About localization:
Live venues on the whole are horrible in this respect. Small differences in location typically lead to considerable differences in sound intensity ('vectorized' SPL, sort of). The result is that normally, from a single seat, by moving ones head, complete flip-flops of the auditory landscape may be perceived. That is, with eyes closed.
Ventriloquist effect to the rescue. Like in the McGurk effect, optical clues trump auditory ones. Our brain constantly calibrates the auditory landscape it reassembles on the basis of visual preceptions. Since these optical clues are present in live venues, we may be under the impression that we can localize the musicians thanks to auditory clues, but we are fooling ourselves. Close your eyes, shake your head, and try again.
When it comes to stereo in the living room, we don't have visual clues about the location of the musicians, and our ears are all we have to go on. One lesson I learned is that the size of a transducer is inversely proportional to our abilility to accurately locate it. This stands to reason. Imagine an entire wall producing sound, versus a small transducer radiating the same sound power. Obviously, when asked where a reproduced sound emanates from, in the case of the wall, we can point our finger at a much wider area than in the case of the small transducer, and still be correct. In short, driver size introduces localization errors. Large speakers don't work well near field. On the whole, drivers should be as small as possible, and should be spaced as closely as possible, on the narrowest possible baffle, in order to maximize our ability to recreate a precise stereo image.
Live venues on the whole are horrible in this respect. Small differences in location typically lead to considerable differences in sound intensity ('vectorized' SPL, sort of). The result is that normally, from a single seat, by moving ones head, complete flip-flops of the auditory landscape may be perceived. That is, with eyes closed.
Ventriloquist effect to the rescue. Like in the McGurk effect, optical clues trump auditory ones. Our brain constantly calibrates the auditory landscape it reassembles on the basis of visual preceptions. Since these optical clues are present in live venues, we may be under the impression that we can localize the musicians thanks to auditory clues, but we are fooling ourselves. Close your eyes, shake your head, and try again.
When it comes to stereo in the living room, we don't have visual clues about the location of the musicians, and our ears are all we have to go on. One lesson I learned is that the size of a transducer is inversely proportional to our abilility to accurately locate it. This stands to reason. Imagine an entire wall producing sound, versus a small transducer radiating the same sound power. Obviously, when asked where a reproduced sound emanates from, in the case of the wall, we can point our finger at a much wider area than in the case of the small transducer, and still be correct. In short, driver size introduces localization errors. Large speakers don't work well near field. On the whole, drivers should be as small as possible, and should be spaced as closely as possible, on the narrowest possible baffle, in order to maximize our ability to recreate a precise stereo image.
You expect him to remain focused on the subject at had? How can he wander off on flights of fancy if he does that?
BTW, I enjoy his tangential excursions
Last edited:
I would ignore everything below the 50% threshold - but that’s only me.
There’s quite an ‘art’ to DOE - I remember my wife doing similar stuff in psychology - there’s a whole range of statistical tests applied (Mann-Whitney u test, Spearman's rho, chi square test, Wilcoxon etc)
Last edited:
Re imaging and localization, would you guys be willing to download the linked test file and tell me what you hear?
http://pmacura.cz/itd.zip
http://pmacura.cz/itd.zip
A slightly different take on localization:
My Quad ESL57's have a light curvature in the vertical plane, and they're tilted back with the stock legs. So the apparent source is behind the speakers, somewhat below floor level. I used to tease the late/great P. J. Walker asking him why he liked his singer crowling on the floor.
When I heard the Magneplanar's at JC's place, I complimented him for having very civil speakers: when I'm sitting, the singer was sitting too, and when I stood up, the singer raise along with me.
My Quad ESL57's have a light curvature in the vertical plane, and they're tilted back with the stock legs. So the apparent source is behind the speakers, somewhat below floor level. I used to tease the late/great P. J. Walker asking him why he liked his singer crowling on the floor.
When I heard the Magneplanar's at JC's place, I complimented him for having very civil speakers: when I'm sitting, the singer was sitting too, and when I stood up, the singer raise along with me.
Thank you for trying, do you hear any left/right move or timbre change between the groups of impulses? There are 6 groups of 3 impulses in each, spaced of 1s between impulses. 3 impulses in one group are all same. The question is if groups of 3 impulses audibly differ. Time interval between groups of 3 impulses is 2s. So, 3 same impulses with 1s spacing, then 2s silence and further group of 3 impulses. There is a longer 4s silence interval between first 3 groups (group 1 - 3) and second 3 groups (group 4 - 6) of impulses.
Last edited:
@ scott wurcer,
Maybe he was talking about the noise distribution of op-amps of the same model, if so he was most likely correct.
@ syn08,
Stimuli were a single/pure tone (fixed frequency) presented at different levels. Level differences step was 1 dB. As said before the range of levels used encompassed the assumed threshold level.
Each level was presented 50 times and the presentation of the various levels was randomly done during the experiment.
After each presentation interval the participant answers the question if the stimulus was present in the interval or not.
So well below the threshold the participant responds with "no" all the time and well above the thresholds the participant responds with "yes" all the time.
In between there is some variability.
The criterion of 50% correct responses as defining the (empirical) threshold is not bound to the method of constant stimuli.
@ Bonsai,
And it could be very misleading, as these are thresholds from onedimensional experiments.
The SDT folks use the term "experimental" thresholds for a distinction to the theoretical thresholds used in the models for the internal judgement processes of the participants.
They point to the fact that sublimal stimuli do often have an impact.
Doing good experimental work on sensory issues is a demanding/complex task and finding the appropriate statistical approach is only part of it.
Maybe he was talking about the noise distribution of op-amps of the same model, if so he was most likely correct.
@ syn08,
Stimuli were a single/pure tone (fixed frequency) presented at different levels. Level differences step was 1 dB. As said before the range of levels used encompassed the assumed threshold level.
Each level was presented 50 times and the presentation of the various levels was randomly done during the experiment.
After each presentation interval the participant answers the question if the stimulus was present in the interval or not.
So well below the threshold the participant responds with "no" all the time and well above the thresholds the participant responds with "yes" all the time.
In between there is some variability.
The criterion of 50% correct responses as defining the (empirical) threshold is not bound to the method of constant stimuli.
@ Bonsai,
And it could be very misleading, as these are thresholds from onedimensional experiments.
The SDT folks use the term "experimental" thresholds for a distinction to the theoretical thresholds used in the models for the internal judgement processes of the participants.
They point to the fact that sublimal stimuli do often have an impact.
Doing good experimental work on sensory issues is a demanding/complex task and finding the appropriate statistical approach is only part of it.
Thanks for posting this. For me, Siegfried Linkwitz was the second step in formalizing a working model of multi-driver loudspeakers. A.N.Thiele and R. Small worked out the electricity>motor>radiating piston gut-level working of drivers, and S.L. clarified how they interact when working together. It was a revolution in thought in the 1970's, and still fundamental.
I'm also glad that the interview caught him undiminished and articulate. His observations came from a place of both deep interest and understanding.
Thanks again,
Chris
Group 1-3 seem a little more focused than 4-6Now you are putting suggestions in my head which may skew the results
No he wasn't , his knowledge of statistics was limited. The usual solution is to give him Rice's paper and tell him to read all of it. Rice, S. O., Mathematical Analysis of Random Noise. Bell System Technical Journal 24
Last edited:
Pavel, on quick listen I hear the sounds as click/spark/explosion impulse sound when played in VLC Player on headphones or loudspeakers. Straight up the first half of the file sounds sort of like the real thing (the starts of the sounds are not immediate), and the second half of the file the sounds are inverted and 'all wrong'/not possible for a pulse sound. I then read your post #21108 and can report that I did not note L/R change or other changes but I did loudly notice the timbre like change characteristic of polarity inversion. I can take a another closer listen but that is what I found with quick run through.
Dan.
Dan.
- Status
- Not open for further replies.