Beyond the Ariel

Is there a big idea here, about sound reproduction in general? Something like, the dipole concept has a problem that negates any/all benefits, so forget that and stick with ___? Or, open baffles have a feature that needs to be addressed with ___? Or?

I think the important thing is not to get all religious about one approach or another - and I confess I am as bad as anyone when it comes to that. What's called for is more rigor in thoroughly analyzing what's going with a loudspeakers, regardless of topology - open baffle, transitional, closed-box, horn loading, what have you.

Several years ago I got in big trouble reviewing Dr. Geddes' book - enough so that I withdrew the review, rather than deal with all the hostility. But my basic point is simple: loudspeaker design still has a long way to go. We have not arrived at the fabled "City On The Hill", and there is much more to be discovered. And regardless of the reaction on various Internet forums, I still feel Dr. Geddes' book has made a significant contribution to the art.

I am very appreciative of all the posters who know way more than I do - Dr. Geddes, John K, Michael, Bjorn Kolbrek, JMMLC, and many others. All of these postings advance the art - sorry, folks, this ain't a construction project, but more of a round-table discussion amongst the most talented people in the business.

What I hope has become evident to the readers is that all approaches have advantages and disadvantages, and the way these are arranged depends on the priorities of the loudspeaker designer. What is given up in one area is gained in another - and this has nothing to do with brand-name awareness or the price of the loudspeaker.
 
Last edited:
John

A question for you, or anyone else.

When I looked at the Orion measurements it was obvious how well it controlled the directivity in the 200 - 500 Hz range, where I am not able to effect much control. Hopefully it is not a coincidence that this is precisely the region where I think the Orion shines. In my book I showed how a rear facing driver could be used to control the directivity in precisely this range while going away at LFs to yield a more efficient monopole. It turns out that this can be done passively, albeit not simply (passive attracts me for obvious reasons). I am going to try this and wondered if you, or anyone else, has actually done this as well and what your experience was (technically please!). This would not be too hard to impliment and may be an audible improvement. (There is no point in not disclosing this since its already public domain anyways.)

Actually I have done something like that. A few years ago I build a little speaker for a DIY gathering called the Bird House. Imagine a speaker enclosure that looked like a steep roofed bird house, as shown below.

An externally hosted image should be here but it was not working when we last tested it.


The tweeter and midrange were dipole format. The woofer was a single woofer firing down as a monopole. The crossover between mid and woofer was an Linkwitz/Riley 2nd order affair. As you can see at my web site such an arrangement yields a smooth transition from dipole through a cardioid at the crossover point, to monopole at low frequency:

An externally hosted image should be here but it was not working when we last tested it.


This figure is for an LR4 crossover but the LR2 behaves in the same manor. It just extend the transition region over a wider frequency range.

It is also the philosophy used in the design of my NaO Mini and applies to the NaO II as well when the woofers are used in sealed box format.

The idea is that the dipole response retains good CD behavior over its useful range. The transition to cardioid around the crossover to the woofer helps control the reflection of the wall behind the speaker in the upper bass lower midrange region. Finally the response at low frequency reverts to a monopole in the sparsely populated modal region.

The Bird House crossover was at 200 Hz. The NaO speakers cross in the 120-130 Hz. My new speaker, the Note, will also have an option to build it in this format using a ported woofer system. It will also include the option for dipole and U-frame cardioid woofer configurations. I am doing it that way so the builder can decide what woofer format he wants to go with.

The problem with doing this passively is that you can not take advantage of the frequency dependent sensitivity of the dipole midrange . As you know, my designed are hybrid with a passive crossover used on the panel and an active crossover between the woofer and panel. The active crossover also contains any needed equalization for the woofer and the dipole midrange.
 
You missed to clearly point out that its even an approximation *below* dipole peak, and you also missed to point out that this is so because OB simply is a CMP system (if you have an other term from "ancient" researchers in that topic - no problem for me :) ).


Michael

Let's not get more confused here. If we are talking about the on axis response and radiation pattern of an acoustic doublet or dipole it has noting to do with CMP, as you call it. There is no requirement that the sources of the doublet be MP. MP is a separate issure.
 
What I hope has become evident to the readers is that all approaches have advantages and disadvantages, and the way these are arranged depends on the priorities of the loudspeaker designer. What is given up in one area is gained in another - and this has nothing to do with brand-name awareness or the price of the loudspeaker.

I also think that we often forget that the real goal is listening to music and making an emotional connection to it. With the right music that can be accomplished with an iPod. The obsession with Hi Fi and high quality reproduction is just that, an obsession and in no way guarantees any greater emotional response will result. I, for one, enjoy listening to music far more when driving than just sitting at home listening.
 
Actually I have done something like that.

John, neither of the speakers in my setup are OB, so it seems to me that it is quite different. A cardiod is created by having the rear facing driver out of phase with the front and at a lower volume velocity. By using a 6 dB HP filter on the rear one (in adition to the LP crossover at 800-1 kHz), the polar transitions from cardiod down to monopole at LFs. I think that this can be done without frequency dependent gain required for the woofers - hence passively.
 
I, for one, enjoy listening to music far more when driving than just sitting at home listening.

Curious (perhaps typical!), I am the exact opposite. I do not connect as much to music in a car, too many distractions, or even an airplane (of which I spend way too much time in!) as I do listening late at night in the quiet and solitude of my theater. In that case I am almost transported away and get completely engulfed in the effect. Cars and headphones just don't get there, for me, although I still like the music.

But I think that your point is, or hope that I understand, that the music and performance are quite a seperate things from the playback and one can make an "emotional connection" to the music quite independently of the sound system.
 
Last edited:
Curious (perhaps typical!), I am the exact opposite. I do not connect as much to music in a car, too many distractions, or even an airplane (of which I spend way too much time in!) as I do listening late at night in the quiet and solitude of my theater. In that case I am almost transported away and get completely engulfed in the effect. Cars and headphones just don't get there, for me, although I still like the music.

But I think that your point is, or hope that I understand, that the music and performance are quite a seperate things from the playback and one can make an "emotional connection" to the music quite independently of the sound system.

It has to be the right music. I'm not driving the back roads of northern New England at 8 or 9/10 listening to Mozart. :) If I'm listen to Mozart I'm probably laying down on the couch, soon to be asleep. Motzart might go better with a hot air ballon ride. Something that everyone should try, the ballon ride that is.

But (about to commit heresy) regardless of the playback quality, as soon as you put a vidio screen between the speakers there is a hole new perspective on music reproduction (event reproduction?). I'll give you an example. I have a DVD of a live perfromance that everyone who has watched it raves about the performance. Take that DVD and turn off the video and it's almost un-listenable. I guess the point is that visual stimulation take precedent over aural stimulation.
 
as soon as you put a vidio screen between the speakers there is a hole new perspective on music reproduction (event reproduction?). I'll give you an example. I have a DVD of a live perfromance that everyone who has watched it raves about the performance. Take that DVD and turn off the video and it's almost un-listenable. I guess the point is that visual stimulation take precedent over aural stimulation.

THIS I agree with, mostly. The video adds another dimension, that is unmistakable, but I tend to agree with George Lucas here, who said that sound is 1/2 of the movie experience. Turn off the sound on that DVD and it is even less appealing than the sound without the video.

But I have found that I can't take video at times. Video is very intense and requires much more "nueral energy" than just audio. I can almost always just listen to music, but not always listen and watch.
 
THIS I agree with, mostly. The video adds another dimension, that is unmistakable, but I tend to agree with George Lucas here, who said that sound is 1/2 of the movie experience. Turn off the sound on that DVD and it is even less appealing than the sound without the video.

But I have found that I can't take video at times. Video is very intense and requires much more "nueral energy" than just audio. I can almost always just listen to music, but not always listen and watch.

I agree that sound can be 1/2 the experience, but sound and sound quality are very different things.
 
John, neither of the speakers in my setup are OB, so it seems to me that it is quite different. A cardiod is created by having the rear facing driver out of phase with the front and at a lower volume velocity. By using a 6 dB HP filter on the rear one (in adition to the LP crossover at 800-1 kHz), the polar transitions from cardiod down to monopole at LFs. I think that this can be done without frequency dependent gain required for the woofers - hence passively.

I'm not sure how you are creating a cardioid. A cardioid requires two sources of equal strength separated by a distance, d, with once source out of phase and with a time delay equal to d divided by the sound speed. This assumes point sources. If directionality is considered it may be possible to do something like that since the rear wave of the front driver would be weaker that the front wave. A rear driver could be added with reduced amplitude so as to cancel the rear radiation form the front driver, but it would still have to have the correct delay to account for differences in propagation distance, and what even the rear driver radiated would have some effect on the front response. And then a rear driver with different directional characteristics could be used....... Can be done? Maybe, but probably difficult to accomplish in a practial manor.
 
I'm not sure how you are creating a cardioid. A cardioid requires two sources of equal strength separated by a distance, d, with once source out of phase and with a time delay equal to d divided by the sound speed.

I think that this is a common misconception, because the delay is not required. The sum of a spherical monopole and a spherical dipole creates a cardiod, no delay is required (this is shown in my book, but uses a spherical source, not point sources). The spherical source model is more accurate than a point source model because the first order enclosure diffraction is accounted for. This diffraction, may add the "delay", I don't know (I'd have to loook at the math in detail). More delay is required to "track" the cardiod directivity as "purely" cardiod, but that is not what I want to do. I want the Cardiod directivity to vanish as the frequency goes lower ending up with a monopole.
 
Let's not get more confused here. If we are talking about the on axis response and radiation pattern of an acoustic doublet or dipole it has noting to do with CMP, as you call it. There is no requirement that the sources of the doublet be MP. MP is a separate issure.

Well *I* am fine with the term CMP.

And yes, we are - or at least I am - talking about on axis response (or wherever you like) of a OB dipole, where you seem - with measurements and simus on the table - to refuse to accept the obvious.

might be I get you wrong - your proposal for correcting OB would possibly better clear the difference in our positions ....

Michael
 
Last edited:
Finally – the last contribution to the fun part of EQing CMP systems.


To prepare for the simu - imagine for a moment what happens when the source is shut down - something like a sine burst of finite length for example.
Until now I've only shown whats needed to keep the signal up and running 100% perfectly !

At shut down of the signal the first and second source get quiet - but the sound of second source still is arriving when the first one is no longer contributing.

Obviously that would result in a tail we would not like.
We actually would listen to monopole first - to dipole then - and then again to monopole (second source) at finish
:)

Best we simply supress that tail by destructive interference.

BUT
then we will have to supress the signal we needed to supress before and after that again and again and again …...


CMP_corr-3_2000Hz_burst.png




To summ up :

I introduced a method to 100% correct for CMP systems.
I introduced a method to 100% correct for CMP systems at deep nulls (100% destructive interference)
I introduced a method to 100% correct for the tail of CMP systems



Now after all this is clarified – lets come to the most serious part here:

I was stuck for a while on the conundrum of the baffle peak of OB systems. Although it is common practice to equalize this away, the EQ is typically done the frequency domain, while the error occurs in the time domain

The most comprehensive solution is correction in the time domain with digital equalization using FIR filters,

To possibly be a little bit more specific - correction is not really possible, neither in DSP nor otherwise.

As we can see the impulse response due to the "back" wave arrives latter than the "front" wave. I really cannot see how a simple minphase equalisation will lead to a grouping of the 2 single pulses!

It's the diffraction that is almost always present in a dipole that is non-minimum phase.

Earl, don't bet the ranch that diffraction from a baffle edge when combined with the direct sound isn't minimum phase. The combined response, direct sound + diffraction genially results in a minimum phase response, be it a dipole system or direct radiator. Kates wrote a paper on this back in '77.

Thanks guys for correcting me.
Yes I was wrong – well – partly wrong – or in other words : not completely right
;)
*But* you are not completely right either.
:D


the question now is :
Who's ranch now belongs to whom?

LOL


Michael
 
Last edited:
I think that this is a common misconception, because the delay is not required. The sum of a spherical monopole and a spherical dipole creates a cardiod, no delay is required (this is shown in my book, but uses a spherical source, not point sources). The spherical source model is more accurate than a point source model because the first order enclosure diffraction is accounted for. This diffraction, may add the "delay", I don't know (I'd have to loook at the math in detail). More delay is required to "track" the cardiod directivity as "purely" cardiod, but that is not what I want to do. I want the Cardiod directivity to vanish as the frequency goes lower ending up with a monopole.

It is not so easy as that. A dipole which has the same amplitude as the monopole creates a cardioid. That means either the either the monopole has to roll off at 6dB/octave or the dipole has to be eq'ed flat (and of course we are limiting the discussion to the response where the radiation pattern is a figure 8 for the dipole. Additionally, the resulting cardioid will have 6dB greater response on axis than either the eq'ed dipole or the monopole. (On axis the monopole will sum in phase with the dipole, at 180 degrees the monopole sums out of phase with the dipole yielding the cardioid null.)

But you don't have a dipole and a monopole. You have a monopole and an out of phase monopole rolled off and somewhat attenuated.

Ignoring the LP filter because it isn't relevant, you have a front driver and a rear driver out of phase separated by, nominally, the cabinet depth, let's say 40 cm. Assuming they are both at the same level then you have a dipole with peak at 430 Hz, and 6dB/octave roll off below. So right away you have a problem because the separation is too great to use it to 800-1000 Hz. So let's say hypothetically that the peak is at 1k Hz. That places the separation at 17.2 cm. Now, assuming the sources are of a size that they individually radiate omni-directional over the frequency range of interest, if the front and rear source are equal strength you have a dipole. If the rear source is reduced in level the response no longer decays to zero amplitude at zero Hz but levels off. Where it levels off it is omni-directional, but above that it is sort of a dipole with shallow nulls at 90 instead of complete cancellation. I don't see any cardioid behavior. The only way this could resemble a cardioid is if the front driver become directional as the frequency rises above the plateau level and the rear driver amplitude (at 180 degrees off the front axis) is shaped and delayed to cancel what the front driver does at 180 degrees. There has to be a delay because the location of the front acoustic source is physically different (further) than that of the rear source.

If you think of a cardioid based on a dipole and monopole then the arrangement is:

(+1)---d/2---(+1)---d/2---(-1)

That does not translated to

(+1)------d------(-r), where r<1.
 
Now that Michael has completed his thesis let me present a different look at reality.

This first figure shows the response of an ideal dipole with first peak at 2500 Hz. Recall that I have repeatedly stated that the useful frequency range of an ideal dipole is limited to about 1 octave below the first peak if constant directivity is to be maintained. That would limit the upper limit of the useful range to 1250 Hz. The upper figure is the frequency response on axis, the lower is the impulse. The double, inverted impulse is clear.

An externally hosted image should be here but it was not working when we last tested it.


The next figure shows the response of the ideal dipole equalized to flat over the useful range, with a -3dB point of 10 Hz. Again the upper figure is the frequency response on axis and the lower is the impulse. As is apparent, just this equalization does a lot to "correct" the impulse response.

An externally hosted image should be here but it was not working when we last tested it.


The last figure compares the dipole response (red) when filtered by text book LR4 HP and LP filters to create a bandpass response from 50 to 1250 Hz with the ideal 50-1250 LR4 band pass. It should be apparent that the filtered dipole response is very close to the ideal band pass in amplitude with only small errors well into the stop band. Similarly, the impulse response of this dipole band pass is just about identical to the ideal LR4 band pass impulse. Recall that the output of any system is its impulse response convolved with the input. Thus, since the impulse is the same (or nearly so) the output to any input will also be the same.

An externally hosted image should be here but it was not working when we last tested it.


Now it can be argued that the tail of the impulse will mean that any input will have a tail that decays in time past the cut off of the input. That is true. But it is also true that this is an artifact of the band pass response and not that it is arrived at trough a dipole.
 
After having had a lot of fun in demonstrating the principle of CMP ( consecutively min phase behaviour ) and its treatment, we possibly should jump to what briefly can be said on the topic in general.

1.) We are facing CMP behavior anytime when we have to deal with constructive and destructive interference caused by virtual sources – be it by reflections, diffraction, looped echos or whatever
2.) CMP behaviour itself is not dependant on frequency
3.) the most basic characteristic of a CMP system is that there is time delay involved
4.) as a direct consequence of the delay time involved, we are facing a frequency response that depends on the time we look at
5.) as a direct consequence of the delay time involved, we are facing a time span - at the beginning and at the end of any sound reproduced - that causes CMP distortion if not corrected
6.) there is a method to 100% correct for CMP distortion
7.) 100% correction of CMP distortion requires both: „unlimited“ resorces and „infinte time“ treatment

Did I miss something?

Michael
 
Last edited: