Hi All,
Something you all will have no doubt played with, is the separation between stereo speakers, which is an important variable in achieving good imaging and sound staging.
The general consensus in the field seems to be that having each speaker about 30 degrees off the centre axis give or take a few degrees, gives the best sound-staging and imaging - room permitting. (If the speakers are aimed down the long axis of the room the room width may not be sufficient to reach 30 degrees without the speakers being too close to the side walls, introducing other deleterious effects like early reflections)
But why exactly is this angle near optimum ? I'd be interested to know what research has been done on why a certain angle gives the best stereo illusion in most circumstances.
As well as the imaging in one exact spot that is down the centre axis of the speakers, I've noticed in my own tinkering that the stability of the image with small movements of the listening position in the order of a foot or so can be very sensitive to the speaker separation.
If the speakers are too close together the imaging may be good at one exact spot but will fall apart even if you move your head a couple of inches to the side, especially the phantom centre. Put them a bit further apart and the image and phantom centre will become a lot more stable and focused.
Anyone know why ? Here's my theory.
Reproducing a phantom image from a left and right speaker is a fundamentally tricky problem, due to the interference patterns formed by two physically separate speakers producing identical signals. If you tried to measure the response a few inches around the ideal listening position with an omnidirectional microphone and both speakers playing you'd notice a lot of comb filtering effects in the treble which changed with the slightest movement of the microphone.
So why don't we hear this ? It seems to me that the stereo phantom centre (and imaging in general from 2 speakers) relies partly on the physical occlusion of the opposite ear by the front of your head/face to achieve a convincing illusion.
When the separation of the speakers is great enough, and therefore the angle of separation is great enough at the listening point, there will be no direct line of sight from the left speaker to the extremities of the right ear, and vica versa, thus channel crosstalk in the highly directional and short wavelength treble will be dramatically reduced, largely mitigating the comb filtering effect in the treble.
If you play mono pink or white noise on a stereo speaker pair, turn your head at right angles in the "ideal" listening position (to listen with one ear with no occlusion) then move your head laterally relative to the speakers you can clearly hear a comb filtering effect as you move. Do the same while facing adequately separated speakers and there will be little or no perceived comb filtering. Try the same test with the speakers a lot closer together and you'll notice some comb filtering even head on.
Broad band variations in the treble of as little as half a dB can cause a significant change in the perception of imaging, (or lack of) so it's no surprise to me that even larger variations caused by destructive interference could conspire to make the sweet spot extremely small and unstable.
I've noticed the threshold in speaker spacing between an unstable image and a stable image can often be quite small - even a couple of inches too close together can ruin the stereo effect, although I'm not quite sure how it can be that fussy when it might amount to only a degree or so change in angular separation.
I'm curious to know what other peoples thoughts are on listener to speaker angle in regards to imaging and image stability.
Something you all will have no doubt played with, is the separation between stereo speakers, which is an important variable in achieving good imaging and sound staging.
The general consensus in the field seems to be that having each speaker about 30 degrees off the centre axis give or take a few degrees, gives the best sound-staging and imaging - room permitting. (If the speakers are aimed down the long axis of the room the room width may not be sufficient to reach 30 degrees without the speakers being too close to the side walls, introducing other deleterious effects like early reflections)
But why exactly is this angle near optimum ? I'd be interested to know what research has been done on why a certain angle gives the best stereo illusion in most circumstances.
As well as the imaging in one exact spot that is down the centre axis of the speakers, I've noticed in my own tinkering that the stability of the image with small movements of the listening position in the order of a foot or so can be very sensitive to the speaker separation.
If the speakers are too close together the imaging may be good at one exact spot but will fall apart even if you move your head a couple of inches to the side, especially the phantom centre. Put them a bit further apart and the image and phantom centre will become a lot more stable and focused.
Anyone know why ? Here's my theory.
Reproducing a phantom image from a left and right speaker is a fundamentally tricky problem, due to the interference patterns formed by two physically separate speakers producing identical signals. If you tried to measure the response a few inches around the ideal listening position with an omnidirectional microphone and both speakers playing you'd notice a lot of comb filtering effects in the treble which changed with the slightest movement of the microphone.
So why don't we hear this ? It seems to me that the stereo phantom centre (and imaging in general from 2 speakers) relies partly on the physical occlusion of the opposite ear by the front of your head/face to achieve a convincing illusion.
When the separation of the speakers is great enough, and therefore the angle of separation is great enough at the listening point, there will be no direct line of sight from the left speaker to the extremities of the right ear, and vica versa, thus channel crosstalk in the highly directional and short wavelength treble will be dramatically reduced, largely mitigating the comb filtering effect in the treble.
If you play mono pink or white noise on a stereo speaker pair, turn your head at right angles in the "ideal" listening position (to listen with one ear with no occlusion) then move your head laterally relative to the speakers you can clearly hear a comb filtering effect as you move. Do the same while facing adequately separated speakers and there will be little or no perceived comb filtering. Try the same test with the speakers a lot closer together and you'll notice some comb filtering even head on.
Broad band variations in the treble of as little as half a dB can cause a significant change in the perception of imaging, (or lack of) so it's no surprise to me that even larger variations caused by destructive interference could conspire to make the sweet spot extremely small and unstable.
I've noticed the threshold in speaker spacing between an unstable image and a stable image can often be quite small - even a couple of inches too close together can ruin the stereo effect, although I'm not quite sure how it can be that fussy when it might amount to only a degree or so change in angular separation.
I'm curious to know what other peoples thoughts are on listener to speaker angle in regards to imaging and image stability.
Last edited:
Ken Kantor would strongly disagree, and his designs are imaging/soundstage champs.
What angles would he advocate then ? Certainly you can use a wider angle than +/- 30 degrees and get good (but different) results, I'm more talking about there being a minimum acceptable angle, based on the geometry of the human head and ears, which I think is not a whole lot less than +/- 30 degrees.
Seems like the argumentation took by DBman was based on true empiricism .
Analysis of the phenomenon is given with mixed information ...seems that the xperimenter is being altered himself by the circumstances .
The same freq.resp.curve of a compression tweeter may sometimes look as a 'comb filter effect' ,and suggestions of how to reduce this distasteful effect may pe precious 🙂
I believe that this test is a litte frode ,since you are talking about tweeters ,it could be : 1)different match of system impedences in l-r channels 2)different mismatch in tweeter manufacturing 3)non anechoic room reflections .
Analysis of the phenomenon is given with mixed information ...seems that the xperimenter is being altered himself by the circumstances .
The same freq.resp.curve of a compression tweeter may sometimes look as a 'comb filter effect' ,and suggestions of how to reduce this distasteful effect may pe precious 🙂
I believe that this test is a litte frode ,since you are talking about tweeters ,it could be : 1)different match of system impedences in l-r channels 2)different mismatch in tweeter manufacturing 3)non anechoic room reflections .
I have to say, I'm really struggling to understand what it is you're trying to say, and how it relates to my original post so I'll respond based on what I believe you are saying, so my apologies if I have misunderstood you.
What is empirical in my original post is my perception of the stability of the imaging and phantom channel especially during lateral movement being significantly worsened when the angle of separation of the speakers is significantly less than +/- 30 degrees. Since perception of stereo imaging and sound field is something that occurs in the mind and can't be measured as such, it can by definition only ever be empirical.
I then put forward the postulation that significant reduction in treble comb filtering due to reduced left-right crosstalk when the angle to the speakers to the listener is wider, may be a factor in this. I don't know this for a fact, and my whole reason for posting was my curiosity to see if there has been any research done on this.
If a compression tweeter did have comb filter-like artefacts in it's individual polar response due to standing wave patterns and/or diffraction effects in the waveguide/horn, then this type of comb filter-like response could not be reduced by the HRTF of the listener, and is a separate issue that may cause image instability regardless of speaker separation. Such drivers would still have the same L-R crosstalk comb filtering to contend with as well.
1) ?
2) A small mismatch in the tweeters FR or sensitivity would not significantly alter the phenomenon, in fact maximum amplitude fluctuations occur in the case of perfectly matched tweeters.
3) While it's true that a single tone in the upper treble range played from either one or both speakers in most rooms will result in an amplitude that varies wildly as you move your head around just a few inches, this is an extremely narrow band phenomenon.
The summed treble reflections in a room are random in phase relationship to the direct signal, causing extremely narrow band variations in amplitude. This shows up on a steady state FR as a noise like residue in the measurement which can have a 10-20dB variation, but the 1/3oct averaged response is (assuming a tweeter with wide dispersion) relatively the same as the original signal.
There is no phase coherence between the direct treble and the treble that has bounced around the room, so there is no possibility for comb filtering. (At least not at treble frequencies. In the lower midrange and below, yes)
On the other hand the direct signal from Left and Right tweeters is phase coherent on a mono signal, and are very nearly in time alignment near the "optimal" eqi-distant listening position.
It can relatively easily be shown by comparing mono and stereo pink noise that it's not due to room reflections.
There is a nice test sound on Linkwitz Lab which includes both mono and stereo pink noise: (pink-alternating3.wav)
Accurate Stereo Performance
Try moving your head left and right (relative to the speakers) both facing the speakers and head turned at 90 degrees, comparing the difference between stereo and mono pink noise.
In the case of mono pink noise with your head at right angles you should hear obvious phasing effects as you move your head a few inches. When facing the speakers head on there will be some variation, but it will be greatly diminished, depending on the angular separation of the speakers, and the speakers themselves.
As a control, the stereo pink noise sample will show that whether your head is forwards or sideways, you can move your head around and you'll experience no phasing effect from comb filtering, despite room reflections.
If side-wall reflections were involved, you would still hear a phasing effect when moving your head on the stereo pink noise test. (This might be possible if the speaker was right against the side-wall as was the listener, close to the speaker, but not with normal speaker positioning)
Nowhere in real life non-reproduced sound can I think of a case where two sound sources with such a widely separated physical position are playing identical phase coherent signals - it's not a natural phenomenon, and the brain isn't equipped to deal with it well. I think we're lucky that our ears are on each side of our head with occlusion from each other or we probably wouldn't be able to be fooled into perceiving a satisfactory stereo image from a 2 channel system at all 😀
Comb filtering due to mixing of two signals with a small time delay (in this case an acoustic time delay) is a well known scientific phenomenon, so there is nothing empirical about that aspect of my post.
What is empirical in my original post is my perception of the stability of the imaging and phantom channel especially during lateral movement being significantly worsened when the angle of separation of the speakers is significantly less than +/- 30 degrees. Since perception of stereo imaging and sound field is something that occurs in the mind and can't be measured as such, it can by definition only ever be empirical.
I then put forward the postulation that significant reduction in treble comb filtering due to reduced left-right crosstalk when the angle to the speakers to the listener is wider, may be a factor in this. I don't know this for a fact, and my whole reason for posting was my curiosity to see if there has been any research done on this.
Don't understand what you're saying there - mixed information ? How have I been altered ? 😕
I'm not entirely sure I see the relevance of this to the discussion, the basis of the effect that I'm describing is comb filtering due to un-equal time delay from two widely (2+ metres) and horizontally separated but otherwise identical signal sources with coherent relative phase.
If a compression tweeter did have comb filter-like artefacts in it's individual polar response due to standing wave patterns and/or diffraction effects in the waveguide/horn, then this type of comb filter-like response could not be reduced by the HRTF of the listener, and is a separate issue that may cause image instability regardless of speaker separation. Such drivers would still have the same L-R crosstalk comb filtering to contend with as well.
Not sure what frode means (!) but in answer to your points:
1) ?
2) A small mismatch in the tweeters FR or sensitivity would not significantly alter the phenomenon, in fact maximum amplitude fluctuations occur in the case of perfectly matched tweeters.
3) While it's true that a single tone in the upper treble range played from either one or both speakers in most rooms will result in an amplitude that varies wildly as you move your head around just a few inches, this is an extremely narrow band phenomenon.
The summed treble reflections in a room are random in phase relationship to the direct signal, causing extremely narrow band variations in amplitude. This shows up on a steady state FR as a noise like residue in the measurement which can have a 10-20dB variation, but the 1/3oct averaged response is (assuming a tweeter with wide dispersion) relatively the same as the original signal.
There is no phase coherence between the direct treble and the treble that has bounced around the room, so there is no possibility for comb filtering. (At least not at treble frequencies. In the lower midrange and below, yes)
On the other hand the direct signal from Left and Right tweeters is phase coherent on a mono signal, and are very nearly in time alignment near the "optimal" eqi-distant listening position.
It can relatively easily be shown by comparing mono and stereo pink noise that it's not due to room reflections.
There is a nice test sound on Linkwitz Lab which includes both mono and stereo pink noise: (pink-alternating3.wav)
Accurate Stereo Performance
Try moving your head left and right (relative to the speakers) both facing the speakers and head turned at 90 degrees, comparing the difference between stereo and mono pink noise.
In the case of mono pink noise with your head at right angles you should hear obvious phasing effects as you move your head a few inches. When facing the speakers head on there will be some variation, but it will be greatly diminished, depending on the angular separation of the speakers, and the speakers themselves.
As a control, the stereo pink noise sample will show that whether your head is forwards or sideways, you can move your head around and you'll experience no phasing effect from comb filtering, despite room reflections.
If side-wall reflections were involved, you would still hear a phasing effect when moving your head on the stereo pink noise test. (This might be possible if the speaker was right against the side-wall as was the listener, close to the speaker, but not with normal speaker positioning)
Nowhere in real life non-reproduced sound can I think of a case where two sound sources with such a widely separated physical position are playing identical phase coherent signals - it's not a natural phenomenon, and the brain isn't equipped to deal with it well. I think we're lucky that our ears are on each side of our head with occlusion from each other or we probably wouldn't be able to be fooled into perceiving a satisfactory stereo image from a 2 channel system at all 😀
Bad solder joint !!!!😛 😱 🙄
Wow I have no words
You really put me in the picture 😉The only thing I can say is that speakers are not supposed to run/play test tones
Or ,if the target of feeding some pink noise is to measure power response ,hearing a system should not be addressed only to one person (sweet spot) ,so the factors involved would be others (directivity in the lows).
Ah I see. 🙂 I see your words, but sometimes the meaning behind them isn't clear.
I'm not trying to put anyone in the picture, just hold a discussion about the topic. 😉Wow I have no words
We seem to be coming from two different angles, all your responses have been about suggesting possible flaws in the speakers that might explain the comb filtering effect, (bad solder joints, mismatched tweeters, tweeters with irregular polar response etc) or room effects like reflections.
My point was that it's simply an inevitable consequence of two widely separated sound sources that are in close phase coherence with nearly equal but varying (with listener position) time delay, and in my last post I was just trying to suggest ways of eliminating other effects such as room reflections when evaluating it with listening tests - with stereo pink noise vs mono being a valuable differentiating test.
Whether comb filtering between two speakers exists wasn't really the point of my original post though - I think it's fairly self evident that it happens, anyone that has tried to measure a real time room response with a mono signal feeding both speakers will have seen it in play.
I was more interested on the effect of the comb filtering on the perceived stereo imaging stability with listener movement and how that relates to speaker positioning.
The target of feeding pink noise to the speakers is to quickly evaluate the extent the comb filtering effect - which can be determined in real time both by ear (the phasing effect is very easy to pick on mono pink noise) and by microphone.
A phasing effect is also evident on some types of music with head movement but is a lot harder to pin down than on a continuous signal such as pink noise.
What the microphone wont tell you (unless you were to use a binaural recording dummy head) is whether there is any significant reduction in comb filtering due to the HRTF.
I agree a speaker shouldn't be designed for one sweet spot - interestingly the comb filter effect is worst near the sweet spot (unless you are exactly in the sweet spot) and once you're more than a foot or two away it ceases to be a problem because now you have the precedence effect taking over on a mono signal.
It's in the transition between sweet spot and further displaced listening positions that comb filtering is problematic, and if wider speaker positioning helps minimize that, so be it. 🙂
Interesting.
Out of curiosity I just calculated the angle that I have my speakers separated by. My current listening room is rather small and crowded, with a lot of objects limiting the possible speaker locations, but their positions were arrived at by listening tests over a few months without regard to aiming for a specific numeric angle. (Now is the first time I've calculated it)
The room is 4.8 metres wide by 3.45 metres deep relative to the listening orientation, with the speakers 2.19 metres apart centre to centre and the listener to speaker cone distance being 2.66 metres.
This works out to an angle of +/- 24.25 degrees FWIW.
I don't think I ever calculated the angle of my previous 4x8 metre listening room of a few years ago, or if I did I didn't make note of it...
Edit: I made a mistake with the distance from listener to speaker, it is 2.66 metres, not 2.36, so the correct angle is 24.25 degrees not 27.5. 😱
Last edited:
Wonder if it has to do with how far away from the side walls the speakers are?
At the moment mine are in the corners and crossed about 1 foot in front of me. That's a 30º angle (from the crossing point) and the center is very stable, very good.
I started out at ~21º pointed straight at me - and that worked well too.
At the moment mine are in the corners and crossed about 1 foot in front of me. That's a 30º angle (from the crossing point) and the center is very stable, very good.
I started out at ~21º pointed straight at me - and that worked well too.
I've been playing with speaker positioning the last couple months after upgrading my Linkwitz Orion+ to the 3.2.1 ASP rev., and my findings are consistent with the OP. From my listening position the distance to each speaker (7.1'/215cm) is very nearly equal to the separation between speakers, with about a 60° included angle (+30°). With that as the basic setup I have found that adjusting the speaker separation by as little as 1" (25mm) can have a surprisingly significant effect on perceived image. Toe-in is also important - I prefer a toe-in angle somewhat less than the included angle to the speakers so that the intersection of a line to the center of each speaker meets about 1' (30cm) behind the listening position.
I also prefer to converge the aiming axis of the speakers either in front or behind the listening position by a foot or so, it seems to give a better more stable image than having it directly pointing towards you.
Again, I don't have a concrete reason for that, but my suspicion is it may be related to baffle diffraction ripples in the on axis response of the speaker, which with symmetrical baffle/driver placement designs is worst dead on axis. Turning the speakers a few degrees to converge behind or in front of the listener should reduce the baffle diffraction ripples in the response, and may lead to a more stable image because of it.
Interestingly I find when it comes to the width of the sweet spot (how far you can move the listener sideways before the image collapses to one speaker) whether converging in front or behind the listener sounds better seems to depend on the directional characteristics of the speaker in the treble.
I've done some listening tests with a 8" full range driver by itself vs the same driver with a ribbon tweeter covering 4Khz+. The full range driver by itself is very directional in the treble, and I find turning the speakers inwards to converge in front of the listener gives the widest listening sweet spot.
I'm guessing this works because as you move left, away from the right speaker, you approach the right speakers on axis line more closely, thus boosting the treble of the further away right speaker, compared to the left speaker which you are closer too, combating the precedence effect.
On the other hand with the ribbon in place I find turning the speakers outwards to converge behind the listener gives a wider sweet spot, and this time I'm presuming that is because turning them to converge behind the listener gives more illumination of the side-walls from the tweeter than converging in front would. It may also help to keep the left and right treble reverberant fields more separated in the room too, especially if you sit near the rear wall.
To be fair, with the ribbon tweeter it sounds good converged in front of the listener as well, but I prefer it converging behind the listener. Converging in front of the listener may be beneficial in the situation where the speakers are quite close to the side-walls though, resulting in too much early reflection.
Again, I don't have a concrete reason for that, but my suspicion is it may be related to baffle diffraction ripples in the on axis response of the speaker, which with symmetrical baffle/driver placement designs is worst dead on axis. Turning the speakers a few degrees to converge behind or in front of the listener should reduce the baffle diffraction ripples in the response, and may lead to a more stable image because of it.
Interestingly I find when it comes to the width of the sweet spot (how far you can move the listener sideways before the image collapses to one speaker) whether converging in front or behind the listener sounds better seems to depend on the directional characteristics of the speaker in the treble.
I've done some listening tests with a 8" full range driver by itself vs the same driver with a ribbon tweeter covering 4Khz+. The full range driver by itself is very directional in the treble, and I find turning the speakers inwards to converge in front of the listener gives the widest listening sweet spot.
I'm guessing this works because as you move left, away from the right speaker, you approach the right speakers on axis line more closely, thus boosting the treble of the further away right speaker, compared to the left speaker which you are closer too, combating the precedence effect.
On the other hand with the ribbon in place I find turning the speakers outwards to converge behind the listener gives a wider sweet spot, and this time I'm presuming that is because turning them to converge behind the listener gives more illumination of the side-walls from the tweeter than converging in front would. It may also help to keep the left and right treble reverberant fields more separated in the room too, especially if you sit near the rear wall.
To be fair, with the ribbon tweeter it sounds good converged in front of the listener as well, but I prefer it converging behind the listener. Converging in front of the listener may be beneficial in the situation where the speakers are quite close to the side-walls though, resulting in too much early reflection.
Last edited:
How about the story of a person, who could not listen stereo because of a brain defect. She always distinguished two sound sources (the two speakers), the brain was not able to integrate two into one.
Now, this person should be called the perfect lister, shouldn`t she ?
Now, this person should be called the perfect lister, shouldn`t she ?
- Status
- Not open for further replies.