Uniform Directivity - How important is it? - Page 4 - diyAudio
Go Back   Home > Forums > Loudspeakers > Multi-Way

Multi-Way Conventional loudspeakers with crossovers

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 21st May 2013, 02:12 PM   #31
Banned
 
Join Date: May 2008
Location: Switzerland
Quote:
Originally Posted by Barleywater View Post
And it is all achievable with:

Click the image to open in full size.
You should give that thing its own thread.
  Reply With Quote
Old 21st May 2013, 02:25 PM   #32
Jay1111 is offline Jay1111  United States
diyAudio Member
 
Join Date: Feb 2013
Location: AZ
Quote:
Originally Posted by Pallas View Post
Well, I'm clearly not the only person who read what you actually wrote (as well as the graph labels) and drew that inference...



Honestly, probably pattern control. The reason being that the pattern is determined by physics and geometry, but if the pattern is fairly uniform over a wide axis, then FR can be controlled with passive/active EQ.

Furthermore, I look at that SEOS graph and I see that a large bit of the variation is a broad peak at 4kHz or so. Assuming that carries through the pattern, basically just one parametric filter will knock that down all over.

Note: no dog in this hunt. I greatly admire what both of you are doing. And went in an entirely different direction (TAD/Pioneer 5" concentric for mids/highs up front in a 3-way + flanking sub setup, 5" KEF Uni-Q's for side and rear surrounds) for my own use. Though I am, admittedly, considering picking up a set of the new plastic SEOS15's, to use with 12" woofers.



That is the opposite of my experience. Speakers with midband pattern discontinuities invariably sound wrong to me.



Only in the extreme nearfield is one listening to the "on axis sound of the speakers."
Are the speakers in a empty room, being equal distance from every boundary? I'm guessing no. At this point your room is not giving you a constant level of reflected sound.

I'm going to bow out before I stir up to many more fanatics....
  Reply With Quote
Old 21st May 2013, 02:58 PM   #33
kevinh is offline kevinh  United States
diyAudio Member
 
Join Date: May 2007
The room and controlling reflections in a small room is the issue, as pointed out earlier the ceiling, floor and sidewall reflections are issues here, ideally if one controls the first reflections so they are >= 10ms and they reflections are reduced in amplitude a lot of 'problems with speakers disappear.

This seems to me to be one advantage of a well designed floor speaker. Another example is an electrostat, which can have a controlled dispersion as long as it isn't placed closer that 4 ft from a wall.
  Reply With Quote
Old 21st May 2013, 03:15 PM   #34
ra7 is offline ra7  United States
diyAudio Member
 
Join Date: Feb 2009
Blog Entries: 1
In Toole's book, Chapters 18 and 19 are very important. He emphasizes smoothness of response over flatness. Also, no particular directivity was preferred in the blind tests. Of the two loudspeakers receiving the highest ratings, one is a conventional direct radiator and another is a horn. So, again, no particular directivity was preferred.

That being said, matching the direcitivty of the drivers near the crossover is a good thing. Peaks in the power response are audible and not good. But beyond that, having constant directivity through the entire spectrum is not necessary. This is what I was calling a fad. Not directivity itself. I like directivity. It allows you to point the sound where you want it to go. I personally own SEOS-12 and SEOS-18 and like them both. In my room, it creates a more believable image. But it's not because they have constant directivity, it is because early reflections are avoided.

What Wayne seems to be saying does have some validity, in that smoothness should be optimized over pattern control. A somewhat narrowing beamwidth is just fine, but ripples in the response are not good. Like Bill said, I don't think the ripples are 5db, and to me at least, there doesn't appear to be much difference in the two waveguides.

Last edited by ra7; 21st May 2013 at 03:24 PM.
  Reply With Quote
Old 21st May 2013, 04:20 PM   #35
diyAudio Member
 
Wayne Parham's Avatar
 
Join Date: Feb 2004
Let me take a monent to say again that I think the SEOS device(s) are useful and interesting and the guys that design with them are great guys. Many of them are friends of mine, and I regret that any animosity developed. I honestly don't feel any ill-will for the designers, we've always been sort of kindred spirits. There are probably other reasons for the rift, and as long as we keep things technical, those do not develop.

I sort of take my lead from Keele's work on CD horns, where he describes horns with the "right size" mouth. The ripple is set by a lot of factors, and while one can generally assume that the larger the mouth, the less ripple there will be - that is not always true. Combine that understanding with a catenary flare to reduce internal reflections and the horn loses its throat diffraction but still acts a lot like the CD horns Keele described.

Beyond that, in my experience with horns, one easy way to get a "horn honk" sound is from standing waves. Pipe modes in the crossover region really make a "cupped hands" horn sound, especially if their impedance peak is manifested in the response curve. That's why I was always so focused on damping - It's like a mantra in my "Crossover Electronics 101" seminar and all my other crossover documents.

I demonstrate this kind of peaking during the talk by showing different crossovers, some that are more vulnerable to impedance peaks and create an audible response peak. We play music through these crossovers and waveguides, and I flip a switch to compare one crossover that can damp the modes with another that cannot. The attendees of the seminar all hear it, as they see it on the transfer function on the overhead projector.

So this is a big focus of mine, and I try to limit those 5dB peaks and valleys because I can really hear them. I do realize that the designers will remove any peaks they see in the response curve using notch circuits. I think that's valid and important. But to me, I try to make the horn flat, and then try to damp it appropriately in the crossover too. It is very important that the system response not have 5dB peaks and valleys in the 1kHz to 4kHz region, because that's right at the apex of the Fletcher-Munson curve where we are all most sensitive. Peaks there stand out there with annoying strength, and I believe that is the think most responsible for what some would call "horn honk."

The diffraction slot in the throat of an constant-directivity horn makes an unnatural treble sound, I think. I've heard some call it "spitty" or "splashy", sort of like the fake cymbals in a drum machine. I don't like that sound either, but I don't describe it as horn honk. I think what people object to as horn honk is a standing wave thing. So I seek to mimimize that in my designs.

I agree with many of you that the words "uniform" and "constant" are somewhat synonymous, and that's why I chose to use "uniform directivity" to describe what I mean. I wanted it to have a similar connotation, but to not use the phrase "constant directivity", which had already been defined in the literature to mean beamwidth that doesn't change with frequency.

Collapsing directivity is a phrase I've often heard to describe a source that has beamwidth that narrows proportional to frequency.

Matched-directivity speakers have collapsing directivity up to the crossover point, and then constant directivity above that. I dunno, it's smooth, and it seems useful, so I just started calling that "uniform directivity." I like using that phrase to me "almost constant directivity." But that's just my nomenclature, and I usually try to describe what I mean as well.

"Blending" is another term I use that really just means summing. I think I borrowed this one from Earl Geddes though. I liked it because it gives the right "feel." He used it when describing the complex summing that occurs when subs and mains are overlapped in the modal region. Im some positions in the room, the summing is constructive, others it is not. Same with the room modes and reflections. This is nicely described as "blending", in my opinion, and so I borrowed his informal terminology.

In the crossover region, the summing isn't nearly as complex as it is in the modal region of a room. But it does have constructive summing in some positions and destructive summing in other positions. The forward lobe is made of constructively summed wavefronts, which then combine to make a nice clean primary lobe.

Movement along the vertical axis creates a path length difference between the listener and the sources. At some point, the summing moves from constructive towards destructive interference. So movement along the vertical plane eventually brings the listener to the edge of the forward lobe, and into the vertical null, where summing is destructive. Then outside these, secondary lobes form. The secondary lobe forms where path length differences create a full cycle shift. If the individual source beamwidths are narrow enough, the secondary lobes will be small. If source beamwidths are large, the secondary lobes will be also.

Movement along the horizontal axis does not create a path length difference, except that caused by the source area, the very reason that the midwoofer becomes directional in the first place. But there is no path length difference change between acoustic centers. So summing is generally constructive out to fairly large angles, where the directivity of the piston sets in. There summing becomes complex, something like the situation with the verticals. So the edge of the pattern in the horizontal is usually a little bit lumpy through the crossover region as a result.

That's why I describe the combined output of the two sources in the overlap region as being "blended." It's also why I like to match their directivities as much as possible, by choosing crossover where the beamwidths of the two subsystems match. I have seen some describe pushing the crossover point so low directivity "matching" is effectively done down where both sources radiation patterns approach omnidirectional. I think this is a bad practice, both because of the summing/blending thing I just described and also because compression drivers are strained when running them too low.
__________________
Visit the π Speakers website
High-quality audiophile loudspeakers and kits

Last edited by Wayne Parham; 21st May 2013 at 04:24 PM.
  Reply With Quote
Old 21st May 2013, 04:36 PM   #36
kevinh is offline kevinh  United States
diyAudio Member
 
Join Date: May 2007
Wayne, some emphasis the importance of matching the horn throat entrance angle to the exist angle of the compression driver and round over of the edges to reduce diffraction. Comments of either?
  Reply With Quote
Old 21st May 2013, 04:55 PM   #37
diyAudio Member
 
Join Date: Apr 2005
Location: Sitting behind the 'puter screen, in Illinois, USA, planet earth
I am not sure one can realistically make a broad brush generalization, one needs to add some qualifiers.

For example, if you listen outdoors and are just one listener then the directivity is essentially* irrelevant as nothing but the direct sound and ground bounce reaches your ears. If you want several people to experience the identical reproduction at the same time, then what you want is the same response vs position.

*The only way how the one loudspeakers radiates anechoic or outdoors that is audible in this case is the degree of homogeneity, if what reaches the R and L ears is substantially identical, there is very little audible clue as to the loudspeakers physical location in depth. If not, one can easily identify the depth location (with eyes closed) and in the latter case one using a pair in stereo will also hear a R&L source as well as the phantom image playing a mono signal, where the former the R&L sources gone / greatly suppressed leaving a stronger mono phantom. This is unrelated to musical enjoyment or pleasure unless you’re into that (I am).

The easiest way to get that simple radiation is with an acoustically small source which automatically has a wide and simple dispersion pattern. A small full range driver on a large flat baffle (eq’d flat) demonstrates this over some bandwidth.
Once one has moved into a room, things change a lot.
loudspeaker directivity is tied to acoustic size while in the market “small is better” with music enjoyment being subjective and many magazines generally a for hire marketing agents and not sources of real information, things are not very clear in hifi land.

BUT elsewhere, if one is concerned with preserving the information in the recording or one needs to understand words, then directivity is very important. Here, no amount of added room stuff brings you closer to the recording than none. It IS reflected sound that fills in the MTF’s which limit the resolution. If you’re not familiar with STIpa or MTF’s, here is the optical MTF analogue.

http://photo.net/learn/optics/mtf/blur2.gif

For information transfer like speech or here for inter-aural cross talk cancellation to work it’s best it was concluded “Loudspeaker directivity is the single most important loudspeaker property for 3D audio with crosstalk cancellation, since room reflections directly degrade the level of crosstalk cancellation.”

3D3A Lab at Princeton University

The same degradation is present with the original signal as well and it is also better preserved at the LP with directivity vs not.

Here is another odd thing in hifi, people obsess about “flatness” and other one meter measurements and they do matter, yet few listen at one meter and what you measure at the LP is certainly more like / is actually what you experience.
It really does seem like the concept in commercial sound that what you experience or measure at the LP is what matters most, that this would / does apply in hifi too and here is where directivity is your friend.
Alternately, among the things which degrade the MTF measurements which comprise the STIpa speech intelligibility prediction is reflected sound.
At work in commercial sound, I have proposed a figure of merit for CD sources too, it would be the difference between the energy in the stated pattern, vs the energy outside the stated pattern. This is something like DI or Q but now includes the intended pattern instead of only a 0 degree reference. This way one can say that say 90% of the total energy is where you intend while 10% is going elsewhere and potentially exciting the reverberant field that half goes where you want and half elsewhere etc.

If preserving the information arriving at the listener is important, then so is directivity if you’re in a room, if enjoyment is the only measure, then anything that works, well, works even 6X9’s on the back deck .

As Bill mentioned, the Synergy horn design is a way to avoid all the interactions from multiple sources. It is a way to drive a horn and radiate ‘as if” there was only one impossible broadband driver even though there may be a large number. One common thing to all of them is that you can walk up and even put your head into the horn and you never hear any more than one sound source floating in front of you.
While for commercial sound the products at work must have a full 3d radiation model (data taken by an independent lab, every 5 degrees over a full sphere in a reflection free area) the format doesn’t plot out the spectrogram style map popular in hifi.
I did ask Sebastian our software guy to take the CLF data and plot a Horizontal map for an SH-50, one of our oldest Synergy horns (I use them at home, a 7 driver, 3 way horn system). Larger than a hifi speaker with pattern control to a lower frequency as a result.

Note this is a full bandwidth plot, 3dB per color, front 180 degrees, extracted from the CLF data, the left frequency boundary is 30Hz not 200Hz.

Wayne, I hope you weren’t too close to the tornadoes, my heart goes out to those who lost and were harmed.
Tom
Attached Images
File Type: png SH-50 H map.png (20.2 KB, 392 views)
__________________
Bring back mst3k and futurama
  Reply With Quote
Old 21st May 2013, 05:21 PM   #38
Banned
 
Join Date: May 2008
Location: Switzerland
Quote:
Originally Posted by Wayne Parham View Post
I demonstrate this kind of peaking during the talk by showing different crossovers, some that are more vulnerable to impedance peaks and create an audible response peak. We play music through these crossovers and waveguides, and I flip a switch to compare one crossover that can damp the modes with another that cannot. The attendees of the seminar all hear it, as they see it on the transfer function on the overhead projector.
Not saying that there is no audible difference but giving people a visual feedback of an audible change they are going to hear will make virtually everybody hear a change even if it's not there.
Probably everybody knows this article: Audio Musings by Sean Olive: The Dishonesty of Sighted Listening Tests
...and this one: McGurk effect - Wikipedia, the free encyclopedia
  Reply With Quote
Old 21st May 2013, 05:41 PM   #39
diyAudio Member
 
Wayne Parham's Avatar
 
Join Date: Feb 2004
Markus, the peaking I'm talking about is a pretty large peak. Put a 33 ohm resistor in series with your waveguide, for example. Don't put in any shunt resistance, so the impedance peaks near cutoff are transformed into the response curve. There's no way you won't hear that.

Subtle stuff needs a double blind, I would agree. But big stuff is obvious. A 5dB peak that's a half octave wide is obvious.

Tom, thanks for the thoughts. We were in the path, but in Tulsa, we didn't get the brunt of it.

Kevin, I matched the H290C throat entrance to the DE250 compression driver exactly. The catenary curve in the waveguide starts at that angle and progresses from there. I think that's the way the SEOS family of waveguides are too, for what it's worth.
__________________
Visit the π Speakers website
High-quality audiophile loudspeakers and kits
  Reply With Quote
Old 21st May 2013, 06:44 PM   #40
diyAudio Member
 
Join Date: Sep 2008
Quote:
Originally Posted by markus76 View Post
It's the same thing People simply don't use precise wording.

What kind of directivity is that - constant or uniform?

Click the image to open in full size.
I assumed that Wayne meant uniform power response in regards to directivity....at least that's how I read it.
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Dumb Directivity Question Patrick Bateman Multi-Way 2 23rd April 2013 02:02 AM
My S13 OB. Uniform polar response to tweeters at last! gainphile Multi-Way 12 24th February 2010 08:03 AM
A question on directivity swak Multi-Way 18 31st July 2005 03:03 PM
Improving on the LCY uniform dispersion ribbon Audiophilenoob Planars & Exotics 34 22nd July 2005 12:01 AM
where is short wiring in an amp important/ not important ? sandro600 Solid State 5 4th April 2004 03:10 AM


New To Site? Need Help?

All times are GMT. The time now is 01:37 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2