As Sean pointed out, most omni's aren't truly omni, since the drivers usually become directional at high frequencies. But the net effect can be the same as for a true omni - namely, a well energized and tonally correct reverberant field.
Just for the record, the Bose 901 was NOT an omnidirectional system. The array of 8 drivers on the rear of the enclosure was highly directional, in stark contrast with Bose's advertisements which depicted a very wide radiation pattern from the rear-firing array and a narrow pattern from the single front-firing driver. In fact, the opposite was the case! The rear-firing array's radiation pattern would have approximated that of a single driver roughly 8" tall by 16" wide (the 45 degree angle down the center of the rear baffle does help somewhat with horizontal dispersion, so maybe the net effect is more like a single 8" by 12" driver). The 901's forward radiation narrowed to a roughly 90 degree angle at about 4 kHz and continued to narrow progressively above that. The rear radiation narrowed to 90 degress at maybe 1.5 kHz, and of course it just got worse at higher frequencies. The selling point of the 901 was the psychoacoustically pleasing effect of a well-energized reverberant field. Note that assuming the equalizer corrects to give more or less flat power response, the on-axis first-arrival sound from that single front-firing driver will be quite tipped up. Bose got away with it because the reverberant energy was so much louder than that first-arrival signal that it dominated the perceived tonal balance. I'm not going into the other design tradeoffs Bose made at this point - they might sue me (I say that only half jokingly). But the 901 is an ingenious application of psychoacoustics, and I tip my hat to Amar Bose on that score.
The Shahinain Diapason on the other hand gets the tonal balance correct in both the direct and reverberant sound fields. The design uses two woofers, four cone midwoofers, two 3" dome mid-tweeters, two 1.5" dome tweeters, and six 1/2" dome supertweeters (the latter to maintain adequate energy in the reverberant field in the top octave). Designer Dick Shahinian takes into account not only the frequency responses of the drivers, but also their physical orientation and inherent radiation patterns in building a psychoacoustically intelligent loudspeaker.
The Ohm F, German Physics, Huff, MBL Radialstrahler, and Wolcott Omnisphere speakers are perhaps more elegant (though not necessarily better sounding) solutions than the Shahinians, as they use fewer and more exotic drivers, but their radiation patterns tend toward the doughnut-shaped at high frequencies rather than the truly omnidirectional. Bi-polar speakers from Mirage and Definitive Technology also do a good job with the reverberant field. Some planars (namely Beveridges and the big Sound Labs) also generate a tonally correct reverberant field, as do cornerhorns like the Klipschorn and Hartsfield, albeit with less reverberant energy (relatively speaking) than an "omni". But the principle of maintaining correct tonal balance in the reverberant field makes sense to me.
Apparently it also makes sense to Seigfried Linkwitz and Jorma (pronounced "Yorrrrma") Salmi. The latter is the designer of the Gradient Revolution, which is a very well thought-out system in my opinion (yup, I sell 'em). The Revolution uses a dipole bass system and a cardioid mid/tweeter module. At first glance it would seem that these very differently-shaped radiation patterns (the figure-8 dipole and heart-shaped cardioid) would give a very disjointed reverberant field, but the audio gods smiled on Jorma. The ear is very tolerant of arrival time differences in the reverberant field; timing is only critical in the first-arrival signal. And it just so happens that (assuming equal on-axis, first-arrival SPL's) the net energy put out into the reverberant field is the same for a dipole and a cardioid! So using fairly conventional drivers (and only a few of them) in a reasonably-sized enclosure, the Revolution has consistent tonality of both first-arrival and reverberant sound. And this matters because the ears take them both into account; a nice smooth on-axis anechoic frequency response curve is incomplete data, from a psychoacoustic perspective.
In case you can't tell, getting the reverberant field right is probably my favorite obscure loudspeaker design consideration. I find that often perceived loudspeaker characteristics are demystified when the reverberant field is factored in, and the most realistic-sounding systems I've encountered are ones that get the reverberant field right.
Now the one justified criticism of omni's (and their wide-pattern cousins) is their typically non-pinpoint imaging, at least in comparison with a good two-way mini-monitor. I'm afraid that to a certain extent this is an inevitable tradeoff. Even from one concert hall to another, there's a tradeoff between precise localization of sound sources and enveloping ambience. The more energy in the reverberant field, the richer the ambience but the less precise the soundstaging. However, it is the early reflections that are the most detrimental to good imaging. By their very nature wide-pattern speakers have more energetic early reflections, and so with wide-pattern speakers it is especially important to treat the first reflection zones if imaging is a high priority.
Getting back to Stan's original post, I wouldn't say that I'm in favor of omnis just for the sake of omnidirectional radiation. Rather, I'm in favor of getting the reverberant sound to have the same (correct) tonal balance as the first-arrival sound - and an omnidirectional or quasi-omnidirectional system is one very effective way of meeting this criteria. And the test is this - with the music playing a bit louder than normal, walk out of the room. If it still sounds like live music through the open dooreway, then those speakers are getting the reverberant field right - much the same as real instruments do.
Duke
Just for the record, the Bose 901 was NOT an omnidirectional system. The array of 8 drivers on the rear of the enclosure was highly directional, in stark contrast with Bose's advertisements which depicted a very wide radiation pattern from the rear-firing array and a narrow pattern from the single front-firing driver. In fact, the opposite was the case! The rear-firing array's radiation pattern would have approximated that of a single driver roughly 8" tall by 16" wide (the 45 degree angle down the center of the rear baffle does help somewhat with horizontal dispersion, so maybe the net effect is more like a single 8" by 12" driver). The 901's forward radiation narrowed to a roughly 90 degree angle at about 4 kHz and continued to narrow progressively above that. The rear radiation narrowed to 90 degress at maybe 1.5 kHz, and of course it just got worse at higher frequencies. The selling point of the 901 was the psychoacoustically pleasing effect of a well-energized reverberant field. Note that assuming the equalizer corrects to give more or less flat power response, the on-axis first-arrival sound from that single front-firing driver will be quite tipped up. Bose got away with it because the reverberant energy was so much louder than that first-arrival signal that it dominated the perceived tonal balance. I'm not going into the other design tradeoffs Bose made at this point - they might sue me (I say that only half jokingly). But the 901 is an ingenious application of psychoacoustics, and I tip my hat to Amar Bose on that score.
The Shahinain Diapason on the other hand gets the tonal balance correct in both the direct and reverberant sound fields. The design uses two woofers, four cone midwoofers, two 3" dome mid-tweeters, two 1.5" dome tweeters, and six 1/2" dome supertweeters (the latter to maintain adequate energy in the reverberant field in the top octave). Designer Dick Shahinian takes into account not only the frequency responses of the drivers, but also their physical orientation and inherent radiation patterns in building a psychoacoustically intelligent loudspeaker.
The Ohm F, German Physics, Huff, MBL Radialstrahler, and Wolcott Omnisphere speakers are perhaps more elegant (though not necessarily better sounding) solutions than the Shahinians, as they use fewer and more exotic drivers, but their radiation patterns tend toward the doughnut-shaped at high frequencies rather than the truly omnidirectional. Bi-polar speakers from Mirage and Definitive Technology also do a good job with the reverberant field. Some planars (namely Beveridges and the big Sound Labs) also generate a tonally correct reverberant field, as do cornerhorns like the Klipschorn and Hartsfield, albeit with less reverberant energy (relatively speaking) than an "omni". But the principle of maintaining correct tonal balance in the reverberant field makes sense to me.
Apparently it also makes sense to Seigfried Linkwitz and Jorma (pronounced "Yorrrrma") Salmi. The latter is the designer of the Gradient Revolution, which is a very well thought-out system in my opinion (yup, I sell 'em). The Revolution uses a dipole bass system and a cardioid mid/tweeter module. At first glance it would seem that these very differently-shaped radiation patterns (the figure-8 dipole and heart-shaped cardioid) would give a very disjointed reverberant field, but the audio gods smiled on Jorma. The ear is very tolerant of arrival time differences in the reverberant field; timing is only critical in the first-arrival signal. And it just so happens that (assuming equal on-axis, first-arrival SPL's) the net energy put out into the reverberant field is the same for a dipole and a cardioid! So using fairly conventional drivers (and only a few of them) in a reasonably-sized enclosure, the Revolution has consistent tonality of both first-arrival and reverberant sound. And this matters because the ears take them both into account; a nice smooth on-axis anechoic frequency response curve is incomplete data, from a psychoacoustic perspective.
In case you can't tell, getting the reverberant field right is probably my favorite obscure loudspeaker design consideration. I find that often perceived loudspeaker characteristics are demystified when the reverberant field is factored in, and the most realistic-sounding systems I've encountered are ones that get the reverberant field right.
Now the one justified criticism of omni's (and their wide-pattern cousins) is their typically non-pinpoint imaging, at least in comparison with a good two-way mini-monitor. I'm afraid that to a certain extent this is an inevitable tradeoff. Even from one concert hall to another, there's a tradeoff between precise localization of sound sources and enveloping ambience. The more energy in the reverberant field, the richer the ambience but the less precise the soundstaging. However, it is the early reflections that are the most detrimental to good imaging. By their very nature wide-pattern speakers have more energetic early reflections, and so with wide-pattern speakers it is especially important to treat the first reflection zones if imaging is a high priority.
Getting back to Stan's original post, I wouldn't say that I'm in favor of omnis just for the sake of omnidirectional radiation. Rather, I'm in favor of getting the reverberant sound to have the same (correct) tonal balance as the first-arrival sound - and an omnidirectional or quasi-omnidirectional system is one very effective way of meeting this criteria. And the test is this - with the music playing a bit louder than normal, walk out of the room. If it still sounds like live music through the open dooreway, then those speakers are getting the reverberant field right - much the same as real instruments do.
Duke