But you should! It's a lot of fun! Just don't do it too often or for too long.
Ok. The reason I asked is that left-right perception in the lower midrange seems to be a lot less precise than many would imagine, especially below 300Hz, and I could easily imagine the source of such a frequency on its own being mistaken between left and right without sufficient other clues.
In the "Flat is not correct?" thread talking about floor bounce I had described how I'd tested a 2 way design with midwoofer at about 80cm from the floor, and added a separate un-baffled woofer right at floor level (directly below the main) low pass filtered to 300Hz (12dB/oct) to experiment with the floor bounce cancellations / vertical modes in the 100-300Hz region.
What I didn't mention is that I only ever had an extra woofer on the left channel.
Despite adding this extra woofer covering from 90Hz to 300Hz and of similar sensitivity to the main speakers only on the left side, there was very little if any apparent image shift to the left, even on male vocals, which surprised me. The difference in tonal characteristic from filling in the previous notches was fairly obvious, but appeared to come from the middle.I'm still interested to see the result of that, however in light of reading some of the references provided later in the thread I think I may have overstated the importance of precedence effect on a random pink noise signal - if there is any at all. (In other words, I'm probably wrong about that as a cause...)
It may well simply come down to modal density of the comb filtering - in the treble region with typical side wall delays the density will be great enough that the critical bandwidth of our hearing should even out any changes, and nothing beyond that may be needed to explain a perceptible lack of treble comb filtering with small moderate movements.
In that case, a fairly steep high pass filtered pink noise sample with variable cut-off frequency should answer the question at what point it becomes perceptible.
perhaps, but in fact You talk about them separately Yourself, when You are talking about imaging that is being destroyed by particular kind of reflections etc.
It doesn't have to be at high levels. With my open baffle I can also sense bass like the bass drum instead of just hearing it like with my previous speakers even at moderate levels.
It's probably due to the larger surface of 12 12" woofers.
those and similar graphs have been already posted many times
I respond by saying that science is not in those results themselves but rather in the discussion of their meaning and scope of relevance, generalizing such results beyond the particular aim and context of the particular test is unfounded, I think we can agree on that
and at the end of the day it turns out that we cannot define any objective standards based on those numerous specific studies, standards of loudspeaker design for music reproduction
I think that this is the reason why in the field of speaker design for music reproduction Toole finally relies on subjective criterion of satisfaction of most listeners
Attachments
? isn't it rather a demo of sound environment simulator software for computer games?
Yes, those are the graphs.
It amazes me how much the stimulus is a factor. I was reading a Griesinger paper the other day about concert hall acoustics and he was saying the same thing: reverberation detectability depended on the stimulus. With castenets or simple orchestration you might strongly hear reverberation in a hall. There are gaps between the notes to reveal the hall's sound. With legato music, slowly varying sound, the reverberation might not be audible at all. Same with Toole/Olive: an impulsive source in a dead environment and the threshold of detectability is quite low and gets even lower as reflection delay gets longer. Use pink noise (which is everpresent and therefore somewhat masking) then longer reflection delay doesn't make reflection detection easier.
Makes it hard to generalize.
Note with all these curves that lower equals more detectable.
David S.
Sure, thereby well knowing that a single listeners judgement
(on a certain aspect under question) can be quite flawed and uncertain.
___
To me any given speaker design is also a kind of question not just
an answer, or even "the last answer".
Finding perceptional detection thresholds for (aspects of) physical
effects in speaker/room interaction and quantifying them is helpful
for sure.
For example floor bounce:
You can decide whether to tackle it in a certain design or not to tackle it.
Having found a good solution to diminish effects
(say "non smooth bass response above Schroeder frequency of most rooms")
a certain resulting design will again rise the question
"Is it an improvement (even audibly) ?"
or
"In how far is it an improvement".
Having tackled floor bounce e.g. by arraying some woofers influences
other aspects of the design, which potentially might cause different
problems as a consequence (say more frequency dependent directivity
in the vertical plane), if not again solutions are found (which in turn will
have their specific consequences ... ad infinitum).
Even the decision not to adress floor bounce in a certain design may be
a valid one, after having analysed some major "decison paths" in the tree
of design options.
Nevertheless awareness of effects in speaker-room-listener interaction
and discussion of their relevance for reproduction quality and fidelity
is valuable, because those of us who "do not know everything yet" get the
chance to adjust their individual weighting of the effects to be considered
in said speaker-room-listener interaction.
___
Since we discussed many aspects of (even of a single) speaker's intercation
with a rooms boundaries dependent from directivity pattern, i would be glad
if someone dropped a line concerning interference between the speakers itselves
in a stereo setup and the relevance (in different frequency ranges, for
different classes of signals) for
- the shape of the listening area where "good" imageing is possible
- possible colorations, artefacts etc. arising from "inter speaker"
interaction solely
Just for visualization:
http://www.falstad.com/interference/
Last edited:
As a supplement to relate my questions to the topic:
In asking
"What is the ideal directivity pattern for stereo speakers?"
the stereo speakers may also be seen as a unit, which
(e.g. with a signal of same amplitude and phase at both speakers)
has a specific directivity pattern varying with frequency.
In asking
"What is the ideal directivity pattern for stereo speakers?"
the stereo speakers may also be seen as a unit, which
(e.g. with a signal of same amplitude and phase at both speakers)
has a specific directivity pattern varying with frequency.
nice! we can see the destructive interference of stereo crosstalk
those problems are related to speaker directivity problem because (Toole, p. 139):
Attachments
As perceived by an omnidirectional mono microphone which is simply summing the two wave-fronts at a point in space without consideration for their angle of arrival.
Quite different to what our binaural ears mounted on either side of our head do, at least at higher frequencies above middle midrange frequencies or so.
Now primarily talking about the frequency range of say >2Khz,
where ILD is assumed being the dominant cue for localization
i'd like to drop the - surely - provoking question
how important is phase coherent radiation in stereo speakers ?
Given fine enough granularity of phase randomness over frequency
and angle ... that "naked ugliness" can be cured to some
extend at least above midrange.
Surely group delay has to stay well controled within sane
limits.
Even diffusivity of boundary reflections in the room will
increase without additional room treatment.
where ILD is assumed being the dominant cue for localization
i'd like to drop the - surely - provoking question
how important is phase coherent radiation in stereo speakers ?
Given fine enough granularity of phase randomness over frequency
and angle ... that "naked ugliness" can be cured to some
extend at least above midrange.
Surely group delay has to stay well controled within sane
limits.
Even diffusivity of boundary reflections in the room will
increase without additional room treatment.
Last edited:
yes, indeed, in case our sense of hearing it is more complicated but in principle this is it
The effects of reduced crosstalk at high frequencies could be simulated by taking the incident angle of each source and feeding it through a HRTF before summing, so in principle it shouldn't be too hard to model, even in a java app like that.
Fortunately for us, cross-talk at high frequencies where there would otherwise be severe lobing/comb filtering effects between left and right speakers (as shown in that example) is not a big problem, provided that there is enough angular separation between the speakers to provide enough occlusion of the opposite speaker from each ear. Less than about 20 degrees separation and it starts to become a real problem. Greater than about 25 degrees and its a noticeable but small problem.
big problem is that we do not know how exactly our sense of hearing weighs all these different cues in such specific case as stereo speakers playing stereo program in acoustically small room
it is probable that in case of stereo in such a room neither (ongoing - phase) time (ITD - IPD) nor intensity (IID-ILD) cues play decisive role in the process because they are per se contradictory, ambigous and generally unreliable under such conditions
perhaps onset (wavefronts arrival times) time cues are more important? perhaps ISD (interaural spectral differences) cues?
it seems that stereo in a room is a puzzle to our sense of hearing
now - perhaps "phase coherent radiation" is not important from perspective of ongoing ITD because such cues are unreliable under discussed conditions
OTOH in may be important from perspective of onset ITD because of wavefronts deformations due to time smear
Last edited:
I don't know but Toole writes about naked ugliness of this problem, perhaps it is not that small
I think the notion is correct but not for normal stereo angles.
I find if you feed mono pink noise to a stereo pair of speakers and set the triangle between you and the speakers to something conventional (usually 45 to 50 degrees seperation of the speakers when viewed from above) that the comb filtering is quite strong. Hearing directivity or head diffraction is not nearly enough to shade the opposite speaker.
Under those conditions you can shift your head left to right and hear the in and out of phase cancelation quite readily. The effect is of treble coming and going in a "swish swish". You can even find the location of exact time allignment for one ear or the other. (Not both at the same time)
The effect is very different if you move up to the midpoint of the speakers such that the angle between is 160 to 180 degrees. Then the phasiness disappears and you get a smooth shift from one side to the other as you shift across the geometric mid point.
In the first case you are hearing the comb filtering as it exists in space due to the different path lengths. In the second case you have enough hearing seperation that the sum is all in your head.
By the way you can listen to binaural recordings with the wide seperation approach. Unfortunately it images inside your hear rather than in front, much as with headphones.
David S.
I'll agree there is still certainly quite a noticeable effect, but compare shifting your head from left to right when facing the speakers, to turning your head at right angles and then moving the same direction and amount in the room - the phasiness in the treble is MUCH worse with your head at right angles because the left and right speaker can sum in the nearer ear more or less like an omni microphone.
If it wasn't for the amount of cross-talk rejection that we do get I think conventional stereo reproduction would be something of a disaster, especially for centre mono signals.
50 degrees really isn't enough to minimize phasiness in the treble with movement, in my last post I was referring to the angle from centre to each speaker, so double those figures. My own speakers are only a little bit under 50 degrees at the moment simply because of small room constraints, and I would consider 50 degrees a bare minimum, certainly not optimal angle at least from the point of view of L-R channel crosstalk and comb filtering.
I'm not sure you need to go quite that far, I find when you get out to about 60-70 degrees total separation treble phasiness with movement on pink noise is mostly gone, and also that the angular change between where you get phasiness and where you don't is quite small.
I'm guessing this could be the angle where the outer ear is fully occluded by the face for the opposite speaker, and if so, this critical angle is very likely different for different people depending on the shape of their face and ears.
Someone with ears that stick out to the sides a lot may need wider speaker separation to get good stereo reproduction than someone with flatter ears
(just a hypothesis)If the angle is too narrow I find it tends to make the imaging unstable (with head movement) so that you only get a good focused image at one particular spot, and moving a few inches either way messes it up. A significantly wider angle and this hyper sensitivity to sweet spot seems to go away. On pink noise to me it sounds like less comb filtering with sideways head movement when the speakers are more widely separated, which directly corresponds to a more stable centre image.
I remember trying that years ago with binaural recordings on a couple of full range drivers - each about half a metre away and directly to the sides 180 degrees from each other. From what I remember it sounded very good, but I don't recall whether the sound localized in my head or externally. Immersion was very good though.
I typically use the sideways technique to find the exact center between two speakers. It works especially good if you connect one out of phase. At least part of the improvement is because facing straight forward you get a different "answer" for each ear.
In a lot of stereo there is more left/right going on than center and we tend to ignore poor center postion or accuracy. Meridian did a balance control using time shift to let you sit down, use the remote for balance, and hopefully get perfect centering. I remember it took a little practice to hear the effect on typical program.
Real three channel would be a nice improvment. You could have a real center image and you could open the left and right speakers to a wider spread as you like, without fear of a "hole in the middle".
David S
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.