What is the ideal directivity pattern for stereo speakers?

[Pano]

If >10 mS can we say >11-12 feet of extra path distance as an approximation?
(3.5M)

 

[audiokinesis]

If >10 mS can we say >11-12 feet of extra path distance as an approximation?
(3.5M)

Yes, for example, six feet out in front of the wall = about 10 ms extra path distance for the reflection.

As we decrease the bipole reflection path distance, the detrimental < 10 ms reflections increase accordingly, and the beneficial > 10 ms reflections decrease slightly. At some point the detriment offsets the benefit, and we are too close to the walls for a bipolar to make sense. Imo, ime, ymmv, etc.

 

[gedlee]

You may or may not be right and maybe the future of loudspeaker design will prove the rest of us wrong.

Yes, well let's wait and see. I have no problem with that. I've been doing this for 45 years, a few more years won't matter. So far, I am very content with the results. Two people differ, the rest agree. I don't have have a problem with that - you might if you are one of the two, I suppose.

 

[Wayne Parham]

Wayne, your corner horn approach does work and my impression was also that it was more like outdoors (few close reflections)

Thanks for the nod, Tom. Means a lot coming from you.

About 10 years ago, several DIYers on my forum were running Pi cornerhorns sans high-horns, putting early Unities on top instead. It was a natural combination.

Duke, I hope you don't mind my referencing your comments about the Pi constant directivity cornerhorn configuration from a few years back:

• Wayne's presentation and speakers @ Midwest Audiofest (2003)

 

[Bob Richards]

I think that a toed-in controlled-pattern dipolar or bipolar speaker can, given sufficient distance to the nearby walls, achieve a < 10 ms reflection density very similar to that of a corresponding monopole, but with nearly double the > 10 ms reflection density once that rear-radiated energy begins to arrive. Not all of it will show up as lateral reflections, but much of it will. (Of course this sort of geometry takes up a lot of real estate in the listening room so it is not always practical, but in some rooms it is).

Does this sound plausible to you?

Linkwitz says he draws the line at 6mS, meaning the dipoles are at least 3 ft. out from any walls. Although more is definitely better, based on my own experience with my open baffle dipoles, I'd agree with Linkwitz's number as a minimum.

 

[graaf]

Sorry to be repetitive, but I think sometimes a little repetition can be helpful.

The constant directivity cornerhorn
...
When in a small room - smaller than a ballroom - you really can't get the speakers far enough away from the walls to be optimal. So in rooms like that, I like the constant directivity cornerhorn approach

I can see clearly the reasons for using cornerhorn - after all it is a kind of ultimate flush mounting plus corner loading, very good reasons

that's why a cornerhorn is a classic design, and I do not mean Klipsch but English pioneers of Hi-Fi like Paul Voigt or Stuart Hegeman

OTOH please note that Hegeman started with cornerhorns like Lowther-Hegeman Reproducer and ended with omni designs for Harman, Eico and finally His own brand

An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.

here is a 1961 interview with Hegeman:
retro vintage modern hi-fi: Stewart Hegeman interviewed by Norman Eisenburg 1961

 

[erjee]

yes, I can estimate it perfectly
BUT can You do the math?

Sure, no problem.

If we have an equilateral listening triangle with sides of 3 meter and a room having a width of 7 meter we can calculate the path length difference as: sqrt(2.6^2+5.5^2) - 3 = 3.1 m. For a sound speed of 340 m/s this would result in a time of 9.1 ms which is 10 ms

The length of the room needs to be at least 6 m, placing the listener 1.7 m from the backwall. So the omni's in this medium sized room should be able to create a good image.

With respect to directivity pattern, it seems to be whatever suits you, as long as it is somewhat smooth.

Well, I think we have to be more specific.

If we assume
1. Early reflections > 10 ms
2. Direction of the reflections is unimportant
3. Equilateral listening triangle of 3 m

This will dictate the minimum room size. For an omni this woulde require a width of 7.3 m and a length of 6 m. Surprising: one needs a medium sized room and one should be able to place the speakers along the long wall

 

[keyser]

Well, I think we have to be more specific.

If we assume
1. Early reflections > 10 ms
2. Direction of the reflections is unimportant
3. Equilateral listening triangle of 3 m

This will dictate the minimum room size. For an omni this woulde require a width of 7.3 m and a length of 6 m. Surprising: one needs a medium sized room and one should be able to place the speakers along the long wall

To most standards that's a pretty large room. It has an unusual ratio to boot!

 

[graaf]

Sure, no problem.

If we have an equilateral listening triangle with sides of 3 meter and a room having a width of 7 meter we can calculate the path length difference as: sqrt(2.6^2+5.5^2) - 3 = 3.1 m. For a sound speed of 340 m/s this would result in a time of 9.1 ms which is 10 ms
...
So the omni's in this medium sized room should be able to create a good image.

yes, about 9 ms but it's not the point, the point is - was this medium sized room - quite a ballroom - 7 m wide actually, perhaps it was closer to 6? was it 3 m equilateral listening triangle actually? perhaps it was more like isosceles for most of the audience, perhaps Pano included?

2. Direction of the reflections is unimportant

but it is important according to most studies quoted in this thread

direction and frequency content

 

[graaf]

at high frequencies our head dramatically reduces channel crosstalk above ~1-2Khz for speaker separations of ~60 degrees and greater.

Are You sure? I would say that a 7 dB for 60 degrees (+/- 30 degrees) is not dramatic enough

 

[Filozof]

I've of course noticed the effect that people's voices sound brighter when you're in the open and at some distance from each other, but I don't understand the above explanation. Once you're in the far-field, as far as I know the ratio of low to high frequencies shouldn't change. What am I missing here?
- Elias

Look at the directivity pattern of human voice, that should explain. For instruments, a good reference: Jurgen (with Umlaut) Meyer-Acoustics and the Performance of Music. Our speech organ is a horn crossed with 4th order BP.

I'm entirely with you on that one. The quote is completely wrong. The reasoin that our hearing has not developed very well above say 8 kHz is because of the extreme amount of HF absortion in the air at these frequencies. Hence, with distance, the HF response tends to fall, not increase. In a large auditorium, air absorption at 10 KHz is greater than the absorption at the walls and almost nothing above 8-10 KHz exists in the reverb field. At some distances even the direct field is strongly affected.

Frequencies above 10kHz are not represented enough in nature for our hearing to support it. I think that the only reson we developed hearing above 10kHz is to have a precise hearing up to 10kHz. Our ability to recognize the pitch difference is almost the same from low freq to very high, around 2Hz.

Range from 4kHz to 8kHz is important because it tells our head how much is tilted. There is a notch on the outer ear that serves just for this purpose, and one of the main reasons why we cannot obtain height information from stereo.

Ideal directivity pattern for stereo speaker is: high directivity at bass and midrange, and low directivity at the treble.

Hm, that should be interesting. Try integrating such loudspeaker in a enviroment and you come up with the same problems as with any other speaker.

---

My thoughts of loudspeakers are that we shouldn't look only at the loudspeaker, but on loudspeaker-room integration. The loudspeaker should be linear in response (because of the music material) and dispersion (because of the reflections), and the room should have a linear reverberation (with tolerancing for lows(+=) and highs (=-)). The amount of reverberation should accompany the type of reproduced music. Any standing waves should be removed for the sake of time introduced correctness. The geometry of a room should redirect reflections fewHz>10ms, and everything after 50ms should not exist in the direct field. In this way we obtain a room that is very dispersive and controled dissipative in the energy domain.

 

[a_tewinkel]

I think that a toed-in controlled-pattern dipolar or bipolar speaker can, given sufficient distance to the nearby walls, achieve a < 10 ms reflection density very similar to that of a corresponding monopole, but with nearly double the > 10 ms reflection density once that rear-radiated energy begins to arrive. Not all of it will show up as lateral reflections, but much of it will. (Of course this sort of geometry takes up a lot of real estate in the listening room so it is not always practical, but in some rooms it is).

Does this sound plausible to you?

This is just about what I've done. I've got a very small room of 5.7 x 2.5 m (I'm a student, so no ballroom here ), with my dipoles placed along the side walls and about 2 m from the back wall (speakers toed in of course). The side wall reflection is pretty weak and I get about 14 ms delay for the back wave. The room is highly reflective and the set-up generates a decent feeling of spaciousness with good recordings, while imaging is reasonably sharp.

 

[DBMandrake]

Now if we are talking about perceived frequency responses of loudspeakers, Salmi, Kates, Queen, Bech and others show that the ear effectively uses a variable time window, short at high frequencies and long at low frequencies. The time window is so short at high frequencies to only let in the direct sound and some nearby cabinet reflections. By the midrange the nearest boundary bounces, such as the floor bounce, will be perceived wile later bounces are excluded. At low frequencies the window is fairly long and most of the room response is perceived.

At least for upper frequencies, the direct sound is very much a reality.

Do we know if any of the commonly used measurement suites has built in a sliding window model which follows their findings yet ? I would love to see a "perceptual" measurement mode in packages such as ARTA, that use a true sliding window time that matches the parameters they have found.

ARTA has a "dual gate" mode which allows you to window the reflection free period, and then smoothly merge that into a "long" window measurement at low frequencies, but it is only two window periods, and it's not explained in the documentation quite how the merging is implemented - my experience with it suggests that it extends the results from the long window measurement too high in frequency relative to the window length of the short window, and it's not a good perceptual match.

Holm Impulse has a more progressive windowing system with more than two window times, but even then I don't think it follows the above research, and is just something that the author came up with empirically that seemed a good idea at the time. (It still doesn't match the perceptual response in a room, for me at least)

 

[Filozof]

I have a question that does not concern loudspeakers in general, but it has something to do with them. Have you ever tried comparing classical music that is nearfield recorded on two types of loudspeakers, ones with constant directivity and ones with narrowing directivity? Does our effort even matter (in producing the perfect loudspeaker when most of the recorded classical material does contain too much HF information (or a screwed up balance)? Sometimes it's corrected by the engeneer, but then bye-bye HiFi. I think there's more than one link in the chain broken. The standards for cinema are perfectly reproducable. On the other hand, such standard in two channel audio doesn't exist in practice.

 

[a_tewinkel]

Now if we are talking about perceived frequency responses of loudspeakers, Salmi, Kates, Queen, Bech and others show that the ear effectively uses a variable time window, short at high frequencies and long at low frequencies. The time window is so short at high frequencies to only let in the direct sound and some nearby cabinet reflections. By the midrange the nearest boundary bounces, such as the floor bounce, will be perceived wile later bounces are excluded. At low frequencies the window is fairly long and most of the room response is perceived.

At least for upper frequencies, the direct sound is very much a reality.

David S.

Interesting stuff. So does this imply that directivity at low/mid frequencies is beneficial, unless you find other ways to eliminate the floor bounce for example?

 

[Elias]

Elias,
visual simulations are great for predicting interactions between sound waves and the environment but should not be used to illustrate how we hear. Indeed if continuous, even sinus waves are played in stereo the listener will have no idea what direction they are coming from. Our brains seem to react best to transients when processing sound information and music happens to be rich of those.

With the same logic we must say frequency response should not be used because it does not illustrate how we hear ! FFT freq response is also based on sinus waves. Bad.

Wavelets are better because they are based on 'transients'.

Cute though your picture may be, it doesn't tell us anything about how sound is perceived by two ears.
...
Be careful about drawing conclusions on how something will sound based on a visual representation of data which does not take into account important aspects of our hearing mechanism.

You too, DBMandrake, you must avoid using frequency response plots from now on because it does not take into account important aspects of our hearing mechanism

And the directivity plot, My goodness, how did it ever include perceptional aspects ?? To be avoided !


- Elias

 

[graaf]

In a large auditorium, air absorption at 10 KHz is greater than the absorption at the walls and almost nothing above 8-10 KHz exists in the reverb field.

yes, and so much for the 20-20 high fidelity requirement

this observation by Dr Geddes harmonizes well with the original high fidelity requirements as defined by H.A. Hartley, first theorist of high fidelity sound reproduction:

His original bandwidth desideratum for music lovers, laid down in the 30s, was perfect reproduction between the limits of 32 and 9,000 Hz. With 30 years worth of hindsight, he expanded the limits from 20 to 12,000 Hz. Only an audiophile, he argues seems to want to reproduce the sound of a triangle more triangular than the real thing.

 

[Elias]

"Only an audiophile, he argues seems to want to reproduce the sound of a triangle more triangular than the real thing."

This was a good one

 

[Filozof]

yes, and so much for the 20-20 high fidelity requirement

this observation by Dr Geddes harmonizes well with the original high fidelity requirements as defined by H.A. Hartley, first theorist of high fidelity sound reproduction:

Actually, I would like to if you can elaborate on real reasons why reproduction range should exceed 12kHz. Not a trick question, just curious.

 

[graaf]

Actually, I would like to if you can elaborate on real reasons why reproduction range should exceed 12kHz. Not a trick question, just curious.

I wasn't ironic this time

I don't think there is any such reason, and I am very glad that Dr Geddes supports with His expert knowledge the opinion that I share

Hartley is my hero, I am in this for Realistic High Fidelity

so You have to ask someone else to elaborate on those reasons why reproduction range should exceed 12kHz