Hi Gang,
I'm afraid I'm about to ask a question that I may not be prepared for the technical responses to, so bear with me.
I have been thinking (but not tinkering) a lot with the idea of line arrays. In part from Tekton speakers, in part from reading Roger Rusell's page on column arrays:
My Experience with Column Systems
To restate Roger's point of view, the per-driver lobing and comb filtering is not an issue with the impulse response of a speaker, and that these effects are only audible with steady state signals.
If that is true, does that not also imply that the idea of controlled dispersion due to having an effectively very large vertical driver is fake? In other words, does the controlled directivity of an array only work during steady-state testing, but not for real music?
Gaaaaaaaah! My head now hurts.
I'm afraid I'm about to ask a question that I may not be prepared for the technical responses to, so bear with me.
I have been thinking (but not tinkering) a lot with the idea of line arrays. In part from Tekton speakers, in part from reading Roger Rusell's page on column arrays:
My Experience with Column Systems
To restate Roger's point of view, the per-driver lobing and comb filtering is not an issue with the impulse response of a speaker, and that these effects are only audible with steady state signals.
If that is true, does that not also imply that the idea of controlled dispersion due to having an effectively very large vertical driver is fake? In other words, does the controlled directivity of an array only work during steady-state testing, but not for real music?
Gaaaaaaaah! My head now hurts.
I don’t see it as a myth since I actually listen to a floor to ceiling three way line array. But you may get a disagreement from people who have lots of education and don’t actually listen to line arrays, but actually to point source speakers.
Hi Zarathu,
I actually am a big fan of line arrays, in terms of listening. I'm just trying to reconcile different bits of data that are not fitting well together.
I think lots of us have seen Keele's pictures:
I'm just wondering, is Russel right? If he's right, then these pictures and testing are pretty meaningless and we should be attempting to simulate the impulse response, not steady state power. If that is true, do we get directivity? If not, how do phase array systems work?
See, I'm loosing it, much like a robot in Asimov's fiction.
Erik
I actually am a big fan of line arrays, in terms of listening. I'm just trying to reconcile different bits of data that are not fitting well together.
I think lots of us have seen Keele's pictures:
I'm just wondering, is Russel right? If he's right, then these pictures and testing are pretty meaningless and we should be attempting to simulate the impulse response, not steady state power. If that is true, do we get directivity? If not, how do phase array systems work?
See, I'm loosing it, much like a robot in Asimov's fiction.
Erik
Russell, me, Jim Griffin, and Don b Keele, are all in our late 60’s. None of us can hear anything above 14kz, if that.
I can only tell you that I have a straight floor to ceiling line array. I has a line of 17- 3.5 inch mids crossed at 2600 to a 30 inch line of dome tweeters with a c-to-c of .85 inches.
I don’t hear any combing at any frequency I can hear.
I am building a new system for down stairs with 25-2.5 inch mids at about 66 inches which is more than 70% of the floor to ceiling at that point. it crosses to a 40 inch line of 3 inch planars at 5000 hz. The mids cross to the mid woof at 500, and the mid woof to the sub woofer at 100. Its a 4 way electronically crossed 24 db L-R active crossover.
The tweeters will be about 1.90 inches from the mids horizontally, and the centers of the teeter line and the mids will be within 25mm of the same plane.
I don’t expect to hear any comb filter distortion, time delay, or phase issues.
Ask me again in September if I do. Building a line array is a long long exercise in time, especially when you enclose each of 50 midranges in its own PVC tube and fill them with 4lb cu ft pink fiberglass.
But the guys with lots of education, and an agenda will most surely disagree with me---and then launch into mathematics and physics that I don’t understand to prove it. Which, for me, would work just as well if they launched in using mandarin Chinese to explain it.
I can only tell you that I have a straight floor to ceiling line array. I has a line of 17- 3.5 inch mids crossed at 2600 to a 30 inch line of dome tweeters with a c-to-c of .85 inches.
I don’t hear any combing at any frequency I can hear.
I am building a new system for down stairs with 25-2.5 inch mids at about 66 inches which is more than 70% of the floor to ceiling at that point. it crosses to a 40 inch line of 3 inch planars at 5000 hz. The mids cross to the mid woof at 500, and the mid woof to the sub woofer at 100. Its a 4 way electronically crossed 24 db L-R active crossover.
The tweeters will be about 1.90 inches from the mids horizontally, and the centers of the teeter line and the mids will be within 25mm of the same plane.
I don’t expect to hear any comb filter distortion, time delay, or phase issues.
Ask me again in September if I do. Building a line array is a long long exercise in time, especially when you enclose each of 50 midranges in its own PVC tube and fill them with 4lb cu ft pink fiberglass.
But the guys with lots of education, and an agenda will most surely disagree with me---and then launch into mathematics and physics that I don’t understand to prove it. Which, for me, would work just as well if they launched in using mandarin Chinese to explain it.
As I understand it, the physics of a line array are similar to phase controlled arrays such as the B&O Beolab 90.
So here's my discontinuity. If Roger is right, then the Beolab's phase control of beaming should not work, but it does. If it works, then why can't we hear significant lobing issues from line arrays?
I'm in a mess all my own.
Of course, I do not wish to detract from your own enjoyment of building and making them. I'm just thinking of two different physical models of operation that don't seem to overlap.
So here's my discontinuity. If Roger is right, then the Beolab's phase control of beaming should not work, but it does. If it works, then why can't we hear significant lobing issues from line arrays?
I'm in a mess all my own.
Of course, I do not wish to detract from your own enjoyment of building and making them. I'm just thinking of two different physical models of operation that don't seem to overlap.
Yes and No,
The reasons you see the lobing of the strait array in the Keele presentation is it is a limited height array in an infinite free space. If you have a floor and ceiling reflection of a full height array you do not have the interference shown. The Beolab speakers are controlling both the frequency response and timing over the frequency range of the drivers. This is the equivalence of both variable power shading and driver position shading. This can do things a CBT cannot.
The reasons you see the lobing of the strait array in the Keele presentation is it is a limited height array in an infinite free space. If you have a floor and ceiling reflection of a full height array you do not have the interference shown. The Beolab speakers are controlling both the frequency response and timing over the frequency range of the drivers. This is the equivalence of both variable power shading and driver position shading. This can do things a CBT cannot.
So you are saying that a full floor to ceiling straight line array placed (against the walls or in the corners?) sounds and measures just like a curved one?
And what is the scale of the integration? Do the beams of the individual drivers form a wide and stable dispersion a few feet from the source or do they need a couple of dozens of feet to sound as they should? It is my understanding that line arrays are great for large venues.
And what is the scale of the integration? Do the beams of the individual drivers form a wide and stable dispersion a few feet from the source or do they need a couple of dozens of feet to sound as they should? It is my understanding that line arrays are great for large venues.
Exactly, in that picture from Keele, you see a 1m finite array being compared to half a CBT array which uses the floor reflection to form the complete beam.
I'm less of a fan of finite arrays, but love a floor to ceiling array as a concept. For a finite array, the concept from Keele works way better.
Did you read this thread: Infinite Line Source: analysis
Even if you miss all the math in this thread, it still explains a lot about the why of a floor to ceiling array as a concept. Worth a read, even if you skip the math parts.
Both ra7 and my thread are full of measured examples, raw measurements as well as those using DSP. What we use is quite different from the fixed EQ of Roger's concept, giving us the tools to adjust it to the room.
Whatever you choose, it is wise to treat first reflection points (all parallel planes to arrays are potential problems, however a lot of other problems are lessened by the multiple drivers).
The floor and ceiling both help make the array "seem" larger than it is. This will lessen the effect of floor bounce (if you can find it at all in measurements).
The left to right directivity for the most part will be similar to that of the driver used if you use a full range driver. If you use a multi-way array the crossover used and line position (e.g. side by side) will further determine the horizontal polars.
Look for the thread by Patrick Bateman to get an idea how the Beolab works...
With even more drivers you could make line arrays that use the Beolab technique for horizontal beam steering.
That's why I recommended treatment of first reflection points on the wall. The floor to ceiling array itself does not control the horizontal polars, the choice of driver determines that.
If you do treat first reflections you'll be able to get pretty amazing results. Even without it the lines are a lot of fun to listen to (but a bit more wild in perception).
With the wall treatment in place (on first reflections) and proper DSP it really gets to be fun! The wall treatment is way easier to accomplish than you'd think. All in all it took me 3 panels in my living room to reduce them far enough. Sure it depends on panel thickness and materials, measurements were my guide. Lots of those in my thread.
By the way, I do have ears and use them too. However with better measurements came improved results. Sometimes it is a tough puzzle, as I demand it sounds better on all material I throw at it, not just "some recordings". I live with these speakers, I didn't build them to have the soundtrack of my life excluded from listening.
I'm less of a fan of finite arrays, but love a floor to ceiling array as a concept. For a finite array, the concept from Keele works way better.
Did you read this thread: Infinite Line Source: analysis
Even if you miss all the math in this thread, it still explains a lot about the why of a floor to ceiling array as a concept. Worth a read, even if you skip the math parts.
Both ra7 and my thread are full of measured examples, raw measurements as well as those using DSP. What we use is quite different from the fixed EQ of Roger's concept, giving us the tools to adjust it to the room.
Whatever you choose, it is wise to treat first reflection points (all parallel planes to arrays are potential problems, however a lot of other problems are lessened by the multiple drivers).
The floor and ceiling both help make the array "seem" larger than it is. This will lessen the effect of floor bounce (if you can find it at all in measurements).
The left to right directivity for the most part will be similar to that of the driver used if you use a full range driver. If you use a multi-way array the crossover used and line position (e.g. side by side) will further determine the horizontal polars.
Look for the thread by Patrick Bateman to get an idea how the Beolab works...
With even more drivers you could make line arrays that use the Beolab technique for horizontal beam steering.
That's why I recommended treatment of first reflection points on the wall. The floor to ceiling array itself does not control the horizontal polars, the choice of driver determines that.
If you do treat first reflections you'll be able to get pretty amazing results. Even without it the lines are a lot of fun to listen to (but a bit more wild in perception).
With the wall treatment in place (on first reflections) and proper DSP it really gets to be fun! The wall treatment is way easier to accomplish than you'd think. All in all it took me 3 panels in my living room to reduce them far enough. Sure it depends on panel thickness and materials, measurements were my guide. Lots of those in my thread
.By the way, I do have ears and use them too. However with better measurements came improved results. Sometimes it is a tough puzzle, as I demand it sounds better on all material I throw at it, not just "some recordings". I live with these speakers, I didn't build them to have the soundtrack of my life excluded from listening
.
Last edited:
No, see the thread I linked in my prior post. These straight arrays all need EQ, but one can get a very good impulse response out of them. I haven't seen that many CBT impulses to compare though.
Here's one of my impulses as measured at the listening spot with DSP applied:
Just for fun, measure at your own listening spot and compare the results.
Much more measurements and different views in my thread, don't want to pollute this thread with it.
As a concept I love the floor to ceiling full range line array in a (relatively small) living room. It can mimic a much taller array that way by using the floor and ceiling reflections.
You can get to within one meter distance and have no obvious degradation, especially on regular program material. You'd need more specific signals to hear any ill effects at that distance.
The ones in Pro sound are used for different reasons.
Last edited:
Erik, don't know if it helps. but you can do a model of a line array in XSim (3D version), I think there's even one in the example files if you want to play with it.
The only difference between steady state and non-steady state is how thin the spectral lines are -- the longer the state is steady, the thinner the line and the more it is possible for it to all be down in comb-notch if it sits in the right place. Transients and momentary notes cover some bandwidth so they aren't likely to be bothered as much by some narrow notches.
As far as comb-notch audibility, we are surrounded by them all the time, every reflection in the room makes some form of a comb notch. Turn off the windowing and smoothing of a measurement system and measure speaker response back where you sit and you'll see lots. Of course related delays are different, but saying comb notches are all audible, all not audible, all good, all bad, is about as meaningful as saying that foods are tasteless. I think we'd have to define things a bit more and test with variations.
The only difference between steady state and non-steady state is how thin the spectral lines are -- the longer the state is steady, the thinner the line and the more it is possible for it to all be down in comb-notch if it sits in the right place. Transients and momentary notes cover some bandwidth so they aren't likely to be bothered as much by some narrow notches.
As far as comb-notch audibility, we are surrounded by them all the time, every reflection in the room makes some form of a comb notch. Turn off the windowing and smoothing of a measurement system and measure speaker response back where you sit and you'll see lots. Of course related delays are different, but saying comb notches are all audible, all not audible, all good, all bad, is about as meaningful as saying that foods are tasteless. I think we'd have to define things a bit more and test with variations.
If you listen to a highly directive speaker off axis you hear very little, regardless of whether the musical content is static or dynamic. Where the polar response shows lobes at some frequencies at angles away from the listening position, the content isn't arriving all within the same cycle. Thus the dynamic response may be of lower magnitude, but longer duration than one might expect. I don't think that renders it inconsequential.
You "could" EQ out every trace of comb filtering at a single spot, if you use REW, the early waterfalls are the best place to look for the combing effect.
Here is an early waterfall plot of an EQ-ed line at ~3 meter:
However, this would only be right at that exact spot of the measurement. This is not the kind of EQ I use in real life. Just an example that it can be done. I was curious enough to find out.
Here's a Dirac pulse with a rough estimate of the same band pass and room curve applied:
At a listening distance there are drivers that sum favourable and those that start to eat away which creates the combing pattern. The more drivers you have, the more the results end up like very close spaced dips and peaks and moves up higher in frequency.
The first signs of that combing with my particular arrays start above 5-6 kHz at that distance.
Driver distances to the ear at an average listening distance of 3m.
Move further back and that combing will move further up in frequency. Get closer driver spacing and you may soften this effect, but only up to a point. As you'd still have drivers higher up the line that will interfere with the driver at ear height, within the combined result, you'd just have more of them that averages out the results.
The average SPL still shows a straight line despite the combing, depending on the smoothing used. I've shown everything from 1/3 octave to 1/48 octave smoothing in my thread. It is much harder to find the same combing in such a measurement. So if you want to find it, go look at early waterfall plots. Look at your current favourite speaker too, at the listening spot, to see what your room does as far as combing goes, created by reflections etc.
For some initial settings use something like:
We, as humans have two ears, spaced apart. That alone creates it's own combing effect that we are very used to and live with in our day to day life. It is all part of how we perceive sound in the real world every day. And as Bill mentioned, every reflection out in the room will also create its own combing patterns. With very different audible results depending on the delay of that reflection.
A lot of what we do hear is in fact determined by our room. Our brain has excellent skills to make us forget about that room though.
For way more info on that last part, the spacing of our ears, look at cross talk cancelation etc.
Disclaimer: to get the results I show here I have used absorbing panels to get it. No DSP is going to clean up the room like this, the DSP I use is *mostly* used to counter the array effects, not the room response. That 'room response' is better fixed acoustically.
* = there's no escaping the room at lower frequencies.
Here is an early waterfall plot of an EQ-ed line at ~3 meter:
However, this would only be right at that exact spot of the measurement. This is not the kind of EQ I use in real life. Just an example that it can be done. I was curious enough to find out.
Here's a Dirac pulse with a rough estimate of the same band pass and room curve applied:
At a listening distance there are drivers that sum favourable and those that start to eat away which creates the combing pattern. The more drivers you have, the more the results end up like very close spaced dips and peaks and moves up higher in frequency.
The first signs of that combing with my particular arrays start above 5-6 kHz at that distance.
Driver distances to the ear at an average listening distance of 3m.
Move further back and that combing will move further up in frequency. Get closer driver spacing and you may soften this effect, but only up to a point. As you'd still have drivers higher up the line that will interfere with the driver at ear height, within the combined result, you'd just have more of them that averages out the results.
The average SPL still shows a straight line despite the combing, depending on the smoothing used. I've shown everything from 1/3 octave to 1/48 octave smoothing in my thread. It is much harder to find the same combing in such a measurement. So if you want to find it, go look at early waterfall plots. Look at your current favourite speaker too, at the listening spot, to see what your room does as far as combing goes, created by reflections etc.
For some initial settings use something like:
We, as humans have two ears, spaced apart. That alone creates it's own combing effect that we are very used to and live with in our day to day life. It is all part of how we perceive sound in the real world every day. And as Bill mentioned, every reflection out in the room will also create its own combing patterns. With very different audible results depending on the delay of that reflection.
A lot of what we do hear is in fact determined by our room. Our brain has excellent skills to make us forget about that room though.
For way more info on that last part, the spacing of our ears, look at cross talk cancelation etc.
Disclaimer: to get the results I show here I have used absorbing panels to get it. No DSP is going to clean up the room like this, the DSP I use is *mostly* used to counter the array effects, not the room response. That 'room response' is better fixed acoustically.
* = there's no escaping the room at lower frequencies.
Last edited:
Don't take my word for it, just try and figure out why I said what I did .
Take a piece of paper and start to draw some lines...
I'm not denying driver spacing has an influence. But eventually it will be dependent on the distances of each of the drivers to your ear (or measurement point). Some drivers will sum, others will subtract... even if you use very small drivers. Closer driver spacing (and as a result there will be more of them) will make it a more gradual process. Much like anti-aliasing.
Look at this example that I typed out in another thread...
The comb filter pattern you get from an array at the listener's position will always be related to driver spacing and distance from that array and the (perceived) length of that array.
Where the combing starts (at what frequency) isn't as clear cut as just assuming it's the frequency based on quarter wave distance from center to center driver spacing. That's all I'm saying. That point where the combing starts to interfere in frequency will depend on more factors than driver spacing alone.
Take a focussed array, even though it has driver spacing it won't exhibit a combing pattern when you measure it in it's exact focus point. However, move out of that focus point and you'll see the comb patterns.
.Take a piece of paper and start to draw some lines...
I'm not denying driver spacing has an influence. But eventually it will be dependent on the distances of each of the drivers to your ear (or measurement point). Some drivers will sum, others will subtract... even if you use very small drivers. Closer driver spacing (and as a result there will be more of them) will make it a more gradual process. Much like anti-aliasing.
Look at this example that I typed out in another thread...
The comb filter pattern you get from an array at the listener's position will always be related to driver spacing and distance from that array and the (perceived) length of that array.
Where the combing starts (at what frequency) isn't as clear cut as just assuming it's the frequency based on quarter wave distance from center to center driver spacing. That's all I'm saying. That point where the combing starts to interfere in frequency will depend on more factors than driver spacing alone.
Take a focussed array, even though it has driver spacing it won't exhibit a combing pattern when you measure it in it's exact focus point. However, move out of that focus point and you'll see the comb patterns.
As I understand it, the physics of a line array are similar to phase controlled arrays such as the B&O Beolab 90.
So here's my discontinuity. If Roger is right, then the Beolab's phase control of beaming should not work, but it does. If it works, then why can't we hear significant lobing issues from line arrays?
I'm in a mess all my own.
Of course, I do not wish to detract from your own enjoyment of building and making them. I'm just thinking of two different physical models of operation that don't seem to overlap.
You are mixing up objective facts with subjective evaluations. Line arrays, will lobe, that's an objective fact, but many people claim not to hear it, that's a subjective evaluation. There is no contradiction since they are both saying different things.
For many aspects of sound reproduction we have studied and correlated the effects to the perceptions. The differences between point sources and line sources has never been quantified to my knowledge so there simply isn't an answer.
I prefer point sources with controlled directivity. That seems to be what most manufacturers prefer - i.e. JBL. Could line sources be just as good or better - it's possible, but I haven't seen that happen yet.
I can't quite make out the scale - is it in milliseconds?
If so, it appears that your design has minimized later significantly discrete reflections that will occur in a room with a point source by spreading the reflections across the time scale -> all the way to the initial impulse. A point source will have a period of dead calm following the initial impulse, which your plot does not show.
Which way is better is an open question. Love to test that some day.
Spaced horizontally, and I am not sure that I would call it comb filtering since the brain processes the ear signals independently, which is how we hear image location. So true "comb filtering" is not going on between our two ears.
Lidia and my study of image perception is showing that delayed signals, i.e. comb filtering, have a substantial effect on both coloration and image quality. So we can't just say that comb filtering is OK - some is some isn't.
@gedlee I did not want to imply we get a comb pattern in our head.
However, our left ear does not hear the left speaker exclusively, it will also pick up part of the right speaker. A delayed part minus the head shading.
Same goes for the right ear. Only listening with one ear dead center would mimic what most microphones pick up.
That delayed signal from the opposite channel in itself creates something very similar to a comb pattern. One should be able to hear that if you play tone signals of different frequencies around ~1000 Hz up to ~5500 Hz. Play tones and slowly increase the frequency and you'll notice some tones appear to come from dead center, while others are a bit harder to place exactly.
That's where the sound from a stereo setup (with a phantom or mono signal) would differ in perception from a speaker playing from dead center.
However, our left ear does not hear the left speaker exclusively, it will also pick up part of the right speaker. A delayed part minus the head shading.
Same goes for the right ear. Only listening with one ear dead center would mimic what most microphones pick up.
That delayed signal from the opposite channel in itself creates something very similar to a comb pattern. One should be able to hear that if you play tone signals of different frequencies around ~1000 Hz up to ~5500 Hz. Play tones and slowly increase the frequency and you'll notice some tones appear to come from dead center, while others are a bit harder to place exactly.
That's where the sound from a stereo setup (with a phantom or mono signal) would differ in perception from a speaker playing from dead center.