Toole's opinion on early reflections is not really that simple. I quote from a studio designer on a Dutch forum who asked Floyd himself:
"Hi Bert.
People who say I dismiss room reflections as unimportant - and there are a few, it seems - simply have not read or understood my book. It is patently obvious that room treatment is necessary, if only to establish conditions suitable for comfortable conversation. This requires reverberation times under 0.5 second. This alone, also pretty much ensures that film dialog will be clearly understood.
The most debated issue relates to first lateral reflections. Some of those arguing vociferously in favor of eliminating them seem to have a conflict of interest, being providers of acoustical materials. Others have more reasoned arguments. I say up front that there can be no universally satisfactory answer because there is no universal scheme for recording stereo or multichannel signals. Only through controlled listening tests can we get useful insights, and these are in short supply. In the meantime opinions reign supreme, and there are many of them.
In the book I show results of several double-blind evaluations, some done by me, some done by others, showing that listeners tend not to be disturbed by lateral reflections, and many even prefer them. I also point out that the professional side of the industry almost universally feels the need to eliminate them. I suggest, respectfully, that humans have a remarkable ability to learn, to adapt, and that recording engineers spending their days adding, adjusting, and removing - at will - delayed sounds from mixes undoubtedly are more highly sensitized to these sounds than are lay listeners. This probably applies to any audio professional, acoustical consultant or enthusiast who focuses enough attention on this task. We learn to hear things and, once heard, they tend not to go away. I well remember that during the resonance detection experiments, we all became extremely skilled at hearing and identifying resonances. During the tests, and for some time afterward, we were hearing little resonances in everyday life that normally would have been totally unnoticed. From such things paranoia is born, and if we had taken this to an extreme, we would have damped our wine glasses.
For stereo listening I have found that it very much depends on the program. Music with lots of decorrelated sounds, classical for example, is sometimes enhanced by reflections, although coincident-mic recordings may benefit from a lack of reflections - letting the direct sounds be more dominant (the Blumlien stereo effects work best in an almost anechoic situation). Pan-potted recordings (the majority of pop) end up delivering essentially monophonic sounds from left and right loudspeakers, and these may well benefit from a bit of spatial enhancement. Otherwise we are left with what really annoys me about stereo: a relatively spatial set of phantom images created by both loudspeakers, and two "anchor" images created by the left and right loudspeakers playing solo. In some recordings we hear a whole string section emerging from a single loudspeaker. Not realistic, and not even pleasant. In the past, I have recommended that serious stereo listeners hang absorbent drapes along each side wall, pulling them out and pushing them back to suit what they are listening to. Our listening room at the National Research Council in Canada had this feature.
In the book, I put more emphasis on multichannel audio, where much of the important sound is delivered by the center loudspeaker, farthest from the side walls. In these situations I conclude that treatment of the side wall reflections is an option. There may be situations in which their effects are audible, but when all 5 or 7 channels are operating, it is improbable that natural room reflections have much of an effect. Other things being equal, the effects of the room are most audible when only a single loudspeaker is operating, and it becomes less so as other loudspeakers (channels) contribute additional uncorrelated sounds.
Of course the degree to which reflections are activated depends on the directional properties of loudspeakers, and the extent to which the loudspeakers are well behaved in their far off-axis responses (side wall reflections can be 50 degrees and more off axis). It has become clear over the years that, with hard side walls, the more uniform the off axis frequency response, the higher the rating of the loudspeaker. So, one has to wonder whether at least some of the dissatisfaction with reflective side walls has to do the misbehaving loudspeakers. Sadly, most manufacturers don't provide us with sufficient data to judge. And that is another, and I would argue much more worthy topic, to argue about.
Best wishes,
Floyd"
Generally, no. I worked on tapered line arrays at McIntosh and the only time domain artifact is that they tend to have some high frequency comb filtering in the top octave. Along with the comb filtering comes some repetition of the highest components of the impulse response. (These two aspects are inseperable.)
A line array has fully independant performance in the horizontal plane and the vertical plane. For the vertical plane the number of elements, their spacing, and any delay and level tapering applied will determine the response.
For the horizontal plane the usual array is only one element wide. This is like saying, in the top view looking down, an array looks like a single driver. As such it has the polar response of a single driver (we are assuming assuming an array of full range drivers). That is, an array of 5" units has the horizontal polar response of a 5" unit. You can create multiway lines and the latest McIntosh systems have a sideways MTM row (3 vertical lines) just in front of a woofer line. The JBL CBT arrays are two way with a tweeter row in front of a woofer row. Again, viewed from above in cross section you would see the lateral displacement between the drivers (rows) and the horizontal polar response is the usual combination of geometry and crossover point.
Don's CBT formula, along with my XRT24 tapering and the preceeding designs (such as Augspurger's) are all about refining the vertical response of the terminated line. If a line is finite in length the near field and farfield vertical response can be made much more uniform if a level tapering profile is applied across its length. Nearfield performance becomes much more important with line arrays since the nearfield typically extends well beyond the listening distance.
Regards,
David S.
Thanks. Very interesting. Also wanted to know more in depth what he meant.
There are certain aspects that are unclear to me though.
-It doesn't seem like Floyd mentiones time or level. It's a big difference whether the sidewall reflections arrive at 4, 10 or 15 ms. And the level is important.
-I can't find any information of how F.Toole treated the room in his experiments. I've seen how the Harman Int. reference room is treated and it seems like Toole favores absorption at both the frontwall and rearwall. This can easily lead to a pretty dead sounding room. So there's a question if his results primarily can be applied to how he specifically treated the room in his researches. The results might have been different with another acoustic concept. In a LEDE concept, one would use diffusion at the rear of the room giving a lateral exponentially decaying diffuse field. And applying absorption surgically to create a RFZ, and a ISD gap with a strong termination will give a much livelier space oppose the Harman Int. reference room.
Last edited:
There are usually many variables at play in just about any research experiment. From my experience, it's rare that all the significant variables are "neutralized" or even specified. For example the psychoacoustic effects of wall reflections being 4mS, vs. 10mS or 20mS. This is a pretty complicated subject, and many things need to be specified in complete detail, for any conclusions to be good enough. When people cite Toole or anyone else, it sometimes makes me wonder if they have done any real research themselves, so they realize how many variables there often are. So they know how to interpret the data. 4mS over what part of the frequency range, etc.?
If you are referring to folks who have actually done research in psychoacoustics, who are these two persons that you have identified?
Good post, a_tewinkel!
I think the quote above really reflects what we are dealing with: there is no single solution to get the best (or most accurate) sound because the available source material is very diverse (and the production conditions are unknown).
I discussed this with keyser today and we both agree on this.
Furthermore, it is my opinion that one should be careful in generalizing conclusions from research to real world rooms. For example, a conclusion about the audible effects of a single reflection (and possible comb filtering involved) in an anechoic chamber might be invalid in a real room with many other reflections.
A good example is the floor bounce. Some say it is audible and problematic, reducing impact of a kick drum for example. Linkwitz states that it is unlikely to be a problem, since it is a natural phenomenon with which the ear/brain can deal, by using spectral information from other reflections to isolate the characteristics of the source from the room.
Now I'm hurt.
No, I haven't done any real research on the audibility of reflections, although I'm up on most of the important studies. I'll let Dr. Geddes speak for himself.
I did mention that I bought an AV receiver to get my home theater running again. I've been experimenting with the various surround modes to see if I could come to some conclusions. Like most AVR's you get three basic options: you can do true 5.1 or 7.1 playback with Dolby or DTS, you can turn 2 channels into 5 with matrixing schemes such as DTS Neo 6 or Dolby Pro Logic, and you can use DSP synthesized modes that simulate a room by adding a variety of artificial reflections to the 2 channel signal.
The 2 channel to 5 channel modes are very different but end up giving a similar result: the sound spreads to fill the 5 channels, the sound gets fuller and the left and right speakers tend to disappear.
I should mention that I've set my 5 channels up all on one long wall. Left and right are in the normal positions, center is under the screen, and left surround and right surround are in wide postions towards the room corners. I like this for music because it gives a wide stage presentation and no musicians behind me. Switching into 5 channel (from 2) gives a whole wall of sound and the effect is quite pleasant. The only downside is that there is a consistent sameness to it.
DTS and Dolby are better in that there seems to be a possibility of both the wall of sound and more individually identifiable sources. More frequently will you notice sound coming from a particular speaker (when the recording calls for it) while the wall of sound effect is still there.
This is all a comparison of 2 channel, 5 channel synthesized and 5 channel discrete, but I think it matches what we find with various speaker directivities: more directional means more sense of the speakers being the singular source of sound. More reverberent floats the sound away from the individual speakers, but since the stationary room becomes a factor the stereo effect becomes homogeneous and "samey".
I can see how some people will prefer one extreme and others the other.
David S.
There has been a vast amount of research on the subject and much of the work has actually been done in real rooms.
Toole's book is best on this subject. Note that he doesn't just talk about his findings on the subject but compares the results of all the major studies such as those of Bech, Ando, Olive, Haas, Deventier and many others. The studies start out as simple single reflection tests but there are also "reflections in the presence of other reflections", simulated rooms (Bech) and numerous real room studies. In general the threshold of audibility of reflections is higher (reflection must be louder) in live rooms, and our heightened sensitivity to reflections ariving much later than the direct sound drops considerably.
Being a "natural phenomenon" doesn't make it more or less audible. The Bech study most directly dealt with this. He simulated, in an anechoic chamber, the first 16 reflections of a typical speaker in a typical room. Since it was electronically simulated he could raise or lower the effect of any reflection and see if it was noticeable. Many of the reflections were below a threshold of detection: rasing them a few dB didn't change the sound. But in two rounds of tests and two published papers he came to the same conclusion that the floor bounce was most likely to be audible of the 16 reflections.
Most the other studies say the same thing: lateral reflections are perceived as spaciousness or image widening. Vertical reflections, floor or ceiling, are spatially inseperable and are perceived as a frequency response error.
Our problem is two ears on the sides of our head and none on the top or the bottom.
David S.
...and/or a distance cue. On the other hand there's also the (acoustically small) room's modal behavior which creates more severe peaks and dips at lower frequencies than any single floor or ceiling reflection.
Last edited:
David, thanks for sharing your experience with the surround synthesis. Does this 'sameness' occur only with the DSP-added reverb or also with the Pro Logic/DTS upmix?
Agreed, Toole does weigh in a lot of factors before drawing a conclusion.
Thanks for explaining the Bech study. Sounds interesting, I really have to dig up that paper this week .
Did the listeners in the Bech study have time to adapt to the situation? Directly comparing two situations for audible differences can be tricky, we always need some time to adapt to a new acoustic environment. This may apply to the floor bounce. I'm interested in your opinion on this.
Agreed, Toole does weigh in a lot of factors before drawing a conclusion.
Thanks for explaining the Bech study. Sounds interesting, I really have to dig up that paper this week
.Did the listeners in the Bech study have time to adapt to the situation? Directly comparing two situations for audible differences can be tricky, we always need some time to adapt to a new acoustic environment. This may apply to the floor bounce. I'm interested in your opinion on this.
I find the floor bounce is pretty dominant when you are fairly close to the speaker, just as the back wall bounce is obvious when the speaker is near the back wall. At greater distances it does get somewhat lost amongst the standing waves.
Here are two speakers being measured close in (1.5 meters or so). 130 and 420Hz, or so, dips from floor bounce.
David
Attachments
I should do some more listening, but my first impressions were that they were surprisingly similar in spite of being totally different approaches. One interesting thing about the artificial reverb, that I have noticed before, is that there is a huge range of audibility as a function of the source. On a lot of music you can add a bunch of reverb and find it just noticable. Switch over to talk radio and the same amount of echo is laughable. Griesinger (Lexicon) talks about this a lot.
It was a "forced choice" comparison using two sounds in seperate trials: a 1 second sample of pink noise and a 3.8 second clip of male speech. In random order one sample was the standard (unmodified) signal and the second had the level of one of the 16 reflections adjusted. You were to choose which was standard (I'm a little unsure how you decide what "standard" means).
He would start each sequence with the particular reflection 10dB louder than normal and gradually lower its level until your choices became random. From this he found 2 parameters: the Threshold of Detection and the Just Noticable Difference. If you compare the Threshold of Detection to the natural level in the room, only reflections typically louder than the threshold would be considered important (detectable). His conclusion from the first paper is that the floor and ceiling bounce are the only potentially audible reflections. A second paper with a more refined setup gave floor and near side wall as audible on noise, and only the floor bounce audible on speech.
David S.
Last edited:
Hi David,
I doubt what you've measured is the floor bounce. Vary the distance of the whole setup from the front wall while keeping height and distance of mic and speaker relative to each other. If the dips change in frequency it is an adjacent boundry effect. If the frequency stays the same and just the magnitude changes, the effect was modal.
It is most certainly audible. If you've ever experimented with a 3 way design where you can adjust the height of the woofer all the way from the floor to just under the midrange driver, the difference is large, and plain to hear.
Of course some of the change is due to room modal effects (mainly the floor-ceiling 1st and 2nd order modes below 200Hz) and some of it is also due simply to proximity to the floor, and also occurs outdoors.
When people speak of floor bounce I think they usually mean only the incident bounce between the speaker and the listener - the time delayed copy which introduces a notch and comb filtering.
At normal listening distances that notch goes high enough in frequency (above 300Hz or so depending on driver height) that the modal density of the room tends to fill in or disguise this notch.
Certainly in my current room I'm unable to measure the notch calculated from the single floor bounce back at the listening position - there are too many other reflections at various random phase angles for it to manifest without some other reflections filling it in.
On the other hand the floor ceiling 1st & 2nd order modes (typically 70Hz and 140Hz for a 2.4 metre / 8 foot ceiling) do tend to manifest at any listening distance. Depending on the length and width of your room the 70Hz notch may be filled in by other modes, but the 140Hz notch will usually remain stubbornly present.
Technically not floor bounce perhaps, but it is still a response aberration whose primary cause are the bounces between floor and ceiling, and are least troublesome when the woofer is very close to the floor, as measured anywhere in the room.
For this reason I think its fair to lump the floor/ceiling modes in with the usual "floor bounce" from a practical perspective.
It may be a natural phenomena for a source of sound which is actually right there where the speaker is located, but now everything played through the speakers has the same signature caused by the floor, (and side/front walls for that matter) which will help localize the speaker.
With speakers we want to be able to produce an entire soundstage with sound sources at a variety of virtual locations (including depth) which are not the same location of the individual speakers.
In fact disguising the precise location of the individual speakers whilst still preserving pin-point imaging of virtual sound sources should be the goal, so anything that unmasks a speakers actual locations when producing a sound stage is a bad thing IMO.
Ironically although Linkwitz says its a natural phenomena that the brain is used to dealing with, he does also admit that he believes that the floor bounce is one of the cues that is used to perceive the distance or depth of the sound source - which I agree with.
The implication being if there is a strong floor bounce signature from the speaker, it will tend to localize the location of the speaker (at least in depth and maybe height) rather than the location of the virtual sound sources.
Last edited:
Good point.
Linkwitz has long stated that he wants to minimize misleading cues, but after the addition of the rear tweeter he has stated that an on-wall speaker misses important spatial cues (ORION++). That seems somewhat strange. On the other hand, one might argue that some cues about the location of the speaker are needed, since listening to stereo in an anechoic situation or with very directional speakers may suffer from in-head or too near localisation of the sound stage.
Linkwitz has long stated that he wants to minimize misleading cues, but after the addition of the rear tweeter he has stated that an on-wall speaker misses important spatial cues (ORION++). That seems somewhat strange. On the other hand, one might argue that some cues about the location of the speaker are needed, since listening to stereo in an anechoic situation or with very directional speakers may suffer from in-head or too near localisation of the sound stage.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.