in studios ... which obviously is not the case.
To get back to the original question; what's the best radiation pattern? I think the weakest link will usually be the way the speakers interact with the room acoustics. Generally it's best to avoid any abrupt radiation differences over frequency due to speaker driver differences/crossover frequencies and such, but every room is different, and that's what would make me want to use one speaker design or another. Some people want the very best imaging at the sweet spot, others want a nice balanced sound anywhere in the room and don't care about having the best of image localization.
If you want to reproduce any imaging cues that are embedded in a recording, room acoustics seem likely to be your biggest enemy, so limited or "controlled" directivity speakers would seem the better choice. If however the recording has little or no imaging info, side wall reflections can synthesize a greater sense of spaciousness, much as dipoles do primarily on the Z axis.
Personally, I'm intrigued by what I've read about wideband line arrays. Roger Russel, head of speaker development at McIntosh in the 1960's - 90's (I think), and primary creator of the ML series speakers in the early 1970's is now preferring 24 3 inch drivers per side, in a vertical line array, with active EQ, for all frequencies. Him and several others have said that wideband line arrays are an astounding experience. I suspect that the main reason is because of how they interact with the room acoustics. Especially in the lower midrange and bass.
If you want to reproduce any imaging cues that are embedded in a recording, room acoustics seem likely to be your biggest enemy, so limited or "controlled" directivity speakers would seem the better choice. If however the recording has little or no imaging info, side wall reflections can synthesize a greater sense of spaciousness, much as dipoles do primarily on the Z axis.
Personally, I'm intrigued by what I've read about wideband line arrays. Roger Russel, head of speaker development at McIntosh in the 1960's - 90's (I think), and primary creator of the ML series speakers in the early 1970's is now preferring 24 3 inch drivers per side, in a vertical line array, with active EQ, for all frequencies. Him and several others have said that wideband line arrays are an astounding experience. I suspect that the main reason is because of how they interact with the room acoustics. Especially in the lower midrange and bass.
See the Pi7 cornerhorn. Made exactly as you say and for those stated reasons.
http://www.pispeakers.com/Pi_Speakers_Info.pdf
http://www.pispeakers.com/Pi_Speakers_Info.pdf
"slight" that is...? "from time to time"?
How these quantities relate the time-frame of localization process? We are talking about <0.001 s here - not about from time to time, and this <0.001 s corresponds to <34.4 cm - quite slight for a head movement
are drilling woodpeckers or are we humans?
First of all, you can't put 'ideal' and 'stereo' in the same sentence without making an oxymoron. The best you can ask for is 'least compromised', but then you have to be clear about least compromised in what respect? Would you like to clarify?
Well, #3 is impossible as soon as you have a room, ditto #4.
I can't see any point in having #6. Why have it?
#7 has to be wrong. Someone justifying their choice of speakers? Try 'sufficiently wide for the audience area'.
#9 is ridiculous. What room? Speaker placed where? It is better to EQ below 200Hz.
You mean, before you put it in a room? That would depend on the room it is destined for, ironically.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.