Any thoughts on the benefits of curving arrays?
I keep wasting hours and hours trying to come up with a sim of a CBT array in VituixCad that doesn't look like crap, but every time I try to do it with 64mm full range drivers, the performance above 1khz is terrible, no matter how long the line is.
The only thing that seems to improve the performace of the array a lot is to be very far away, like four meters or more. But I live in a 4500 square foot house, and despite the very large size, the furthest I can sit from the speakers is about two meters. (Like a lot of newer homes, there's a lot of rooms, but the actual size of the rooms isn't much different than a 2000' 3br/2ba home.)
I keep wasting hours and hours trying to come up with a sim of a CBT array in VituixCad that doesn't look like crap, but every time I try to do it with 64mm full range drivers, the performance above 1khz is terrible, no matter how long the line is.
The only thing that seems to improve the performace of the array a lot is to be very far away, like four meters or more. But I live in a 4500 square foot house, and despite the very large size, the furthest I can sit from the speakers is about two meters. (Like a lot of newer homes, there's a lot of rooms, but the actual size of the rooms isn't much different than a 2000' 3br/2ba home.)
Those 3D surface charts really help illustrate the response. You can create them in Excel, and there is support for 3d viewports in .NET WPF applications, but they aren't available in .NET Forms, unless you buy a scientific charting component from Nevron, Telerik, or some others, with price tags in the $1000 range. I had given up on ever being able to do 3D charting for .NET Forms a while back, but I just found Chart Director. It looks promising, and it is only $99. If it works out, I might try modeling the CBT.
I enjoy focused arrays......................
https://www.diyaudio.com/community/threads/mr-bates-your-focused-arrays.221676/page-2
but that has its own set of benefits / drawbacks.
https://www.diyaudio.com/community/threads/mr-bates-your-focused-arrays.221676/page-2
but that has its own set of benefits / drawbacks.
There is some great content from "evilnui" in this thread here, that relates to Dave's thread:
https://www.diyaudio.com/community/threads/software-to-predict-directivity.313575/
In particular, some great explanations of weighting that don't go too far into the weeds with the math
https://www.diyaudio.com/community/threads/software-to-predict-directivity.313575/
In particular, some great explanations of weighting that don't go too far into the weeds with the math
I found the software in The Wayback Machine and attached it. I did a virus scan using Windows Defender before uploading it.
Here's a description of the software:
"The Vertical Polar Response: Line Array is a design estimation tool to predict Vertical Dispersion, Polar Patterns and resulting Frequency Response in discreet driver grouping Line Arrays and Bessel Arrays. Program can be used to predict long, variably sized, arrays of Circular Piston and Rectangular Drivers in the Near and Far Fields with different inter-driver spacing. Saves 2D Response and 3D Frequency/Angle surface pictures in GIF format, saves Frequency Responses in FRD File format.
Current Version 0.40 Date: Feb 24th 2002
Copyright - Yavuz Aksan 2001, 2002"
Attachments
Doesn't this depend on implementation? With DSP, you can easily get very large delays. These boards each have 24 channels of DSP (3 ADAU1701 chips) with 20 small class D amps (the woofer amps are separate).
Hi Neil
It’s not an issue of how much delay is needed but in what “direction”. Electrical delay is much like pushing the elements away from the observer. The problem is that as you rotate the array the delayed elements are still effectively pushed away from the observer.
Imagine an arc with the ends recessed. From the front it is convex. Turn it 180 degrees and it becomes a concave arc. With the case of electrical delay and 180 turning it becomes convex again. The ends are always pushed away.
For small rotations the differences are small.
It’s not an issue of how much delay is needed but in what “direction”. Electrical delay is much like pushing the elements away from the observer. The problem is that as you rotate the array the delayed elements are still effectively pushed away from the observer.
Imagine an arc with the ends recessed. From the front it is convex. Turn it 180 degrees and it becomes a concave arc. With the case of electrical delay and 180 turning it becomes convex again. The ends are always pushed away.
For small rotations the differences are small.
You can delay the drivers on the end to implement a convex array or delay the drivers in the middle to have a concave array. It's the relationship of those delays that determine the direction.
I can change the effective curvature on my line arrays using a Bluetooth app that manages those delay relationships, and the maximum delay I use is about 16". It makes a huge difference in the sound quality, but it's hard to make sense of the comparison because the frequency response changes as the drivers become "focused", and I don't have any provision to change the response with curvature right now. But that's one of the reasons I might try modeling them--to better understand those interactions. I had given up on my old Active Speaker Design code with the 2D charts, but I may try one more variant.
I can change the effective curvature on my line arrays using a Bluetooth app that manages those delay relationships, and the maximum delay I use is about 16". It makes a huge difference in the sound quality, but it's hard to make sense of the comparison because the frequency response changes as the drivers become "focused", and I don't have any provision to change the response with curvature right now. But that's one of the reasons I might try modeling them--to better understand those interactions. I had given up on my old Active Speaker Design code with the 2D charts, but I may try one more variant.
Here's a 3D sim of a straight array modeled using Aksan's VPR spreadsheet.
Both sims feature twenty five 64mm drivers with a spacing of 6mm between each; the only difference is the distance of the sim. One sim is at ten meters and one sim is at one meter.
It's interesting to see how much the output narrows as you get further and further from the array.
Also, this sim seems to make an interesting argument for SHORTER not LONGER arrays. IE, the reason that the response of the array is so chaotic up close is because the pathlengths from your ear to the ends of the array are so LONG. A convex array like the CBT only makes this worse (and may be why it sims so badly.) A concave (focused) array solves this issue, but only solves it at one distance.
Neil's idea of being able to set the focal point of the array might be where it's at. My fav speaker of all time is still the Beolab 90. I love being able to change how the loudspeaker sounds using an app. It's a game changer; if you have a recording that sounds crummy, you can go for the omni mode. If you have a very good recording, you can "focus" the loudspeaker.
Both sims feature twenty five 64mm drivers with a spacing of 6mm between each; the only difference is the distance of the sim. One sim is at ten meters and one sim is at one meter.
It's interesting to see how much the output narrows as you get further and further from the array.
Also, this sim seems to make an interesting argument for SHORTER not LONGER arrays. IE, the reason that the response of the array is so chaotic up close is because the pathlengths from your ear to the ends of the array are so LONG. A convex array like the CBT only makes this worse (and may be why it sims so badly.) A concave (focused) array solves this issue, but only solves it at one distance.
Neil's idea of being able to set the focal point of the array might be where it's at. My fav speaker of all time is still the Beolab 90. I love being able to change how the loudspeaker sounds using an app. It's a game changer; if you have a recording that sounds crummy, you can go for the omni mode. If you have a very good recording, you can "focus" the loudspeaker.
Attachments
I stayed up until 2am last night fiddling with this stuff. Some things that hit me:
1) This way of weighting things is a bit like the Snell "eXpanding Array" that you did, which is also similar to the Horbach Keele array. Basically the idea being that the speaker gets larger and larger as the frequencies get larger and larger. In a Snell eXpanding Array, the tweeter is the only element playing at 10khz, but as you drop down to 5khz, the mids start to come in (but very much attenuated) and then by 2500Hz the mids AND the tweeter are all playing at nearly the same volume. Of course, the behavior that I just described is true of just about every MTM. But what's interesting about the eXpanding Array (and the Horbach Keele array) is the rigorous attention to the slope of the filter and the impact of those filters on the off-axis response. I would argue that MTMs from the early 90s, like the Focal Aria 7, were basically optimized for on-axis response only, and that they had some nasty off-axis nulls because the woofers were too large.
2) To me, the REALLY tantalizing thing here, is that it just might be possible to make one of these arrays with ONE tweeter and an array of small woofers. That would be REALLY nice, because the hardest thing to deal with in a line array is the terrible response above 5khz. An obvious solution is to use a vertical line of tweeters, but that basically doubles the cost of the array, just to improve the performance above 5khz. Not really a great tradeoff, and the CBT36 array was kinda infamous for having a tweeter that wasn't great, because the array required something like a hundred and fourty four tweeters.
If I'm understanding David's post correctly, the "path" here would be to treat the array like an "array inside of an array" or even "an array inside of an array inside of an array."
Again, I may be doing this wrong (I'll know soon enough), but I think the path would look like this:
1) The first step is to do the high frequencies. That might be an array of five drivers
2) next step is the midrange frequencies. As shown above, that might be eleven drivers
3) last step is the low frequencies. As shown above, that would be twenty three drivers.
The "catch" is that the eleven drivers are a subset of the twenty three drivers, and the five drivers are a subset of the eleven drivers. IE, the filters would be constructed in such a way, that only ONE of the 23 drivers is playing full range, and there's a lowpass filter on the other twenty two drivers, so that they are "brought in" as frequencies go lower and lower and lower.
From a practical standpoint, if it's possible to put a tweeter in the center of this thing, that will make life a heck of a lot easier. Because the center element is probably the only one that should be playing to 20khz (I think - I may be wrong.) So if there's a full range as the center element, your maximum output will be severely curtailed because a typical 2.5" full range is capable of about 90-95dB total. So that would set a hard upper limit on the output of the entire array, unless you're willing to live with an array that doesn't have flat response. On the other hand, if you can stick a small tweeter in the center, you could probably use a compression driver or a ribbon to get the output up to 110dB or even higher.
The last thing - I believe it's not as simply as using a series of low pass filters, because there's a weighting to the elements.
Here's the 23 element array, but this time with the shading referenced in David's post here : https://www.diyaudio.com/community/goto/post?id=7116457
Note how compared with the flat array, the off axis response is better behaved
Maximum output is substantially lower of course, due to the shading
Note how compared with the flat array, the off axis response is better behaved
Maximum output is substantially lower of course, due to the shading
Attachments
There is practical difficulty to that approach in that the central driver needs to have a similar directivity vertically to the next drivers either side of it. The spacing of the pairs or groups of drivers either side generates more vertical directivity.
There are of course ways to do that, Dave's own XA90 waveguide for the tweeter, using a planar or other taller driver that has it's own increased vertical directivity or the Perlisten approach.
I think a tweeter in the center of a bunch of mids is a valid improvement. Knowing typical driver dimensions you can also put 3 tweets in the center and get the first crossover point center-to-center distances to a little a little less. Some automotive tweets are useful here if they have smaller mounting flanges than a typical box mounted tweeter. Of course you would want to use units that aren't messing up the response with poor mounting flange design. I found that putting a directivity flange on the tweeter was always useful to get its directivity a better match for the natural array directivity. We did that at Bose also for the TV soundbars. these are typically a 3 axis system with a left firing and a right firing (as well as center) array. You always put a fairly directional element on the left and right ends of the bar (at +-90 degrees) to achieve the top octaves directivity that you had trouble getting solely from the array.
I've seen some mentioning of single central tweeters limiting output capability of the system. I've never run into that. Using a central 3 tweeters can give more than enough excursion, if that is an issue. Otherwise, the thermal requirements are also adequate as long as your program material doesn't put out 0dBFS signal strength in the top Octave (not ever found in natural material). Not something to worry about.
I think we need to define 2 types of progressive or expanding array. Both have similarities along with their pluses and minuses.
Arbitrarily:
Type one uses all drivers of the same size, say a number of full range 4" units. We make them progressive by low pass filtering the outermost units so that no frequencies that would cause lobing are allowed. As frequency goes up we raise the cuttoff frequency so that effective length follows wavelength. The units are not filtered at the low end (except for possibly for the addition of subs) and achieve LF output capability from using all drivers together for low frequencies. Units can be linearly spaced or log space. Adding a center tweeter as suggested by Patrick would be a subclass of type one.
Type two would be a symmetrical array with progressively larger units as you progress from the center to the edge. i.e. the WMTMW types. Examples of such arrays would be the Horbach/Keele or my XA series. The main distinguisher here is that each element needs both high pass and low pass filtering. as you move from center to ends you progressively shift each unit's bandpass, ideally in proportion to its driver's center to center spacing. Driver sizes are scaled to suit the power handling needed for their respective frequency ranges.
I did some modeling at Bose (with their proprietary software, left behind...) that showed that both approaches can give near ideal constant directivity and lobe free performance.
Finally, as to fractal designs, I think that is a great way to describe them. I'm not sure who originated the term but I know I used it to describe the Snell systems, back in the day. It made sense that if you could come up with a formula that worked well for the central 3 elements, then you could think of the group of 3 as a single element that was in the middle of a second group of 3 (appropriately scaled), and so on, and so on.
Probably not truly a fractal but not a bad way to describe expanding arrays.
David
I've seen some mentioning of single central tweeters limiting output capability of the system. I've never run into that. Using a central 3 tweeters can give more than enough excursion, if that is an issue. Otherwise, the thermal requirements are also adequate as long as your program material doesn't put out 0dBFS signal strength in the top Octave (not ever found in natural material). Not something to worry about.
I think we need to define 2 types of progressive or expanding array. Both have similarities along with their pluses and minuses.
Arbitrarily:
Type one uses all drivers of the same size, say a number of full range 4" units. We make them progressive by low pass filtering the outermost units so that no frequencies that would cause lobing are allowed. As frequency goes up we raise the cuttoff frequency so that effective length follows wavelength. The units are not filtered at the low end (except for possibly for the addition of subs) and achieve LF output capability from using all drivers together for low frequencies. Units can be linearly spaced or log space. Adding a center tweeter as suggested by Patrick would be a subclass of type one.
Type two would be a symmetrical array with progressively larger units as you progress from the center to the edge. i.e. the WMTMW types. Examples of such arrays would be the Horbach/Keele or my XA series. The main distinguisher here is that each element needs both high pass and low pass filtering. as you move from center to ends you progressively shift each unit's bandpass, ideally in proportion to its driver's center to center spacing. Driver sizes are scaled to suit the power handling needed for their respective frequency ranges.
I did some modeling at Bose (with their proprietary software, left behind...) that showed that both approaches can give near ideal constant directivity and lobe free performance.
Finally, as to fractal designs, I think that is a great way to describe them. I'm not sure who originated the term but I know I used it to describe the Snell systems, back in the day. It made sense that if you could come up with a formula that worked well for the central 3 elements, then you could think of the group of 3 as a single element that was in the middle of a second group of 3 (appropriately scaled), and so on, and so on.
Probably not truly a fractal but not a bad way to describe expanding arrays.
David
Just starting to play with it. Very interesting. It seems to add simulated driver directivity which is an enhancement but makes it hard to compare to my results. Note that he uses power ratios to define element strength and I used pressure ratios. So if, for example, you wanted to try a Bessel array the strengths would be 0.25, 1.0 1.0, -1.0, 0.25 (try in 2D at 75 degrees).
Indeed very illustrative, thank You!
From the perspective of lowering of the level of vertical reflections the short Bessel array looks as useless as expected, and the weighting of the long array does seem not worthy of the effort it takes, doesn't it?
a compression driver tweeter on a waveguide in the center of a bunch of mids describes a synergy horn. The problem is that the synergy gets very crowded when you try to make it small enough to fit this design paradigm. Inexpensive small full range drivers have limited output. I've been dissuaded from looking small dome mids by their cost.
Just the waveguide itself with a 1.4" exit CD allows you to cross over at 650 Hz or so for home use. The ATH thread in this forum provides the tools to design and 3D print the waveguide. This is with 300 mm diameter waveguide and 6.5" woofers above and below, not an array but an MTM. How much better can we hope to do vertical directivity wise with vertical woofer array?
