Acoustic reflectors and piston drivers

Status
Not open for further replies.
we need to have a basic understanding of how our senses work, including sense of hearing ...
If our understanding is "basic" only, our ability to come to better solutions will stay "basic" too. 🙄
... our hearing sense doesn't only measure, it interprets as well, hearing is not only quantitative process but also qualitative
Even the qualitative aspect boils down to numbers. They only are more difficult to derive. 🙂
this paper looks more interesting:
Nice find, but rather OT. Where does it tell us anything about the source of ASW and LEV or how to control those effects?
The best I can read from it in the context of this thread is:
- the individual interpretation of ASW against LEV is highly ambigious
- the "felt values" of ASW and LEV differ much from one individual to another.[/QUOTE]

Rudolf
 
If our understanding is "basic" only, our ability to come to better solutions will stay "basic" too. 🙄

yes, very true

Even the qualitative aspect boils down to numbers. They only are more difficult to derive. 🙂

highly controversial philosophical question I would say - no room for discussing it here
suffice to say that we don't know how to boil it down 😉 how to boil down the interpretative/adaptative functions of our senses

Do You really believe that anechoic and echoic spaces are just quantitatively different? And that music can be reduced to clicks, noises and sinusoids?

You must be kidding

Nice find, but rather OT. Where does it tell us anything about the source of ASW and LEV or how to control those effects?
The best I can read from it in the context of this thread is:
- the individual interpretation of ASW against LEV is highly ambigious
- the "felt values" of ASW and LEV differ much from one individual to another.

I can't believe that this is really the best You can read from it 😉

what about size of environment factor?
 
However I think it would be fair to say that the direct sound only dominates our perception at high frequencies when there is some minimum direct to reflected ratio, and that once you're far enough into the reverberant field of a large fairly reflective room so that the direct sound is less than the reverberant field, this effect goes away and we start to hear the total power response of the room, including reflections from the floor and ceiling, and that the apparent image locations become a fused conglomerate of the various sources. I'm not sure what the actual figures are for this threshold, but the transition is certainly there.

But this is precisely not what most studies are showing, rather that our perception of sound balance is much more biased by the direct component even if simple measurements show reverberent field dominance.

Most people look at measurements taken with omni mikes and at a certain distance (critical distance) the revereberent field equals then surpasses the direct field. They assume that perception follows this and whichever field is stronger will dominate. Studies have shown that the direct field can be up to 15dB below the reverberent field level and still determine frequency balance.

On music ? Can you be sure you're not just hearing changes in the vertical off axis response of the speakers, especially at the crossover frequencies ?

Sure because the audible effect is of a comb filter, i.e. a quickly changing pitch tied to listening height. It is easiest to hear with pink noise but you will hear it on music if there is a steady broadband HF component.
 
But this is precisely not what most studies are showing, rather that our perception of sound balance is much more biased by the direct component even if simple measurements show reverberent field dominance.

That's the problem with simple measurements, they're sometimes too simple and inadequate. I found that it's easy to have percussive sounds from a center speaker move towards a speaker 60° to the left (or the right) when playing normal 2 channel stereo recordings. The delay of the side speakers was about 10ms and had the same level as the center speaker.

More meaningful measurements should probably show single reflections and their level, delay, angle and spectrum.
 
But this is precisely not what most studies are showing, rather that our perception of sound balance is much more biased by the direct component even if simple measurements show reverberent field dominance.

Most people look at measurements taken with omni mikes and at a certain distance (critical distance) the revereberent field equals then surpasses the direct field. They assume that perception follows this and whichever field is stronger will dominate. Studies have shown that the direct field can be up to 15dB below the reverberent field level and still determine frequency balance.
True I suppose, but that assumes a certain minimum time delay of the first reflection, and there may be some variation in how far below the reverberant field the direct field can go based on the delay time. What minimum delay time for first reflection was the 15dB figure based on ?

It would be interesting to calculate if a sufficiently long delay is necessary that a much further than typical listening distance could bring the delay of first reflection below the required figure. Depending on the room it could also be that by the time you were far enough away to get the direct signal 15dB below the reverberant field there was insufficient time delay between direct and first reflection to be able to sort between direct and reflected signals anyway.

One other thing to consider is that whilst what you say might be true for perception of spectral balance, does it necessarily follow that the direct field could be 15dB below the reverberant field with no effect on the localization of sound sources or imaging ? I find that a bit harder to believe, and I think the critical distance could be more applicable to image localization shift than to perceived spectral balance.

Sure because the audible effect is of a comb filter, i.e. a quickly changing pitch tied to listening height. It is easiest to hear with pink noise but you will hear it on music if there is a steady broadband HF component.
Hmm. With what kind of speaker design ? Have you tried it with just a single speaker playing and do you still get the same effect ?

I would have thought that in the treble region the only audible source of comb filtering would be the interference pattern between left and right speakers, as no room reflections are close enough in time delay to the direct signal in a typical listening position to produce comb filtering in the treble that isn't extremely tightly packed and thus inaudible on broad spectra signals.

With a dual mono pink noise test signal you can easily hear comb filtering in the treble as you move your head about near the left-right equi-distant point, but once the distance differential is more than about 10cm or so this effect starts to go away. Perhaps you're inadvertently moving slightly sideways relative to left and right speakers as you go up and down ?

Or perhaps it's the effects of baffle diffraction on the tweeters response as you go off the vertical axis. I just find it unlikely that an audible comb filtering effect at treble frequencies with movement is going to be due to the ceiling or floor, when there are so many dozens of wavelengths between them, with the resultant extremely high modal density.

I can't say I've ever noticed such an effect, although I have been using speakers with narrow vertical directivity in the treble for years now, which would tend to reduce any such effect.
 
Last edited:
True I suppose, but that assumes a certain minimum time delay of the first reflection, and there may be some variation in how far below the reverberant field the direct field can go based on the delay time. What minimum delay time for first reflection was the 15dB figure based on ?

The -15dB figure comes from a study of PA system EQ. The direct field could be up to 15dB below the reverberent field and it was still best to EQ the direct sound to flat. Of course the direct field preceeds all of the reverberent field reflections.

One other thing to consider is that whilst what you say might be true for perception of spectral balance, does it necessarily follow that the direct field could be 15dB below the reverberant field with no effect on the localization of sound sources or imaging ? I find that a bit harder to believe, and I think the critical distance could be more applicable to image localization shift than to perceived spectral balance.

Beyond perceived balance, many studies have shown that the first incidence of a sound event dominates and masks reflections and their direction. This is the precedence effect. Coherent delayed sound is masked by the inital arrival until the reflections delay and level are considerable. The sound source location will be determined by the first arrival even if the delayed arrival is stronger.

Hmm. With what kind of speaker design ? Have you tried it with just a single speaker playing and do you still get the same effect ?

I would have thought that in the treble region the only audible source of comb filtering would be the interference pattern between left and right speakers, as no room reflections are close enough in time delay to the direct signal in a typical listening position to produce comb filtering in the treble that isn't extremely tightly packed and thus inaudible on broad spectra signals.


No, I've heard this effect many times with many different speakers. The easiest way to first hear it is to put pink noise to a speaker and then walk (backwards) away from the speaker. You will hear a pitch of ascending frequency. The floor and ceiling reflections get closer in time to the direct sound as you recede. You can hear the same thing with waves at the seashore: walk away or crouch down and the perceived pitch goes up as the direct and first reflection get closer in time. With comb filtering I believe you tend to hear the pitch of the edge before the first response null, and perhaps the following peak.

David S.
 
Regarding the direct field to reverb field data and perception, the missing info here is, as Simon alluded to, the time delay to the onset of the reverberation field. For a test using a PA system this is going to be completely different than it will be in a small room and I would contend that as a result those results are inapplicable to the small room problem.

The precedence effect refers to the domiante source direction location and does not consider tonal effects, only the perceieved direction. AT least those studies that I have seen, and Blauert mentions this in his book. The precedence effect is often taken too be too broadly applicable.
 
Regarding the direct field to reverb field data and perception, the missing info here is, as Simon alluded to, the time delay to the onset of the reverberation field. For a test using a PA system this is going to be completely different than it will be in a small room and I would contend that as a result those results are inapplicable to the small room problem.

The -15dB reference was a generalization for PA systems in the typical auditorium setting, as stated. More recent studies by Kates and Salmi have gone to modeling of the hearing process and tend towards models that use variable length time windows, long for low frequencies and short for high frequencies. As such the effective level of any reflection is reduced by an amount related to its arrival delay. Such a model, or a measuring system based on such a model will be usable in rooms of any size.

The Kates and Salmi papers are directly aimed at the domestic listening room and the effects of room reverberation on perceived frequency balance. They come to the same conclusion: that steady state response is only relevant for low frequencies and that the direct field dominates perception at high frequencies.

The common notion that steady state (RTA) measurements are a foolproof indicator of perceived balance (DBMandrake's comment of "far enough into the reverberent field") is wrong and also explains why most DSP EQ schemes have to fall back on "room curves". The ear has a canny ability to hear through the reverberent field back to the direct or early sound.

The precedence effect refers to the dominant source direction location and does not consider tonal effects, only the perceieved direction.


I did say reflections and their direction.

David S.
 
Last edited:
Hi Dave (I was just through Toronto on my way to Montreal)

I don't see where we are in any disagreement, "dominates" sounds stronger than I might like (makes it sound like the reverberation field is imperceptable) and the HF caveat is important as well. Remember that we hear broad band and "dominance" in one frequency range is not necessarily dominance of perception. The reverberant field matters, but yes the direct field is more important. Bottom line is that it all has to be right.

I think that most of us agree on the rank order of things: direct field is the most important, a clear direct field (no very early reflections) is next most important, and finally that the reverberant field can be a factor (depending on room, its size, listener location, etc. We probably disagree on the exact details and weightings though.
 
Hi Dave (I was just through Toronto on my way to Montreal)

I think that most of us agree on the rank order of things: direct field is the most important, a clear direct field (no very early reflections) is next most important, and finally that the reverberant field can be a factor (depending on room, its size, listener location, etc. We probably disagree on the exact details and weightings though.

As a lazy speaker designer I tend to get excited about any papers that suggest that a measuring system can be devised that mimics the action of the ear: that gives a psychoacoustically correct response curve. This would make matters much easier and lets us sidestep the debate over how much to weight the direct field and the reverberent field, the importance of wall bounces etc.

Feel free to stop by if you are passing through Toronto again!

David S.
 
I think that most of us agree on the rank order of things: direct field is the most important, a clear direct field (no very early reflections) is next most important, and finally that the reverberant field can be a factor (depending on room, its size, listener location, etc. We probably disagree on the exact details and weightings though.

What about the spectral balance of early reflections?

If the spectrum of early reflections is not very important, but the direct field and the reverberant field are, then (ignoring how you're going to make sure you don't get VER) the two main performance criteria for loudspeakers would have to be flatness and smoothness of response in the expected listening window and a smooth power response. Or am I putting things too bluntly here?
 
What about the spectral balance of early reflections?

If the spectrum of early reflections is not very important, but the direct field and the reverberant field are, then (ignoring how you're going to make sure you don't get VER) the two main performance criteria for loudspeakers would have to be flatness and smoothness of response in the expected listening window and a smooth power response. Or am I putting things too bluntly here?

I would have to say that "flat" or "smooth linear" response is assumed in all cases. To me this really goes without saying as I cannot see any argument that would claim that direct, reflected or reverb signals should not all be "flat" (or "smooth linear", near flat - I think it is agreed that a subtle downward pitch of a few dB / decade in these responses is desirable.)
 
No, I've heard this effect many times with many different speakers. The easiest way to first hear it is to put pink noise to a speaker and then walk (backwards) away from the speaker. You will hear a pitch of ascending frequency. The floor and ceiling reflections get closer in time to the direct sound as you recede. You can hear the same thing with waves at the seashore: walk away or crouch down and the perceived pitch goes up as the direct and first reflection get closer in time. With comb filtering I believe you tend to hear the pitch of the edge before the first response null, and perhaps the following peak.
I still don't see how this is so - on the one hand you're adamant that the precedence effect is so powerful that it's the direct signal which almost completely dominates the perceived frequency response (above a few hundred Hz) despite the presence of time delayed reflections, and in the next breath you're ascribing a perceived comb filtering effect in the treble to considerably delayed floor and ceiling reflections, which by your first assertion should be windowed out by the brain. There is a major contradiction here. 😕

You say you hear this effect when walking towards and away from the speakers, but surely when doing so your ears are at roughly twice the height of the speakers design axis ?

Therefore you're a long way off the vertical axis of the speaker, and by walking forward and back you're also changing that angle. Depending on the midrange/tweeter driver spacing, crossover frequency and slope, at such a steep angle you'll very likely be introducing a travelling comb filter effect through driver phase interaction alone, without the help of any floor or ceiling reflections.

This will cause a comb filter like effect in the direct signal, which will most definitely be heard without or despite any room reflections.

Another possible contributing factor may be tweeter diffraction off the top edge of the baffle at steep vertical angles, which would change with angle, and have a short enough time delay to fuse with the direct signal.

The distance of the reflection from the ceiling or floor is simply too far away to cause a discrete, walking comb filter effect as you describe at treble frequencies.

To prove this one way or another you would need to try one of the following things -

(a) take the speakers outside with no ceiling, add a few absorbent mats for the treble ground reflection and do the same walking to and fro with your ears at the same standing height (eg off the vertical axis) to see whether it still happens, or

(b) put your speakers on stands within the room so that their design axis is exactly at standing ear level so you can walk directly towards and away from them without changing your angle to the drivers or cabinet face, thus eliminating changes in driver phase and diffraction with distance.

If you do (a) my bet is you will still find a moving comb filter effect whilst (b) will eliminate it despite room reflections still being present.

Edit: both these tests would need to be done with a single speaker not a stereo pair, to avoid comb filtering between left and right.
 
Last edited:
Part of your confusion is because you have misunderstood the definition of the "precedence effect" . Look up the definition and think about it. Hopefully that will clear up things.
No, I don't think I'm confused about precedence effect, unless you can explain how you think I might be ?

Typical wall/floor/ceiling reflections to most listening positions easily exceed the minimum 2ms required for precedence effect to work, which is what allows the direct signal's frequency response to be the dominant factor in perceived spectral balance in the first place. (Something that most speaker designs rely on, since they primarily optimize on-axis response)

I would also make the observation that even if precedence effect didn't exist, for comb filtering to be audible with movement the path length difference between direct and reflected sound needs to be on a similar order to the wavelength of the sound.

At lower midrange frequencies this is the case for typical floor and wall reflections but it is not the case at treble frequencies. The path length differential in wavelengths between direct and reflected sound of a wall/floor reflection is so great at treble frequencies that there are dozens if not hundreds of modes per 1/3 octave, which again, even if the precedence effect didn't exist, would be smoothed out by the ears perception mechanism.

The only time you notice such reflections is on a steady state infinitely narrowband signal like a single sine-wave. Play a 10Khz sine wave and move your heard around just a few inches and you'll hear it go up and down in amplitude dramatically due to room reflections. Without 1/3 octave averaging (and on a steady state signal) we notice the dense interference patterns of the reflections plain as day. (A single sine-wave can't be 1/3 octave averaged by our ears, but pink noise can be...)

Do the same with pink noise and you can move your head around a foot or two and you won't notice any difference in amplitude or frequency response in the treble. (Provided that you're not playing phase coherent noise on both a left and right speaker)

I've tried both tests many many times before, so I'm not just guessing how it will sound.

If I play pink noise on only one channel and move directly towards and away from the speaker with my ears exactly at the design axis height (half way between mid and tweeter) I hear absolutely no comb filtering effect in the treble moving towards or away from the speaker, even though the tweeter is only about 70cm from a bare reflective side wall and 90cm from the floor.

What I do definitely hear though, as I move from about 1 metre away to 2 metres away is a change in comb filtering in the lower midrange around approximately 200-400Hz, which is quite audible as a "sweeping" of pitch of the noise, as Dave describes. It is most definitely not in the treble though, and a quick calculation of the distances involved will show that 200-400Hz is where the effect will be happening for the wall/floor reflection.
 
Last edited:
I still don't see how this is so - on the one hand you're adamant that the precedence effect is so powerful that it's the direct signal which almost completely dominates the perceived frequency response (above a few hundred Hz) despite the presence of time delayed reflections, and in the next breath you're ascribing a perceived comb filtering effect in the treble to considerably delayed floor and ceiling reflections, which by your first assertion should be windowed out by the brain. There is a major contradiction here. 😕

The only issue in dispute is what frequency range we are talking about. The conversation started with the audibility of floor and ceiling bounces. My comment was that beyond the primary floor bounce dip at 150 to 300Hz (typical and obviously tied to the particular geometry) that I hear similar effects in the treble range. Yes, clearly the time difference between direct and reflected sound determines the spacing of the comb filter nulls and the frequency of the first null. Less clear is why you hear the effect as a ptich and what the perceived pitch is tied to. I believe it is tied to the peak below the first null but be aware that a reflection will have an infinite series of peaks and you might perceive more than one of them.

From my experience I would definitely describe this as an upper midrange to low treble effect, typically in the 1 to 5kHz range. When I described the sea shore counterpart this was for small waves that break on the beach with a "fizz" sound. They have a continuously changing pitch that I can only assume is related to a reflection and the varying reflection delay as the wave climbs the shore. Again, deep knee bends change the frequency. My recollection of this is that it is in the same range or higher.

I understand your scenario of crossover comb filtering and it vaying with angle. Although that could happen that is not what I am describing. As a working speaker designer I would be very aware of any interdriver cancelation at any useable listening angle (by design no system would have it and if it did it would be measured as a matter of course). The effect is more universal than that i.e. it would occur with any speaker of a given tweeter height in any sound room which tended towards it. It was the McIntosh sound room where I first noticed it. The floor was carpeted but not with overly thick carpet. Since then I have observed treble reflections off of most carpets. I was considerable distance from the sytem (say 4 meters) and the pitch increased as I receeded.

Another interesting pitch phenomanon I've observed is the reflection from picket fences (or any fence or wall with a slat type structure). Impulsive noise (balls whacked at the tennis court, for example) give a quick sweep reflection, kind of a "bowweep" sound. I assume the reflection train has reflections that bunch together later in time (or was it the other way around?).

I know this seems inconsistant with the notion that the ear is a time filter that is shorter at high frequencies. The question is how steep is the time window and how strongly does it reject? As you point out, the reflections that are audible can only be a few wave periods delayed otherwise the null spacing would be too narrow.

Again, these are effects I have heard numerous times, so you disbelief of whether they should exist gets trumped by my repeated observations.

David S.
 
Status
Not open for further replies.