Spherical Speaker Arrays: Anyone have any thoughts or knowledge about them?

[JonW]

I was just thinking about some non-traditional speaker designs. In particular I’m wondering about the idea of spherical speaker arrays. Basically, they are speakers with arrays of drivers arranged in a sphere (or something close). I can find a fair number of examples on the web but not much regarding any theory or pros/cons. Or how they might sound. Anyone have any ideas about them?

Possible pros:
-Maybe they’d provide something analogous to an omnidirectional sound. Could be different and interesting. Or not.

Possible cons:
-Sound from the multiple drivers could cancel each other out, phase would be tough to align, etc.
-Room interaction issues (this problem also exists with open baffles, etc.)

Some variables:
-I wonder how you’d take measurements and develop the crossover (with a 2-way or 3-way speaker).
-I don’t know how would you wire up the drivers to keep impedance manageable.

Most designs use only full range drivers. Maybe if you made a 2-way (or 3-way) version it could yield something interesting. I have no idea, really. I suspect that this style of speaker will be shot down with one, simple post by someone more knowledgeable than I. Anyways, just some curiosity rattling around in my skull at the moment.



Images not available

[Face]

Most designs use only full range drivers.

Not all. :E

http://www.edesignaudio.com/product_info.php?t=1&products_id=607



FYI: it's doubled in price since it was initially released.

[Hank]

Wasn't there an Italian company that had a globe enclosure with one driver and it was hung from the ceiling on a wire? I think it got quite a bit of press.

[dlneubec]

One of the most successful I have heard about is the Raal Requisite "Eternity", which some of the members probably heard at RMAF2009. It has an array of subs, array of mids and a tweeter array as well, I believe.

http://www.raal-requisite.com/pages/eternity.html

They show a lot of the construction details in an article at 6moons:

http://6moons.com/industryfeatures/roadtourserbia/raal.html

[eb15]

They come and they go, along with the other omni-directional speaker designs.

If you use multiple drivers in the same range there are potential comb filtering effects to be considered, along with usual room reflections and their effect on the sound.

If you choose very high quality drivers, the costs goes up dramatically as you multiply the numbers needed for the design. You also end up having to connect various things in series and in parallel combination groupings to keep impedance within reason.

Since low frequencies become omni directional as they increase pass the baffle diffraction size, and most people don't believe that trying to reflect treble off the floor beneath the speakers is of any benefit, most designers concentrate on just the mid-bass on up for their "omni" speakers. I think Daniel Queen's cylindirical horn/waveguide (you can find the patent online) could be interesting, with modern crossovers and driver selection, or applying a Beveridge-like acoustical lens (info available in the online patent searchs) to alternative planar driver types which could be then be placed up against the walls of a room, (perhaps even hung on) would have more potential for success commercially or as an interesting DIY project, than the "point a lot of drivers everywhere" approach, shown above...

There was one site that was going to do Shahinian obelisk and diapason clones but I don't think they ever finished them. Someone here did something like the Hegeman/Morrison point the drivers at a reflecting ball, but it didn't measure well. (I don't know if the real things measure well, either. Shahinian's multiple drivers pointing multiple ways didn't for stereophile.)

[mazurek]

The keyword to search for is beamforming

[JonW]

Not all. :E

Now-that's-a-lotta-bass.

Wasn't there an Italian company that had a globe enclosure with one driver and it was hung from the ceiling on a wire? I think it got quite a bit of press.

I think that’s Gallo?

The keyword to search for is beamforming

How interesting. Never heard of it. Thanks. Kind of looks beyond the DIY realm, though.

http://en.wikipedia.org/wiki/Beamforming

http://www.tombutcher.com/blog/2009/08/beamforming-and-speaker-arrays/

They come and they go, along with the other omni-directional speaker designs.

If you use multiple drivers in the same range there are potential comb filtering effects to be considered, along with usual room reflections and their effect on the sound.

If you choose very high quality drivers, the costs goes up dramatically as you multiply the numbers needed for the design. You also end up having to connect various things in series and in parallel combination groupings to keep impedance within reason.

Since low frequencies become omni directional as they increase pass the baffle diffraction size, and most people don't believe that trying to reflect treble off the floor beneath the speakers is of any benefit, most designers concentrate on just the mid-bass on up for their "omni" speakers. I think Daniel Queen's cylindirical horn/waveguide (you can find the patent online) could be interesting, with modern crossovers and driver selection, or applying a Beveridge-like acoustical lens (info available in the online patent searchs) to alternative planar driver types which could be then be placed up against the walls of a room, (perhaps even hung on) would have more potential for success commercially or as an interesting DIY project, than the "point a lot of drivers everywhere" approach, shown above...

There was one site that was going to do Shahinian obelisk and diapason clones but I don't think they ever finished them. Someone here did something like the Hegeman/Morrison point the drivers at a reflecting ball, but it didn't measure well. (I don't know if the real things measure well, either. Shahinian's multiple drivers pointing multiple ways didn't for stereophile.)

Thanks for the detailed thoughts. Yes, measurement (and subsequent crossover design) would be tricky, I think. Phase issues also concern me. Acoustic lenses are a very interesting idea, but maybe left more for people who do this for a living. I’m just a weekend hack, if I could even deem myself that. (Precious few weekends are available for such projects.) Comb filtering makes sense but can, I’d think, be avoided by logical placement of the drivers, relative to each other. Maybe not, though.


Yeah, it could get expensive. But I do wonder how different things will sound versus a more conventional speaker. Maybe a more modest project with, say, only 4-5 cheap tweeters and 4-5 cheap midwoofers per side might be a worthwhile experiment. See my next post, below, for the basic idea. Worth the effort or will it sound like garbage?

[JonW]

One of the most successful I have heard about is the Raal Requisite "Eternity", which some of the members probably heard at RMAF2009. It has an array of subs, array of mids and a tweeter array as well, I believe.

http://www.raal-requisite.com/pages/eternity.html

They show a lot of the construction details in an article at 6moons:

http://6moons.com/industryfeatures/r...rbia/raal.html

Hi Dan,

You’re the omni guru here so I was hoping you might post or I was going to send you a PM. Yeah, that RAAL system gets rave reviews. Not heard it myself. How do you think the multi-driver approach compares to (your) single driver (for a given frequency range) approach?

Just in the spirit of experimenting a little, how does an idea like this sound…?

Take 4 midwoofers and mount them each from a central post of some sort, each pointing out 90 degrees from the other, parallel with the floor. It would look kind of like the mids of the RAAL’s you linked to. But only 4 around (rather than 6- a cheaper experiment) and only a single row (avoids comb filtering). Then maybe have a 5th midwoofer pointing to the ceiling, in the middle of the other 4 drivers. So 5 drivers in total.

Then make a similar 5 driver cluster of tweeters. Place this tweeter cluster near/above the woofer cluster for a 2-way speaker.

Some things I wonder about at the moment are:
-How to get the phase to align with so many drivers doing their thing.
-Would having 2 midwoofers in the same cabinet, but pointing away from each other, present a problem? (I could isolate them if need be.)
-Measurements might be tricky with so many drivers (and then troubles with xover design)

Might this be silly or worth a shot for something interesting? I’d value your opinion here. I’d be happy to discuss it here, by email, or even phone. I’ll soon be putting back together my measurement rig for another project so I could get double use out of it if I have a second system to measure.

[neuro]

A project along these lines would be quite interesting. I think you would still get comb filtering in theory, just in the horizontal plane rather than the vertical as happens in a line array. At the listening position though, signal from the side drivers would be pretty far off axis. Guess that depends of how you orient the driver square with respect to that listening position. Sitting 45 degrees off axis from each of two drivers ought to put you right in the middle of comb filter land. Some more thought would have to be applied as to whether that comb filter could be used as part of the design... Not sure what the frequency response of a comb filter is off hand.

[JonW]

A project along these lines would be quite interesting. I think you would still get comb filtering in theory, just in the horizontal plane rather than the vertical as happens in a line array. At the listening position though, signal from the side drivers would be pretty far off axis. Guess that depends of how you orient the driver square with respect to that listening position. Sitting 45 degrees off axis from each of two drivers ought to put you right in the middle of comb filter land. Some more thought would have to be applied as to whether that comb filter could be used as part of the design... Not sure what the frequency response of a comb filter is off hand.

Makes sense. The way I'm thinking about it... If you have 4 drivers (or a tweeter and woofer pair) per side of the cabinet, then point one face/driver right at the listener (as normal), then the comb filtering happens 45 degrees off the listening axis. And we won't sit 45 degrees off axis. Plus due to off-axis response drops (especially at high frequencies) the magnitude of the comb filtering will be less than normal, even if you were sitting in that region.

OK, I've been thinking about this potential project some more. I've got some of the issues/questions from above worked out in my skull. A few remaining things to think about are:

a) Will 4 drivers, one of each side of the cabinet, get the job done? Or might a fifth one pointing up help? Or go to a 5 or 6 sided cabinet with a driver on each side?

b) Is this experiment likely to yield something interesting, with a pleasing sound, and be worth the time and money to put into it? Hmmm...

To be honest, I was expecting someone to give a reason why such a driver arrangement is a silly idea, from a sonic perspective. Maybe there still is a flaw in the approach.

[Bear]

Funny, I was thinking of a variation on a theme... Having an upward firing woofer, but with tweeters on four sides -- taking the Pluto idea to a full omni.

[dlneubec]

Hi Dan,

You’re the omni guru here so I was hoping you might post or I was going to send you a PM. Yeah, that RAAL system gets rave reviews. Not heard it myself. How do you think the multi-driver approach compares to (your) single driver (for a given frequency range) approach?

Just in the spirit of experimenting a little, how does an idea like this sound…?

Take 4 midwoofers and mount them each from a central post of some sort, each pointing out 90 degrees from the other, parallel with the floor. It would look kind of like the mids of the RAAL’s you linked to. But only 4 around (rather than 6- a cheaper experiment) and only a single row (avoids comb filtering). Then maybe have a 5th midwoofer pointing to the ceiling, in the middle of the other 4 drivers. So 5 drivers in total.

Then make a similar 5 driver cluster of tweeters. Place this tweeter cluster near/above the woofer cluster for a 2-way speaker.

Some things I wonder about at the moment are:
-How to get the phase to align with so many drivers doing their thing.
-Would having 2 midwoofers in the same cabinet, but pointing away from each other, present a problem? (I could isolate them if need be.)
-Measurements might be tricky with so many drivers (and then troubles with xover design)

Might this be silly or worth a shot for something interesting? I’d value your opinion here. I’d be happy to discuss it here, by email, or even phone. I’ll soon be putting back together my measurement rig for another project so I could get double use out of it if I have a second system to measure.

I've actually thought about doing this kind of design. I think the concept should work just fine. It is similar in many ways to the approach I took with two drivers firing at each other (one up, one down), except it will have the 45º nulls at the 45º off axis as discussed, rather than having true 360º radiation. I don't believe phase will be an issue, since you will have comb filtering cancellation at 45º, which suggests you won't hear the side or rear firing drivers directly on axis, but only as reflected sound. I think you end up with all drivers wired in phase. It seems to me you would probably have a something like a 4 leaf clover shaped horizontal polar response.

The up and downfiring driver concept has a problem with interaction of the drivers with the opposing baffle that it is firing at, causing reflections. In my case, I went with an MTM configuration and I wanted the drivers close to time aligned, which meant putting the tweeter baffle between the up and down firing mids. It worked very well, but was very constrained in terms of how far the mid baffles had to be apart to avoid negative interaction and the size shape and position of the tweeter and its baffle was also a key to getting it to work. Also, the flexibility in crossover point was not very flexible due to all these issues.

By firing drivers in 4 directions, you get lots of in room spl (a challenge with only one up firing driver, hence using one up and one down in most of my designs), but have much more flexibility in driver CTC, etc. Each driver is stressed less for a given spl level, and xmax limitations are less a factor as are distortion at higher spl levels, kind of like a line array, but with horizontal rather than vertical comb filetering. The design might be much more like a conventional speaker to design, since you will likely fire one set of drivers at the listener, pointing nulls to the sides. A WMT arrangement, or example, would be very flexible in terms of driver placement, vertically. You may not have to do any BSC, because of the close to 360º radiation pattern. I haven't thought through, however, what happens to the power response as wavelengths get shorter than the baffle widths and each driver becomes more directional and starts to beam. Does comb filtering disappear? Do the nulls get larger simply due to wavelengths? Lots of questions to be thought about.

Sorry for the rambling response, this is all just coming off the top of my head without careful consideration. However, I say go for it! If I was not $ challenged lately, I might have jumped into this already myself. I think it is a fascinating concept and the reviews and comments on the RAAL suggests it has a lot of potential. :T

[JonW]

Funny, I was thinking of a variation on a theme... Having an upward firing woofer, but with tweeters on four sides -- taking the Pluto idea to a full omni.

Interesting idea. Or what about having both T and W on all sides, maybe even firing up? See my thoughts, below.


-=-=-=-=-

Hi Dan,

Thanks for the helpful info. I was just reading through your older thread that has a lot of good things to learn:

http://www.htguide.com/forum/showthread.php4?t=22445

I've actually thought about doing this kind of design. I think the concept should work just fine. It is similar in many ways to the approach I took with two drivers firing at each other (one up, one down), except it will have the 45º nulls at the 45º off axis as discussed, rather than having true 360º radiation. I don't believe phase will be an issue, since you will have comb filtering cancellation at 45º, which suggests you won't hear the side or rear firing drivers directly on axis, but only as reflected sound. I think you end up with all drivers wired in phase. It seems to me you would probably have a something like a 4 leaf clover shaped horizontal polar response.

That is exactly how I’ve been thinking about it. Good to see that I’m on the right track. Has this been done already? I haven’t seen anything yet but I’d think that the idea must have been around for decades. I wonder if it could turn out well or not.

The up and downfiring driver concept has a problem with interaction of the drivers with the opposing baffle that it is firing at, causing reflections. In my case, I went with an MTM configuration and I wanted the drivers close to time aligned, which meant putting the tweeter baffle between the up and down firing mids. It worked very well, but was very constrained in terms of how far the mid baffles had to be apart to avoid negative interaction and the size shape and position of the tweeter and its baffle was also a key to getting it to work. Also, the flexibility in crossover point was not very flexible due to all these issues.

Good points. I get it.

By firing drivers in 4 directions, you get lots of in room spl (a challenge with only one up firing driver, hence using one up and one down in most of my designs), but have much more flexibility in driver CTC, etc. Each driver is stressed less for a given spl level, and xmax limitations are less a factor as are distortion at higher spl levels, kind of like a line array, but with horizontal rather than vertical comb filetering. The design might be much more like a conventional speaker to design, since you will likely fire one set of drivers at the listener, pointing nulls to the sides. A WMT arrangement, or example, would be very flexible in terms of driver placement, vertically.

Also how I’ve been thinking about it. But, in the grand scheme, higher SPL, etc. is probably not a big deal. Getting a project like this working and yielding something with a unique, pleasing sound is the big issue. Although I do quite like the idea of lower distortion from multiple drivers. I’ll probably skip a WMT at this point and just go MT for the sake of simplicity and cost. If the general idea works well then we can consider a 3 way with better drivers.

You may not have to do any BSC, because of the close to 360º radiation pattern.

Not sure on this one. My guess is that the end result of the 4 drivers is that when you measure you will only get data from the main driver pointing right at you/the mic. And you will have to gate it such that you do not measure any of the bounces, ambient response, etc. So then you just make a pretty standard crossover with BSC and such.

I haven't thought through, however, what happens to the power response as wavelengths get shorter than the baffle widths and each driver becomes more directional and starts to beam. Does comb filtering disappear? Do the nulls get larger simply due to wavelengths? Lots of questions to be thought about.

I’m not the guy to ask such seemingly complex questions. But, as you mentioned above, my guess is that comb filtering would not be an issue here. Although it could with a larger array, like the RAAL’s you mentioned.

My latest idea for what to try is in the next post, below...

[JonW]

OK, so what do you think of this latest, crazy plan:

Build a regular looking cabinet. Have a tweeter and a woofer on each of the 4 sides. And then also have a tweeter and a woofer on the top side, facing up. Then put switches in the wiring such that I could turn a given driver on or off.

That way I could see if, say, having the 5th woofer and/or tweeter firing up adds anything nice to the other 4 that are already working. Heck, I could even compare the sound of the standard single MT primary source versus 4 or 5 MT pairs all playing. Because of how I’d need to take gated measurements and make the crossover accordingly, as described above. My guess is that a pretty standard crossover will be the way to go, as if it’s only a single MT. Then switch on the other drivers to see if it sounds better, but no need to adjust the crossover. OK or unreasonable? Many seem to say that the tweeter need only be a single point source not omni, but a few seem to disagree. So I don’t know, ahead of time, if having more than one tweeter in the system is a good thing.

Beyond the front firing MT, the other drivers are not the prime source of sound, they are only adding “ambience.” Although if the other drivers are not adding much to the presentation I cannot help but wonder if they’ll make much of an improvement at all- or even detract from all the room bounce.

Too crazy?

[joecarrow]

There was actually a professor of music at my school, Rensselaer Polytechnic Institute (mostly engineering there), and he used a couple of these spherical speakers to provide a more instrument-like presence to his purely electronic instruments. The thinking was that traditional electronic instruments like synthesizers do not have the same psychoacoustical presence and visceral presence in the room compared to real physical instruments, because real instruments have a complex spherical pattern of radiation that gives the ears a lot of cues about the instrument.

If I recall correctly, his spherical speakers were not playing the same thing through all drivers, with different filters and effects being faded between the different drivers. I saw him perform in person a couple of times, and there really did seem to be something to that. Of course, this was not about accurate reproduction of sound, but about more physically involving creation of sound from a digital instrument.

I just googled him, and it appears that he's still doing a lot of interesting work: http://www.arts.rpi.edu/crb/

[Dennis H]

John Murphy is using 24 of the Dayton RS90 full-range in his line arrays. Zaph liked the Aurasound version. How about 12 of them on a dodecahedron? It wouldn't cost all that much and wouldn't be that big with 3.5" drivers. Your 'crossover' would just be a passive (or active) EQ.

http://www.trueaudio.com/array/index.htm

http://www.parts-express.com/pe/showdetl.cfm?Partnumber=290-210

http://en.wikipedia.org/wiki/Dodecahedron

[JonW]

There was actually a professor of music at my school, Rensselaer Polytechnic Institute (mostly engineering there), and he used a couple of these spherical speakers to provide a more instrument-like presence to his purely electronic instruments. The thinking was that traditional electronic instruments like synthesizers do not have the same psychoacoustical presence and visceral presence in the room compared to real physical instruments, because real instruments have a complex spherical pattern of radiation that gives the ears a lot of cues about the instrument.

If I recall correctly, his spherical speakers were not playing the same thing through all drivers, with different filters and effects being faded between the different drivers. I saw him perform in person a couple of times, and there really did seem to be something to that. Of course, this was not about accurate reproduction of sound, but about more physically involving creation of sound from a digital instrument.

I just googled him, and it appears that he's still doing a lot of interesting work: http://www.arts.rpi.edu/crb/

Interesting. What you say about how a real instrument interacts with the room differently than a speaker makes sense. And it makes me wonder if some sort of multi- or omni-directional speaker might make for a pleasant sound. On the other hand, I can also see where lots of room bounce issues would detract. Hmmm...

John Murphy is using 24 of the Dayton RS90 full-range in his line arrays. Zaph liked the Aurasound version. How about 12 of them on a dodecahedron? It wouldn't cost all that much and wouldn't be that big with 3.5" drivers. Your 'crossover' would just be a passive (or active) EQ.

http://www.trueaudio.com/array/index.htm

http://www.parts-express.com/pe/showdetl.cfm?Partnumber=290-210

http://en.wikipedia.org/wiki/Dodecahedron

Yeah, the dodecahedron thing is what I had in mind. So any idea how something like that sounds? Worth a shot at some experimenting? Although I would tend to favor 2 or 3 ways designs over full ranges, in general. If there might be something to be gained from spherical/dodecahedral designs, I'd imagine that you could just put, say, a tweeter and a woofer on each face.

That John Murphy link is, ummm, interesting. It looks like a regular sealed line array just put in a corner? He's flown plenty of red flags to raise my skepticism- naming the project after himself, repeating how picky he is about sound quality, saying how amazing the RS90 driver is, using his own guitar playing for distortion tests, his odd way of posting on the web and then claiming it's disclosed publicly... But maybe he has a point and something interesting? Hard for me to tell.

[---k---]

Jon, I think you're being nutty. :P I think you need to use the TB W4-1337 because it has a silver reflective cone. Then cover the remaining surfaces of the 'dodecahedron-thingy' with mirrors and you got a nice disco ball! :T

More serious: I don't know, it could sound great. But, if I recall, you didn't like the larger open baffle speakers you heard, because you described the sounds as "a wall of sound" and that they lacked imaging and precision. I would anticipate that is what you'll get with a ball speaker - don't know.

Where are you planning on putting this speaker and what are you thinking you would use it for? What ever happens, I look forward to hearing it some day.

[JonW]

Jon, I think you're being nutty. :P I think you need to use the TB W4-1337 because it has a silver reflective cone. Then cover the remaining surfaces of the 'dodecahedron-thingy' with mirrors and you got a nice disco ball! :T

Ooooo, disco ball. I didn’t think of that! I like it. Of course the speaker would have to be rotating. And the only music it could play would be disco-riffic.

More serious: I don't know, it could sound great. But, if I recall, you didn't like the larger open baffle speakers you heard, because you described the sounds as "a wall of sound" and that they lacked imaging and precision. I would anticipate that is what you'll get with a ball speaker - don't know.

Where are you planning on putting this speaker and what are you thinking you would use it for? What ever happens, I look forward to hearing it some day.

Yes, exactly. I have not found open baffle speakers to be exactly to my liking for the reasons you mentioned. At least the few I’ve heard. (Which isn’t to condemn them, just maybe not to my taste or I haven’t heard the right one for me yet.)

I’m just wondering if this different approach might yield something with an interesting sound. Something I’ve never quite heard before. Could be good. It might not. I’m hoping to see if there are any other opinions out there. I don’t have too much free time so I have to limit how many projects like this I take on. I’m still thinking about it. It sounds like a fun experiment. Especially if I could make a speaker where you have 5 T’s and 5 W’s per speaker and all could be turned on and off to hear the results. If there’s a chance of making something with a pleasing sound… I don’t know… Hmmm…

[JonW]

So will this whole idea of switching various drivers on and off actually work?

Let’s say that each speaker has 5 T’s and 5 W’s, all wired together and able to play at the same time. Probably series-parallel for the first 4 and then maybe series or parallel for the fifth. And we want to be able to turn each one on or off, to compare the sound of various combinations. There would be 10 switches per speaker.

If you were to just disconnect a given driver with a switch, the crossover network is messed up. Suddenly the impedance will not be the same or the next driver in series will not work. We’d need to account for that, I think.

Maybe a possible solution would be to have a resistor of the same value as whatever the driver measures (e.g., 4 or 8 ohms). And have a single pole double throw (SPDT) switch selecting between running through the driver or a resistor. Kind of like in the picture below.

Would that work? Or am I missing something completely?

[Dennis H]

Hey Jon,

I'm inclined to cut Murphy some slack. He's a long-time designer with a lot of commercial designs under his belt. He was also a judge at the Dayton DIY along with Joe D'Appolito and others. And he's been listening to his DIY 2-way arrays for almost 30 years so I guess he must think this will be a step up (not that we'd all agree they're the best thing since sliced bread).

http://www.trueaudio.com/array/MCLA_MLA_8_16.htm

[JonW]

Hey Jon,

I'm inclined to cut Murphy some slack. He's a long-time designer with a lot of commercial designs under his belt. He was also a judge at the Dayton DIY along with Joe D'Appolito and others. And he's been listening to his DIY 2-way arrays for almost 30 years so I guess he must think this will be a step up (not that we'd all agree they're the best thing since sliced bread).

http://www.trueaudio.com/array/MCLA_MLA_8_16.htm

Hi Dennis,

Thanks. The guy could be a total teddy bear and a genius, for all I know. And he’s got to know at least 1,000 times more than I do about speaker design. I never even heard of him, which only shows how little I know. I was just put off by the presentation. Let’s just say that, in my day job, when I see someone present their work in such a way, well, it’s rare that the work behind it is decent. And they don't usually last very long in the game. We’re all here for hobby fun. So I’ll try to leave the work attitude behind.

So about this switching drivers on and off idea…? And whether or not such a speaker with many drivers might sound good or not…?

[dlneubec]

So will this whole idea of switching various drivers on and off actually work?

Let’s say that each speaker has 5 T’s and 5 W’s, all wired together and able to play at the same time. Probably series-parallel for the first 4 and then maybe series or parallel for the fifth. And we want to be able to turn each one on or off, to compare the sound of various combinations. There would be 10 switches per speaker.

If you were to just disconnect a given driver with a switch, the crossover network is messed up. Suddenly the impedance will not be the same or the next driver in series will not work. We’d need to account for that, I think.

Maybe a possible solution would be to have a resistor of the same value as whatever the driver measures (e.g., 4 or 8 ohms). And have a single pole double throw (SPDT) switch selecting between running through the driver or a resistor. Kind of like in the picture below.

Would that work? Or am I missing something completely?

JonW,

I don't think the switch concept would work because a driver's impedance is complex and changes with frequency, so the load is not consistent like a resistor is. That is one of the reasons why textbook crossovers don't work well, because they assume a simple driver impedance that is fixed.

BTW, I'm not sure if the upfiring driver would add anything or not to those in the horizontal array. It would take some experimentation, but I suspect it would simply complicate things and might add vertical comb filtering issues to sort out along with the horizontal ones. 8O

As far as instruments go, that is obviously very complex. Some have horn shapes that face the audience, some face up, some face down, percussion and piano probably goes all directions, etc. etc., so an omni solution might be just as likely to simulate real instruments as a monopole. I suspect the success of any speaker has a lot to do with how the source material was recorded. Consider that what you listen to was recorded with mics in fixed positions. Also, what you hear when you go to an amplified concert is speakers, not instruments, so we have speakers simulating speakers, I guess.

Dan

[JonW]

Jon,

I don't think the switch concept would work because a driver's impedance is complex and changes with frequency, so the load is not consistent like a resistor is. That is one of the reasons why textbook crossovers don't work well, because they assume a simple driver impedance that is fixed.

Hi Dan,

Thanks for the further thoughts. That’s an excellent point. When I was making the Spassvogels, I put some effort into keeping the impedance fairly flat, in the crossover design. I did a little listening with versus without the impedance flattening LCR section and I dare say I think it sounded a little better flattened. As you can see in the plot below it only varies between about 6-9 ohms in the range of 80-20,000 Hz. So maybe here if I were to flatten the impedance with this new speaker as well I might be able to, say, put a 7 ohm resistor in there? Ahead of time I can’t tell how flat I’d be able to get the impedance. And even if it only varied 6-9 ohms, I’m not sure if a single resistor value would be enough to keep things sufficiently consistent.

Alternatively, to still allow switching between different driver configurations, I could have each driver connected to a pair of binding posts on the outside of the cabinet. And an external crossover. Then to get various configurations of drivers playing I could use jumper cables to connect the the drivers in various ways. That would take a while to change configurations. And not be nearly as slick as flipping a switch. But could get the job done.

BTW, I'm not sure if the upfiring driver would add anything or not to those in the horizontal array. It would take some experimentation, but I suspect it would simply complicate things and might add vertical comb filtering issues to sort out along with the horizontal ones. 8O

Makes sense. I think it might be interesting to hear if there is an effect and if it’s a good or bad one. My first impression is that the upfiring drivers would benefit from a half sphere or some other such diffuser, much like you did with your omnis. If there is something to a spherical array then we’d need the upfiring driver(s). But I still don’t know if this whole approach has any major benefits.

As far as instruments go, that is obviuosly very complex. Some have horn shapes that face the audience, some face up, some face down, percussion and piano probably goes all directions, etc. etc., so an omni solution might be just as likely to simulate real instruments as a monopole. I suspect the success of any speaker has a lot to do with how the source material was recorded. Consider that what you listen to was recorded with mics in fixed positions. Also, what you hear when you go to an amplified concert is speakers, not instruments, so we have speakers simulating speakers, I guess.

Dan

Yup. I can convince myself that an omni or spherical array approach has strong merits for the reasons you state. And I can also talk myself into thinking that it’s all going to go awry because of room bounce, interference, etc. I don’t know…

[dlneubec]

Well, not having an electronics background, my take may be totally wrong here, but it seems to me that when you flatten the impedance in a crossover, part of what you are doing is dealing with a complex, non-flat driver impedance.

In other words, your flattening efforts are partly in response to the varying impedance that the driver presents over it usable range. Now when you stick a static load, like a resistor, in place of the driver, the crossover will no longer yield a flat impedance, since the drivers impedance varies with frequency and the resistors does not.

[dlneubec]

Yup. I can convince myself that an omni or spherical array approach has strong merits for the reasons you state. And I can also talk myself into thinking that it’s all going to go awry because of room bounce, interference, etc. I don’t know…

Again, my theory may be bad here, but I think the problem with reflected sound is when it does not match the direct sound or is not delayed enough and smears the direct sound. This happens with a typical monopole speaker because the off axis sound is not the same as the on axis. In a true omni, the sound is the same in all directions. Therefore your concern is not that sound is reflected, but that the reflected sound is delayed enough so that the brain does indeed interpret it as reflected sound, not direct, which in turn gives you the sense of space, soundstage width and depth, etc.

I believe as long as you get the omni source far enough away from the walls so that the reflected sound is delayed by about 6ms or more (I think that is the number Linkwitz refers to), it will be interpreted by the brain properly, just as it does with an instrument in a space where reflections come off all the walls. The other approach is to use constant directivity to reduce reflections at the listening position, which is Geddes approach.

[---k---]

I think I'm loosing it. I could have sworn that I wrote a post earlier recommending Toole's book again to you.

Did I get moderated?

[JonW]

I think I'm loosing it. I could have sworn that I wrote a post earlier recommending Toole's book again to you.

Did I get moderated?

I think you emailed that to me. Does the book address this omni/multi directional topic?

Which isn’t to say that you should not be moderated. :W :P

Again, my theory may be bad here, but I think the problem with reflected sound is when it does not match the direct sound or is not delayed enough and smears the direct sound. This happens with a typical monopole speaker because the off axis sound is not the same as the on axis. In a true omni, the sound is the same in all directions. Therefore your concern is not that sound is reflected, but that the reflected sound is delayed enough so that the brain does indeed interpret it as reflected sound, not direct, which in turn gives you the sense of space, soundstage width and depth, etc.

I believe as long as you get the omni source far enough away from the walls so that the reflected sound is delayed by about 6ms or more (I think that is the number Linkwitz refers to), it will be interpreted by the brain properly, just as it does with an instrument in a space where reflections come off all the walls. The other approach is to use constant directivity to reduce reflections at the listening position, which is Geddes approach.

Hi Dan. Good info. That all makes sense.

A stupid thought that I should probably not even post, but here goes: if you’re trying to, say, simulate the sound of an instrument (e.g., a cello) in a room, how about a single omnidirectional speaker? Not a pair. Probably a bad idea...

Well, not having an electronics background, my take may be totally wrong here, but it seems to me that when you flatten the impedance in a crossover, part of what you are doing is dealing with a complex, non-flat driver impedance.

In other words, your flattening efforts are partly in response to the varying impedance that the driver presents over it usable range. Now when you stick a static load, like a resistor, in place of the driver, the crossover will no longer yield a flat impedance, since the drivers impedance varies with frequency and the resistors does not.

I see what you’re saying. And I probably know less about electronics than you do. (I didn’t even know how to solder prior to building my first speakers.) I’m thinking about it this way, which may be totally wrong… The picture below is the Spassvogel crossover. Ignore everything except for the tweeter and the L5, R8, C6 section. The LCR is there to flatten the impedance. (Let’s call it “flat” for the moment.) I’m thinking that the tweeter and the LCR work in convert to give the flat impedance plot. Where one part would make a peak, the other a trough. So with the tweeter and the LCR together, you’re always flat. And so a single resistor could substitute.

I don’t know if what I said is correct. Also, I don’t know if the real world 6-9 ohm “flatness” of the plot is close enough to be approximated by a single resistor value and yield a good result.

At any rate, if switches won't work I think that I still have the option of having each of the 10 drivers/side connected to their own pair of binding posts. Then use jumper cables and an external crossover to get the various driver combinations. At least I think that might work…?

[Dennis H]

Your Zobel idea would work but you'd need one on each driver if you want to switch them in and out.

[---k---]

Jon,
Could have sworn I wrote a longer post. Probably wrote it and closed my browser rather than hitting post. doh!

Anyway, Toole's book touches breifly on Dipoles and omnis - and I didn't think it was positive, and talks a lot about line arrays - positive. The book is mostly about how speakers interact with rooms. Like I said, I don't think it is directly applicable to your design, but it might get you help your thinking about how the speaker will interact with the room. If nothing more, it has a lot of graphs to look at. Come to Chicago, and you can borrow my copy.

[JonW]

Jon,
Could have sworn I wrote a longer post. Probably wrote it and closed my browser rather than hitting post. doh!

Anyway, Toole's book touches breifly on Dipoles and omnis - and I didn't think it was positive, and talks a lot about line arrays - positive. The book is mostly about how speakers interact with rooms. Like I said, I don't think it is directly applicable to your design, but it might get you help your thinking about how the speaker will interact with the room. If nothing more, it has a lot of graphs to look at. Come to Chicago, and you can borrow my copy.

Clicking to get the book on Amazon might be easier than driving to Chicago. Although I certainly would like to pay you another visit. It’s been a while. If there were only more time in life…

Yeah, I can certainly see how room interactions might be a good or a bad thing. Idunno.

[JonW]

Your Zobel idea would work but you'd need one on each driver if you want to switch them in and out.

Thanks, Dennis. OK, good to know that I’m thinking about it correctly. I was first thinking about a resistor on the switch for each driver- easy enough. But if using a switch requires an entire LCR for each driver… With 5 tweeters and 5 woofers per speaker that becomes impractical.

OK, let me think aloud (with my typing fingers) how to make a speaker with 5 MT pairs and an easy way to switch them around so you can listen with 1 pair (front) versus 4 (front, rear, left, right) versus 5 (all) versus 2 (front and top). Hmmm…

I could take measurements with 1, 2, 4, and 5 drivers connected. Although because of the required gating I suspect that only the 1 MT pair data will be of any use. (The impedance data may differ more than the frequency responses.) Then develop a crossover.

Maybe go back to the idea of having each driver connected straight to binding posts on the cabinet. Connect the desired number of MT pairs using jumper cables. Have the crossover external. I think that might get the job done. And I *think* that one crossover may be OK with all driver combos. Especially if you’re mostly using data only from the front MT pair. Any opinions to the contrary?

I don’t like the binding post idea because it will not allow for rapid switching between configurations. But that still might be OK. I’d have to use the correct series/parallel wiring to keep impedances as close as possible to each other case. For example, the 4 MT case wired in series-parallel would have the same resistance as the 1 MT case.

If I can come up with a clever way to do this, it could be a fun project- seeing how 1 versus 4 versus 5, etc. MT pairs sound in a given room. Might be worth the effort, might not. I’m still thinking…. (Danger Will Robinson!)

[---k---]

Jon,
I was wondering through PE and came across this speaker by Neil Davis:

http://techtalk.parts-express.com/showpost.php?p=1595034&postcount=21

Then I remembered a post he made a while back describing it. I think he does a lot of what you are thinking about, but with active crossovers. I don't remember. You might want to email him.

[Dennis H]

I could take measurements with 1, 2, 4, and 5 drivers connected. Although because of the required gating I suspect that only the 1 MT pair data will be of any use. (The impedance data may differ more than the frequency responses.) Then develop a crossover.

Maybe go back to the idea of having each driver connected straight to binding posts on the cabinet. Connect the desired number of MT pairs using jumper cables. Have the crossover external. I think that might get the job done. And I *think* that one crossover may be OK with all driver combos. Especially if you’re mostly using data only from the front MT pair. Any opinions to the contrary?

I don't think one passive crossover would work because it would be feeding a different impedance depending on how many drivers you had. How about using an active crossover for your experimenting? It wouldn't care about impedance and you could use switches. You could do it with a DCX or with LspCAD's emulation thingy where it uses your sound card. It still wouldn't be a simple A/B test though because it would get louder with more drivers and louder always sounds better.

[JonW]

Jon,
I was wondering through PE and came across this speaker by Neil Davis:

Then I remembered a post he made a while back describing it. I think he does a lot of what you are thinking about, but with active crossovers. I don't remember. You might want to email him.

Ryan-

Excellent find! Yes, I'll drop him a line (when I get back from a short trip that I leave on tomorrow morning.) It would be very cool to hear such a speaker. Heck, it could save me a lot of work. Maybe I could even settle for some listening impressions.

I don't think one passive crossover would work because it would be feeding a different impedance depending on how many drivers you had.

Hi Dennis,

Good point. I was thinking that maybe 4 drivers wired in series-parallel have the same resistance. So then maybe they also have the same impedance. But I guess not. (I still have a lot to learn.)

How about using an active crossover for your experimenting? It wouldn't care about impedance and you could use switches. You could do it with a DCX or with LspCAD's emulation thingy where it uses your sound card. It still wouldn't be a simple A/B test though because it would get louder with more drivers and louder always sounds better.

Excellent idea. I forgot that actives don't care about the impedance. And I happen to have a DCX already. I have only played with it a little. Setting crossover points and slopes is really easy. More complex than that I haven't learned. Now that I've been through one passive crossover design I feel more comfortable using passive components to alter frerquency response, phase, and impedance. But maybe I'll think about active for this project. If I start this project, that is. I must admit to thinking about this one more than I thought I would...

[Dennis H]

I was thinking that maybe 4 drivers wired in series-parallel have the same resistance.

Sure, that would work. I was thinking you might get in trouble with 2 or 5 or whatever weird number of drivers.

[JonW]

Sure, that would work. I was thinking you might get in trouble with 2 or 5 or whatever weird number of drivers.

Right. 1 versus 4 MT pairs will be OK. 1 versus 5 or 1 versus 2 might get tricky.

But I really like your going active idea. Build the cabinets with 5 MT pairs, each driver to a binding post pair. Then an active crossover. Use spade jumpers to get the different configurations and have a listen.

I'll be traveling today and tomorrow and that will give me some time to think about this project without much for distractions. Hmmm...

[CraigJ]

Raal Instinct

While the Canadian's and the Californian are working on the new and improved Ardent, I took JonW's challenge and made an attempt at cloning the Raal Instinct/Requisite omni speaker.



First, I used the traditional lost wax method to cast the bronze base;




Weeks of hand sanding, and here one is (woofer gasket material needs trimming);



So far, it sounds good for mono. Measurements planned for this weekend.

[Carl V]

Whoa....that is impressive.

[BOBinGA]

Now that's different! Got a better description, measurements yet?

-Bob

[CraigJ]

Now that's different! Got a better description, measurements yet?

-Bob

It all kinda started in this thread here; https://www.htguide.com/forum/showth...ght=aleksandar when JonMarsh was commenting on a Raal omni design he had heard at RMAF. I subsequently asked Aleksandar Radisavljevic (Raal) to join our conversation, and Dan N. and I each made separate omni designs from Aleksandar's suggestions. My attempt is here;



Dan's design became the Contrappeso Project, and can be found here; https://www.htguide.com/forum/showth...ght=aleksandar I enjoyed the soundstage from the omni speaker and motivation from JonW and the following article in 6Moons; http://6moons.com/industryfeatures/r...rbia/raal.html got me going. In the article, Aleksandar talks about his 15" Eternity and a smaller 8" named Instinct. Since I only have 8-12" woofers (should have purchased more Aurasounds), I attempted to "clone" a large Instinct.

Here is an initial measurement in room at one meter without the bass unit;



It's all fun and an easy way to recycle stuff

[JonW]

Craig-

Those look super interesting!! :T :T So cool. I’d love to hear them. Although, at this point, I guess that you would, too- in stereo with crossovers. I really wonder how they sound, especially compared to “normal” monodirectional speakers.

I still think about trying out a spherical system or something of the sort. Got a few ideas kicking around the old noggin. But hobby time has been so scarce that I dare not start up any new projects. Got to keep at the ones already started. (Will post an update in a moment.)

You should start a thread on this project. Cool. Good luck! :T