So another friend/client has asked me to build him a set of on wall stereo mains for his new 86” OLED tv which is currently wall mounted. There will be no center channel so wide horizontal dispersion is critical for a phantom center to work.
I have 46” of height to work with. As wide an enclosure as needed but I’d like to keep it slim. Enclosure depth is limited to 5.5” to keep the profile the same as the screen offset from the wall. The enclosures will butt up right on the outer edges of the screen so in effect, the screen will be a continuation of the baffle.
My first thought was to use a line of 8 ribbon tweeters but that’s the most costly solution. My second thought was a line of 24 3/4” dome tweeters…..specifically the Dayton ND20FB. Here’s Zaph’s measurement page
http://zaphaudio.com/tweetermishmash/compare.html
So we can see excellent frequency response and extension as well as great off axis response due to the small size. I can space these less than 2” center to center and work within my 46” limit. I can also get the center to center spacing close to whatever midwoofer I choose so crossing these at 3.5k won’t be an issue at all…..and with 24 of them, I don’t see a problem with HD either…..
……but I wonder what happens to the response of the driver vertically…..given the center to center spacing of around 1.75” inches or 45mm, I’m expecting a rolloff to begin at around 8000hz from the driver to driver wavefront interference….am I correct in this or is it at 1/2 wavelength?……16k? Also what would be the expected slope of the rolloff?……1st order?….2nd order?
Thanks in advance. Midwoofer selections and considerations will come after a tweeter solution only given the need for phantom center coverage is the limiting design parameter.
I have 46” of height to work with. As wide an enclosure as needed but I’d like to keep it slim. Enclosure depth is limited to 5.5” to keep the profile the same as the screen offset from the wall. The enclosures will butt up right on the outer edges of the screen so in effect, the screen will be a continuation of the baffle.
My first thought was to use a line of 8 ribbon tweeters but that’s the most costly solution. My second thought was a line of 24 3/4” dome tweeters…..specifically the Dayton ND20FB. Here’s Zaph’s measurement page
http://zaphaudio.com/tweetermishmash/compare.html
So we can see excellent frequency response and extension as well as great off axis response due to the small size. I can space these less than 2” center to center and work within my 46” limit. I can also get the center to center spacing close to whatever midwoofer I choose so crossing these at 3.5k won’t be an issue at all…..and with 24 of them, I don’t see a problem with HD either…..
……but I wonder what happens to the response of the driver vertically…..given the center to center spacing of around 1.75” inches or 45mm, I’m expecting a rolloff to begin at around 8000hz from the driver to driver wavefront interference….am I correct in this or is it at 1/2 wavelength?……16k? Also what would be the expected slope of the rolloff?……1st order?….2nd order?
Thanks in advance. Midwoofer selections and considerations will come after a tweeter solution only given the need for phantom center coverage is the limiting design parameter.
You are going to get a lot of "comb filter" effects that may or may not be offensive. You will hear changes in amplitude as you move your ears up and down or use a different measurement distance, but obviously, we don't do that in normal listening. The model below was created for this thread, where there is an explanation of how it works and some confirmation of its results. The code was written in Visual Basic (based on some very old code), but it still compiles and runs in Visual Studio. This isn't exactly your design, as it is a 51.75" array with 23 drivers and 1.5" spacing between the drivers (not CTC). You would have to change the code to model your proposed design more accurately, but the response is representative of your design. The model allows changing the curvature and shading of the drivers, which can be very effective at smoothing the response. Note: X is frequency, Y is vertical listening angle, and the amplitude is color-coded.
Thanks for sharing Neil but i doubt the accuracy of your simulation......at less than 2" center to center spacing, i can't see how i'd hear any phase shift, particularly below the 1/4 wavelength frequency.
And not sure what the X axis is in your graph.....is that frequency?............and if so. why a scale to 100khz?
And not sure what the X axis is in your graph.....is that frequency?............and if so. why a scale to 100khz?
Like Neil said, comb filtering. Combing dulls, mutes, VHF info.
Line array math is well established. c2c spacing limits how high a line works as a line without combing.
Whoops--that model was actually for the midrange array (7 6" drivers), not the tweeters. I forgot that I hadn't finished the crossover and tweeter calculations, so the data I entered for the tweeters was ignored. The original thread that "speaker dave" started was just focused on arrays of full-range drivers, and I only took the code far enough to address that concern. I added the GUI for a 2-way model, but I forgot about where I had left the code. I might go back and finish the code someday to model 2-way line arrays, but it's not high on my to-do list.
I modified the code to allow using smaller midrange drivers, and the first sim below is for 25 1" drivers with .5" between each driver. The results I got with this model agreed with the spreadsheet that Dave was using, so I'm confident that the model is correct. It simply sums all of the outputs at a fixed measurement point, keeping track of the phase due to different travel lengths--it's just ray summation over multiple frequencies and measurement angles. You are going to get interference patterns due to the travel distances from the multiple drivers--see the discussion in that earlier thread for why that is the case. I believe this model is representative of what you are going to get in the 24-driver array you describe in the original post.
The second model run below is using Legendre shading for the drivers, using the approximation offered by Keele. Adding the right curvature results in even smoother off-axis results, depending on the measurement distance. More importantly, with curvature the array takes on the properties of a constant beamwidth transducer, which you can only see in a model that takes into account different measurement distances.
Yes, the X-axis is frequency, and the powers-of-10 axis is how log charts are displayed in ChartDirector by default. The data stops at 20KHz, but I'd have to check whether ChartDirector can be set up to show 20/200/2000/20000.
If anyone wants to play with this code, I'll gladly zip it up and send it via PM. The model has a lot of promise to model line array behaviors, but it's not something I want to support anymore.
I modified the code to allow using smaller midrange drivers, and the first sim below is for 25 1" drivers with .5" between each driver. The results I got with this model agreed with the spreadsheet that Dave was using, so I'm confident that the model is correct. It simply sums all of the outputs at a fixed measurement point, keeping track of the phase due to different travel lengths--it's just ray summation over multiple frequencies and measurement angles. You are going to get interference patterns due to the travel distances from the multiple drivers--see the discussion in that earlier thread for why that is the case. I believe this model is representative of what you are going to get in the 24-driver array you describe in the original post.
The second model run below is using Legendre shading for the drivers, using the approximation offered by Keele. Adding the right curvature results in even smoother off-axis results, depending on the measurement distance. More importantly, with curvature the array takes on the properties of a constant beamwidth transducer, which you can only see in a model that takes into account different measurement distances.
Yes, the X-axis is frequency, and the powers-of-10 axis is how log charts are displayed in ChartDirector by default. The data stops at 20KHz, but I'd have to check whether ChartDirector can be set up to show 20/200/2000/20000.
If anyone wants to play with this code, I'll gladly zip it up and send it via PM. The model has a lot of promise to model line array behaviors, but it's not something I want to support anymore.
Thanks for taking the extra time to clarify! I can see it now……which doesn’t deter me from the plan as the drive to drive unit interference is pretty benign. Now I’ve gotta just figure on the acoustic low pass from the destructive interference…..if it comes on as I suspect at 10k like a first order, I’m good.
Yep, it's interesting to look at the vertical response of a short line array, but I'm not convinced that the information you get is all that useful. When you are sitting in a fixed location, all those variations in the response curve are somewhat irrelevant. These simple models don't relate the variations of amplitude verses frequency versus listening angle on establishing a coherent stereo image, and they don't help visualizing the variations in amplitude versus measurement distance. That's why I'm not too interested in finishing off this model--it just doesn't tell us enough about how a specific line array design will relate to our listening experiences.
But like Dave (RIP), I believe that line arrays are the best way to fill a room and are often the best choice for serious listening. I had some nice point source 3-way designs for our main TV for years, but after switching to line arrays I wouldn't consider anything else. I even put some in my wife's greenhouse (see second photo).
But like Dave (RIP), I believe that line arrays are the best way to fill a room and are often the best choice for serious listening. I had some nice point source 3-way designs for our main TV for years, but after switching to line arrays I wouldn't consider anything else. I even put some in my wife's greenhouse (see second photo).
A bit long but educating thread of long(high) lina arrays https://www.diyaudio.com/community/...ers-a-25-driver-full-range-line-array.242171/
Many people are happy with their fsigle fullrange, multi-horn, 2-way, 4-way, dipole etc. whatever speakers.... despite of how they really work objectively.
Using infinite number of small tweeters doesn't really help... the total height and listening distance determine vertical directivity above certain freq.
Many people are happy with their fsigle fullrange, multi-horn, 2-way, 4-way, dipole etc. whatever speakers.... despite of how they really work objectively.
Using infinite number of small tweeters doesn't really help... the total height and listening distance determine vertical directivity above certain freq.
Juhazi answers the question about the roll-off, which wasn't evident from the charts that I posted. If you rotate the chart view to see just the frequency versus amplitude without the vertical angle, you can see the overall shape of the roll-off more clearly. The first chart below is a 37" array with 25 1" drivers with a 1.5" CTC spacing. The roll-off will be 3dB per octave at a distance of 8 feet, starting around 2500Hz. So, some EQ will be needed to compensate for that drooping response.
As the array gets larger, the 3dB per octave roll-off is more obvious. The chart below is for a 150" 100-driver array with the same drivers.
And, as Juhazi also shows, the roll-off frequency also depends on the measurement distance. As the measurement distance increases, the roll-off frequency also increases. Here is the 150 inch array at 16' away:
About 6 years ago, Werewolf posted an excellent tutorial on the math behind the ideal (infinite) line array, and these ray-tracing models for large arrays are consistent with the math model. But you need to use one of the ray-tracing models for shorter arrays to see where the roll-off will start and to view the amount of lobing on the vertical axis. The relatively short array in the OP will have roll-off starting around 2500Hz, and it will have some lobing in the vertical response, but the vertical response doesn't appear to be too "rough". The roll-off at higher frequencies is enough that it will need to be addressed with EQ, but the vertical lobing is probably not going to be an "issue" for most fixed listening positions.
If you look at Werewolf's conclusions about the line array response, you will see what he finds most interesting about the line array: ..."as we increase distance from the Infinite Line Source, the response falls off at only 3dB for every doubling of distance". He notes that 3dB attenuation with distance is "a HUGE plus" (his emphasis) for stereo listening, as it will result in a wider sweet spot. Unfortunately, I don't know how to model sweet spots 🙁
As the array gets larger, the 3dB per octave roll-off is more obvious. The chart below is for a 150" 100-driver array with the same drivers.
And, as Juhazi also shows, the roll-off frequency also depends on the measurement distance. As the measurement distance increases, the roll-off frequency also increases. Here is the 150 inch array at 16' away:
About 6 years ago, Werewolf posted an excellent tutorial on the math behind the ideal (infinite) line array, and these ray-tracing models for large arrays are consistent with the math model. But you need to use one of the ray-tracing models for shorter arrays to see where the roll-off will start and to view the amount of lobing on the vertical axis. The relatively short array in the OP will have roll-off starting around 2500Hz, and it will have some lobing in the vertical response, but the vertical response doesn't appear to be too "rough". The roll-off at higher frequencies is enough that it will need to be addressed with EQ, but the vertical lobing is probably not going to be an "issue" for most fixed listening positions.
If you look at Werewolf's conclusions about the line array response, you will see what he finds most interesting about the line array: ..."as we increase distance from the Infinite Line Source, the response falls off at only 3dB for every doubling of distance". He notes that 3dB attenuation with distance is "a HUGE plus" (his emphasis) for stereo listening, as it will result in a wider sweet spot. Unfortunately, I don't know how to model sweet spots 🙁
Hi,
It's actually fun to imagine what makes the roll-off on highs, because it's not technically roll-off off highs but roll-on off the lows 😀 The lows sum constructively due to long wavelength and sources at different distance from observation point being more in phase, within say 90deg, until the slope starts. The shorter the wavelength gets more and more sound from all the sources just are in various phases and do not interfere only constructively but in with all kinds of phase relationships and some of it cancels, power reduces.
If this was not a full height floor to ceiling array to make "infinite image sources" in room, or if it was in anechoic environment, the response would look yet different if observation height was near the array end. Compared to near center of the array observation point the furthest sound source would be further away, basically making the roll off start bit lower in frequency again, just like with shrinking listening distance the furthers sources get further compared to the closest one at ear height.
I have no line sources at home so cannot do listening tests, but some of most memorable sounds I've heard have been live sound with line arrays. Line arrays outside can have the kind of woosh-woosh sound to it, if there is wind for example, which could change phase relationship of various parts of the line making the interference change, which makes it audible. Inside a room there should not be such thing I assume. Those who have arrays know better than me, just speculating.
It's actually fun to imagine what makes the roll-off on highs, because it's not technically roll-off off highs but roll-on off the lows 😀 The lows sum constructively due to long wavelength and sources at different distance from observation point being more in phase, within say 90deg, until the slope starts. The shorter the wavelength gets more and more sound from all the sources just are in various phases and do not interfere only constructively but in with all kinds of phase relationships and some of it cancels, power reduces.
If this was not a full height floor to ceiling array to make "infinite image sources" in room, or if it was in anechoic environment, the response would look yet different if observation height was near the array end. Compared to near center of the array observation point the furthest sound source would be further away, basically making the roll off start bit lower in frequency again, just like with shrinking listening distance the furthers sources get further compared to the closest one at ear height.
I have no line sources at home so cannot do listening tests, but some of most memorable sounds I've heard have been live sound with line arrays. Line arrays outside can have the kind of woosh-woosh sound to it, if there is wind for example, which could change phase relationship of various parts of the line making the interference change, which makes it audible. Inside a room there should not be such thing I assume. Those who have arrays know better than me, just speculating.
Last edited:
The roll-off of a line is basically same phenomenom as the effect of a single driver's diameter. Arrival time difference (-> phase which rotates easily at high Freq) from different spots on the radiator surface. Infinite number of virtual point sources would make ripples (interferences) disappear.
One can simulate this with Edge by changing the mumber of "Source density" when using a rectangular line driver. or by using different size and number of drivers.
In real life application in a room with reflective boundaries lots of this theoretical simulation will get corrupted...
One can simulate this with Edge by changing the mumber of "Source density" when using a rectangular line driver. or by using different size and number of drivers.
In real life application in a room with reflective boundaries lots of this theoretical simulation will get corrupted...
Actually it would be really nice to see how diiferent a line's room response is compared to "normal" speakers... A line's horizontal dispersion is not so different...
https://trueaudio.com/array/
https://www.audiosciencereview.com/forum/index.php?threads/murphy-corner-line-array.25090/
https://trueaudio.com/array/
https://www.audiosciencereview.com/forum/index.php?threads/murphy-corner-line-array.25090/
Just thinking about measuring cases of drivers to match them into groups, the sheer amount of wiring needed, and making all of those holes is triggering PTSD. They certainly look nice.
My biggest concern with this approach would be getting sufficient vertical dispersion, but I'm one that wants reasonable frequency response for TV use whether I'm standing, sitting, lying in the floor a few feet from the speakers, etc.
Yeah, I long ago reached the point where I 'needed' corner systems that were acoustically centered at > ~71% baffle height.
Not sure that I agree with this, unless I haven't understood it..
So a tweeter with a flat response isn’t exactly ideal….although a simple contour filter would be relatively easy
But then there’s a stacked planar solution as well
Response above 13k being ignored and still crossing over at 3k or so, a ‘short’ line of 4 elements would flatten this native rising response. Each planar is 200mm so this can work for the design……
But then there’s a stacked planar solution as well
Response above 13k being ignored and still crossing over at 3k or so, a ‘short’ line of 4 elements would flatten this native rising response. Each planar is 200mm so this can work for the design……
I did extensive line array simulations in Vituix using diffraction models for the drivers. These are easy to do.
Dayton makes a nice family of full range drivers suitable for use from 1.5" up. A line of the 1.5s next to a line of 5" woofers would likely do it. OTOH many have been satisfied with a single line of TC9s, which your concern about Horizontal dispersion argues against.
Wesayso and I found that the vertical listening angle can be extended with shading in groups of 4-6 drivers for example. But TV implies seated listening only so you are good so long as you measure and equalize at ear height.
Combing becomes a non-issue if you are listening from far enough away as the actual path length difference isn't the CTC of individual drivers but the difference between the hypotenuse of triangles, one leg of which is the CTC but the other leg is the direct path length.
Dayton makes a nice family of full range drivers suitable for use from 1.5" up. A line of the 1.5s next to a line of 5" woofers would likely do it. OTOH many have been satisfied with a single line of TC9s, which your concern about Horizontal dispersion argues against.
Wesayso and I found that the vertical listening angle can be extended with shading in groups of 4-6 drivers for example. But TV implies seated listening only so you are good so long as you measure and equalize at ear height.
Combing becomes a non-issue if you are listening from far enough away as the actual path length difference isn't the CTC of individual drivers but the difference between the hypotenuse of triangles, one leg of which is the CTC but the other leg is the direct path length.