Allison on Soundfields

[Howard Ferstler]

The home page "recent updates" sidebar now has an article listed called "Allison on Soundfields" that will be of interest to fans of Acoustic Research. Just click on the title to get the draft to download. The home page, of course, is at:

http://www.classicspeakerpages.net/

The info this draft contains should be of interest to any fans of the "classic" AR speaker lines. The version that appeared in the JAES is a bit later was shorter than this original that Roy sent to me some time ago. This is an almost definitive statement about the AR-3a as its performance relates to real-world soundfields and power response, and it has info in it that Roy used later on to design the LST, as well as the Allison Acoustics line. There is also info in it that relates to the AR-ax.

I have received permission from the AES to post this artilce. Note that it is a large (over 25 MB) PDF (Adobe) file and will take some time to download. If the text and illustrations are hard to work with, it would be simple to print them out. Lots of paper required, unfortunately, and the photos will print slower than the text section.

Howard Ferstler

[Zilch]

Thank you for your effort to accomplish this, Howard.

I'm downloading it on dialup, alas (2 hours,) but it'll doubtless be a fun and worthy read....

[kkantor]

The home page "recent updates" sidebar now has an article listed called "Allison on Soundfields" that will be of interest to fans of Acoustic Research. Just click on the title to get the draft to download. The home page, of course, is at:

http://www.classicspeakerpages.net/

The info this draft contains should be of interest to any fans of the "classic" AR speaker lines. The version that appeared in the JAES is a bit later was shorter than this original that Roy sent to me some time ago. This is an almost definitive statement about the AR-3a as its performance relates to real-world soundfields and power response, and it has info in it that Roy used later on to design the LST, as well as the Allison Acoustics line. There is also info in it that relates to the AR-ax.

I have received permission from the AES to post this artilce. Note that it is a large (over 25 MB) PDF (Adobe) file and will take some time to download. If the text and illustrations are hard to work with, it would be simple to print them out. Lots of paper required, unfortunately, and the photos will print slower than the text section.

Howard Ferstler

This is one of the landmark papers in the field of loudspeaker design. Thanks!

-k

[Shacky]

Thank you for your effort to accomplish this, Howard.

I'm downloading it on dialup, alas (2 hours,) but it'll doubtless be a fun and worthy read....

Double kudos to Howard!!!

Zilch - how the heck to you get 20K+ posts with dial up?

[Zilch]

Zilch - how the heck to you get 20K+ posts with dial up?

It didn't happen in the last couple of months, is how; I've been at this a while....

*******

On first read, Howard, the text is virtually identical to that published in the AES Anthology (1972).

I do see additional in-room measurement curves and photos of rooms shown, i.e., more room data, as well as the dummy head curves, and the illustrations in the PDF are shown with higher resolution than in the print version.

Other than that, what is it you consider to be substantively different between the two versions as you earlier suggested was the case?

[soundminded]

The home page "recent updates" sidebar now has an article listed called "Allison on Soundfields" that will be of interest to fans of Acoustic Research. Just click on the title to get the draft to download. The home page, of course, is at:

http://www.classicspeakerpages.net/

The info this draft contains should be of interest to any fans of the "classic" AR speaker lines. The version that appeared in the JAES is a bit later was shorter than this original that Roy sent to me some time ago. This is an almost definitive statement about the AR-3a as its performance relates to real-world soundfields and power response, and it has info in it that Roy used later on to design the LST, as well as the Allison Acoustics line. There is also info in it that relates to the AR-ax.

I have received permission from the AES to post this artilce. Note that it is a large (over 25 MB) PDF (Adobe) file and will take some time to download. If the text and illustrations are hard to work with, it would be simple to print them out. Lots of paper required, unfortunately, and the photos will print slower than the text section.

Howard Ferstler

I've read it through once already and will probably read it over several more times. It was a great contribution to the knowledge of the art at the time it was published and said many interesting and valuable things that were novel. I can't find anything explicitly wrong with it but it strikes me as incomplete and inadequate. Also, some of the conclusions are taken somewhat out of a broader context. One important point it makes, the importance of both the direct and reverberent field being reasonably flat is not reflected in speaker design as there are no provisions to adjust them independently to any degree in any speaker I'm aware of (my own designs have these provisions but so far they are designed for specific locations in specific rooms with known frequency selective acoustics.) An interesting thing happens if the top octave of the direct field is removed while the indirect field which in my case is over 95% of the radiated field occurs. On some material especially with sibilant and other hf explosive sounds, there is a strange sound similar to the phase shift effect deliberately used for some pop recordings. Again, the signal filtering of all high frequencies does not accurately reflect the phenomena of hf rolloff in concert halls because that is a dynamic process where the relative spectral content of each note changes gradually with time as the sound decays while the filtering suggested for this type of sound system is a steady state filter affecting all sounds from the earliest to latest arriving equally. To duplicate the effect on the subjective timbre of instruments the way a concert hall affects them, you'd have to filter different time delay components differently. (DSP-1 which I contend was an infringement on my patent makes provisions for this.) The conclusion that the interference effects between drivers at the crossover frequencies is not particularly important is almost surely correct. Not only does every multiway speaker ever produced exhibit this characteristic, it is likely most musical instruments do the same live. This is because from the audience's point of view they are not point sources and early arriving sounds travel multiple trajectories from different points on the instrument arriving at the listener simultaneously. Nor is it surprising that the FR effects of the outer ear irrelevant. Lots of thoughtful work here, too bad this is about the point where valid and pertinent research stopped. Toole's papers pay lip service to accuracy and then abandon it for market research.

Allison made the same point I did about where microphones are placed for recordings compared to where the listener sits in the audience I did. This has a profound effect on both timbre and the relative quantity and quality of the reverberant field captured vis a vis the direct field and early reflections.

The Acoustic Energy Field Transfer Function theory differs from this model and it differs from the pulsating sphere model. Within their limited context the Transfer Function Theory incorporates what they leave out but it also incorporates much they don't even try to address. It gives an exact result for any point in any room with any source, it isn't an approximation. The method of measurement it suggests would give entirely different and much more informative data. Unfortunately the equipment to perform this measurement doesn't exist yet.

[soundminded]

I've read it through once already and will probably read it over several more times. It was a great contribution to the knowledge of the art at the time it was published and said many interesting and valuable things that were novel. I can't find anything explicitly wrong with it but it strikes me as incomplete and inadequate. Also, some of the conclusions are taken somewhat out of a broader context. One important point it makes, the importance of both the direct and reverberent field being reasonably flat is not reflected in speaker design as there are no provisions to adjust them independently to any degree in any speaker I'm aware of (my own designs have these provisions but so far they are designed for specific locations in specific rooms with known frequency selective acoustics.) An interesting thing happens if the top octave of the direct field is removed while the indirect field which in my case is over 95% of the radiated field occurs. On some material especially with sibilant and other hf explosive sounds, there is a strange sound similar to the phase shift effect deliberately used for some pop recordings. Again, the signal filtering of all high frequencies does not accurately reflect the phenomena of hf rolloff in concert halls because that is a dynamic process where the relative spectral content of each note changes gradually with time as the sound decays while the filtering suggested for this type of sound system is a steady state filter affecting all sounds from the earliest to latest arriving equally. To duplicate the effect on the subjective timbre of instruments the way a concert hall affects them, you'd have to filter different time delay components differently. (DSP-1 which I contend was an infringement on my patent makes provisions for this.) The conclusion that the interference effects between drivers at the crossover frequencies is not particularly important is almost surely correct. Not only does every multiway speaker ever produced exhibit this characteristic, it is likely most musical instruments do the same live. This is because from the audience's point of view they are not point sources and early arriving sounds travel multiple trajectories from different points on the instrument arriving at the listener simultaneously. Nor is it surprising that the FR effects of the outer ear irrelevant. Lots of thoughtful work here, too bad this is about the point where valid and pertinent research stopped. Toole's papers pay lip service to accuracy and then abandon it for market research.

Allison made the same point I did about where microphones are placed for recordings compared to where the listener sits in the audience I did. This has a profound effect on both timbre and the relative quantity and quality of the reverberant field captured vis a vis the direct field and early reflections.

The Acoustic Energy Field Transfer Function theory differs from this model and it differs from the pulsating sphere model. Within their limited context the Transfer Function Theory incorporates what they leave out but it also incorporates much they don't even try to address. It gives an exact result for any point in any room with any source, it isn't an approximation. The method of measurement it suggests would give entirely different and much more informative data. Unfortunately the equipment to perform this measurement doesn't exist yet.

Here is an obvious error in the paper, an assertion that is easily disproved by example. It's at the bottom of page 3 and top of page 4;

"Aside from the effects of standing waves, the reverberant field SPL is (theoretically)

determined only be two factors: the power level of the source, and the

room constant!. ! is a factor expressing the average energy absorption coefficient

of the room surfaces and the total surface area. Thus the room can be

thought of as an acoustic energy sink: the source of energy (the loudspeaker)

pours energy into the sink at a given rate, and the steady-state amplitude of

this energy is determined by the rate at which energy is absorbed by the room

surfaces and furnishings. The £teady-state energy, which is the reverberant

3.

field, is independent of the directivity of the source or the position it occupies

in the room, at least above 1,000 Hz. Below 1,000 Hz the reverberant field in a

living room is not dependent on the directivity of the source but it is dependent

to some extent on the position of the source, because position affects room mode

excitation."

Consider a room where one wall is covered by a heavy drape or tapestry, the wall facing opposite it is covered by a mirror. Place one pair of Bose 901s in front of the draped wall, another in front of the mirrored wall. Bose 901 is a speaker whose directivity index is close to zero within its frequency range, it generates sound fields that are almost entirely reverberant. (Or use AR3as or any other higher Q speaker facing the walls instead.) With identical settings and power inputs, the speakers will sound very different depending upon which pair you are listening to. The concept of time averaging and space averaging to explain acoustics is therefore inherently flawed. In the extreme case I have given, the reverberant field would be almost entirely absorbed by the drape. If it is the first object the field is reflected off of, there will be little left for subsequent reflections. On the other hand, if the first reflection is the mirror, almost all of the sound will be reflected on the first bounce and only a small portion will reach the draped wall on the next bounce. The RT and SPL will be much higher. This demonstrates that reverberant fields are dynamic vector fields which cannot be simplified by viewing them as steady state (time and space averaged) static fields.

Now cut out a section of drape and a section of mirror of equal size and swap them so that the section of mirror is directly facing the speaker on the draped wall while the drape is facing the speaker on the mirrored wall and repeat the experiment. The R factor will not change because the total absorption of the room remains unchanged. The high Q speaker facing the section of mirror on the draped wall will far more effectively bounce its first reflection off that wall with the mirror section on it than the Bose 901 will while on the opposite wall, its sound will be absorbed more effectively than the Bose's field will proving that not only is the detailed arrangement and type of reflective surfaces a factor in RT and SPL but so is speaker directivity and the position of the speaker with respect to the boundary values of the room. The hundred year old concepts developed by Sabine was an advance in its day, the first ray of scientific knowledge in acoustics but it is not adequate to accurately predict the precise nature of reverberant sound fields in concert halls or in small rooms. It's a gross oversimplification.

This paragraph in Allison's paper is simply wrong.

[Carlspeak]

Hey, you guys aren't the only ones who've argued about loudspeaker/room interactions. Check out the 13 pages of posts from a more international perspective here:

http://www.diyaudio.com/forums/showthread....mp;pagenumber=1

[Zilch]

Thanks for the post to that believable data. This is the kind of response curves that I typically see when I measure loudspeakers - they are terrible.

[soundminded]

There is another error, too, since we are in the mood to nitpick. In the phrase "only be two factors" located in the first sentence, the word "be" should be "by." And the original text has the icon reproduced as an underlined R and not an explanation point. The latter looks kind of odd with your scanned copy.

Getting a bit more serious, in the last sentence of the section you reprinted Roy indicates that the position of the source at frequencies below 1 kHz is important "to some extent." That means that it is not necessarily going to be the case with some radical speaker locations and rather unusual decor considerations, and even above 1 kHz the kind of locations and reflecting and absorbing boundaries you outlined are rather extreme. In typical rooms and with typical placement, what he said works fine.

Howard Ferstler

"and even above 1 kHz the kind of locations and reflecting and absorbing boundaries you outlined are rather extreme."

I gave an extreme example to illustrate the point clearly. But even that example is not far from the situation sometimes encountered. For example. Consider a room where the only wall available is the front wall of a room where there are windows with heavy drapes. Bose 901 placed on that wall would be substantially affected and to a different degree by that placement from an alternative to say a horn type speaker like a Radio Shack Mach 1 with its bullet horn narrow dispersion tweeter. This points out a simple and stark fact that demonstrates just how badly thought out the concept of current sound systems is. Of all the sound systems in the world, no two sound exactly alike, even with the same equipment arranged differently in the same room. What's worse, no engineering provisions are incorporated in most of them to compensate for that fact, no such provisions are even wanted by audiophiles. Therefore no matter what the recording engineer heard when he mastered the recording at the studio, the home listener will hear something different, even if it is the engineer himself using identical equipment he mastered the recording with.

I've encountered Berkovitz's mistakes before. In the challenge phase of my patent, the examiner threw up Berkovitz's patent to me. Berkovitz said that the reverberent field after 100 ms is indistinguishable from random noise. My reply was had he thought so, he should have incorporated a random noise generator in his design. It is obviously false.

I've only gotten this far on my second read through of this paper. I'll continue to report any other significant errors I think I've discovered. The search for simplicity in analyzing acoustics such as one number R to describe a room strikes me as naive as trying to find one number to describe something relatively simpler such as power output of an audio amplifier. It was a start, and a good one but it was only a start. A journey of a thousand miles begins with a single step the Chinese say but sooner or later, if you don't take more steps, you never will reach the destination.

[kkantor]

I've encountered Berkovitz's mistakes before. In the challenge phase of my patent, the examiner threw up Berkovitz's patent to me. Berkovitz said that the reverberent field after 100 ms is indistinguishable from random noise. My reply was had he thought so, he should have incorporated a random noise generator in his design. It is obviously false.

Of course, plenty to debate on this subject, but you are misquoting Berkovitz, and accusing him of an "error" that I do not believe he made. Here is what he actually wrote:

"The pattern of reflections in a concert hall can be considered to have two aspects: the early reflection pattern formed by low-order reflections arriving at a listener within a time period, for an average size concert hall, of the order of 100 msec. following arrival of the direct sound; and the "reverberation" pattern comprising large numbers of temporally closely spaced reflections arriving during a period following the early reflections. In most concert halls, the density and randomness or incoherence of reflections in the reverberation pattern is sufficient to resemble band-limited noise with spectral characteristics similar to those of the original sound, but without distinct time- or direction-ordered components."

-k

Footnote: Berkovitz' description above is consistent with the "Linear Prediction" model of audio signal identification.

[soundminded]

Of course, plenty to debate on this subject, but you are misquoting Berkovitz, and accusing him of an "error" that I do not believe he made. Here is what he actually wrote:

"The pattern of reflections in a concert hall can be considered to have two aspects: the early reflection pattern formed by low-order reflections arriving at a listener within a time period, for an average size concert hall, of the order of 100 msec. following arrival of the direct sound; and the "reverberation" pattern comprising large numbers of temporally closely spaced reflections arriving during a period following the early reflections. In most concert halls, the density and randomness or incoherence of reflections in the reverberation pattern is sufficient to resemble band-limited noise with spectral characteristics similar to those of the original sound, but without distinct time- or direction-ordered components."

-k

Footnote: Berkovitz' description above is consistent with the "Linear Prediction" model of audio signal identification.

Just because Berkovitz can't measure it due to his inadequate modeling and measuring methods doesn't make it so. In fact IMO it is dead wrong. It is probably based on the same scalar model of sound fields in general and of reverberant sound fields in particular we argued about some time ago. If the scalar model of sound were correct, binaural sound reproduction would work...and sonar wouldn't. Insistance on this simplistic and incorrect idea is one big reason the state of the art is frozen. When you hit your head against a brick wall and it doesn't budge, eventually it is a good idea to stop, step back, taken another hard look at it from a different perspective to try to find a way around it. This conversation ended the last time and I will not resurrect it again. From my side it is not open to debate.

[speaker dave]

Let me comment on SoundMinded's post Number 10.

In your experiement with with 2 pairs of 901, one mounted in front of a reflective wall and the other in front of an absorbtive wall, it is correct that the level in the room would be different. Since the absorbtive case has much of the energy absorbed by the very first surface then, yes, the reverberent field level will decrease. The effective room RT would not. Whatever percentage of sound that did bounce off the realatively dead surface, plus the radiation that hit all the other surfaces, plus radiation from the front firing driver would build up a diffuse field that decays in the normal manner.

The reference to reverberent level in the room being determined by sound power of the source and average absorption of the room is usually refered to as the Hopkins Stryker law or equation. It is a fundamental in room acoustics and not something we need to debate. It does only apply to rooms with a diffuse field or at least to the frequency range where the field is diffuse (above 160 Hz in my living room, by measurement). It won't necessarily apply to contrived rooms with atypical surfaces (one wall only made absorbtive). Acousticians also know its limitations for nonstandard room shapes. For example, with extreme aspect ratios, such as large office areas with a relativley low ceiling it isn't typically used. For concert halls, after the research of Baron, "revised" theory gives a reverberent field with a small drop over distance (about 1dB every 10 meters) rather than constant throughout the room.

That the systems sound different is probably true, but is removed from what Allison and Berkovitz are explaining. If you believe that the steady state response (I.e. the revereberent field level in the room) determines the perceived response, then even though the RT of the room doesn't change with your two experimental cases, the reverberent field spectrum can be very different. It is probable that the absorbtive wall has some frequency dependent parameters (absorbs more with higher frequencies) and also likely that the sound that misses that absorbtive patch, that which diffracts around the patch, plus the front firing drivers contribution, has its own spectrum. This would give a different steady state response in the room.

Most researchers these days do not believe that the steady state response (the "room curve") is a very good indicator of your perception of the loudspeaker's sound. I've seen this repeatedly given as an assumption in this forum, and I know that Allison stated it at the time, but all modern research on the subject has pointed to the ear perceiving a time windowed response, that is shortest for high frequencies and longest for low frequencies. Generally this means that if you want a response curve that reflects what you perceive then you should measure the direct sound (only) for high frequencies, a time window for mid frequencies long enough to capture the floor and ceiling bounces, and a window long enough at low frequencies to let in most of the effects of the room. Read the papers of Soren Bech, Kates and others such as the Lipshitz and Vanderkoy paper on modifying the power spectrum of a loudspeaker in the room, and you'll see. Certainly with your two cases of Bose 901, the time spectrum of the the response would be very different. With one front firing unit and 8 rear units, you have created a case similar to having a variable pot on the rear firing units. Turning them down (placing absorbtion behind) will knock down the early reflections and much of their contributions to reverberent level. To the extent that early arriving sound has its spectrum altered, the perceived balance of the system would change.

Spaciousness, a parameter apart from perceived frequency balance, would clearly change as well.

David

[soundminded]

Let me comment on SoundMinded's post Number 10.

In your experiement with with 2 pairs of 901, one mounted in front of a reflective wall and the other in front of an absorbtive wall, it is correct that the level in the room would be different. Since the absorbtive case has much of the energy absorbed by the very first surface then, yes, the reverberent field level will decrease. The effective room RT would not. Whatever percentage of sound that did bounce off the realatively dead surface, plus the radiation that hit all the other surfaces, plus radiation from the front firing driver would build up a diffuse field that decays in the normal manner.

The reference to reverberent level in the room being determined by sound power of the source and average absorption of the room is usually refered to as the Hopkins Stryker law or equation. It is a fundamental in room acoustics and not something we need to debate. It does only apply to rooms with a diffuse field or at least to the frequency range where the field is diffuse (above 160 Hz in my living room, by measurement). It won't necessarily apply to contrived rooms with atypical surfaces (one wall only made absorbtive). Acousticians also know its limitations for nonstandard room shapes. For example, with extreme aspect ratios, such as large office areas with a relativley low ceiling it isn't typically used. For concert halls, after the research of Baron, "revised" theory gives a reverberent field with a small drop over distance (about 1dB every 10 meters) rather than constant throughout the room.

That the systems sound different is probably true, but is removed from what Allison and Berkovitz are explaining. If you believe that the steady state response (I.e. the revereberent field level in the room) determines the perceived response, then even though the RT of the room doesn't change with your two experimental cases, the reverberent field spectrum can be very different. It is probable that the absorbtive wall has some frequency dependent parameters (absorbs more with higher frequencies) and also likely that the sound that misses that absorbtive patch, that which diffracts around the patch, plus the front firing drivers contribution, has its own spectrum. This would give a different steady state response in the room.

Most researchers these days do not believe that the steady state response (the "room curve") is a very good indicator of your perception of the loudspeaker's sound. I've seen this repeatedly given as an assumption in this forum, and I know that Allison stated it at the time, but all modern research on the subject has pointed to the ear perceiving a time windowed response, that is shortest for high frequencies and longest for low frequencies. Generally this means that if you want a response curve that reflects what you perceive then you should measure the direct sound (only) for high frequencies, a time window for mid frequencies long enough to capture the floor and ceiling bounces, and a window long enough at low frequencies to let in most of the effects of the room. Read the papers of Soren Bech, Kates and others such as the Lipshitz and Vanderkoy paper on modifying the power spectrum of a loudspeaker in the room, and you'll see. Certainly with your two cases of Bose 901, the time spectrum of the the response would be very different. With one front firing unit and 8 rear units, you have created a case similar to having a variable pot on the rear firing units. Turning them down (placing absorbtion behind) will knock down the early reflections and much of their contributions to reverberent level. To the extent that early arriving sound has its spectrum altered, the perceived balance of the system would change.

Spaciousness, a parameter apart from perceived frequency balance, would clearly change as well.

David

"Most researchers these days do not believe that the steady state response (the "room curve") is a very good indicator of your perception of the loudspeaker's sound."

I discarded that notion 35 years ago. If you read my other postings, you will find my explanation of why the spectral content of the sound field reaching a listener is a dynamic event and the subjective timbre is inseparable from the reverberant field. That means if you don't reproduce the reverberant field correctly, you can't reproduce the timbre. This is why the tonality of musical instruments heard at a concert cannot be duplicated with today's concept of a sound reproduction system.

The figure of RT is a simplistic notion which depends heavily on how the measurement is made and how the data is interpreted. Technically it is the time required for sound to decrease by a factor of one million or 60 db which is why it is sometimes referred to as the RT60. The measurement is skewed by among other things the distance to the source. Unless the source sound is excluded from the data, measurements close to the source will show very low RTs because the direct field will overwhelm the reverberant field. Also, the data can be skewed if there is a flutter echo condition. Data I have for various concert halls gives the example of Carlton College Concert Hall in Northfield Minnesota built in 1970 and later renovated. It followed the traditional shoebox design but due to some peculiarity of this hall (lack of "sculpted elements" among them) this strange flutter echo was one problem, excessive reverberation another. Classical reverberation onset didn't occur until 300 ms. This caused a wide discrepency between the measured and perceived RTs. At mid frequencies, the actual RT was in the range of 2.15 seconds but the perceived RT was 3.0 seconds. "There is a strange pulsating quality to the reverberant sound for short duration sources. This is believed to result from the coincidence of flutter intervals with a repetition rate of 800 to 1000 msec." This was prior to remodeling begun in 1981.

"It is probable that the absorbtive wall has some frequency dependent parameters (absorbs more with higher frequencies)"

Not probable, certain. Allison and Berkovitz saw this too; "There is one obvious contaminant left in the data: the effect of the average room absorption with frequency." The reflection/absorption coefficients of all materials is frequency dependent. Curves for many materials are easily obtainable, especially for materials used in office construction such as office partitions, sheet rock, and ceiling tiles. Noise transmission especially in open landscape offices is a serious concern in modern office design. The unpublished method of analysis and measurement I developed I call acoustic energy field transfer theory which my patented and other inventions are based on does not have these limitations. At the current state of the art, computation of the exact nature of the reverberant sound field is not possible. The only approach I've seen which may have long range promise is called Computational Flow Dynamics (CFD) which is used in part to predict precise air flow in a room with multiple sources to assess the efficiency and efficacy of Air Conditioning designs and the probability that they will be adequate in simulated failure modes. This is applied for among other things, mission critical data centers. The model takes into account the entire geometry of the room including restrictive air passageways of equipment cabinets, the volume, velocity, location etc. of the flow sources and predicts the air flow at all points within the room. These sources would be effectively the equivalent of DC sound sources since they are unidirectional. The best available software is very expensive, very difficult to use, and not considered particularly reliable or accurate. But even if it were perfected, it's a long way from that to being able to predict the sound field in one part of a room resulting from the sound field generated in another part. Ultimately when such software is developed, my method of analysis will marry up to it.

Meanwhile, here's another one of Berkovitz's and Allisons mistakes in their paper that comes straight out of my analysis method;

"Clearly, the total energy output of this system is more easily predicted by the

2 pi anechoic measurements of the individual speakers than by the 4pi rmeasurements

of the complete system cum cabinet."

The 4pi free space radiation measurement is the only way to obtain the accurate energy propagation signiture of a sound source. This in not only true because it is far more descriptive and detailed but because more than one 4pi radiation signiture can result in the same two pi radiation measurements. The observed bass boost in 2pi (steradian) space is due to the fact that speakers are quasi-omnidirectional at low frequencies and comparatively highly directional at high frequencies and this includes AR3a. In 4 pi space, this results in much energy radiated towards the sides, back, below, and above at low frequencies but little at high frequencies. In 2pi space, that energy is funneled into a smaller solid angle and in 1pi space into a smaller angle yet. Also, a minor contribution is made from the fact that most materials reflect sound to a greater degree at low frequencies. So if a pulsating sphere which had a flat response in all directions were placed in 2 pi space in a room where there were no differences in the reflectance of sound as a function of frequency, the spectral balance would remain unchanged but the sound at any point in the room would be louder. There would of course be the expected interference patterns affecting FR at any given point due to the reflections from behind.

"but all modern research on the subject has pointed to the ear perceiving a time windowed response, that is shortest for high frequencies and longest for low frequencies."

Interesting question and one I think not so easy to test. One aspect of the DSP-1 which I claim as an infringement on my patent is that it not only innovated RT as a function of frequency which was novel at the time, but the relative RT at high frequencies to mid frequencies can be adjusted in 10% increments from 10% to 100% of mid frequency RT although the inflection point is not given in the manufacturer's specs. My own experience with with long RT simulations such as cathedrals with RTs in the range of 5 to 8 seconds is that it is audible but who knows if that is the result of changes at the end of the hf perception window. In order to test this, a device which simulates long RTs without a regenerative loop such that the FR of delays beyond a certain point could be altered to determine if there are audible differences would have to be devised. Even today, this would be a very expensive and elaborate device to build because it would not only require a long string of linear delays but a large number of pick off points each equalized and mixed into the sum field separately. OTOH, it might be possible to write a software program to simulate that.

"It won't necessarily apply to contrived rooms with atypical surfaces (one wall only made absorbtive)."

Don't some audiophiles build live end / dead end rooms? I guess not.

[Zilch]

Most researchers these days do not believe that the steady state response (the "room curve") is a very good indicator of your perception of the loudspeaker's sound. I've seen this repeatedly given as an assumption in this forum, and I know that Allison stated it at the time, but all modern research on the subject has pointed to the ear perceiving a time windowed response, that is shortest for high frequencies and longest for low frequencies. Generally this means that if you want a response curve that reflects what you perceive then you should measure the direct sound (only) for high frequencies, a time window for mid frequencies long enough to capture the floor and ceiling bounces, and a window long enough at low frequencies to let in most of the effects of the room. Read the papers of Soren Bech, Kates and others such as the Lipshitz and Vanderkoy paper on modifying the power spectrum of a loudspeaker in the room, and you'll see.

I read Allison and Berkovitz as saying the same: "UT, oh, frequency selective in-room absorption has screwed us up here; our 'total energy, max dispersion' thesis is not applicable above the transition frequency (which is quite low), and we need anechoic measurements of the direct field over the listening window axes to describe what we actually hear there." Toole's synthesis of modern research reiterates this perspective.

If we integrate the listening window polars shown in the paper, it is easily seen to be THAT response which is most accurately reflected in the averaged in-room curves, certainly not the power response as measured in AR's reverberant chamber. There is no evidence that a dominant reverberant field exists at higher frequencies in these typical listening spaces.

From this I conclude that the wide dispersion touted for the drivers themselves is all but moot in the systems (due to interference and diffraction effects) and in situ (due to differential absorption of successively higher frequencies), and that its only remaining contribution of significance lies in the generation of ASW spaciousness cues via early reflections.

From this perspective, then, it's clear that wide-dispersion for its own sake is self-defeating, and attempting to put the room in charge of spectral balance above the transition frequency is not only ineffective, but may also generate undesirable artifacts. Geddes goes further and says it absolutely does, and the result is "disaster;" Toole apparently agrees....

[soundminded]

I read Allison and Berkovitz as saying the same: "UT, oh, frequency selective in-room absorption has screwed us up here; our 'total energy, max dispersion' thesis is not applicable above the transition frequency (which is quite low), and we need anechoic measurements of the direct field over the listening window axes to describe what we actually hear there." Toole's synthesis of modern research reiterates this perspective.

If we integrate the listening window polars shown in the paper, it is easily seen to be THAT response which is most accurately reflected in the averaged in-room curves, certainly not the power response as measured in AR's reverberant chamber. There is no evidence that a dominant reverberant field exists at higher frequencies in these typical listening spaces.

From this I conclude that the wide dispersion touted for the drivers themselves is all but moot in the systems (due to interference and diffraction effects) and in situ (due to differential absorption of successively higher frequencies), and that its only remaining contribution of significance lies in the generation of ASW spaciousness cues via early reflections.

From this perspective, then, it's clear that wide-dispersion for its own sake is self-defeating, and attempting to put the room in charge of spectral balance above the transition frequency is not only ineffective, but may also generate undesirable artifacts. Geddes goes further and says it absolutely does, and the result is "disaster;" Toole apparently agrees....

"it's clear that wide-dispersion for its own sake is self-defeating, and attempting to put the room in charge of spectral balance above the transition frequency is not only ineffective, but may also generate undesirable artifacts."

And that's just for a speaker. Imagine what it does to the sound of a grand piano.

[speaker dave]

From Zilch: "From this perspective, then, it's clear that wide-dispersion for its own sake is self-defeating, and attempting to put the room in charge of spectral balance above the transition frequency is not only ineffective, but may also generate undesirable artifacts. Geddes goes further and says it absolutely does, and the result is "disaster;" Toole apparently agrees.... "

Come on Zilch, now your just stirring the pot.

Wide dispersion does increase spaciousness and acoustic source width. I'm saying it has minimal impact on the perceived frequency response which is dominated by the direct field. "attempting to put the room in charge of spectral balance" doesn't really pertain. I'm not sure what Geddes has said on the subject but Floyd, in his book, speaks of side wall reflections increasing spaciousness with minimal negative impact on perceived balance.

From Howard: "How on earth can you say that the room curves Allison and Berkovitz ran do not mainly reflect the overwhelming importance of the power response?"

"Overwhelming importance" implies a perceptual impact equal to the measured effect. My point is that measuring an in-room response with an omni microphone is not the same as our perception, with binaural hearing and an ability to discriminate based on arrival time.

If we just looked at high resolution measurements of speakers taken in a live room we would give up. Luckily what we hear isn't nearly as confused as it looks.

David

[Pete B]

I'd say that the "sense of spaciousness" comes from the diffraction notches in the

frequency response providing cues in the sense of HRTFs.

I have a pair of speakers known for pin point imaging (SPICA TC-50s); I gave them a

good listen before taking any measurements and something was seriously wrong. It

sounded like surround sound, from just the two speakers, as if something or the two

channels were out of phase. I checked this of course, several times, several different

ways. When I finally measured them, they had a deep notch due to tweeter problems,

around 20 dB at 3kHz. Fixed the tweeters and then I heard the imaging that these

speakers were known for.

I believe that the diffraction notches cause this to a lesser degree, providing what some

may call spaciousness and find pleasing, in vintage AR speakers. I'd say that it is this

far more than the wide dispersion, or anything else, that is the source of the sense of

spaciousness.

[Zilch]

From Zilch: "From this perspective, then, it's clear that wide-dispersion for its own sake is self-defeating, and attempting to put the room in charge of spectral balance above the transition frequency is not only ineffective, but may also generate undesirable artifacts. Geddes goes further and says it absolutely does, and the result is "disaster;" Toole apparently agrees.... "

Come on Zilch, now your just stirring the pot.

Not at all. As you see, Howard steadfastly maintains the data supports his contention that a reverberant field dominates, and all that implies, as initially suggested by the Beranek analysis presented in the paper, despite the "UT, oh, it doesn't" clearly stated therein.

There's four of us now here agreeing on this point. I cannot expect Howard to also agree, given that he is wed to his Allisons, but clearly, in light of this, he might more productively be looking elsewhere for the actual basis of what he perceives in listening to them.

I believe that the diffraction notches cause this to a lesser degree, providing what some

may call spaciousness and find pleasing, in vintage AR speakers. I'd say that it is this

far more than the wide dispersion, or anything else, that is the source of the sense of

spaciousness.

Geddes and Toole both teach that it's early reflections with low IACC, suggesting that wide dispersion is not necessary to generate enhanced ASW, and in fact, creates an array of problems of its own....

[Shacky]

There's four of us now here agreeing on this point. I cannot expect Howard to also agree, given that he is wed to his Allisons, but clearly, in light of this, he might more productively be looking elsewhere for the actual basis of what he perceives in listening to them.

I'm not sure I see 2 people agreeing to much here. Just to be clear, who are the 3 you count in your court?

[Zilch]

I'm not sure I see 2 people agreeing to much here. Just to be clear, who are the 3 you count in your court?

If there was even an inkling apparent that you had any interest whatsoever in understanding or appreciating the content of this thread, I would reply, Shacky, but given that your purpose is solely to crap upon it, instead, I'm afraid I have no resources sufficiently disposable to squander on formulating a considered response to your inquiry....

[speaker dave]

While the strength of the reverberant field in relation to the direct field would vary, depending upon crossover points and driver diameters (and driver arrangements if multiples were used for the same bandwidth coverage), for the most part the reverberant field would still dominate.

So, it looks to me as if one is going to do direct-field measurements of the upper frequencies you are going to have to do a whole lot of them to get a meaningful result, and driver interference and diffraction effects would be a problem. In that case, doing independent driver curves as Allison mentioned (and as was done both at AR when he was there and with his later Allison Acoustics company) would be a better technique than doing full-system measurements.

Howard Ferstler

My point is that the reverberant may dominate in a simple measurement but it does not dominate our perception. Likewise, the fact that edge reflection or crossover effects vary greatly with angle doesn't mean that they can be discounted, just that their audible effect varies with angle. If a speaker has a different (anechoic) frequency response at every angle then it has a different perceived sound.

Try this, feed pink noise into a speaker and have a helper place a ruler edgewise on the baffle adjacent to the tweeter (create a reflective ledge). As the helper varies the spacing from tweeter to ruler and "wiggles" the ruler (varies the angle) I guarantee you will hear a varying comb filter effect. This will be audible at any reasonable listening distance, even at distances where the room response swamps it and it is hard to measure. This is an angle specific response abberation that has minimal, if any, effect on radiated power response.

It is very audible. Please try it.

Response variation with measurement angle is not some impossible challenge to deal with. Modern loudspeaker designers have learned to design cabinets with rounded corners and improved grille designs to reduce the driver response variations. In-phase or Linkwitz-Riley crossover designs give minimal response variation through crossover transitions. That Vilchur and Allison didn't know of, or use, these techniques 50 years ago by no means diminishes their great contributions to loudspeaker design. Certainly, the later AR designs did take advantage of this evolution of design practice.

"We stand on the shoulders of the giants that came before us."

David

[speaker dave]

Howard,

Now, thats an interesting argument about the relative strength of the direct and reverberent fields. Even if our ear canal is 0.1% of the spherical area the loudspeaker sees, sound in the other 99.9% doesn't rattle around the room for a while and then all end up at the ear drum. That reverberent energy has the same size target to hit (the ear canal) and the vast majority of it will decay away, sadly unheard. (Warming up our curtains and seat cushions, I guess.)

We can calculate the direct sound level and also the reverberent field level. The first is a function of source Q and distance, the later is a function of room RT and volume (or simply total absorbtion in Sabins. Look up "Hopkins Stryker" in an acoustics book for the equation) In most rooms, at most typical listening distances, the reverberent field is a little higher than the direct field, at least for low frequencies. For higher frequencies, just as likely, the direct field is stronger. The distance where the two fields are just equal is called the critical distance.

Regarding pink noise, I suggested using that because the effect is so obvious. In my experience though, any abberation that you can hear with pink noise you can also hear with music with a little careful listening. I know from experience that you can certainly hear the ruler test with music.

You are going to try it , aren't you?

David

[Shacky]

I'm afraid I have no resources sufficiently disposable to squander on formulating a considered response to your inquiry....

You could save us all some time and just say no

I'm beginning to believe your main objective is to create more and more opportunity for you to post on the internet. You obfuscate everything till the cows come home. Is posting the only thing you do all day!

[Zilch]

Is posting the only thing you do all day!

You KNOW what I'm doing online; I've shipped 20 pairs of Karma Indignas pro bono in the last week.

Bite me, Shacky....