@dynomike
Agree. The height of the array will determine the bandwidth in which directivity is controlled, i.e., the array acts like a line source emitting cylindrical wavefronts. Shorter arrays have difficulty maintaining the constancy of the beamwidth beyond certain frequencies. In this regard, floor-to-ceiling arrays achieve the best performance. The CBT gets around the problems of shorter arrays by using shading and delay to improve the performance. Its small form factor is definitely a plus, not something to be sniffed at.
Your comments about far field versus near field are so true. This is discussed in Dave Smith's original line array paper, which is an awesome read. Have you read it? It is here on diyaudio somewhere.
Juhazi,
I have The Edge but I could never get an accurate estimate of an array response with it. I might give it another go. Can it do mic positions in two dimensions? That is, can it correctly calculate response from a combination of drivers at a fixed distance away from the baffle? Its a been a while since I played with it, but doesn't it have some kind of nearfield/farfield switch? Maybe I have an older version.
Agree. The height of the array will determine the bandwidth in which directivity is controlled, i.e., the array acts like a line source emitting cylindrical wavefronts. Shorter arrays have difficulty maintaining the constancy of the beamwidth beyond certain frequencies. In this regard, floor-to-ceiling arrays achieve the best performance. The CBT gets around the problems of shorter arrays by using shading and delay to improve the performance. Its small form factor is definitely a plus, not something to be sniffed at.
Your comments about far field versus near field are so true. This is discussed in Dave Smith's original line array paper, which is an awesome read. Have you read it? It is here on diyaudio somewhere.
Juhazi,
I have The Edge but I could never get an accurate estimate of an array response with it. I might give it another go. Can it do mic positions in two dimensions? That is, can it correctly calculate response from a combination of drivers at a fixed distance away from the baffle? Its a been a while since I played with it, but doesn't it have some kind of nearfield/farfield switch? Maybe I have an older version.
Last edited:
ok.. you sold me on it!
nice hybrid lines there too. i like the waveguide tweeter for matching dir.
You can set the mic location in 3 dimensions using the mic distance input and moving the mic symbol with your mouse. Its a nice simple program
Yep, my procedure
- set baffle dimensions
- set driver count
- set driver shape and dimensions
- set mic distance
In graphic window
- move drivers to locations
- move mic to location
Save first measurement (toggle cloud)
Move mic - save measurement
- repeat
Change mic distance - save measurement
- repeat
Move drivers or remove/add drivers - save measurement
-repeat
etc. Change driver size, change baffle shape/size etc etc.
Very fun and educative! But we must remember that this is just simple simulation of measurement at a fixed point. Relevance to audible sound must be learned by also making respective prototypes and listening to them.
Room reflections and modes really do play significant role. Unfortunately in Edge we can't add floor, ceiling or walls. Also all drivers are in same z plane and in same phase.
ps. works in Win10 too! Sometimes it gets locked, then you must delete program file (just one .exe) and clean windows registrys with regedit. Then new installation!
- set baffle dimensions
- set driver count
- set driver shape and dimensions
- set mic distance
In graphic window
- move drivers to locations
- move mic to location
Save first measurement (toggle cloud)
Move mic - save measurement
- repeat
Change mic distance - save measurement
- repeat
Move drivers or remove/add drivers - save measurement
-repeat
etc. Change driver size, change baffle shape/size etc etc.
Very fun and educative! But we must remember that this is just simple simulation of measurement at a fixed point. Relevance to audible sound must be learned by also making respective prototypes and listening to them.
Room reflections and modes really do play significant role. Unfortunately in Edge we can't add floor, ceiling or walls. Also all drivers are in same z plane and in same phase.
ps. works in Win10 too! Sometimes it gets locked, then you must delete program file (just one .exe) and clean windows registrys with regedit. Then new installation!
Last edited:
Nice work on your arrays. I've designed and built many arrays and listened to other designs as well. With single driver / 1-way arrays I've always found the top end to be lacking so 2-ways are what I prefer. The power response also presents a problem, especially with narrow baffles and small drivers. You can equalize for a flat on-axis response but the horizontal off-axis tends to roll off too much. There are some smaller drivers that do a better job off-axis but still sound inferior to a standard dome or ribbon tweeter. The best 1-way arrays I've heard are CBT designs though you could possibly achieve similar results in a straight array with multiple amps, delay, and shading.
As far as the comb filtering and driver sizes, etc. you can easily test for the response effects if you have some time and several channels of amplification. If each channel has a level control you can even experiment with power tapering. A bonus would be having the ability to add delays. It would be interesting to see the results of a floor-to-ceiling array to determine how much impact there might be from adding shading and delays.
One thing to consider is that even if all of the drivers are working in phase the gain will never be uniform across the entire frequency range because of the off-axis response of the drivers that are farther away from ear level.
As far as the comb filtering and driver sizes, etc. you can easily test for the response effects if you have some time and several channels of amplification. If each channel has a level control you can even experiment with power tapering. A bonus would be having the ability to add delays. It would be interesting to see the results of a floor-to-ceiling array to determine how much impact there might be from adding shading and delays.
One thing to consider is that even if all of the drivers are working in phase the gain will never be uniform across the entire frequency range because of the off-axis response of the drivers that are farther away from ear level.
Last edited:
I'd love to see the IR and STEP responses (first 20 ms) of a full range CBT array as well as an early CSD plot (about 3 ms with a 0.1 ms rise time). I'm curious to see what it does there. Do we have any CBT arrays like that here on DIYAudio?
I'd like to see it from the 2 way CBT array as well. I've never seen those plots as far as I can recall. Preferably taken at the listening position in a reasonable/good room.
Due to changes in my baffle damping, using butyl rope, I'd rather not operate on my own arrays to find out. 🙂
I hope someone is willing to show that kind of plots from a CBT, be it a full range or 2 way design.
I cannot (and will not) deny the effects of comb filtering. But there's quite the difference in sound as well as measurements between an IIR corrected floor to ceiling line array and one corrected with (relative short window) FIR. I still see the comb filtering effects in the waterfall plots, but the difference in the waterfall plots is big between those two. Much bigger than the difference in the FR plot alone would indicate.
You can see some of the difference in the APL_TDA plots as well:
Correction with IIR to flat FR, even though the title of the jpg says "no cor"
compared to:
Corrected with FIR
The normalization in these plots removes the comb effects, but there is a clear difference in high frequency timing visible. For that to show up in a plot at those frequencies it is a huge timing difference, one that is easily heard. Both plots were made at the listening position with left and right speaker playing.
Some examples of early waterfall plots can be found here: http://www.diyaudio.com/forums/full-range/242171-making-two-towers-25-driver-full-range-line-array-137.html#post4437663
In all honesty, I did not dare post the waterfall plot showing the higher frequencies due to comb filtering showing there. I've said something along the lines: if you come in for a listen, I'll show the plots afterwards 🙂. I've looked at a lot of early CSD plots from other type of speakers to compare and while there is comb filtering clearly showing, much more clear in an early CSD than in any other plot, it's still pretty decent compared to other speakers. But a single tweeter is cleaner up there, until you get to the crossover point.
I did not want to start another comb filter debate on my thread so I posted the mid range plots only.
I'd like to see it from the 2 way CBT array as well. I've never seen those plots as far as I can recall. Preferably taken at the listening position in a reasonable/good room.
Due to changes in my baffle damping, using butyl rope, I'd rather not operate on my own arrays to find out. 🙂
I hope someone is willing to show that kind of plots from a CBT, be it a full range or 2 way design.
I cannot (and will not) deny the effects of comb filtering. But there's quite the difference in sound as well as measurements between an IIR corrected floor to ceiling line array and one corrected with (relative short window) FIR. I still see the comb filtering effects in the waterfall plots, but the difference in the waterfall plots is big between those two. Much bigger than the difference in the FR plot alone would indicate.
You can see some of the difference in the APL_TDA plots as well:
Correction with IIR to flat FR, even though the title of the jpg says "no cor"
compared to:
Corrected with FIR
The normalization in these plots removes the comb effects, but there is a clear difference in high frequency timing visible. For that to show up in a plot at those frequencies it is a huge timing difference, one that is easily heard. Both plots were made at the listening position with left and right speaker playing.
Some examples of early waterfall plots can be found here: http://www.diyaudio.com/forums/full-range/242171-making-two-towers-25-driver-full-range-line-array-137.html#post4437663
In all honesty, I did not dare post the waterfall plot showing the higher frequencies due to comb filtering showing there. I've said something along the lines: if you come in for a listen, I'll show the plots afterwards 🙂. I've looked at a lot of early CSD plots from other type of speakers to compare and while there is comb filtering clearly showing, much more clear in an early CSD than in any other plot, it's still pretty decent compared to other speakers. But a single tweeter is cleaner up there, until you get to the crossover point.
I did not want to start another comb filter debate on my thread so I posted the mid range plots only.
Last edited:
Thanks! You are lucky to have been exposed to so many arrays.
Why do narrow baffles / small drivers present horizontal off-axis problems? That is counter-intuitive to me. I was surprised at the well-behaved off-axis of the TC6 drivers, and certainly larger drivers will have a greater drop in power response at high frequencies.
Could you describe how such a test would be carried out?
This explains why the response from the line array of 2" drivers does not have the same satisfying treble as a single 2" driver (which I've usually found to be quite balanced once EQ'd - no problematic difference between on-axis and power response).
The smaller the driver, the less of a problem this would be. ie. for some tweeters like the 3/4" XT19, directivity is not a practical concern.
Sorry - it was late at night and I didn't do a very good job of explaining myself. The power response comment is more related to the step response of the narrow baffle and limited bass extension of the smaller drivers (2"-3" in diameter as well as some of the 3.5"-4" with higher Fs ratings. Some of the step loss is compensated for due to the inherent reduced coupling as you increase in frequency. But even after adding EQ to smooth out the response and to increase the bass extension these arrays always tend to sound "thin" and poorly balanced.
For testing all you need to do is take a response curve from a single driver and then add drivers while overlaying the response curves. This way you can observe the total amount of gain as well as where the phase cancellations occur. Do this at a few different distances and heights with the microphone. The amount of amplifier channels will depend on the number of drivers and how they are wired.
One problem with adding equalization is the possibility of trying to smooth out a dip that's due to a phase cancellation. That's why it's so important to measure at different angles.
For the people still getting hung up on driver size and CTC spacing:
24x75mm:
12x150mm:
Note that the dip-peak-dip-peak at 2.5, 3.5, 5, 6kHz is pretty much exactly the same between both arrays. I was not super careful with placing the drivers perfectly on the baffle so this explains some small differences in this area.
The top end (>8kHz) becomes a bit different because of baffle edge diffraction - changing the driver size changed the relationship between the width of the drivers and the width of the baffle. You'd experience the exact same problem when using a single driver in a small box so it's not an array consideration but just a speaker design consideration in general.
24x75mm:
12x150mm:
Note that the dip-peak-dip-peak at 2.5, 3.5, 5, 6kHz is pretty much exactly the same between both arrays. I was not super careful with placing the drivers perfectly on the baffle so this explains some small differences in this area.
The top end (>8kHz) becomes a bit different because of baffle edge diffraction - changing the driver size changed the relationship between the width of the drivers and the width of the baffle. You'd experience the exact same problem when using a single driver in a small box so it's not an array consideration but just a speaker design consideration in general.
Last edited:
Compelling evidence. I will play with The Edge and report back. Is it modeling the performance correctly? Observations, or rather simulations, are not matching theory. So, the theory needs to be changed to explain the simulation or we don't understand the modeling or there is an error in the simulation.
Last edited:
Edge simulates a finite line array which will beam and have high frequency rolloff. The theory behind a floor to ceiling array is that if the floor and ceiling are perfectly reflective acoustically and you include the image arrays (the ones you would see in the mirrors if the floor and ceiling were covered in mirrors), you have an infinite line array which generates a cylindrical wavefront, which does not have the high frequency rolloff. In reality, they are not perfectly reflective, but it is a good approximation.
Marc
Marc
Dennis Murhy's MCLA measurements are average of 4 measurements. Is there a reason why he doesn't show a single sweep measurement? I don't believe in cylindrical wavefront "theory", it is flawed.
https://www.trueaudio.com/array/MCLA_array_test_results.htm
https://www.trueaudio.com/array/MCLA_array_test_results.htm
I see and experience a lot of advantages, but even I don't buy the cylindrical wave front explanation. At least not for this full range type of array.
Yet I do not regret playing with this type of speaker, on the contrary. The Edge simulations are fun in itself but do not convince me, especially the comparison between the large and small drivers on the previous page.
No takers to show STEP and IR of a CBT? Seriously, I'm curious to see them. As well as the early waterfall plots.
Yet I do not regret playing with this type of speaker, on the contrary. The Edge simulations are fun in itself but do not convince me, especially the comparison between the large and small drivers on the previous page.
No takers to show STEP and IR of a CBT? Seriously, I'm curious to see them. As well as the early waterfall plots.
Edge and floor to ceiling arrays
I have never used Edge but it seems to me that if you want to model a FTC array you'll need to include the images by at least tripling the physical height. This will capture one floor and one ceiling reflection. If you can copy the physical array and paste it, then you can try more reflections easily?
I have never used Edge but it seems to me that if you want to model a FTC array you'll need to include the images by at least tripling the physical height. This will capture one floor and one ceiling reflection. If you can copy the physical array and paste it, then you can try more reflections easily?
Regarding the cylindrical wavefront theory. Maybe cylindrical wavefront is not a good way to think about it. Think of a single 15" woofer. At wavelengths approaching the dimensions of the driver, the sound becomes directional. The same thing happens when you make the driver very large in a particular direction. It becomes directional for wavelengths approaching its dimension. It's the same with vertical ribbons. Call it what you want, but the vertical directivity control is the same phenomenon that causes beaming in drivers.
Juhazi said,
"Dennis Murhy's MCLA measurements are average of 4 measurements. Is there a reason why he doesn't show a single sweep measurement? I don't believe in cylindrical wavefront "theory", it is flawed.
https://www.trueaudio.com/array/MCLA...st_results.htm"
Dennis Murphy is the famous crossover expert/speaker design (Salk Sound and Philharmonic Audio) while John Murphy is known for his trueaudio.com site and corner line array design.
Two different guys with different interests.
Jim
__________________
"Dennis Murhy's MCLA measurements are average of 4 measurements. Is there a reason why he doesn't show a single sweep measurement? I don't believe in cylindrical wavefront "theory", it is flawed.
https://www.trueaudio.com/array/MCLA...st_results.htm"
Dennis Murphy is the famous crossover expert/speaker design (Salk Sound and Philharmonic Audio) while John Murphy is known for his trueaudio.com site and corner line array design.
Two different guys with different interests.
Jim
__________________
I need to learn more about how you are measuring for the FIR correction. I did some auto-correction with IK ARC2, but it really fuzzed out the step response, like it was trying to correct for the line array itself. I didn't like the look or the sound of it.
Thanks for sharing your plots, which certainly look much better with FIR. I'm sure you have more information on your thread about how you did the correction, but a TL;DR version would be great! Specifically, I don't get how you are getting a clean impulse response from a line array at all...
- Status
- Not open for further replies.