Thanks for explaining, Markus. And yes, my REW window must have been 7 ms or longer.
For the curious I add the 3 ms (left) and 10 ms (right) gated measurements for the woofer part, middle-high part and combined of the Swing dipole:
View attachment 349282
The result should be better, I know.
Rudolf,
These plots indicate target crossover of 500Hz? If so, then basic alignment is way off.
I review again data from: http://www.diyaudio.com/forums/mult...ns-beaten-behringer-what-144.html#post3400487
Problems persist with levels and excess phase.
How are you establishing timing reference for individual driver measurements?
ah! sure. Dipoles are too wide for that. I need to experiment with some cheap behringer monitors. Or careful absorbtion on the sides..
I know. That H frame is too large as a source to work precisely at 500 Hz.
As of now the woofers are geometrically 1.2 m in front of me and almost 45° down from my ears. The perceived bass position is 2.2 m away and about 20° down. I'm sure that I don't hear exactly the same level difference as I measure.
Not sure I understand exactly what timing you are referring to. Reference points are the listening position and the 90° dipole null.
Rudolf
Rudolf,
A timing reference such as a second channel that is hard wired as loop for comparison to acoustic path of speaker and microphone that allows calculation of flight time between acoustic centers of driver and microphone. Microphone is fixed, thus flight time differences between drivers is determined as delay/excess phase.
Design choices are yours; IMO very poor.
10ms measure also reveals no dipole EQ <100Hz.
Between design and implementation, lobe behavior must be very great. At 1.2m listening position spatial information is so garbled that primary image is flat, and glued to wall.
A timing reference such as a second channel that is hard wired as loop for comparison to acoustic path of speaker and microphone that allows calculation of flight time between acoustic centers of driver and microphone. Microphone is fixed, thus flight time differences between drivers is determined as delay/excess phase.
Design choices are yours; IMO very poor.
10ms measure also reveals no dipole EQ <100Hz.
Between design and implementation, lobe behavior must be very great. At 1.2m listening position spatial information is so garbled that primary image is flat, and glued to wall.
Thanks for explaining. I haven't looped back the amplifier outputs yet. Probably should do that. But on-axis time alignment will not be my main concern for sure.
I'm working within self-imposed limits. If you choose others, that's perfectly ok with me.
You need to widen the gate to see the dipole EQ below 100 Hz.
Horizontal lobing? Vertical? What do you expect?
Exciting. So you are sure that in my case spatial depth could be even more improved than what I have today?
Rudolf
Thanks for explaining. I haven't looped back the amplifier outputs yet. Probably should do that. But on-axis time alignment will not be my main concern for sure.
I'm working within self-imposed limits. If you choose others, that's perfectly ok with me.
You need to widen the gate to see the dipole EQ below 100 Hz.
Horizontal lobing? Vertical? What do you expect?
Exciting. So you are sure that in my case spatial depth could be even more improved than what I have today?
Rudolf
For 1.2m listening distance, image stage should start 1.2m behind plane of speakers, and extend well beyond front wall.
Markus,
Broadband burst with no temporal cues for spaciousness played as mono is convolved with distance cues inherent to speaker/listener distance. With single speaker sound appears to come from speaker. With two speakers two sets of speaker distance cues leads to triangulation that doubles apparent distance to image between speakers.
Broadband burst with reverberant cues played through single speaker adds distance cues of burst with physical distance cues of speaker resulting in sound from direction of speaker but with added depth space of reverberation effect.
Paired lateral reflections are same type of effect. When these are strong enough they lead to more spacious presentation. Unfortunately contralateral pairs tend to null stereo presentation so added spaciousness is detail limiting in presentation of width.
Strong front wall reflection with dipole adds distance cues for an image that is behind front wall. When poorly done this tends to flatten image into front wall.
Broadband burst with no temporal cues for spaciousness played as mono is convolved with distance cues inherent to speaker/listener distance. With single speaker sound appears to come from speaker. With two speakers two sets of speaker distance cues leads to triangulation that doubles apparent distance to image between speakers.
Broadband burst with reverberant cues played through single speaker adds distance cues of burst with physical distance cues of speaker resulting in sound from direction of speaker but with added depth space of reverberation effect.
Paired lateral reflections are same type of effect. When these are strong enough they lead to more spacious presentation. Unfortunately contralateral pairs tend to null stereo presentation so added spaciousness is detail limiting in presentation of width.
Strong front wall reflection with dipole adds distance cues for an image that is behind front wall. When poorly done this tends to flatten image into front wall.
I hope you are aware of the work of Bronkhorst regarding the influence of lateral wall reflections on perceived source distance:
http://webistem.com/acoustics2008/a...-sevilla/forumacusticum/archivos/psy03003.pdf
Without a better knowledge of the angular distribution of reflected energy in my room you certainly can't make valid assumptions about imaging "glued to wall".
Rudolf
Rudolf,
I read your link, last two sentences reveal paper to be exceptionally clueless:
D/R ratio is summary statistic that obviously crushes polarization information that is the fundamental directional information about a propagating sound. D/R isn't a directional cue; direction of sound propagation is a directional cue.
Second part of first sentence brings second receiver online. Obvious theme in nature of this with the whole bilateral symmetry thing going on in sound sensing organisms in the animal kingdom. Single perceptual being with two simultaneous perspectives on tap to help determine direction.
Not much left to physics of shading and delay to get bearings. Sonar and radar in both active and passive implementations are well understood.
Enjoyment of audio is low on greed list compared to military and medical applications of principles in imaging.
Methodology of study is its own basis of limited data looking for low budget academic support. It makes excuses for itself by introduction of virtual sources; physics of the real deal are already well worked out.
Simple experiment of binaural cues in both direction and depth is ability to hear a fly buzzing about one's head and to track motion with eyes closed. This isn't about the number of lateral reflections of either a virtual source, or of lateral reflections from a real source.
HRTF libraries and convolution allow synthesis of very effective binaural recordings that maximally demonstrate perception of binaural cues.
Paper is crappy regurgitation claiming obscurity of already well delineated science.
In my living room/listening space exists a familiar object: an inexpensive large faced, hanging, battery powered quart clock in a plastic housing. Not much of a challenge at all to find in total darkness with two ears. More difficult, not impossible to find with one ear, for perspective changes during the search.
It's not my clock; respectfully I shut it off for much of my listening, and reset it when I'm done. It hangs on brick wall to left of listening position, at night it's reflection is clearly visible in glass window panes that comprise near entirety of front wall. With eyes closed and facing any direction I can slowly turn head and face directly at the clock.
Likewise I can look directly at front wall of glass in direction of my own reflection. So somewhat peripherally I see clock on wall and virtual image of clock in reflection from glass window. Closing eyes and making small horizontal rotations of head, direction of clock jumps around. Momentarily clock jumps to different locations in space, sometimes all the way to my right periphery; but soon location settles in manner that becomes fixed, and slow rotation of head with eyes still shut I can point head such that when I open my eyes, I am looking directly at clock.
Raising right hand to right ear while maintaining direct gaze on clock its spatial and timbrel qualities distort a little. Making very small horizontal head motions within several degrees of directly looking at clock, spatial and timbrel distortion disappear, and clock direction and sound are identical in perception to conditions of not cupping right ear. With cupping condition maintained, head is slowly returned to gazing at front window. At a certain point along rotation direct path between clock and right ear enters into shadow of head. Perception of clock location becomes unstable; with continued rotation toward center of front wall, perception of clock location jumps to direction of clock's reflection in front wall glass, and remains fixed there. Repeating this with both ears cupped strengthens perception of virtual clock sound location.
When it's dark out, reflections of both speakers with radial tweeter array are visible with gaze centered on front wall. With cupped ears all perceived sound jumps to other side of front wall. I can move directly between speakers. Without ears cupped its like giant headphones, standard stereo recordings just don't work well this way, but cupping ears towards front wall and clean image appears completely on other side. Turning completely around, back wall becomes front wall and image jumps completely to other side of it with cupped ears. With radial tweeter array and ceiling firing woofer, no toe-in changes are needed or are possible. With speakers set in plane midway between front wall and back wall, image scale is identical; only change is left for right when facing one way v facing 180 degrees with cupped ears.
With speakers on midway plane of front/back walls, sweet spot extends to substantial portion of vertical plane bisecting speaker locations.
I need no further knowledge of your room nor make assumptions, other than that your own words are accurate description of what I've heard myself many times, with many different systems with poor alignment: "image glued to that front wall":
http://www.diyaudio.com/forums/multi-way/235200-ob-speakers-room-acoustics-12.html#post3488669
When standing fairly close to vertical plane passing through speakers with uniform radiation in horizontal plane, ears may be cupped forward for one image, backward for a second image, and up at ceiling for yet a third. Then of course there is image without ears cupped at all.
I read your link, last two sentences reveal paper to be exceptionally clueless:
D/R ratio is summary statistic that obviously crushes polarization information that is the fundamental directional information about a propagating sound. D/R isn't a directional cue; direction of sound propagation is a directional cue.
Second part of first sentence brings second receiver online. Obvious theme in nature of this with the whole bilateral symmetry thing going on in sound sensing organisms in the animal kingdom. Single perceptual being with two simultaneous perspectives on tap to help determine direction.
Not much left to physics of shading and delay to get bearings. Sonar and radar in both active and passive implementations are well understood.
Enjoyment of audio is low on greed list compared to military and medical applications of principles in imaging.
Methodology of study is its own basis of limited data looking for low budget academic support. It makes excuses for itself by introduction of virtual sources; physics of the real deal are already well worked out.
Simple experiment of binaural cues in both direction and depth is ability to hear a fly buzzing about one's head and to track motion with eyes closed. This isn't about the number of lateral reflections of either a virtual source, or of lateral reflections from a real source.
HRTF libraries and convolution allow synthesis of very effective binaural recordings that maximally demonstrate perception of binaural cues.
Paper is crappy regurgitation claiming obscurity of already well delineated science.
In my living room/listening space exists a familiar object: an inexpensive large faced, hanging, battery powered quart clock in a plastic housing. Not much of a challenge at all to find in total darkness with two ears. More difficult, not impossible to find with one ear, for perspective changes during the search.
It's not my clock; respectfully I shut it off for much of my listening, and reset it when I'm done. It hangs on brick wall to left of listening position, at night it's reflection is clearly visible in glass window panes that comprise near entirety of front wall. With eyes closed and facing any direction I can slowly turn head and face directly at the clock.
Likewise I can look directly at front wall of glass in direction of my own reflection. So somewhat peripherally I see clock on wall and virtual image of clock in reflection from glass window. Closing eyes and making small horizontal rotations of head, direction of clock jumps around. Momentarily clock jumps to different locations in space, sometimes all the way to my right periphery; but soon location settles in manner that becomes fixed, and slow rotation of head with eyes still shut I can point head such that when I open my eyes, I am looking directly at clock.
Raising right hand to right ear while maintaining direct gaze on clock its spatial and timbrel qualities distort a little. Making very small horizontal head motions within several degrees of directly looking at clock, spatial and timbrel distortion disappear, and clock direction and sound are identical in perception to conditions of not cupping right ear. With cupping condition maintained, head is slowly returned to gazing at front window. At a certain point along rotation direct path between clock and right ear enters into shadow of head. Perception of clock location becomes unstable; with continued rotation toward center of front wall, perception of clock location jumps to direction of clock's reflection in front wall glass, and remains fixed there. Repeating this with both ears cupped strengthens perception of virtual clock sound location.
When it's dark out, reflections of both speakers with radial tweeter array are visible with gaze centered on front wall. With cupped ears all perceived sound jumps to other side of front wall. I can move directly between speakers. Without ears cupped its like giant headphones, standard stereo recordings just don't work well this way, but cupping ears towards front wall and clean image appears completely on other side. Turning completely around, back wall becomes front wall and image jumps completely to other side of it with cupped ears. With radial tweeter array and ceiling firing woofer, no toe-in changes are needed or are possible. With speakers set in plane midway between front wall and back wall, image scale is identical; only change is left for right when facing one way v facing 180 degrees with cupped ears.
With speakers on midway plane of front/back walls, sweet spot extends to substantial portion of vertical plane bisecting speaker locations.
I need no further knowledge of your room nor make assumptions, other than that your own words are accurate description of what I've heard myself many times, with many different systems with poor alignment: "image glued to that front wall":
http://www.diyaudio.com/forums/multi-way/235200-ob-speakers-room-acoustics-12.html#post3488669
When standing fairly close to vertical plane passing through speakers with uniform radiation in horizontal plane, ears may be cupped forward for one image, backward for a second image, and up at ceiling for yet a third. Then of course there is image without ears cupped at all.
Is that how you perceive the two speaker case? The distance of the phantom image doubles? I don't perceive it this way. The phantom image is still as far away as in the single speaker case.
Barleywater,
I appreciate your detailed description of hearing those clock reflections in your room and understand your findings - really not that different from what I would experience in my room. But in stereo reproduction of music our brain has to react to a multitude of reflections returning in a rather small time window from different directions - especially with a loudspeaker system like yours. I thought the paper dealt with such circumstances.
Didn't you read my second sentence in that link? I'm in the comfortable situation that I can control the amount of front wall reflections somewhat in intensity and direction. I reported that I could make a "bad" choice with imaging "glued to that wall with minimal depth". And I could make a "good" choice with growing depth extending into the next room.
I did not know that I have to explain that I didn't stick to the "bad" choice.
Rudolf
I can relate to the discussion in the paper. D/R is linked to how I perceive stage distance. When using an 8" full range driver the distance is very close. About only an armlength away. The same speaker can be made to project a more distant sound stage when the cross-firing setup is used that I had described some pages ago.
But in no way it's as simple as measuring the D/R ratio and the paper acknowledges that. The detailed structure of reflections and especially their delay and direction is probably most important. I think even more important is how our hearing processes that information in context with the room's acoustics which acts as a frame of reference.
Markus,
Some clarification: Speaker location from listener is vector that is fixed as real source. Distance to speaker as virtual source with second speaker becomes projection from listener to vertical plane passing through both speakers. Listener is constrained to plane perpendicular to vertical plane passing through both speakers. With intensity panning, center image is 2x listener distance to plane of speakers. This is the effect of stage moving closer to listener as listener moves toward vertical plane of speakers. All forward depth for listener collapses when head is directly between speakers in reflection free space, or in perfectly symmetrical setup.
In your case with Nathans, symmetrical setup would be with speakers midway between front wall and back wall; and with them either pointing directly at each other, or directly at side walls with backs pointing directly at each other. In case with speakers directly pointing at each other, and head directly centered in between, contralateral reflections are strong, yet no forward spaciousness is generated. From same locations, if speakers are rotated to point at reflective front wall, a forward image forms. This is behavior of any directional, forward firing speaker. This image is strengthened with cupping ears, effectively modifying HRTF vector of highest hearing sensitivity. Image is also strengthened by moving small distance back from vertical plane of speakers; also effectively moving speaker closer to natural HRTF vector of highest hearing sensitivity.
All this reveals presence of ITD mechanism in conjunction with peak hearing sensitivity that is naturally forward of vertical plane passing through ears.
Also demonstrated is sound field synthesis of ITD cues with intensity panning between real sources.
Markus,
Then psychoacoustics is crushing all forms of reverberant energy into directionless scalar much as Rudolf's reference to study of D/R statistic as directional cue. This is purely one-dimensional reverberation added to single source and subsequent convolution with a multiple dimension basis. At simple level this is just increasing apparent bulk volume of listening space. Without equal treatment of all reflecting surfaces, perception can't be of increasing bulk reverberant volume, but takes on attributes of width, depth, and height.
You claim to look for realistic spaciousness, and keep circling back to one dimensional echos convolved with vector space of listening room. It is merely added effect.
Commonly produced psychoacoustic literature isn't providing you with any useful information for improving perception of spaciousness in your listening as more realistic, because most of this literature concludes that more study is needed. Truly one dimensional thinkers. Reality is most of these studies trample real physics into statistical mush, so of course more study in needed.
Your preferred setup with no reflections other than strong contralateral reflection is identical to adding second pair of speakers at desired virtual reflection point in free anechoic space with delay and attenuation control. Analysis is easy. With virtual reflections turned off, all spaciousness is in stereo program. With main speakers off, and virtual contraleral reflection speakers on, vertical plane through speakers is closer to listener, with left and right channels of program swapped. This means virtual stage is closer to listener, and subtends a greater angle in space, thus more width, but left and right are swapped. Now, with only one main speaker on along with corresponding virtual contralateral reflection speaker on, two sources have 100% identical waveform other than scalar magnitude. This is exactly intensity panning in plane parallel to vertical plane through the main speaker and virtual reflection speaker with depth determined by listener distance from vertical plane through speaker pairing. Accounting for stereo setup, a centered relatively monophonic cloud of spaciousness is imposed on primary stereo image.
Oh well. Directional hearing plus directional speaker leads to such possible listening setup.
Easy solution for dealing with recordings that seem lacking in spaciousness is to switch in a reverberation device.
As Soundtrackmixer points out, wonderful multiple channel digital reverberation engines exist, and primarily use is production phase of recording and in live shows.
Then psychoacoustics is crushing all forms of reverberant energy into directionless scalar much as Rudolf's reference to study of D/R statistic as directional cue. This is purely one-dimensional reverberation added to single source and subsequent convolution with a multiple dimension basis. At simple level this is just increasing apparent bulk volume of listening space. Without equal treatment of all reflecting surfaces, perception can't be of increasing bulk reverberant volume, but takes on attributes of width, depth, and height.
You claim to look for realistic spaciousness, and keep circling back to one dimensional echos convolved with vector space of listening room. It is merely added effect.
Commonly produced psychoacoustic literature isn't providing you with any useful information for improving perception of spaciousness in your listening as more realistic, because most of this literature concludes that more study is needed. Truly one dimensional thinkers. Reality is most of these studies trample real physics into statistical mush, so of course more study in needed.
Your preferred setup with no reflections other than strong contralateral reflection is identical to adding second pair of speakers at desired virtual reflection point in free anechoic space with delay and attenuation control. Analysis is easy. With virtual reflections turned off, all spaciousness is in stereo program. With main speakers off, and virtual contraleral reflection speakers on, vertical plane through speakers is closer to listener, with left and right channels of program swapped. This means virtual stage is closer to listener, and subtends a greater angle in space, thus more width, but left and right are swapped. Now, with only one main speaker on along with corresponding virtual contralateral reflection speaker on, two sources have 100% identical waveform other than scalar magnitude. This is exactly intensity panning in plane parallel to vertical plane through the main speaker and virtual reflection speaker with depth determined by listener distance from vertical plane through speaker pairing. Accounting for stereo setup, a centered relatively monophonic cloud of spaciousness is imposed on primary stereo image.
Oh well. Directional hearing plus directional speaker leads to such possible listening setup.
Easy solution for dealing with recordings that seem lacking in spaciousness is to switch in a reverberation device.
As Soundtrackmixer points out, wonderful multiple channel digital reverberation engines exist, and primarily use is production phase of recording and in live shows.
The study doens't claim that a directionless scalar is the end of the story. In fact the paper shows that it's not. As with many psychoacoustic studies it merely points in the right direction. More (much more) research is needed. By the way, physics alone won't help you much in explaining human perception.
Don't know why you keep revisting that point over and over again. Of course is adding room reflections at the reproduction end an effect. And yes, utilizing room reflections is like adding additional sources/speakers. In fact there are upmixing algorithms that do just that, e.g. Audyssey DSX. The reproduction becomes less accurate but it also adds spaciousness. Spaciousness is something we perceive in the real world with natural sound sources each and every day. Any reproduction technique incapable of reproducing spaciousness has to sound less realistic (which is certainly the case for 2-speaker stereo). Now pick your poison, more accurate or more realism. For stereo reproduction I currently pick "realism" over accuracy. For multichannel the whole discussion is moot anyway.
Last edited:
The "reverberation device" is not enough if its output is still only coming from the two stereo speakers radiating the direct signal.
You really don't seem to know the sensation of spaciousness. Did you ever listen to a stereo recording and its upmixed version in a very dry environment like a control room adhering current standards?
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.