Beyond the Ariel

IIRC in line sources there is also a falloff of SPL with frequency at -3dB/decade, which would overlay with the proposed changed room reflection pattern, and of course the line source character depends on the length of the line vs. frequency, so I am not sure if this effect would be so clearcut. In any case line sources seem to be one of those terrible to measure but pleasant to listen to things...

One possibility that I see in this design is that the source size would vary more strongly with frequency than in the usual speaker. If a ribbon is not used, and the transitions between the drivers are kept at somehow conventional frequencies, quite possibly it would not act as a line source in any frequency band (line too short for line source character at any frequency, by eyeball). It would, however, gradually increase in source size as frequency goes down.

Now say from 100 to 10000 Hz the wavelength varies 100-fold. Assuming a 1" CD as the smallest element and all 12" acting in unison at the largest, we'd have a 30 to 40-fold variation in source size, quite a bit better then the usual 1" to 8" or 10" change and close enough to 100x. This might end up sounding more natural, since low frequency sources in nature are usually physically large. Incidentally humans judge apparent source width through the envelope of the LF part of the sound, that's how we can localize LF sources and ascribe a size to them.

Michael btw that graph of attenuation with humidity was illuminating - I thought I was hallucinating but I feel a strong change in sound character depending on the weather. Some days things just sound noticeably duller than others.

Finally, a suggestion re: cosmetics: Another of my thought experiment speakers :)rolleyes:), never built, was an OB with a trapezoid baffle shape, but folded back around a square cross-section "wireframe" column (using wood of course). The areas of the column that should be left open would be covered with black speakercloth. That would give the appearance of column speakers while preserving the OB character, and with a trapezoid foldback you'd hardly have much cavities in the back to watch out for.

Ugly sketch attached.
 

Attachments

  • picture1.jpg
    picture1.jpg
    8.9 KB · Views: 1,118
re: aesthetics. I was thinking of making a more visually appealing shape in wood or colored baffle and using clear lexan for the rest of the baffle, or in front of the solid baffle. This way you have acoustically the better shape while visually a more pleasing shape.

Its not my idea really as I have seen this done by someone on another forum and the results were quite nice.
 
Geometry of Sound Source

This section of the Lenard Audio page discusses the subjective impression of large sound sources - and matching emission areas to the wavelengths involved. The author specifically mentions the perception of a flat wavefront midway in the theatre that matches the image size of the screen.

With an interesting and personal view of the pro audio side of things, there's a lot of information here. There's a good section on theater horns and the history of their development in the cinema.

The decade, or three-octave, rule for horns makes sense. If the crossover from the 12" driver is at 1.2 kHz (woofer cone is one wavelength across), the horn should ideally cover a range from 1.2 ~ 12 kHz (1.2" wavelength), with a small supertweeter helping out above that.

Below 240~200 Hz (five to six foot wavelength), the bass array grows 2~4X larger, and is close-coupled to the floor reflection. Below 80 Hz, stereo servo subwoofers are located at the floor/sidewall boundary.

Contrast that to the typical audiophile speaker with its single 7" woofer and 1" tweeter, and the very large range of wavelengths they emit - no wonder the size and scale of instruments doesn't come across as realistic.
 
Administrator
Joined 2004
Paid Member
Re: Geometry of Sound Source

Lynn Olson said:
Contrast that to the typical audiophile speaker with its single 7" woofer and 1" tweeter, and the very large range of wavelengths they emit - no wonder the size and scale of instruments doesn't come across as realistic.


You ain't kiddin'! :D


But it does start to get hard to squeeeeeze a symphony orchestra into the living room. Large shoehorn needed.
 
About the serial R for attenuation, I found the same dull effect on various speakers. So I avoid them at all costs.

One of my friend claimed that he got excellent results from autotransformer on his Klipschorn. For the 3way passive xover in such system, there's almost invetible serial R for attenuation. So does the loss in liveness.

He found that in the very old version of Klipschorn, there're the autotransformers for mid-high attenuations, and the vividness in sound is much better than the newer one with normal resistors.

I myself have no experiences on this. Anyone?
 
The Klipsch autoformers are easily available as replacement parts and only cost about $20 or so. Remarkably, PWK started using them back in 1959, partly as a way to use much smaller values of caps (somewhat different crossover than I sketched out).

Altec and JBL, of course, were too proud to ever use autoformers, and stuck with textbook 12 dB/octave crossovers made from mylar caps and Radio Shack-grade L-Pads. I've looked inside the crossover boxes for Altecs and was dismayed they'd spent more on the beautiful precision-cast aluminum box than the parts inside - and this was in late-Sixies production for a professional-model A5 theater speaker.

I know this is going to enrage the Church of Altec devotees, but that was the difference between Altec+JBL vs Klipsch. PWK would use junky-looking PA drivers with phenolic diaphragms and surprisingly refined crossovers - and this goes back to the late Fifties. Back then, Altec+JBL made drivers that were Museum of Modern Art-quality works of art, but with generic crossovers with no driver compensation or correction at all. Out of sight, out of mind.

When a company doesn't bother to optimize the crossover to the drivers, builds it from the cheapest parts they can find, and spends a fortune on elegant machining of the drivers, cabinetry, crossover boxes, and advertising, that tells you a lot about priorities.

Altec+JBL continue to make good, even superb, drivers, with some of them true classics by any standard. But if PWK - and the BBC in England - were doing crossover/driver optimization in 1959, and Altec+JBL weren't doing it 10~15 years later, something ain't right. Maybe the engineers at Altec+JBL felt that system optimization wasn't worth the trouble, and all the glamour and high-tech went into the drivers. The difference could be summed up in two speakers: the JBL Ranger-Paragon vs any BBC monitor. Completely different set of esthetics, both visually and sonically.

I am sure the Altec+JBL bias against crossover optimization is why the US industry was ten years late to the table, when the Brits had settled on BBC/KEF optimization throughout the Seventies. JBL finally started making mirror-imaged speakers by the early Eighties, ten years after the Brits did.

Maybe part of the reason is that Los Angeles is a company town, with a strong Not Invented Here attitude towards non-Los Angeles technology. I went to college in LA during the late Sixties, and Altec+JBL pretty much owned the place, with a firmly established duopoly in the broadcast, recording, and movie business. Back then, Brit or Euro-design speakers of any type were very hard to find in Los Angeles, and the Altec 604 Duplex and the studio version of the JBL Century 100 ruled the roost. I was very glad to see competition start to appear in the Seventies - it was long overdue.
 
Administrator
Joined 2004
Paid Member
Re: Geometry of Sound Source

Lynn Olson said:
This section of the Lenard Audio page discusses the subjective impression of large sound sources

Thanks for the link, Lynn. A very interesting site.
Though I understand what he is trying to convey about sound, and agree with him, I have to take a technical quibble with his inverse square law.

a17_inv-sq-law-screen-spk.gif


Once passed thru a lens, light no longer falls off in the inverse square. Not at all. If it did, you'd lose the major benefit of using a lens! Surprisingly, most technicians don't understand this.

Does the same apply to sound passed thru a "lens" such as a horn? Maybe. From what I've read and the diagrams I've seen of wavefronts in horns - it seems to have some bearing.

Light coming from a lens does not radiate the same way as light coming from a point source, nor does sound coming from a horn radiate in the same way as sound coming from a point source.
 
Does the same apply to sound passed thru a "lens" such as a horn? Maybe. From what I've read and the diagrams I've seen of wavefronts in horns - it seems to have some bearing.

I don't think horns act as lenses do. A horn acts as an impedance transformer (increasing efficiency, not relevant in this context) and a waveguide (restricting radiation into a fraction of full 4 PI space, possibly relevant).

The inverse square law comes from the idea that the sound intensity will spread out spherically and pass through ever increasing surface areas:

- At a distance r from the source, the sphere area the sound will pass through will be 4 PI r^2.
- At a distance 2r from the source, the sphere area will be 4 PI (2r)^2 = 4 PI 4r^2, or 4 times larger, but receiving the same initial power.
- Hence, 1/4 the power (-6 dB) for twice the distance.

If the point source is not radiating spherically, but into an axially symmetrical waveguide AKA horn, the area receiving the initial sound power will be smaller by a fraction f. So the source will radiate through a surface of area (4 PI / f) r^2. Since the r^2 term has not changed, the inverse square law should also not change, whether in a horn or a waveguide.

Special case line sources, here one assumes two parallel walls (floor and ceiling), so the surface of the wavefront is no longer proportional to r^2 but to r * h (h: room height). While r varies with distance from the line source, h does not, so the surface the sound radiates into is proportional to r, and a doubling of r gives you a halving of power per surface, in other words, -3 dB per distance doubling.
 
Actually now that I just wrote this - the interesting point comes in from the fact that we don't have point sources in real life. Therefore to some extent none of these formulas really apply, and the benefits of large speakers precisely come from the fact that they are less of a point source, in general, even ignoring line sources.

In addition to that in a room reflection and reverberation prevent some of the power losses (through restricted radiation space) and that again makes the SPL falloff smaller than predicted by point source theory.
 
cinema "sound" reality check

Hi


I haven't read all I must admit, but for sure a really nice overview from Lenard Audio about cinema though with some flaws as panomaniac already has pointed to.

To give an even more correct explanation – the inverse square law IS valid also with lenses in use. At double the distance you always get ¼ the illumination given the SAME lens applied. The size of the picture gets 4 times bigger then of course. (no cylinder wave front with light !)

Where Lenard is definitely wrong is that there is no such virtual origin of light ( the projector in his explanation) that corresponds to the perception of a BIG picture. With very few exceptions screen materials have pretty uniform reflection directivity. Lenard's explanation would imply a mirror like screen ( with respect to reflection directivity ) which is not really worth your ticket if you didn't catch the only sweet spot in your cinema.

The subjective perception of a BIG picture is based on two factors only :
1) the size of the screen with respect to the viewing distance ( the field of view covered by the screen, in other words )
2) the absence of other light sources than the picture on the screen itself

One thing to keep in mind whenever drawing parallels to cinema "sound" systems is that this is just the ART of NOISE and by no means the ART of SOUND that Dolby, THX and Co are providing.

Many, many cinemas are equipped with 2 way passive main speakers - a double 15" in an undampened box and a 2" CD on a really BIG horn.
Not kidding this is 90% of cinema reality and even at Dolby Studio I was surprised to find this horrible speakers behind the screen.
Beside that the predominant majority of movies in Cinema or at Home Cinema

playing Dolby Digital ( AC3 ) sound format which is nothing better than iPod or MP3.


#########

IIRC in line sources there is also a falloff of SPL with frequency at -3dB/decade
What is the explanation behind that?


Greetings
Michael
 
Administrator
Joined 2004
Paid Member
MBK said:
I don't think horns act as lenses do. A horn acts as an impedance transformer (increasing efficiency, not relevant in this context) and a waveguide (restricting radiation into a fraction of full 4 PI space, possibly relevant).


I believe that's true. A horn and an optical lens are not the same, and there are many types of lenses. But it does seem that there is more going on with a wavefront in a horn than simple wave guidance.

The question would be, is the shape of wave front coming out of a horn significantly different from the shape of a wavefront coming from a point source at the distance of the compression driver?

And if so, is there an audible difference? Is the shape of the wave from a given diameter horn different from that of wave from a cone speaker of the same size? Can the difference be heard?

Reflectors are a different matter, and work much the same in light as sound. Hemispherical and parabolic reflectors can make significant changes in radiated sound. But that's another subject.
 
MBK said:
IIRC in line sources there is also a falloff of SPL with frequency at -3dB/decade, which would overlay with the proposed changed room reflection pattern, and of course the line source character depends on the length of the line vs. frequency, so I am not sure if this effect would be so clearcut. In any case line sources seem to be one of those terrible to measure but pleasant to listen to things...


The discussion on "natural sound" from large format speakers seems to apply to long linesources as another solution to getting realistic radiating area by gowing vertical instead of horizontal.

Tweeter ribbon = 90" long and 0.5" wide = 45 in^2
covers 3K - 30K Hz ~8" speaker area

Midrange ribbon = 90" long and 3" wide = 270in^2
covers 80 - 3K Hz ~18" speaker area

Woofer line array = four 15" woofers
covers 20 - 80 Hz


I believe it is critical to get the entire human voice range on one speaker and selected a narrow source to control beaming plus a large radiating area to keep Xmax small.

I live with the phase issues of steep analog LR8 crossover slopes while waiting for full digital crossovers with room equalization. Room equalization for the bass seems more valuable than a flat frequency servo woofer.

The -3dB/decade falloff of SPL with frequency in the nearfield mentioned by MBK makes linesources seem more dynamic then their 96db/m SPL spec.
 
Wide Angle View

The 100th page is a good place to step back and look at what we're trying to do. To start, let's compare a monopole, say, a 12" driver with a HF horn, with a similar dipole. For now, let's set aside Baffle Step Compensation, and just look at the difference between the monopole and the dipole.

Relative to the monopole, there's a peak around 400~500 Hz, and response starts dropping off at a 6 dB/octave rate below 220 Hz. By playing around with the baffle shape, edge treatment, and size, we can smooth out the peak to some degree, but the onset of the 1/f rolloff region, at some frequency, is inevitable.

Rather than let the baffle become really large, we can transition to a monopole subwoofer at an arbitrary frequency. Since much of the benefit of a dipole is lack of box standing-wave modes, that says we should avoid W and H-baffles, and the subwoofer box should be small in comparison to the wavelengths at the crossover frequency. Picking a number out of thin air, the 12" sealed Rythmik subwoofer in a 1.5~2 cu. foot box crossed over at 80 Hz looks like a good compromise.

So we need to cover the range between 80 and 220 Hz. That's 1.5 octaves. 110 to 220 Hz is 1 octave. With the 1/f rolloff rate of 6 dB/octave, the dipole needs to make up between 6 and 9 dB, relative to the monopole, which is flat through the same frequency range.

6 to 9 dB doesn't sound like much, until you consider the impact on amplifier power, and most important, IM distortion from the driver. In the monopole, driver excursion is increasing at a rate of 12 dB/octave. In the dipole below the 1/f frequency, excursion is increasing at a rate of 18 dB/octave. In a horn, diaphragm excursion increases at a rate of 6 dB/octave. (Note that a full-sized, non-truncated straight bass horn in the 80 to 220 Hz region is gigantic, since the diameter of the mouth, and the length of the horn, need to equal an 80 Hz wavelength - about 14 feet! Folding a horn reduces its size but also drastically reduces its bandwidth.)

This comparison of monopole, dipole, and horns shows us that a conventional equalized dipole is taking us in the wrong direction, towards more distortion, not less. If we care about distortion, the area of the driver must increase when the driver goes below the 1/f frequency.

How much? Well, the true (power) efficiency of a driver doubles when the number of drivers doubles. Compared to one driver, two are 3 dB more efficient, four are 6 dB more efficient, eight are 9 dB more efficient, and so on. Not surprisingly, when the areas of the drivers approach horn dimensions, efficiencies become comparable as well.

The prospect of eight drivers per channel isn't very appealing. Fortunately, if we mount them next to a boundary, an image will appear, doubling the number of drivers. This gets us down to four drivers - large, but within reason. Raising the crossover frequency of the subwoofer to 110 Hz, the requirement drops to two woofers - but they both need to be at the floor boundary, otherwise you'll need three drivers, one at listening height, two at the floor boundary.

The Visaton NoBox BB is about the simplest possible way to get from here to there, with the least number of drivers. The wideband driver is 8", the bass driver is 15", and mounted close to the floor. The presence of the big 15" driver is not quite what it looks like - remember, this dipole system has the about the same excursion performance as a single 8" driver in a sealed box. In terms of excursion performance, the NoBox represents the lower bound of what we'd like to do.

So if we want "conventional" dimensions, we must raise the frequency of the subwoofer, equalize and boost the power going into the 1/f region, and accept (much) more distortion than the equivalent monopole system. Sorry, that's how it is. The harmless-looking requirement for a 6~9 dB of equalization has a very substantial tradeoff of increased driver area or increased IM distortion.

Bastani, in his Apollo and Prometheus systems, chose a very high subwoofer crossover of 220 Hz to meet the 12" widerange driver. The Visaton NoBox chooses a different tradeoff, with excursion and distortion comparable to the 8" widerange driver. Linkwitz chooses moderately small driver areas, audiophile drivers selected for low distortion (with excursion limits) and fairly heavy equalization.

Horn systems are different. The distortion is (much) lower, thanks to the 6 dB/octave increase in excursion vs frequency, compared to the monopole 12 dB/octave and the dipole 18 dB/octave. But most all-horn systems cheat - once the frequency gets below 200~300 Hz, wavelengths get really awkward and big, so we see folded and truncated bass horns, with passbands of little more than an octave.

Many so-called "horn" systems run the bass horns right through and below horn cutoff, the Altec A7 being the most famous example, although the Klipschorn does as well. When you run a horn below its cutoff, the excursion performance is no better than the sealed-box direct-radiator equivalent - but no worse either, which is still pretty respectable for a prosound 15" driver.

We obviously can't expect a dipole of any reasonable size, short of an entire wall, to equal what horns do. But we can surely ask that dipoles equal the distortion performance of an equivalent monopole, instead of settling for less. That's why I've been nudging the system in the direction of large bass arrays, with a variable-geometry overlapping crossover to maintain uniform directivity and constant power efficiency. There's been a temptation to use 18" drivers, but nearly all of them are intended for professional subwoofer use, with very heavy diaphragms and outright bad performance above 1 kHz.

By comparison, there seem to lots of 12 and 15-inch professional midbass drivers, with light cones, low distortion, and quite respectable midrange performance. Arrays of 12 and 15-inch drivers have a long and distinguished track record in stacks of bass guitar amps on stage and full-sized movie theater speakers.

Not a bad starting point for the bass array of a dipole speaker, which needs all the help it can get from low-distortion, linear-excursion, and well-behaved midbass drivers. The less crossover trickery and fancy equalization, the better it's going to sound. The unique dipole virtue of "hearing the speaker as it really is" - without box or horn coloration - applies to the bass array as well as the rest of the range.
 
Progressive-Loss Mesh ... Not A New Idea

Right here in the original Rice and Kellog patent of 1925, a cabinet with lossy mesh on the sides and rear.

Excellent article on the history of loudspeakers - highly recommended. This underlines the point of interesting ideas that are lost over time - the original patents and papers make for interesting reading, with all sorts of subtle design details.
 
Lynn,

I think you are on the right track. However, some of your baffle calculations I believe are in error.

Relative to the monopole, there's a peak around 400~500 Hz, and response starts dropping off at a 6 dB/octave rate below 220 Hz. By playing around with the baffle shape, edge treatment, and size, we can smooth out the peak to some degree, but the onset of the 1/f roll-off region, at some frequency, is inevitable.
My prototype OB has a 48" by 18" flat baffle. I calculate Feq to be 140 Hz, based on a baffle area of 864 Sq inches (.56 sq Meters). Without doing the math, the baffles in the thread are considerably larger than that. Feq is the Frequency where a monopole and dipole have equivalent output. This is about an octave from your planned crossover point of 80 Hz. With the floor bounce, the output of a Qts ~ .7 driver is relatively flat to about 70 Hz, rolling off after that. By using a moderate Q notch filter on the DCX2496 to remove the 6 db "hump" above Feq, I found that I needed only about 3 db of equalization at 40 Hz to be subjectively flat. This lowers the required driver displacement from insane to merely large. :)
My sense it that 2 * 15 will be in proportion to the rest of the design.

Regards,

Doug
 
Re: Re: Geometry of Sound Source

panomaniac said:


Thanks for the link, Lynn. A very interesting site.
Though I understand what he is trying to convey about sound, and agree with him, I have to take a technical quibble with his inverse square law.

a17_inv-sq-law-screen-spk.gif


Once passed thru a lens, light no longer falls off in the inverse square. Not at all. If it did, you'd lose the major benefit of using a lens! Surprisingly, most technicians don't understand this.

Does the same apply to sound passed thru a "lens" such as a horn? Maybe. From what I've read and the diagrams I've seen of wavefronts in horns - it seems to have some bearing.

Light coming from a lens does not radiate the same way as light coming from a point source, nor does sound coming from a horn radiate in the same way as sound coming from a point source.
I sort of wonder about the sound image = screen image thing. One thing I am sure of is that most of the time the video changes views, and recordings normally don;t change accordingly.
 
DougL said:
My prototype OB has a 48" by 18" flat baffle. I calculate Feq to be 140 Hz, based on a baffle area of 864 Sq inches (.56 sq Meters). Without doing the math, the baffles in the thread are considerably larger than that. Feq is the Frequency where a monopole and dipole have equivalent output. This is about an octave from your planned crossover point of 80 Hz.

With the floor bounce, the output of a Qts ~ .7 driver is relatively flat to about 70 Hz, rolling off after that. By using a moderate Q notch filter on the DCX2496 to remove the 6 db "hump" above Feq, I found that I needed only about 3 db of equalization at 40 Hz to be subjectively flat. This lowers the required driver displacement from insane to merely large. :)

My sense it that 2 * 15 will be in proportion to the rest of the design.

Regards,

Doug

I'm largely in agreement with your post. With the 1/f set to 220 Hz, I was being extremely conservative about the 1/f power requirements - as you mention, with larger baffles, the transition frequency goes lower, decreasing the area requirements - although the area requirement always remains larger than the equivalent monopole.

What happens if the bass radiating area is "excessive"? You get a little closer to the horn condition, with its far lower distortion, but without the frightening size requirements, or the single-octave limitations of folded horns.

Most of the dipoles we've seen so far are underpowered compared to their direct-radiator monopole equivalents, much less horns. I'm advocating going just a little way in the opposite direction - a system that has a bit more capability than the equivalent monopole, not less. If we have too much bass efficiency we can always series-connect the LF drivers or use EQ to decrease output, for a change. It is quite a luxury to be able to crank back on the bass power, reducing excursion at the same time.

There is a broad spectrum of bass arrays we can build - heavy EQ with drivers that are too small, a lightly equalized array with performance broadly equivalent to a monopole, or a lightly equalized array with performance a bit better than a monopole. No matter what we do, dipoles still have to contend with that 18 dB/octave excursion vs frequency rolloff, while monopoles are at 12 dB/octave, and horns are at 6 dB/octave (above cutoff).

The tradeoff of size+cost vs distortion performance and bass headroom is ours.