We've done our share of straight array, CBT array simulation and analysis starting in this thread by @nc535 : https://www.diyaudio.com/community/...-wall-or-corner-placement.337956/post-6170250
The simulation I linked to is a CBT with 1.5" drivers placed close together and even with that small size there's still comb filtering due to the growing distance of drivers as seen from the listening position. The examples from Keele aren't exactly honest for practical purposes as they were done with idealized point sources of very small proportions. For the CBT array he included the ground floor image to lengthen the CBT, the straight array did not receive the same courtesy. So I regard those idealized simulations as "somewhat colored" and I've done plenty of footwork to see if they have a shot in practice, The only array that has no combing, albeit only at a single spot is with the curve reversed, a focused array.
My conclusion was easy to make, the next best thing wasn't a CBT for me, but a frequency shaded straight array. Maybe easier to see in some plots.
Here's a post from the past with all types simulated in Vituixcad with my specific 3.5" driver size and 85.5 mm spacing between drivers, they all use the same shading (except the unshaded array):
https://www.diyaudio.com/community/...ver-full-range-line-array.242171/post-7492870
The only way to battle the combing even further (with arrays consisting of full range drivers) is with either a smaller driver size or a larger listening distance to the array. It will move combing up in frequency.
The focused array (with shading) has one spot where it's ideal, move a hair up or down and you're out of the beam. Some faint lobes can be seen due to driver spacing.
The CBT array (with shading) does have a wider vertical "beam" but the combing isn't solved with these size drivers. We also see lobes due to the driver size/spacing.
The straight array (with shading) has a more bundled vertical beam, obviously because that's what I've worked on, and less combing than the CBT. Again, lobes due to driver spacing of this size (3.5" in all cases).
The unshaded array: a bit of everything and fun to play with in real life. No where near as bad as the graphs may indicate. Just look at the DI (Directivity Index), the red line in the middle left graph of all simulations and may help to form your own opinion/ draw your own conclusions on their respective performance.
One honorable mention when talking about arrays has to be @bbutterfield with his fractal arrays: https://www.diyaudio.com/community/...raight-cbt-with-passive-xos-and-no-eq.330031/ Quite the accomplishment.
The simulation I linked to is a CBT with 1.5" drivers placed close together and even with that small size there's still comb filtering due to the growing distance of drivers as seen from the listening position. The examples from Keele aren't exactly honest for practical purposes as they were done with idealized point sources of very small proportions. For the CBT array he included the ground floor image to lengthen the CBT, the straight array did not receive the same courtesy. So I regard those idealized simulations as "somewhat colored" and I've done plenty of footwork to see if they have a shot in practice, The only array that has no combing, albeit only at a single spot is with the curve reversed, a focused array.
My conclusion was easy to make, the next best thing wasn't a CBT for me, but a frequency shaded straight array. Maybe easier to see in some plots.
Here's a post from the past with all types simulated in Vituixcad with my specific 3.5" driver size and 85.5 mm spacing between drivers, they all use the same shading (except the unshaded array):
https://www.diyaudio.com/community/...ver-full-range-line-array.242171/post-7492870
The only way to battle the combing even further (with arrays consisting of full range drivers) is with either a smaller driver size or a larger listening distance to the array. It will move combing up in frequency.
The focused array (with shading) has one spot where it's ideal, move a hair up or down and you're out of the beam. Some faint lobes can be seen due to driver spacing.
The CBT array (with shading) does have a wider vertical "beam" but the combing isn't solved with these size drivers. We also see lobes due to the driver size/spacing.
The straight array (with shading) has a more bundled vertical beam, obviously because that's what I've worked on, and less combing than the CBT. Again, lobes due to driver spacing of this size (3.5" in all cases).
The unshaded array: a bit of everything and fun to play with in real life. No where near as bad as the graphs may indicate. Just look at the DI (Directivity Index), the red line in the middle left graph of all simulations and may help to form your own opinion/ draw your own conclusions on their respective performance.
One honorable mention when talking about arrays has to be @bbutterfield with his fractal arrays: https://www.diyaudio.com/community/...raight-cbt-with-passive-xos-and-no-eq.330031/ Quite the accomplishment.
I'm tempted to try a two-way CBT array, using little 2" full-rangers for the mid-treble, above 300Hz maybe, and then maybe 6" woofers below. The woofers could be unshaded, to get better sensitivity and headroom. The smaller full-range units would push combing concerns another octave further up. And perhaps you wouldn't need frequency-dependent shading on the full-rangers - just amplitude shading. No? (Or would it be problematic crossing from unshaded woofer array to amplitude-shaded full-range array?)
Last edited:
If you're not trying to get bass out of the full rangers, then you don't need frequency dependent shading.
Lobing interferences happen always with line arrays even curved, delayed or shaded. It is a matter of resolution of measurements/simulations. Audibility is another question...
Interferences can be seen in impulse response and eg. CSD/decay with appropriate settings. This applies to large panels/planars too.
kimmosto's VDAD2 simulation are very precise showing these, but again simulation parameters can hide them (too much smoothing, too long gating). sesayso's sims with VCAD2 are fantastic, and if I remember he was a bit shocked when he realized the amount of vertical problems with "normal" long arrays some yeas ago. The legend of constant response in vertical is just wrong, it comes from nearfield measurements done 20-30 years ago, and that legend lives on...
Don Keele's simulations have quite lot of smoothing in my eyes and he doesn't tell exact settings of them. I have read his original papers/posters many times and he seems to skip some difficult issues like horizontal directivity changes and asymmetry...
Interferences can be seen in impulse response and eg. CSD/decay with appropriate settings. This applies to large panels/planars too.
kimmosto's VDAD2 simulation are very precise showing these, but again simulation parameters can hide them (too much smoothing, too long gating). sesayso's sims with VCAD2 are fantastic, and if I remember he was a bit shocked when he realized the amount of vertical problems with "normal" long arrays some yeas ago. The legend of constant response in vertical is just wrong, it comes from nearfield measurements done 20-30 years ago, and that legend lives on...
Don Keele's simulations have quite lot of smoothing in my eyes and he doesn't tell exact settings of them. I have read his original papers/posters many times and he seems to skip some difficult issues like horizontal directivity changes and asymmetry...
Agreed. Like I said, I didn't do extensive simulations. I did build them and I did extensive measurements.
At any particular spot there was always some comb filtering. But averaged across a number of positions, the overall response was quite smooth. I did feel the comb filtering was audible to a certain extent, and in my opinion manifested most clearly in the impulse response. I felt you could hear the very slight delay between drivers, and you can see it in the original post. You can see a trail of 3-4 lower amplitude impulses after the first impulse.
However the audibility of this was not that objectionable, and would be less IMO if I had used a paper cone driver like the Dayton PS95 instead of an aluminum cone FE85. The PS95 still has a good high end but a softer more buttery sound and would be a more agreeable, less analytical character. Measurements and simulations don't really tell you that kind of stuff.
At any particular spot there was always some comb filtering. But averaged across a number of positions, the overall response was quite smooth. I did feel the comb filtering was audible to a certain extent, and in my opinion manifested most clearly in the impulse response. I felt you could hear the very slight delay between drivers, and you can see it in the original post. You can see a trail of 3-4 lower amplitude impulses after the first impulse.
However the audibility of this was not that objectionable, and would be less IMO if I had used a paper cone driver like the Dayton PS95 instead of an aluminum cone FE85. The PS95 still has a good high end but a softer more buttery sound and would be a more agreeable, less analytical character. Measurements and simulations don't really tell you that kind of stuff.
Last edited:
If you use DSP to "reconstruct" the impulse at the listening position, using only the direct sound for as much as possible, you could tame that large and big sound the arrays have.
But.... be aware that it just might be one of the likable aspects of the sound. In the end we hear the direct sound plus the room response. Speakers that light up the room just enough always are a little more forgiving for a lot of recordings. At least, that's what I've found and reason for me to use ambience speakers after removing early reflections.
I hate to say anything about CBT since there has been so much water under my bridge since I did those simulations. Ultimately, I did an unshaded straight array using 2.5" drivers and liked it very much. Only after that did I start looking at shading because I found too much vertical response variation, even though I enjoyed the array both seated and standing.
What I found was that shading reduces the ripple in the FR but if carried too far also narrows the vertical beam. @wesayso directed his shading efforts towards a design that kept the vertical window wide enough so that it was best for seated listening but not too bad or not quite as good for standing, while I was focused at the time on optimizing for seated only - just out of curiosity not because I wanted that. I saw the problem with combing in the CBT and tried (in simulation) arrays of 1.5" drivers. But the small drivers don't extend low enough alone, are low efficiency, and can't take much power. Also, I had no wish to do a 2-way array, even though I modeled one. I think the 3, 3.5" drivers are kind of in a sweet spot in terms of compromise between robustness and combing. Mathematically at least, though, if you hear combing you simply need to listen further away, which may not be possible in your room, so you need to be aware of the issue.
In that thread I moved onto other configurations looking for a solution with a wide vertical beam. You can get that with a constant directivity waveguide or coaxial driver (which solves vertical lobing due to CTC), but then you also have the floor reflection to contend with, which I believe isn't the problem it appears to be in simulation and can be addressed to some extent in the system design. @perrymarshall's other threads opened my eyes to OB and took them away from cardioid for a while. I'm now leaning towards cardioid-like because I get a solution with a smaller footprint that actually has a cabinet in which I can embed a dsp Hypex plate amp, which helps with WAF for me. If I had a bigger room it would be OB for sure because it's easier and has less design risk. Either way, I will use an 8" coax at the top because beaming of the coax mid with larger drivers is starting to become an issue as you approach XO to the coax CD/tweeter.
I like this CBT design/implementation. Perry's artistry with passive crossovers once again shines through, resulting in one of the most practical CBT's I've seen. Contrast it with one of Don Keele's designs which, IIRC, was 2-way with something like 90 1" tweeters, or my simulations with multiple channels of DSP and amplifiers.
What I found was that shading reduces the ripple in the FR but if carried too far also narrows the vertical beam. @wesayso directed his shading efforts towards a design that kept the vertical window wide enough so that it was best for seated listening but not too bad or not quite as good for standing, while I was focused at the time on optimizing for seated only - just out of curiosity not because I wanted that. I saw the problem with combing in the CBT and tried (in simulation) arrays of 1.5" drivers. But the small drivers don't extend low enough alone, are low efficiency, and can't take much power. Also, I had no wish to do a 2-way array, even though I modeled one. I think the 3, 3.5" drivers are kind of in a sweet spot in terms of compromise between robustness and combing. Mathematically at least, though, if you hear combing you simply need to listen further away, which may not be possible in your room, so you need to be aware of the issue.
In that thread I moved onto other configurations looking for a solution with a wide vertical beam. You can get that with a constant directivity waveguide or coaxial driver (which solves vertical lobing due to CTC), but then you also have the floor reflection to contend with, which I believe isn't the problem it appears to be in simulation and can be addressed to some extent in the system design. @perrymarshall's other threads opened my eyes to OB and took them away from cardioid for a while. I'm now leaning towards cardioid-like because I get a solution with a smaller footprint that actually has a cabinet in which I can embed a dsp Hypex plate amp, which helps with WAF for me. If I had a bigger room it would be OB for sure because it's easier and has less design risk. Either way, I will use an 8" coax at the top because beaming of the coax mid with larger drivers is starting to become an issue as you approach XO to the coax CD/tweeter.
I like this CBT design/implementation. Perry's artistry with passive crossovers once again shines through, resulting in one of the most practical CBT's I've seen. Contrast it with one of Don Keele's designs which, IIRC, was 2-way with something like 90 1" tweeters, or my simulations with multiple channels of DSP and amplifiers.
I watched Don's youtube-series on his CBT yesterday and it really got me more hyped on building one of these one day. Of course he had all of those small tweeters helping out, but as far as I've understood combing ain't that big of an issue in practice from a 3-3.5" either, both from what I've read before and from what I read here, right? Ah I really wish there where more of these built so it would be easier to find some to listen too, I really like the theory of lines, but is a slightly worried about that combing.
What happens if you'd raise one of these arrays from the ground a bit btw? Place them on a sub or something to get a bit of elevation?
Btw has there been any experiments on some kind of diffusion ta tame that combing?
What happens if you'd raise one of these arrays from the ground a bit btw? Place them on a sub or something to get a bit of elevation?
Btw has there been any experiments on some kind of diffusion ta tame that combing?
In my opinion, the advantage of having the array on the ground is that the floor becomes an extension of the speaker and eliminates “floor bounce”. Also, subjectively they don’t really sound like the tweeters are on the floor.
That said you could make a symmetrical array where the hot zone is in the middle of the curve and pointed straight at you.
That said you could make a symmetrical array where the hot zone is in the middle of the curve and pointed straight at you.
Ah yeah of course you have that mirror effect with the arrays on the floor, was just wondering how much of it you loose if you'd choose to raise it up for some reason. My thought was to have these at home and then from time to time bring them to smaller events where it would be beneficial to have higher up to not be obscured by people.
Toole'’s work suggests it is not near as audiable as they look. From Taylor: http://p10hifi.net/planet10/TLS/downloads/taylor-line-array.pdf
dave
If we calculate the bending radius of the focused array and the distance to the listener, then it will be ideal and this will be the sweetest place for one listener, but we don’t need more, we’re not going to dance and listen to Beethoven’s 9th Symphony at the same time
.I think that's a great test.
Particularly for large scale line arrays hanging in/at venues etc.
Walking around, moving both horizontally and varying the distance to speakers can give pink noise a whisshy sound pretty easily heard.
Often easy to hear with music too, if there is significant indoor airflow,.... and very easy when outdoor wind picks up..
Indoors, on smaller home audio lines, my ears aren't really good enough to hear the phasey sound itself, unless something's wrong and I'm listening to pink.
(who the heck wants to listen to pink, lol)
The way I do think I hear comb filtering with home audio lines, is a bit of a muted HF/VHF, in terms of sparkle, air, and crisp transients.
The only explanation I can think of for what I perceive I'm hearing, is that the c2c math simply takes its toll on the high end...by limiting how high a line can still behave as a line..