I am sorry for the late reply. I have had way too many deadlines to meet for a retired guy.
"Salvation lies within."
Originally Posted by xianthax
Here is the conclusion from the linked article:
1. In virtual reverberant simulations, perceived lateral angles of sources more than 45* from the median plane are biased toward the median plane, an effect that grows with increasing distance (as D/R decreases). More medial sources tend to be biased laterally; this lateral bias decreases with increasing distance.
2. The localization bias caused by reverberation is greater for low-frequency sounds than for high-frequency sounds.
3. Listeners do not always weight low- and high-frequency localization cues optimally in a reverberant space. In the most reverberant stimulus condition (1.7 m simulated source distance), localization accuracy, as measured by RMS error, is poorer for noises containing both low-frequency and high-frequency components than for a narrowband, high-frequency noise alone, and better than for a narrowband, low- frequency noise alone.
Compare to Toole's conclusions.
Warden Samuel Norton, from the movie, Shawshank Redemption.
It would be my pleasure to make that comparison. So that everyone follows along, please allow me to state your case first.
The point being discussed here is the location of sound sources which in our case means speakers in our rooms. So let’s start at the top with our speaker configuration and for the sake of discussion let’s use ITU/AES recommendation for surround system:
Let’s cover some important terminology that is used in this space including the paper we are discussing, namely the fact that angles are relative to a listener looking forward. The center channel is therefore zero degrees. The angles increase to the right. In the ITU recommendations then right front speaker is at 30 degrees. Left front is –30 degrees. Surrounds as mentioned are +- 110 degrees.
Now let’s take the left front (LF) speaker. The prosecution’s case is simple: said speaker spits out sound that hits your ear but it also sends sound all over your room. The right wall for example will receive some of that sound energy and bounce it back at you. Ditto for the ceiling, floor, etc. The point he is trying to make is that all of these reflections serve to confuse you, making it hard to tell the direction of the sound. After all, there are “little speakers” all around the room now courtesy of the sound bouncing around. Seems like a very convincing argument, no? We almost don’t need a study but he provides one anyway. So let’s examine if he has sufficiently made his case that these reflections have a detrimental effect.Defense’s Case:
First thing we need to do is dig into what the paper analyzed and its full set of conclusions. We could be falling in the ditch trusting the summary put forward.
The way the study methodology is based on measuring the binaural response of a room that “had a thin carpet on the ﬂoor and a suspended ceiling composed of acoustically treated tiles. Several pieces of ofﬁce furniture were in the room, including a whiteboard on the wall nearest to KEMAR, causing some modest asymmetries in the reverberant energy for sources in the left and right hemiﬁeld.”
So what we have in plain language is an ordinary (office) room that has a lot of reflections in it. The sound source was a Bose sound-cube (cue up all the Bose jokes
Impulse response was measured for distances ranging from 0.15 meters (0.5 feet) to 1.7 meters (5.5 feet). Source sound angles ranged from -90 degrees to +90 degrees in 15 degree increments. The impulse response gives us the room+speaker transfer function which can then be applied to the stimulus signal. Since we measured them for each ear independently, we can then run the tests using headphones which indeed is what they did (testers then wore Sennheiser HD 580 headphones).
At the risk of stating the obvious this test is based on speakers playing in a room so therefor it is “amplified sound.” So the notion that such research is for non-amplified speech is non sequitur.
So the quick summary is that a real room was measured and then simulated with noise coming from different angles and lengths to the listener and they were asked to tell which direction they thought it came from relative to what was simulated.
The test signal was noise with either low or high frequency spectrum. Tests were performed with either one or both. For convenience, here is a graph for the composite noise. The point I am about to make had the same result in the other test where each noise signal was played independently:
The X axis is the angle of the real sound. The Y axis is the error in degrees between what the listener heard and what the reality was. As an example, if the source is at 45 degrees and the listener picks 45, then it shows up as zero in that graph. If he picked 30, then it would show up as 15 degrees (of error). You see multiple lines there. Each represents a different distance from the listener ranging from 0.5 feet to 5.5 feet as mentioned. If you are having a hard time telling them apart, the 5.5 feet one is on the bottom. And 0.5 on top.
With that primer out of the way, let’s examine what the results say. Our center speaker is at 0 degrees. If we look that up we see that the degree of error is zero. As I post directly from the article in my last reply, we are most precise when a sound is coming toward us on a line bisecting our ears. So the reflections in the room did not have an impact there even though our center speaker is also sending sound waves everywhere in addition to the direct sound being sent forward.
Now let’s look at the case of our left and right speakers. Per ITU spec they are at + and – 30 degrees. So let’s go to the right to where it says 30 degrees on the X axis. Now, most of us probably sit farther than 5.5 foot limit of this study so there is no specific data point to use. Looking at the curves, we see that error gets smaller and smaller as we move from 0.5 feet to 5.5 feet. So it seems safe to use the bottom curve. Once more we see that the error is essentially zero! Once again we see absence of a problem. We were able to tell the direction of the “left and right” speakers nearly as well even though we had all of those reflections to deal with .
How about our surround speakers? They are located past 90 degrees. This test does not address that situation and indeed, the listeners were told the range was + and – 90 degrees. So we can’t determine anything about that but I will come back to this in a moment.
Let’s pause for a moment and realize what we just covered. In a room full of reflections, at distances of 5.5 feet, our ability to track the location of a source was still quite good. None of what we intuited happened regarding the brain/ear getting confused. It is as if the reflections were filtered out.
Of note, the prosecutor put forward a summary which talked about angles > 45 degrees. It also did not note the “1.7m” (5.5 foot) limitation of the study. So even without reading the report we should have questioned why such data was being put forward regarding applicability to our situation. Two days later, no one did.Did we "win?"
You might think this would be time for a victory dance. But it is not. The whole premise of wanting to know the precise location of the speakers is false! Your center speaker may be below an 8 to 10 foot projection screen. Imagine two people talking on said screen. Exactly who wants the sound to come completely from the location of the center speaker? How would that be realistic and appropriate target for our reproduction system? How would that create the illusion that you are inside the movie rather than a plain room of your home theater?
How about when we play music? Do we want the sound to come precisely from those speakers themselves ala 60’s stereo rock music where the singer is in the left speaker and the guitar in the right? How about when listening to a classical concert? We want sounds to come directly from the speakers or a larger soundstage? We can be so anal about precise reproduction in our home that we manufacture criteria such as “pin point” localization where in reality what the emotional brain *desires* is not that.
Not saying we want the front speaker sound to come out of the back speaker but the notion that we want the sound to come directly from the cone of the speaker is something we may wish we had, but if we closed our eyes, chances are that is not what we would prefer. Folks go out of their way to deploy these so called “dipole” surround speakers so that they can’t tell sound sources from behind. If you believe that, why would accuracy of a point source be of value to you as angles get more acute (where we did lose accuracy)?
Let’s say you did have a target of having the sounds come directly from the speaker. How do you know that is the precise angle of where it is supposed to come from? Some people believe in 30 degree model, others say 25 or some other variation. Who knows what was used to record all the content you are listening to? How about your distance to the speakers vs. what was used when the talent heard the same thing? You don’t really know what was “there.” So you could accurately reproduce the wrong angle!Explaining the Results
How on earth did all of those reflections not confuse us in our scenario? Is the study right? Well, think about it. When you are at home and someone calls you, do you have trouble telling which direction to turn to? I assume not. If you are watching TV, do you have any problems knowing where the sound is coming from? Go ahead and test it, we can wait
. How about the last time that you ran the calibration tones on your AVR or processor? Surely you did not think you had a “problem” there as far as telling where your speakers there. Sure there can be some error but nothing remotely to what one thinks should be happening.
The reason we didn’t have a “problem” here is that our brain is not working the way we think it is. If it did, then we would have completely different outcomes given our reflective our rooms and everyday lives are. The hearing process has two parts to it. The first occurs at the onset of the sound event. We note it immediately (first phase) but then a longer (“integration”) process (second phase) starts where the brain collects more acoustic energy (“data”) over tens of milliseconds. Once it has accumulated enough it will then start a process to determine what it has heard. This determination occurs at a cognitive (higher) level than the parts of the brain that perform the normal hearing function. The measurement engineers in the crowd know this concept of performing multiple measurements and averaging them to reduce noise and extract more usable signal. Similar thing here.
The above is why we don’t hear echoes in the room even though that is precisely what they are: later reflections of the original sound. The brain is smart and is attempting to extract more knowledge from the environment to better hear what is going on. In doing so, it is sampling each reflection and adding it (“integrating it” in research lingo) as if it was part of the direct sound and thereby, increasing its total power.
Getting back finally to the question that was asked of me regarding Dr. Toole’s conclusions in this matter, the answer is plain view in section 2.1. It explains exactly what I just explained regarding mechanism behind sound localization (direction of sounds). He doesn’t throw this stud at you because his paper came out in 2006 and this study in 2011. Yet his view completely holds since it was based on many other listening tests showing the same phenomena.
The kind of analysis I just performed is precisely what Dr. Toole does in his classes, book and papers. He connects the dots. He takes research such as the current paper that is full lingo that is hard to understand and connects them to our scenario at home. He has a unique ability in doing so and hence the reason I use his writing so much. I could refer to other work but you see how much it takes to get through the topic using raw data that way.
The prosecutor hopes that you put aside this science which admittedly can be complex and hard to follow and instead follow your instincts that says reflections must be interferences that degrade the source. If you think about it, this is no different than someone wanting you to buy an expensive cable because it is thicker or has fancier construction. You are quick to not believe that. Why differently here? Science has to be our guide in all of this and not a part-time occupation.
Bottom line is this: our client, the reflection, is innocent:). Importantly we showed that not using our evidence but that of the prosecutor!
I started this reply with that line from the jail warden in my favorite movie, Shawshank Redemption. There, he tells that to character of Tim Robbins, Andy Dufresnem, to ostensibly believe in the bible. Little did he know that Andy had buried a pick-axe inside it which he used to escape the prison! Salvation was indeed within. The cover of the book told you nothing and such is the case with the summary of the paper quoted. Andy was innocent and so is our client.
P.S. What happened to quoting the other article on surround movies? Was the abstract all you had?