Don't forget that ITD and IID need to both change in a similar manner so that we can get the largest and best sweet spot.
It's true that there are so many competing properties that things get really confusing.
It's true that there are so many competing properties that things get really confusing.
Tom Danley said:
We're keeping the horizontal width at 15" and raising the crossover frequency to where the 7.5" vertical retains control? Or are we now using a smaller 7.5" round or square symmetric 70° waveguide with that higher crossover?
Your 1.6:1 derivative of large performance spaces translates to 56.25° with a 90° horizontal; it's angels on a pin we're doing here. 18-Sound XT1086 is 80° x 60°, 1.33:1. Let's use that and pretend we hear a difference.
soongsc said:
Define these, please for the benefit of the readership.
Even if it only half-*** works, it's an improvement over head-in-a-vice listening, and for those who want to fully exploit it, defined directivity horns have been made previously.
Hi all, Zilchlab (I like that, it reminds me of the days of high voltage haha)
“We're keeping the horizontal width at 15" and raising the crossover frequency to where the 7.5" vertical retains control? Or are we now using a smaller 7.5" round or square symmetric 70° waveguide with that higher crossover?
With the 67 degree case, the height is the same but the width is half or equal to the height.
The onset of loss of pattern loss is lowered about a half octave because the governing angle and dimension has increased from 45 to 67 degrees while mouth size in that plane is the same and with equal angles in both planes, pattern flip at pattern loss is avoided entirely..
Also, when I say pattern flip, I mean that from a horn that appears to produce a wide pattern, it is narrow instead.
This exchange of pattern shape begins at the first (highest) frequency where pattern loss point is reached.
One can see this same effect in a single horn however depending how you look at it.
For work, our speakers are measured by a third party every 2.5 degrees over an imaginary sphere in an anechoic environment.. This is needed to produce an acoustic model that can be used in room design programs like EASE and others.
As a side benefit of this method, if you add a power capacity measurement to the data, provides a nice way to view the spec’s and radiation pattern.
http://www.danleysoundlabs.com/technical downloads.html
If you download the CLF data file for the SH-50 and the viewer you can see what I mean.
Look at the radiation balloon at the top right, grab it with the cursor and view the balloon head on like the speaker is facing you. Go to the frequency box right below and set it to 400Hz, the early part of pattern control. Step up to say 1-2 KHz, now the pattern takes on sort of a diamond shape, “flipped” relative to the horn shape.
Grab the balloon with the cursor and look at it from more of a side view.
Now, step up to say 4KHz or above and you see the horn takes on the expected pattern orientation.
If you go to the bottom right, you can select electro acoustic data, then the H or V -6dB points and see a directivity graph or click polar’s or one can click balloon spectra and then move the H and V cursor through the balloon and see the frequency response at any angle H and V.
If you examine the same data file for the SH-96, a larger but asymmetric (60X90) horn used in movie theaters etc, one can see the pattern has flipped at 315Hz and this is about as much change as I can get away with for our commercial stuff.
Also probably why the horns you mention are in the same neighborhood angle ratio wise.
Also, there are no crossover lobes or nulls produced in the design so what your seeing is only horn shape and dimension related..
I am showing this, warts and all in a couple products I designed so that it is clear I am not attacking anyone or "there horn", but my point would be that this stuff is present in any real horn measured with sufficient resolution.
I don’t think one can produce a square radiation pattern without having this and the more rectangular you go, the more difficult it becomes to do. Unfortunately for us, we don’t decide where the audience seats are.
Hope that helps.
Best,
Tom
“We're keeping the horizontal width at 15" and raising the crossover frequency to where the 7.5" vertical retains control? Or are we now using a smaller 7.5" round or square symmetric 70° waveguide with that higher crossover?
With the 67 degree case, the height is the same but the width is half or equal to the height.
The onset of loss of pattern loss is lowered about a half octave because the governing angle and dimension has increased from 45 to 67 degrees while mouth size in that plane is the same and with equal angles in both planes, pattern flip at pattern loss is avoided entirely..
Also, when I say pattern flip, I mean that from a horn that appears to produce a wide pattern, it is narrow instead.
This exchange of pattern shape begins at the first (highest) frequency where pattern loss point is reached.
One can see this same effect in a single horn however depending how you look at it.
For work, our speakers are measured by a third party every 2.5 degrees over an imaginary sphere in an anechoic environment.. This is needed to produce an acoustic model that can be used in room design programs like EASE and others.
As a side benefit of this method, if you add a power capacity measurement to the data, provides a nice way to view the spec’s and radiation pattern.
http://www.danleysoundlabs.com/technical downloads.html
If you download the CLF data file for the SH-50 and the viewer you can see what I mean.
Look at the radiation balloon at the top right, grab it with the cursor and view the balloon head on like the speaker is facing you. Go to the frequency box right below and set it to 400Hz, the early part of pattern control. Step up to say 1-2 KHz, now the pattern takes on sort of a diamond shape, “flipped” relative to the horn shape.
Grab the balloon with the cursor and look at it from more of a side view.
Now, step up to say 4KHz or above and you see the horn takes on the expected pattern orientation.
If you go to the bottom right, you can select electro acoustic data, then the H or V -6dB points and see a directivity graph or click polar’s or one can click balloon spectra and then move the H and V cursor through the balloon and see the frequency response at any angle H and V.
If you examine the same data file for the SH-96, a larger but asymmetric (60X90) horn used in movie theaters etc, one can see the pattern has flipped at 315Hz and this is about as much change as I can get away with for our commercial stuff.
Also probably why the horns you mention are in the same neighborhood angle ratio wise.
Also, there are no crossover lobes or nulls produced in the design so what your seeing is only horn shape and dimension related..
I am showing this, warts and all in a couple products I designed so that it is clear I am not attacking anyone or "there horn", but my point would be that this stuff is present in any real horn measured with sufficient resolution.
I don’t think one can produce a square radiation pattern without having this and the more rectangular you go, the more difficult it becomes to do. Unfortunately for us, we don’t decide where the audience seats are.
Hope that helps.
Best,
Tom
Wayne Parham said:
I think this might be a point of confusion. Correct me if I am wrong but increasing directivity means it beams more or the directivity narrows. Wider directivity = directivity decreases. Your lingo seems to flip this.
JoshK said:
You're right, I misspoke, thanks for catching that. What I meant to say was the coverage angle increases below the passband, not the directivity.
I think you, Tom, Earl, Zilch - most of the "regulars" on this thread understand this, but for anyone unfamiliar with the directivity issues, I'll clarify briefly:
Below the frequency where the horn gains pattern control, its coverage angle becomes wider. Another way of looking at it, the directivity collapses as frequency rises until the horn/waveguide gains pattern control.
Once the horn has gained control, it is able to set the pattern. If it's a CD horn, then the coverage pattern is uniform (spectrally balanced) through the passband.
ZilchLab said:
IID: Interaural Intensity difference.
ITD: Interaural Time Difference.
These two would best be maintained so that they shift in a consistent manner so that the image does not shift with frequency. Thus a good wide sweet spot can be maintained. I wonder how deeply this has been explored.
I've done a lot of exploration to discover what setups worked best for imaging over a wide area. I must admit though, my process has been empirical, perhaps even subjective. A little bit of my personal history in looking at this is in the post called "Pi horn design philosophies".
When I first noticed my cornerhorns had such good imaging, I attributed it to corner placement and first thought it was purely the directivity. After a few installs (and friends that bought them moving from place to place and putting them in different environments), I noticed that the best sounding setups were those in rooms that were rectangular with the speakers in corners closer together than wide. The rooms were slightly narrower than deep. Not by a huge margin, certainly not more than 2:1. Something like 1.4:1 to 1.6:1 depth-to-width. This forces the forward axis to cross in front of the listener just by virtue of room layout. It also tends to bring the time difference down between speakers when the listener off-center, simply because the separation is not so great.
I've also heard them in very large rooms and very small. I've had these speakers since the late seventies, so I've had a chance to hear them in a large number of environments. If the room is too large or too small, the imaging doesn't work. In large rooms, I think everything is just too widely spaced and too distant. In very small rooms, you're basically unable to move far enough away from the speakers for the self-balancing illusion to work. If you can't be behind the crossing of the axis, you have to sit directly betweeen them for proper stereo balance. Spectral balance is good in both cases, they sound fine but more like a mono signal.
It's the stereo imaging that is affected when you're setup wrong, with too much separation or too little, or with the listening spot not behind the forward axis or too far off to one side. Imaging is good over a wider area when this is done right, but the whole room doesn't get good imaging even though it will have good spectral balance. In fact, many people have commented that my speakers sound balanced even when you leave the room, go down the hall. This is because the reverberent field is uniform, a result of constant directivity. But the stereo image is best in a smaller area, roughly outlined in the illustrations in the post called "Imaging, placement and orientation".
That's why some rooms don't really lend themselves to constant directivity cornerhorns, but still benefit from the same kind of setup/orientation, just not from corner placement. If the room doesn't have the right corners, it's better to use a traditional speaker designed for CD, and place them flanking the audience with toe-in that crosses in front of the listening area. I'd much prefer the CD cornerhorn arrangement, where possible, because it offers the additional benefit of CD all the way down to the Schroeder frequency, enforced at the low end by the walls. But in some rooms, that's not an option.
Empirically, I find that the minimum separation is about 10 feet. Seems like maximum separation is about 30 feet. I'm shooting from the hip here, remembering all the installs I've done or seen done, and what I thought of the sound. The best were always somewhere between about 3 meters and 10 meters apart and crossed in front of the listeners.
I've written about this pretty extensively on my forum, in an effort to document what I've found for reference. There are a LOT of posts on this subject, so you might do some searches there. There you'll find not only my expereinecs but also those of many others that have setup their speakers, talking about their placement choices, orientation, room treatments, etc.
There are four posts on that general suject in the "imaging" post linked earlier, and I"ll include it here again for convenience.
It would be nice to see a more formal study done too, I agree. I wonder how you would perform such a study though. I suppose it could be done by interviewing listeners to find which setups provided the best illusion of imaging. That's sort of what I've recorded over the years, albeit informally, by virtue of feedback from users on my forum.
When I first noticed my cornerhorns had such good imaging, I attributed it to corner placement and first thought it was purely the directivity. After a few installs (and friends that bought them moving from place to place and putting them in different environments), I noticed that the best sounding setups were those in rooms that were rectangular with the speakers in corners closer together than wide. The rooms were slightly narrower than deep. Not by a huge margin, certainly not more than 2:1. Something like 1.4:1 to 1.6:1 depth-to-width. This forces the forward axis to cross in front of the listener just by virtue of room layout. It also tends to bring the time difference down between speakers when the listener off-center, simply because the separation is not so great.
I've also heard them in very large rooms and very small. I've had these speakers since the late seventies, so I've had a chance to hear them in a large number of environments. If the room is too large or too small, the imaging doesn't work. In large rooms, I think everything is just too widely spaced and too distant. In very small rooms, you're basically unable to move far enough away from the speakers for the self-balancing illusion to work. If you can't be behind the crossing of the axis, you have to sit directly betweeen them for proper stereo balance. Spectral balance is good in both cases, they sound fine but more like a mono signal.
It's the stereo imaging that is affected when you're setup wrong, with too much separation or too little, or with the listening spot not behind the forward axis or too far off to one side. Imaging is good over a wider area when this is done right, but the whole room doesn't get good imaging even though it will have good spectral balance. In fact, many people have commented that my speakers sound balanced even when you leave the room, go down the hall. This is because the reverberent field is uniform, a result of constant directivity. But the stereo image is best in a smaller area, roughly outlined in the illustrations in the post called "Imaging, placement and orientation".
That's why some rooms don't really lend themselves to constant directivity cornerhorns, but still benefit from the same kind of setup/orientation, just not from corner placement. If the room doesn't have the right corners, it's better to use a traditional speaker designed for CD, and place them flanking the audience with toe-in that crosses in front of the listening area. I'd much prefer the CD cornerhorn arrangement, where possible, because it offers the additional benefit of CD all the way down to the Schroeder frequency, enforced at the low end by the walls. But in some rooms, that's not an option.
Empirically, I find that the minimum separation is about 10 feet. Seems like maximum separation is about 30 feet. I'm shooting from the hip here, remembering all the installs I've done or seen done, and what I thought of the sound. The best were always somewhere between about 3 meters and 10 meters apart and crossed in front of the listeners.
I've written about this pretty extensively on my forum, in an effort to document what I've found for reference. There are a LOT of posts on this subject, so you might do some searches there. There you'll find not only my expereinecs but also those of many others that have setup their speakers, talking about their placement choices, orientation, room treatments, etc.
There are four posts on that general suject in the "imaging" post linked earlier, and I"ll include it here again for convenience.
It would be nice to see a more formal study done too, I agree. I wonder how you would perform such a study though. I suppose it could be done by interviewing listeners to find which setups provided the best illusion of imaging. That's sort of what I've recorded over the years, albeit informally, by virtue of feedback from users on my forum.
I have started to search for information related with imaging, most of which studies were done on Binaural recording and playback. My intial thinking is that this could probably be expressed in terms of ITD and IID vector changes. But since no consistent way has been developed to control directivity pattern, study is sort of in the back seat.
My personal findings:
The cross-channel delay from one speaker to the other ear (and vice versa) is ~140us for the nominal stereo triangle (all sides same length). This is regardless of absolute size but when to close (<1m or so) the level difference start to change, favoring the ear closer to the speaker, opening an "hole in the middle" of the panorama.
I find stereo imaging (especially depth sensation) begins to detoriate once the speakers produce more than 1/4 of those 140us, say some 40us of offset. Which happens to equal a delay of ~2 samples at 44.1kHz, or a corresponding path length difference of 344[m/s]*40u=1.4[cm]. Which in turn means the sweet spot must be hit with a lateral precision "window" of about 3cm for the most accurate results. Of course any head turn also dircetly influences this. Those 40us are BTW also the known level of interchannel group delay distortions (different GD "ripple" per channel) which start to be audible. I think this is not a pure coincidence, or is it? (@Jakob)
And from 10cm or more I cannot say I'd continue to have a "stereo" experience with wideband signals. For bandlimited signals a properly trading effect (by use of specific directivity) can obtain some compensation, though.
The only way to widen the sweetspot beyond this 10cm window for me is to use trinaural (aka optimum linear matrix) projection which may widen it to 40cm or so and also has less degradation with head turn.
- Klaus
The cross-channel delay from one speaker to the other ear (and vice versa) is ~140us for the nominal stereo triangle (all sides same length). This is regardless of absolute size but when to close (<1m or so) the level difference start to change, favoring the ear closer to the speaker, opening an "hole in the middle" of the panorama.
I find stereo imaging (especially depth sensation) begins to detoriate once the speakers produce more than 1/4 of those 140us, say some 40us of offset. Which happens to equal a delay of ~2 samples at 44.1kHz, or a corresponding path length difference of 344[m/s]*40u
And from 10cm or more I cannot say I'd continue to have a "stereo" experience with wideband signals. For bandlimited signals a properly trading effect (by use of specific directivity) can obtain some compensation, though.
The only way to widen the sweetspot beyond this 10cm window for me is to use trinaural (aka optimum linear matrix) projection which may widen it to 40cm or so and also has less degradation with head turn.
- Klaus
I posted pics of the asymmetric JBL horns designed to do it earlier in this thread. The large ones used in the first Everest project are well documented in the JBL Pro spec sheets for the defined directivity speakers built with them, and the Keele paper describing their development.
Why aren't these more popular? Best I can figure is a center channel does it easier and better, and it's a hard sell, so JBL quit making them....
Why aren't these more popular? Best I can figure is a center channel does it easier and better, and it's a hard sell, so JBL quit making them....
"Why aren't these more popular? Best I can figure is a center channel does it easier and better, and it's a hard sell, so JBL quit making them...."
Hello Zilch
Well those horns are so damn big for the Everest I can see that being a problem. That whole system was really big. Could have fit a bunch of Bose cubes in just one cabinet.
The S2600 and S3100 were more manageable but also large speakers.
Rob🙂
Hello Zilch
Well those horns are so damn big for the Everest I can see that being a problem. That whole system was really big. Could have fit a bunch of Bose cubes in just one cabinet.
The S2600 and S3100 were more manageable but also large speakers.
Rob🙂
Robh3606 said:
Always on the lookout, I have several smaller ones here that don't work worth a whit.... 😛
Hi Zilchlab
You asked "Why aren't these more popular? Best I can figure is a center channel does it easier and better, and it's a hard sell, so JBL quit making them."
Part of what made the previous asymmetric horns less than ideal were that they were only covering a small part of the frequency range you needed to make a "happy speaker".
Still the idea of projecting more in one direction than another if far from dead.
We make a variation on a GH-60 horn, which does something like what the JBL was supposed to do.
In the case below we had a speaker that had to be positioned over the audience, not in front of it.
We nicknamed it the bird house.
Here is a photo of them, the speakers are the funny things hanging off the edge of the deck.
http://www.prosoundweb.com/photos/photo/797/
The seats directly below it are at the rear of it’s “Vertical” coverage, were about 20 feet away while the seats at the first row were 110 feet away, about 50 degrees of “Vertical” coverage, 90 degree horizontal, proportioned to avoid pattern flip..
The amplitude shading was such that the installer reported the seats at 100 feet were 1 dB louder than the seats directly below with only one dB variation over that area and at the edge of the dirt track, it was already -15 dB down
Best,
Tom
You asked "Why aren't these more popular? Best I can figure is a center channel does it easier and better, and it's a hard sell, so JBL quit making them."
Part of what made the previous asymmetric horns less than ideal were that they were only covering a small part of the frequency range you needed to make a "happy speaker".
Still the idea of projecting more in one direction than another if far from dead.
We make a variation on a GH-60 horn, which does something like what the JBL was supposed to do.
In the case below we had a speaker that had to be positioned over the audience, not in front of it.
We nicknamed it the bird house.
Here is a photo of them, the speakers are the funny things hanging off the edge of the deck.
http://www.prosoundweb.com/photos/photo/797/
The seats directly below it are at the rear of it’s “Vertical” coverage, were about 20 feet away while the seats at the first row were 110 feet away, about 50 degrees of “Vertical” coverage, 90 degree horizontal, proportioned to avoid pattern flip..
The amplitude shading was such that the installer reported the seats at 100 feet were 1 dB louder than the seats directly below with only one dB variation over that area and at the edge of the dirt track, it was already -15 dB down
Best,
Tom
ZilchLab said:
Have you heard them? I must admit that I never have. They look interesting but my fear is their size would cause problems like the round horns do - large CTC spacing making a narrow forward lobe with closely spaced nulls. I'm not sure how useful the pattern is either, seems like room coverage and imaging would be better with a 90 degree pattern toed in 45 degrees than the hybrid pattern from the Everest. Just some guesses though, haven't really thought about them.
I've always had problems with wide separation between sound sources (above the Schroeder frequency). To me, the comb filtering notches make it important to keep sound sources physically close, so there's a reasonably wide arc between null angles, places where the path length differences are 1/2 wavelength, causing notches in response.
On the same subject but going to the other extreme, I also don't care much for systems that try to pack stuff too tightly together. By itself, that's a good thing but there always seems to be a compromise that's a deal breaker for me. Coaxial and coentrant horns solve the problem of path length distances by being tightly spaced but they intoduce other problems as a result. Either the tweeter horn is too small to work properly or it shades the midrange by being out in front. Other designs pack drivers along the walls, making all sorts of reflections and interference.
Every coaxial or coentrant horn system I've seen had ripples (from too small a horn) or high Q notches (from reflections and interference), some have both. The high Q notches can be smoothed in the measurement system so you usually won't see them but they're there. Look for a resolution or smoothing figure on the chart, most are smoothed 1/3rd octave (33%) which completely hides sharp notches (like off-axis nulls and self-interference reflections), displaying them as shallow valleys by virtue of being averaged over a wider frequency range. Even 1/6th octave (16%) smoothing reduces the depth of a high Q notch quite a bit, making it look more like a wider/shallower dip than a spiked notch, betraying the fact that it is caused by interference. Run a high-res sweep with no smoothing and check through the pattern and you'll find nulls everywhere.
I'm not entirely opposed to smoothed charts because they show the general spectral balance of a loudspeaker and hide some useless detail, essentially noise in the measurement system. But they can also hide some important details as well. The problem is most manufacturers use smoothed charts so to compare apples-to-apples, you have to compare similarly smoothed charts taken at similar resolution. But still, my point is that if you measure a loudspeaker with high resolution and don't apply any smoothing, that's the best way to find interference notches. They'll show up with the tell tale high Q spike downward. You will not see interference notches if smoothing is applied.
Wayne Parham said:
Yes, those pics were taken in ZilchLab.
The effect is quite disconcerting at first; we are used to the image collapsing with lateral movement. Once past that, engagement sets in, and it's very pleasurable.
There's no getting around the fact that there's only one "true" sweet spot, but that's not a deal breaker, and there's certainly no harm in liking the option, anyway.
Yah, those other factors come into play, as well, and while there may be academic merit in pursuing their resolution, a center channel would appear to be a better approach.
Tom illustrates the utility of the concept in other applications, and there are more, one of which I'm exploring....
Kevin Haskins said:
My opinion is that these may not be bad, but only a poor fit to some potential uses. For example something might be well suited to playing music for a large outdoor crowd but perform poorly in a typical living room.
Also for many uses "colloration" is OK. Let's go back 300 years. Instrument makes would purposely design the soundboards of violins so as to "distort" the perfect sine wave of a vibrating sting so that it had the sound we all know to be a violin. If you measure the soundboard using hifi terms it adds HUGE amounts of harmonic distortion, 2nd, 3rd and 5th. Soundboards were the only method of amplification available back then and they did what they could with the technology. Sure enough there are many people who LIKE the sound of these highly distorting mechanical amplifiers
Today we have more options for making sound. Some produce a clean signal others don't. I like the cleaner sound for many genre but for others either it does not matter or many even improves it.
One more observation. The room distorts the sound of any speaker. So I think we don't want "perfect" speakers. We might prefer speakers that distort the sound in the exact opposite way a room does, so the two cancel. Many times in the live sound world you see this done. Some "sound guy" will walk around the venue with a hand held measurement device and the techs will work the EQ of every driver and maybe even move or add speakers. I'm sure in the end some of those drivers are anything BUT "flat". So,... point is that likely "flat" is almost never the best. But if you are selling mass produced speakers where all must be alike then "flat" is a good standard. So,... it just might be that one of these "poor" horns sound good to one person in his listening room even if the measurement predict poor performance.
Wow, that response was from an old post. I tend to agree with with the sentiment expressed by Kevin Haskins though, in that the comparison and critique of the JBL horn was unfair. I mean, it is really hard for one manufacturer to describe another manufacturer's products objectively, especially if the design goal and target market is the exact same. But I still thought it was wrong.
What I thought was the most funny was the description of the vertical performance:
"The JBL horn has a substantial amount of internal reflection and diffraction as is evident throughout its frequency response curves. It is not a CD device as it only holds CD (reasonably well) in the horizontal plane. In all other planes it appears to be highly directional. Since it is not CD only one plane can be EQ’d and the others just have to follow. This results in a far less than desirable vertical response for example."
What I thought was the most funny was the description of the vertical performance:
"The JBL horn has a substantial amount of internal reflection and diffraction as is evident throughout its frequency response curves. It is not a CD device as it only holds CD (reasonably well) in the horizontal plane. In all other planes it appears to be highly directional. Since it is not CD only one plane can be EQ’d and the others just have to follow. This results in a far less than desirable vertical response for example."
Wayne Parham said:
http://www.diyaudio.com/forums/showthread.php?postid=1518869#post1518869
Earl's reply was indeed amusing with respect to his definition of constant directivity, when seen in the light of subsequent developments:
http://www.diyaudio.com/forums/showthread.php?postid=1519005#post1519005
The "Critique" was (is) merely a recitation of the Geddes mantra, for the most part unrelated to, and unsubstantiated by, the actual data; it took over a year to get at the truth of it....
There's that, and then there's also the matter of the vertical nulls. All speakers with drivers stacked vertically on a baffle has them, so the key thing is putting them where they aren't harmful.
It makes no sense to talk about the horn/waveguide in isolation since it covers only part of the bandwidth. Constant directivity in the vertical is meaningless if vertical nulls cut the vertical pattern in half (or more).
That's what I thought was kind of funny. I have long argued that tall vertical coverage is a bad thing for home hifi, so I don't agree that axisymmetry is a benefit, at least it is not a universally accepted benefit. A tall mouth dimension also increases CTC spacing, reducing the vertical null angle. This, I believe is universally accepted as being undesirable. So as I see it, the vertical performance of a round horn is its biggest weakness, at least when combined on a baffle with other drivers in a loudspeaker system.
The real trick is getting a loudspeaker using a horn horn/waveguide with pattern that remains as much as possible within the nulls. The balance is how much vertical pattern loss is acceptable at the low end verses how widely spaced the nulls can be made. It's a balancing act and I think it is nearly as important as the horizontal pattern, at least insofar as the nulls must be far enough apart to present a useful forward lobe. Then what's outside those nulls - the outer lobes - should be limited as much as possible too.
Like all things, loudspeaker design is a balancing act to be sure, one that no speaker gets everything totally right. You can't optimize one feature without compromising another. But you also don't want to make a compromise that may put nulls in the target listening area.
It makes no sense to talk about the horn/waveguide in isolation since it covers only part of the bandwidth. Constant directivity in the vertical is meaningless if vertical nulls cut the vertical pattern in half (or more).
That's what I thought was kind of funny. I have long argued that tall vertical coverage is a bad thing for home hifi, so I don't agree that axisymmetry is a benefit, at least it is not a universally accepted benefit. A tall mouth dimension also increases CTC spacing, reducing the vertical null angle. This, I believe is universally accepted as being undesirable. So as I see it, the vertical performance of a round horn is its biggest weakness, at least when combined on a baffle with other drivers in a loudspeaker system.
The real trick is getting a loudspeaker using a horn horn/waveguide with pattern that remains as much as possible within the nulls. The balance is how much vertical pattern loss is acceptable at the low end verses how widely spaced the nulls can be made. It's a balancing act and I think it is nearly as important as the horizontal pattern, at least insofar as the nulls must be far enough apart to present a useful forward lobe. Then what's outside those nulls - the outer lobes - should be limited as much as possible too.
Like all things, loudspeaker design is a balancing act to be sure, one that no speaker gets everything totally right. You can't optimize one feature without compromising another. But you also don't want to make a compromise that may put nulls in the target listening area.
Hi guys
Zilchlab, I am not sure what your apparent “thing” is with Earl but I would ask, have you ever measured / listened to a plain simple off axis horn like the JBL?
These really DO have problems if built like they were back then.
Horns similar to this were where I started when trying to make the amplitude shaded system.
ChrisA
One thing that is often a goal in building a loudspeaker is to make it a faithful re-producer. This is quite different than if one is making a guitar of musical instrument speaker where the speaker is part of the sound.
With a reproducer, one assumes the creative part is done and then accurately reproducing the sound is the goal. While this means some things will sound good and others not, an accurate reflection of the recording is the goal not maximum enjoyment on all recordings.
This approach is pretty much necessary when you are supplying speakers to others who’s tastes are different than yours, the best one can ask for then is do it’s best with any program material.
Rooms fwiw don’t cause distortion as in harmonics etc unless one has things that are rattling which is why distortion measurements also shouldn’t be done indoors.
What rooms do is provide portions of the original signal, reflected and arriving at various times AFTER the initial direct arrival, these add causing peaks and dips in the response and greatly effecting your ability to hear what ever stereo image was present in the recording.
You can’t fix much of this with EQ because much of it is caused by delayed partial signals NOT an non-flat amplitude response.
Eqing things which are not minimum phase (like these cancellation notches in a room) may make the amplitude look better with pink noise but if you measure phase or time, one see’s the EQ made things worse. Conversely, EQ can fix other things in an essentially perfect way, both magnitude and phase, a double edged sword that must be used with caution.
Wayne
“Other designs pack drivers along the walls, making all sorts of reflections and interference.”
Boy that line is getting tired.
You can build them that way, but if one follows the rules for our systems, one can avoid all those interferences as well as the acoustic phase shift and lobes and nulls that are normally produced at crossover.
Just as with the placement of subwoofers and for exactly the same reasons, if you place multiple sources less than about ¼ wavelength apart, they feel each others radiation pressure, add coherently into and radiate as one source.
Measure a group of close spaced subwoofers (less than a quarter wavelength apart at the highest frequency in question), do you see a complex interference pattern, do you see any directionality? NO you don’t, THEY ADD COHERENTLY.
It is the exact same thing in the horn where multiple drivers combine or where two ranges combine, they are ALL less than a quarter wavelength apart.
ALL of this is acoustic size dependent, not dependent at all on names like subwoofer and midrange. Sound doesn’t care, it works exactly the same way at 30Hz as it does at 300Hz as it does at 3000Hz, when you see that, you might finally see how our speakers can and do work.
“Every coaxial or coentrant horn system I've seen had ripples (from too small a horn) or high Q notches (from reflections and interference), some have both. The high Q notches can be smoothed in the measurement system so you usually won't see them but they're there. Look for a resolution or smoothing figure on the chart, most are smoothed 1/3rd octave (33%) which completely hides sharp notches”
Yes, there is a real problem with coax drivers in general, the transition from center pole to cone as a horn, has too large a discontinuity and / or too rapid a change in angle at too large a dimension.
These is one commercial coax driver I found that has a minimal problem with it’s geometry and it’s upper and lower section were close enough to the right locations that I could make a synergy crossover work with it so there is no phase shift related to the crossover. Given this was the best behaved one I could find, I used in a couple products at work. If you look at the CLF file for the SH-100, you can see the balloon has some structure in its upper range due to the cone / center pole transition.
Even so, this is Vastly better than many I tried, it is surprising how bad some coax drivers are. Also, notice how low in frequency the speaker has forward directivity . This is a “low Q” speaker used in commercial sound. These are often overhead or on side walls, the intention is to have about a 90 degree conical pattern. The system has a shallow horn extension which adds gain to the cone and pattern control.
So far as smoothing, you would be hard pressed to find any major company with unsmoothed data. How many offer an unspecified curve, how many subwoofers have no curves at all?
I suppose the argument for smoothing is that since at best your ears have a detection bandwidth of about 1/6 octave, that there is little reason to show things you can’t hear or detect.
I think most of our data sheet curves were taken about 1/6 octave smoothing.
1/3 is too wide I think for any engineering response information but it is an existing standard for many things like EASE and other room prediction / design software.
Also keep in mind when one sets the measurement bandwidth, one is also setting resolution or in effect smoothing. With the TEF machine, smoothing is used because you set the measurement resolution which is retained through out the bandwidth. That is to say if you swept from 20Hz to 20KHz with a 5Hz resolution, one would have a quarter octave resolution at 20Hz and 1000 times more resolution at 20KHz, No one needs .00025 octave bandwidth.
Don Keele I think it was, considering a log sweep for TDS, had observed that loudspeakers tend to have a constant Q resonances such that a resolution limited to a Q of 10 was more than enough to reflect what a speaker can do.
Systems like Smaart use a sliding bandwidth, also providing a “smoothing” by adjustment of resolution.
This measurement noise problem can be seen when one measures a woofer with a limited HF bandwidth at a distance and gets a measurement with significant ups and downs, “noise” which is not present when measured very close nor a signal such a system can produce. Another clue to noise limiting is measurements system fail to track the systems lf response slope as it falls below say –30dB midband becoming dominated by noise.
In any case, since a narrow cancellation notch is much harder to hear than a narrow peak of the same bw/volume (as any high resolution room response curve shows), peaks are usually what one would be more concerned with if one were “fixing” problems
Best,
Tom
Zilchlab, I am not sure what your apparent “thing” is with Earl but I would ask, have you ever measured / listened to a plain simple off axis horn like the JBL?
These really DO have problems if built like they were back then.
Horns similar to this were where I started when trying to make the amplitude shaded system.
ChrisA
One thing that is often a goal in building a loudspeaker is to make it a faithful re-producer. This is quite different than if one is making a guitar of musical instrument speaker where the speaker is part of the sound.
With a reproducer, one assumes the creative part is done and then accurately reproducing the sound is the goal. While this means some things will sound good and others not, an accurate reflection of the recording is the goal not maximum enjoyment on all recordings.
This approach is pretty much necessary when you are supplying speakers to others who’s tastes are different than yours, the best one can ask for then is do it’s best with any program material.
Rooms fwiw don’t cause distortion as in harmonics etc unless one has things that are rattling which is why distortion measurements also shouldn’t be done indoors.
What rooms do is provide portions of the original signal, reflected and arriving at various times AFTER the initial direct arrival, these add causing peaks and dips in the response and greatly effecting your ability to hear what ever stereo image was present in the recording.
You can’t fix much of this with EQ because much of it is caused by delayed partial signals NOT an non-flat amplitude response.
Eqing things which are not minimum phase (like these cancellation notches in a room) may make the amplitude look better with pink noise but if you measure phase or time, one see’s the EQ made things worse. Conversely, EQ can fix other things in an essentially perfect way, both magnitude and phase, a double edged sword that must be used with caution.
Wayne
“Other designs pack drivers along the walls, making all sorts of reflections and interference.”
Boy that line is getting tired.
You can build them that way, but if one follows the rules for our systems, one can avoid all those interferences as well as the acoustic phase shift and lobes and nulls that are normally produced at crossover.
Just as with the placement of subwoofers and for exactly the same reasons, if you place multiple sources less than about ¼ wavelength apart, they feel each others radiation pressure, add coherently into and radiate as one source.
Measure a group of close spaced subwoofers (less than a quarter wavelength apart at the highest frequency in question), do you see a complex interference pattern, do you see any directionality? NO you don’t, THEY ADD COHERENTLY.
It is the exact same thing in the horn where multiple drivers combine or where two ranges combine, they are ALL less than a quarter wavelength apart.
ALL of this is acoustic size dependent, not dependent at all on names like subwoofer and midrange. Sound doesn’t care, it works exactly the same way at 30Hz as it does at 300Hz as it does at 3000Hz, when you see that, you might finally see how our speakers can and do work.
“Every coaxial or coentrant horn system I've seen had ripples (from too small a horn) or high Q notches (from reflections and interference), some have both. The high Q notches can be smoothed in the measurement system so you usually won't see them but they're there. Look for a resolution or smoothing figure on the chart, most are smoothed 1/3rd octave (33%) which completely hides sharp notches”
Yes, there is a real problem with coax drivers in general, the transition from center pole to cone as a horn, has too large a discontinuity and / or too rapid a change in angle at too large a dimension.
These is one commercial coax driver I found that has a minimal problem with it’s geometry and it’s upper and lower section were close enough to the right locations that I could make a synergy crossover work with it so there is no phase shift related to the crossover. Given this was the best behaved one I could find, I used in a couple products at work. If you look at the CLF file for the SH-100, you can see the balloon has some structure in its upper range due to the cone / center pole transition.
Even so, this is Vastly better than many I tried, it is surprising how bad some coax drivers are. Also, notice how low in frequency the speaker has forward directivity . This is a “low Q” speaker used in commercial sound. These are often overhead or on side walls, the intention is to have about a 90 degree conical pattern. The system has a shallow horn extension which adds gain to the cone and pattern control.
So far as smoothing, you would be hard pressed to find any major company with unsmoothed data. How many offer an unspecified curve, how many subwoofers have no curves at all?
I suppose the argument for smoothing is that since at best your ears have a detection bandwidth of about 1/6 octave, that there is little reason to show things you can’t hear or detect.
I think most of our data sheet curves were taken about 1/6 octave smoothing.
1/3 is too wide I think for any engineering response information but it is an existing standard for many things like EASE and other room prediction / design software.
Also keep in mind when one sets the measurement bandwidth, one is also setting resolution or in effect smoothing. With the TEF machine, smoothing is used because you set the measurement resolution which is retained through out the bandwidth. That is to say if you swept from 20Hz to 20KHz with a 5Hz resolution, one would have a quarter octave resolution at 20Hz and 1000 times more resolution at 20KHz, No one needs .00025 octave bandwidth.
Don Keele I think it was, considering a log sweep for TDS, had observed that loudspeakers tend to have a constant Q resonances such that a resolution limited to a Q of 10 was more than enough to reflect what a speaker can do.
Systems like Smaart use a sliding bandwidth, also providing a “smoothing” by adjustment of resolution.
This measurement noise problem can be seen when one measures a woofer with a limited HF bandwidth at a distance and gets a measurement with significant ups and downs, “noise” which is not present when measured very close nor a signal such a system can produce. Another clue to noise limiting is measurements system fail to track the systems lf response slope as it falls below say –30dB midband becoming dominated by noise.
In any case, since a narrow cancellation notch is much harder to hear than a narrow peak of the same bw/volume (as any high resolution room response curve shows), peaks are usually what one would be more concerned with if one were “fixing” problems
Best,
Tom
- Status
- Not open for further replies.