Electrostatic speakers?

[olieb]

do you haver the link?

DuckDuckGo can help

First time ESL builder

Thx bolsert ! SO this is why audioststic had suck large openes ? Because they 3 mm D/S spacing

www.diyaudio.com

EDIT: These very nice looking pictures are simulations though, and while there is a lot of interesting stuff in the linked thread I could not find any measurements. These might present some surprises (see below in @Arindal's comment).

 

[Arindal]

do you haver the link?

I think this was referring to a DIY project employing a horizontally segmented stator architecture ESL. A little bit like a line source variant of Quad ESL63/2905 stator. Have read about it some years ago.

From transducer theory this is a proper solution to achieve broader horizontal listening window and reduce the ´head-in-vice´-problem. On the other hand you might have more problems with limited diaphragm area for frequency bands in which several segments are already inactive (upper midrange in particular), and in theory you lose diaphragm control hence increase resonance issues. As the planar foil is still one piece with the central parts responsible for upper midrange and treble are ´floating´ with no surround/stiff suspension, one might expect a lot of pretty funny, unpredictable issues with breakup, bending waves and resonances.

Just from theoretical point I would prefer the Martin Logan solution with arc-shaped diaphragm and vertically segmented suspension to prevent long-wavelength resonances.

 

[Newman]

...I had read the new posts starting a few days ago, and mainly only saw discussions about radiation patterns...that didn't appear to have much if any myth content.

Heaps of sighted listening reports as if they are in the air waves, though. And did you notice how the radiation pattern realities are pegged to the sighted listening reports as explanations? The myth being that first they are hearing something real, then correctly identifying the radiation pattern as the cause. Whereas in truth it's more like they already know about the radiation pattern, then they start 'hearing' the presumed sonic attributes of such radiation patterns. Placebo effect.

...When it comes to electrostats radiation patterns, I have pretty strong ideas. I see them as operating as line arrays in both vertical and horizontal dimensions. I think Harry Olson's classic line array work describes them well under that viewpoint.

Far from it. See below where you fleshed out your thoughts. (BTW 'line array' describes a string of separate drive units: a large/long stat panel is a line source.)

I was just looking for what you considered to be myth within what was recently discussed.

See first paragraph above.

Must say, reading your full myth buster reply, I can't get on board with parts of it.

That's okay, let's dig a little deeper into your areas of concern.

Stat's vertical beaming is an asset imo. Operating as a line array with the tightest c2c spacing possible for VHF drivers, lets the same sound arrive at any vertical height up to the top of the panels.

The large stat panels we see in practice are nothing like proper line arrays, either vertically or horizontally. In order to do so they would need to touch both the floor and the ceiling (and for horizontally, both side walls). Instead, they behave as truncated line arrays, which are one of the worst configurations in all of audio. And this is the problem: get it 100% right and a proper line source is excellent, but get it 90% right (eg 90% of the room height but, nonetheless, truncated) and it flips and becomes poor.

And limits energy heading to the ceiling for destructive reflected comb filtering.

Hang on, Toole has shown that comb filtering via room surfaces is little to no concern. (Contrastingly he says that comb filtering within the direct sound itself is a very big concern.)

Horizontal beaming...stat's Achilles heel. . A horizonatal line array simply calls for an extremely precise listening position, ime. (Yes, I've turned straight line arrays on their sides)
I've never heard Quads, and know nothing about the rings. I've heard plenty of flat, segmented, and curved panels which all require precise listening focus, some more than others.
I postulate that the vertical line array action (better way to describe than beaming, as beaming only occurs above the height of the panels),
and the same line array action in the horizontal plane that requires precise perpendicular orientation to the horizontal lien,
is what helps give stats their unique sound....along with no crossover, and being a full range membrane.

Full range stats are in the minority, so the others have crossovers, and yet are still described as unique.

Hey, here is a thought that fully explains it all: the sighted listening effect! Why run around with postulates of cause and effect (what I have been known to describe as 'mythattribution'), when the most blatant potential cause has not been accounted for and controlled out of the listening tests? Occam's Razor and all that.

We will forever get nowhere as long as audiophiles deduct from their keyboards and not from valid tests.

Sorry, but imnsho this is just flat wrong. Square waves are an excellent test, better and more precise than Fourier analysis really.
A perfect impulse, an excellent flat mag and flat phase transfer function..... produce near perfect square waves...I say near perfect because I've found square waves to be more discerning than FFT
Anyone who disses the legitimacy of square waves, is basically just saying they don't place value on phase and/or time alignment.
The argument about high rez peering into the leading edge and corners is a copout imo...the lack of rez is simply due to the limited bandwidth of audio. Square waves must be evaluated compared to what approaches perfect within the audio bandwidth.

I have argued before that a square wave test is a great investigative tool, but to perfect the square wave reproduction is not an appropriate audio goal .

In summary:

- square wave reproduction is a useful test for analysing the performance of some audio gear.

- square wave reproduction is NOT a useful goal for the performance of audio gear.

- square waves do not occur in music, in air.

- even if a square wave was impossibly created in air, by magic, we could only hear part of it. And that part could be precisely simulated using sine waves of 20 Hz to 20 kHz.

- slew rate maximisation would be desirable if we could hear magic square waves, but in the real world each component has a 'sufficient' max slew rate, which it is not worth the trouble of exceeding by more than a decent 'padded margin' for comfort.

The pro-square wave argument hinges on whether the inaudible part of a square wave (10 kHz as an example) affects the audible part in a way that is audible. If the answer is 'yes', then a hifi component needs to reproduce square waves accurately to sound accurate. If the answer is 'no', it doesn't. As far as I know, the answer is 'no'.

The 1 kHz square wave output of a Quad ESL-63 looks like this:

© Stereophile

Does that leading edge look impressive? Shows fast transient response? Well, this graphs shows how, even with only the first, third and fifth fundamental (in this case 1, 3 and 5 kHz sine waves), you can build a square wave that has edges nearly as fast as having every fundamental up to the 101st (in this case, 101 kHz). It is really hard for an amateur to read a single square wave in isolation and tell if it required many fundamentals or few (ie much bandwidth or little) to get the 'look' of it.

Another example of the problem: the argument that better square waves matter is an argument that high-res digital audio matters. It is clear-as-crystal that high-res digital trounces 16/44 in square wave reproduction, so, if the principle is true that square waves matter, then high-res audio will sound better. I dare you to demonstrate that!

Consider this: a speaker can trounce another speaker for square wave performance, yet be much worse in controlled listening tests. Toole and Olive can tell you all about such speakers. This alone should tell us that it is very emphatically the wrong test. We need to focus on measures for loudspeaker performance that correlate to preference, not ones that we have a logical argument should be important but can't prove experimentally (and actually have partial evidence of negative correlation).

Yep, add all the above together, and we have a showroom stunt. None of which denies that it is a useful development tool to analyse problems with audio gear. Just don't make it a goal, or 99% of your performance and cost will go into inaudible things. Square wave analysis of a component is an excellent tool for examining the performance of a piece of gear, but it ('replicating' a square wave from input to output) is absolutely not a good goal for audio gear, i.e. many people get into the mistaken mindset that if component A can output a better square wave than component B, then component A is a better audio component. Pursuing this goal can lead to weird, unnecessary and even dysfunctional designs/formats/constructions. Beyond a certain point you can only make a square wave better by dysfunctional choices, i.e. adding more and more ultrasonic information until the audible content becomes a minute part of the signal, and the cost of delivering more and more ultrasonic information is completely wasted, and potentially dysfunctional - you end up trying to build speakers and analog power amps that deliver infinite power at infinite frequencies with no side effects, and very little of its output power is in the audible range.

I think open-baffle subs that have more dipole like radiation, cardoiod like to the sides, may make a decent improvement matching stats that radiate front and rear. Dunno, but it makes some sense.

Lots of things make some sense, but when put to the test, do they pass? Isn't it more rational to be sceptical until we see it tested? That way, we don't lead ourselves astray with logic in the absence of evidence (which, BTW, is a natural temptation).

Gotta define room correction. If you mean acoustical, totally agree. If you mean DRC, disagree other than for knocking down modes with a few judicious PEQs.

What I mean is this: the notion that stats don't integrate well with subwoofers (and the idea that dipole subs will be the answer because it 'makes sense') is at least 40 years old, so the experiences being gathered at that time, and for decades thereafter, were definitely involving subwoofers playing all the room modes full bore. Which is the perfect way to create 'slow bass', among other ills. No wonder no sub at all sounded better.

fwiw, you seem to sometimes kill both myths and facts together...which puts one leg into the subjective camp as surely as believing myths..
That said, I do appreciate your myth busting work

Happy to be corrected where I 'kill facts' ie throw the baby out with the bath water. Not seeing that above...yet.

cheers

 

[Newman]

...Supressing early reflections and doing so in a frequency-independent manner, is always an audible advantage even if it comes at a price.

Which science are you referring to?

Toole's description of the experiments that found listener preference increases when side wall reflections are allowed, for stereo playback.

 

[Newman]

this is so crazy. a shame audiophiles love to show of their system.....else they wouldn't insist on wide dispersion and we would have better options of speakers that limit room interaction.

Quite the converse. It has been demonstrated that comb filter effects off of the room boundaries look bad on a graph, but have little correlation to listener preference.

The crazy thing is that audiophiles continue to be inappropriately influenced by graphs like the one you are responding to.

 

[Newman]

I think this was referring to a DIY project employing a horizontally segmented stator architecture ESL. A little bit like a line source variant of Quad ESL63/2905 stator. Have read about it some years ago.

Correct. It was a theoretical beam width chart, not actual measurements.

From transducer theory this is a proper solution to achieve broader horizontal listening window and reduce the ´head-in-vice´-problem. On the other hand you might have more problems with limited diaphragm area for frequency bands in which several segments are already inactive (upper midrange in particular), and in theory you lose diaphragm control hence increase resonance issues. As the planar foil is still one piece with the central parts responsible for upper midrange and treble are ´floating´ with no surround/stiff suspension, one might expect a lot of pretty funny, unpredictable issues with breakup, bending waves and resonances.

Just from theoretical point I would prefer the Martin Logan solution with arc-shaped diaphragm and vertically segmented suspension to prevent long-wavelength resonances.

Oh no, from a theoretical point the Martin Logan solution is relatively ineffective.

 

[Arindal]

And did you notice how the radiation pattern realities are pegged to the sighted listening reports as explanations? The myth being that first they are hearing something real, then correctly identifying the radiation pattern as the cause.

FYI i have been conducting several blind listening tests comparing different radiation pattern variants of one and the same speaker with no other parameters changed. Admittingly this did not involve stats but one aspect was lower midrange directivity (direct-radiating midwoofer vs. cardioid) and that is pretty similar to what dipoles and planer stats are showing as a major difference to conventional concepts in terms of directivity.

Instead, they behave as truncated line arrays, which are one of the worst configurations in all of audio ... and it flips and becomes poor.

Why so? What exactly becomes poor when a truncated line provides vertical directivity?

Me thinks the majority of popular concepts in sound reinforcement belong to this category as well, and they got overwhelmingly popular for good reasons.

Toole has shown that comb filtering via room surfaces is little to no concern. (Contrastingly he says that comb filtering within the direct sound itself is a very big concern.)

That is kind of a contradiction, at least for living rooms, as early reflections from side walls, ceiling or floor easily come within a time window which is influencing direct sound hence defining tonality and localization.

I doubt he has specifically shown that comb-filtering is irrelevant. Could you link this statement please?

here is a thought that fully explains it all: the sighted listening effect!

That might deliver an explanation for high end listeners who are familiar what had been said about stats, admittingly.

But I have been involved in blind listening tests with stats vs. conventional speakers with several trained listeners taking part to whom stats were unfamiliar, so bias can be ruled out. While many claims (for example about impulse response, localization precision and natural ambience) about stats could not have been backed up, claims of midrange transparency, imaging stability and reduced proximity were found to be true.

the notion that stats don't integrate well with subwoofers (and the idea that dipole subs will be the answer because it 'makes sense') is at least 40 years old, so the experiences being gathered at that time, and for decades thereafter, were definitely involving subwoofers playing all the room modes full bore. Which is the perfect way to create 'slow bass', among other ills. No wonder no sub at all sounded better.

That might have been the case some 30 years ago. In the meantime, subwoofers with room correction and EQ have appeared, some even as part of hybrid stat concepts (Martin Logan was at the forefront here). Can say from own experience that this did not help much as even post-EQ the difference in upper bass/lower mid decay was audible. The picture changes dramatically if you use a dipole or cardioid woofer. Admittingly these are not classic subwoofers but the x-over freq is higher so lower midrange decay could have contributed to the problem.

Toole's description of the experiments that found listener preference increases when side wall reflections are allowed, for stereo playback.

I was not aware that such experiments were specifically asking for a verdict on imaging, localization stability and alike. Listeners asked for ´overall preference´ might have subjectively preferred reflection-laden reproduction with deteriorated imaging/localization, but that really does not say anything about the reproduction quality. It might have just sounded more pleasant (confirming Dr. Bose´s claims from 1967 leading to the 901 reflecting speaker) or reduced recording properties such as annoyingly thin monaural phantom sources.

If you postulate that what listeners in an overall preference test described as ´preferred´, is scientifically righteous, you have to grant the same to every potential aspect of reproduction which people subjectively like, even if it might qualify as a flaw from technical perspective.

from a theoretical point the Martin Logan solution is relatively ineffective.

Why so? The vertical directivity is retained, the horizontal window is controlled and as narrow as it is reasonable for a more ´beaming´ speaker in exactly the frequency range this is desired. The main difference of the arc-shaped panel is that instead of indefinitely narrowing down the beam towards higher frequencies (which a flat panel would deliver), the arc is ensuring a constantly moderate listening window of +-10deg, particularly above the frequency when the panel is wider than the wavelength (above 2K in this case).

FYI the more current models also segment the planar foil vertically by clamping it between some suspension clamps. So vertical resonances of longer wavelengths and breakup is pretty well attenuated, and the whole thing behaves more like a line array than a line source.

 

[Newman]

That is kind of a contradiction, at least for living rooms, as early reflections from side walls, ceiling or floor easily come within a time window which is influencing direct sound hence defining tonality and localization.

Have you studied any of his books in detail?

I doubt he has specifically shown that comb-filtering is irrelevant. Could you link this statement please?

Get the books. Jumping to his conclusion: "... the often-mentioned comb filtering engendered by early reflections is not a problem. None of these (controlled experiment) listeners heard it, or at least they didn’t comment on it. If it is reflected in their subjective ratings, it appears to have had a positive effect." - Sound Reproduction, first edition, pg 261

 

[gnarly]

It is not only a matter of positioning and angling. Both vertically and horizontally narrow beaming caused by a planar diaphragm forming an even wavefront, also poses problems as it is pretty frequency-dependent.

I've found it helpful to fundamentally think about stats/planars operating as line sources in both the vertical and horizontal dimensions. I think classic line source math that takes into account the length of the line and the tightness of the inter driver spacing, can be applied...... to vertical for sure, as the vertical line is usually straight. Application to horizontal for flat panels is easy too, whereas curved or segmented panels get tricky of course.

In my mind, beaming is not an appropriate way to describe a stat's vertical radiation, which is that of a truncated line source. Any listening/measuring height within the height of the speaker will have the same high-mid/HF/VHF performance...especially so given the near continuous interdriver spacing a stat or ribbons can provide.

If the flat panel stat's diaphragm is full range, the only frequency dependent aspect of an even wavefront would seem to be simply based on the diaphragm's dimensions.

I still have a pair of Acoustat-X and Martin Logan CLS, although both are in the closet at the moment, as I'm too busy comparing different DIY MEH versions.
I can't say the curved CLS does any better job with horizontal dispersion than the 3 angled segments of the Acoustat-X....but honestly, it's kind of a big who cares as both are single seat tight focus speakers. It's just what they are, are part of their sound.

Thanks for the info on the Quads....twas new to me.

From a theoretical standpoint, yes, but what is this good for in terms of reproduction quality? It is almost useless, if not misleading. The resulting graph is mostly a result of the reproduced frequency range plus phase shift or group delay issues between the sine frequencies contained in the squarewave.

No, it is not about disputing phase issues as a whole, but about asking which alterations are audible and which not. A squarewave or single step response, are almost useless for this. Audible changes (like low frequency group delay distortion) are not visible, and visibly pretty disturbing alterations to the signals (such as a low-oder allpass or a simple crossover somewhere in the middle of a the spectrum) are not audible because the phase shift is below the thresholds.

I think this, like @newmans reply which I'll get to next, miss my point. Squares waves form an identity with our FFT measurements.
As speaker's transfer functions and impulse responses improve, so do measured square waves. That simple.

To the extent one believes that improving speakers transfer functions and impulse responses has audible benefit, whether they realize it or not, the same can be said for square wave response.

I'm not advocating the use of square waves, as the art of reading them has largely passed in our era of FFT. And FFT is a damn sight easier to glean info from.
I'm just saying we have to be real, and acknowledge square waves are a very legitimate, albeit somewhat antiquated, test of speakers.
You just better have excellent speakers if you play with square waves, or you'll swear they have no value !!!

Personally, I really dislike when technical measurement discussions turn into questions of audibility. Audibility is an entirely separate issue.
And frankly, discussions about phase audibility in home audio I most often call BS, because they are almost always within the context of the enormous masking that takes place in small small rooms, and they are often without having speakers that are worth of comparisons to begin with. Or comparisons made with headphones, which are universally inappropriate for low freq tests.

 

[Fidji]

Have you studied any of his books in detail?


Get the books. Jumping to his conclusion: "... the often-mentioned comb filtering engendered by early reflections is not a problem. None of these (controlled experiment) listeners heard it, or at least they didn’t comment on it. If it is reflected in their subjective ratings, it appears to have had a positive effect." - Sound Reproduction, first edition, pg 261

Proper control of early reflections is crucial part of acoustic design for high performant systems.

 

[Jazzman53]

do you haver the link?

Yes, I do.

It took a while to find it among the huge number of threads about electrical segmentation on the DIY Audio Forum.
Here's that specific post: https://www.diyaudio.com/community/threads/first-time-esl-builder.246846/page-7#post-4163636

Some background info on the author of that post, and on how electrical segmentation became the cutting edge for DIY ESL builders:
Steve Bolser (a.k. a. "Bolserst" on the DIY Audio Forum) is a respected ESL guru / mathematician / aerospace engineer in the Seattle area. Steve created the 'Segmented ESL Calculator' (Excel Spreadsheet Program) that I and many other ESL builders use to design and tailor the dispersion of our segmented panels.

Another major contributor is ESL guru & physicist, Dr. Rod White in New Zealand (a.k.a. 'Golfnut' on the DIY Audio Forum).

My understanding is that Steve Bolser developed the ESL Calculator in collaboration with Dr. White, based on Dr. White's 2009 paper published in the Journal of the Audio Engineering Society, which is available here: https://www.researchgate.net/public...c_Loudspeaker_with_a_Zero-Free_Polar_Response

Consequently; most ESL builders on the DIY Audio Forum are now building segmented wire-stator or segmented PCB-stator designs (perf-metal stators can't be easily segmented)-- enabled by Steve Bolser and Dr. White freely sharing their knowledge on the forum.

I encourage anyone who's interested in this topic to search threads by 'Bolserst' and 'Glolfnut' on the DIY Audio Forum.

Jazz

 

[tengiz]

Just from theoretical point I would prefer the Martin Logan solution with arc-shaped diaphragm and vertically segmented suspension to prevent long-wavelength resonances.

I’ve always wondered whether better horizontal directivity could also be achieved by using a diaphragm with variable horizontal “sensitivity” - highest at the center and gradually reduced toward the edges, like applying a weighting window. I think it might be possible to control this by varying the electrical resistivity of the membrane material. This would be in addition to the arc-shaped panel and vertical segmentation. Of course, I have no idea how feasible it would be to actually manufacture such a diaphragm.

 

[Duke]

Just from theoretical point I would prefer the Martin Logan solution with arc-shaped diaphragm and vertically segmented suspension to prevent long-wavelength resonances.

SoundLab electrostats offer an arguably well thought-out combination of attributes. Disclaimer - I'm a dealer. And in sighted listening I think they sound good. So please everyone, discount my opinion accordingly.

SoundLabs use a segmented horizontal curve which gives a relatively wide and uniform radiation pattern, the horizontal angle between the panels being small enough that there is no "picket fence" effect at audible frequencies.

Each vertical facet is divided into "cells" which vary in size such that their drum-head resonances are spread out in frequency in a way that helps to offset dipole cancellation. I've measured SoundLabs in-room and they have low-end extension down into the low 20's. Of course their large size helps with that.

Here is a photo of the back of a SoundLab panel. You can see that the horizontally-segmented curve is arranged to converge to the rear and diverge forward. You can also see the gradation in the vertical size of the individual cells. The entire diaphragm is a single membrane (receiving a full-range signal), and is clamped by the frame, effectively dividing it into the many individual cells:
.

(The small vertical lines you see in the above photo are the insulated stator wires. There is also a hard plastic "waffle" in each cell for structural rigidity. The large rectangles - formed by the 7 horizontal divisions and the multiple vertical divisions which grow larger towards the top and bottom - are what I'm referring to as "cells".)

SoundLabs seek to approximate a line source, and the recommendation is that their height be a little bit less than the height of the room, such that the floor and ceiling reflections effectively extend the listening distance to which they approximate line source behavior.

SoundLab offers two radiation pattern widths: 90 degrees, and 45 degrees. The photo above is of a 45-degree panel. I started out with the 90 degree panels 25 years ago, as that was the only choice in the full-sized models. Comparing the two makes for a pretty close to "apples to apples" comparison of the different radiation pattern widths, though of course the line-source-approximating, dipole SoundLabs are not representative of different radiation pattern widths in "conventional" speakers. The pattern of course widens to the familiar dipole figure-8 at frequencies too low for the specific panel geometry to control the pattern width.

Roger West of SoundLab has an online white paper describing their technology, if anyone is interested. He is not particularly focused on the fact that his approach results in relatively little spectral discrepancy between the direct sound and the reflection field, but imo that is among their best attributes, and largely distinguishes them from their competition.

 

[Jazzman53]

I think there may be some confusion about usage of the term "segmented" (or "segmentation") in this thread-- as it's being appled to two completely different concepts in several posts. Specifically; I feel the need to differentiate between what I would call physical segmentation and electrical segmentation.

An example of physical segmentation:
The previous post describes a Soundlab panel as "a segmented horizontal curve which gives a relatively wide and uniform radiation pattern".
in this example, "segmented horizontal curve" accuarately describe of a panel wherein the diaphragm is vertially sectioned into many narrow, flat facets that join together to form a horizontal a curve. The resulting wavefront is curved because the radiator itself is curved.

An example of electrical segmentation:
In another post, a "segmented" flat panel ESL is described wherein the driving stator-conductors (wires) are "segmented" into separate smaller groups, each receiving a different band of progressively attenuated frequencies-- to collectively project a wide and uniform radiation pattern.
In this case, a flat diaphragm projects a wavefront that's curved electroincally by the progressively attenuated phasing and frequency bands fed to the driving wire groups.

 

[Jazzman53]

I’ve always wondered whether better horizontal directivity could also be achieved by using a diaphragm with variable horizontal “sensitivity” - highest at the center and gradually reduced toward the edges, like applying a weighting window. I think it might be possible to control this by varying the electrical resistivity of the membrane material. This would be in addition to the arc-shaped panel and vertical segmentation. Of course, I have no idea how feasible it would be to actually manufacture such a diaphragm.

Kudos for thinking outside of the box but ESLs have such poor sensitivity to start with that progressively decreasing sensitivity from the panel center to its edges might be a bridge too far.

 

[misterdog]

The Quad 63 was first released in 1981, some 44 years ago. I can take a panel from those first speakers and install them in my 989's or some 2812's, they are basically the same speaker.

11 Jul 2017 — If Stereophile gave an award for Loudspeaker We've Most Frequently Reviewed, the hands-down winner would have to be the Quad ESL.The list of past and present Stereophile contributors who've written about the ESL's various incarnations includes John Atkinson, Martin Colloms, Anthony H. Cordesman, Art Dudley, Larry Greenhill, J. Gordon Holt, Ken Kessler, Dick Olsher, Herb Reichert, William Sommerwerck, Steven Stone, and Sam Tellig. The ESL-63 was John Atkinson's personal "Editor's Choice" in 1992, and the ESL-989, a successor to the ESL-63, was Stereophile's 2003 Loudspeaker of the Year and Product of the Year.
So over 44 years how many thousands or tens of thousand of these speakers have been sold ?
Are all the people who purchase them fools, or swayed by marketing?
yet some can choose quotes from various writers and sources of audio science to prove without doubt that these speakers have flaws...

Show me a perfect speaker, there isn't one.

What does " Eamonn Forde, long-time music industry journalist and book writer" have to say about them I wonder given that he obviously is not in favour of vinyl, my main source for 35 years.

Again vinyl is a flawed medium yet it can sound on a par with digital if you are prepared to spend £ 10,000 plus on a front end.
Perhaps certain contributors should start a thread entitled, the least flawed loudspeaker design, where arguments and debates can go on for decades.

 

[Newman]

Yes, I do.

It took a while to find it among the huge number of threads about electrical segmentation on the DIY Audio Forum.
Here's that specific post: https://www.diyaudio.com/community/threads/first-time-esl-builder.246846/page-7#post-4163636

Yes, that's the post I was familiar with and referring to when I said above that those are not measurements, they are simulations from theory.

Which is not to say I doubt they translate into measurements, but I never see anyone publishing such measurements. It would be nice to see what is happening in practice with real designs.

Of special note, and importance to some of the discussion of Martin Logan speakers in this thread, is the middle of the three simulations, being for a curved stat panel. It shows just what a poor solution it is to curve the panel: slightly better than the flat panel's laser beam, but nowhere near as effective as the segmented simulation. @Arindal

Some background info on the author of that post, and on how electrical segmentation became the cutting edge for DIY ESL builders:

Steve Bolser (a.k. a. "Bolserst" on the DIY Audio Forum) is a respected ESL guru / mathematician / aerospace engineer in the Seattle area. Steve created the 'Segmented ESL Calculator' (Excel Spreadsheet Program) that I and many other ESL builders use to design and tailor the dispersion of our segmented panels.

Another major contributor is ESL guru & physicist, Dr. Rod White in New Zealand (a.k.a. 'Golfnut' on the DIY Audio Forum).

My understanding is that Steve Bolser developed the ESL Calculator in collaboration with Dr. White, based on Dr. White's 2009 paper published in the Journal of the Audio Engineering Society, which is available here: https://www.researchgate.net/public...c_Loudspeaker_with_a_Zero-Free_Polar_Response

For people who don't want to deal with ResearchGate, Rod has published his paper on AES with free access, link.

Consequently; most ESL builders on the DIY Audio Forum are now building segmented wire-stator or segmented PCB-stator designs (perf-metal stators can't be easily segmented)-- enabled by Steve Bolser and Dr. White freely sharing their knowledge on the forum.

I encourage anyone who's interested in this topic to search threads by 'Bolserst' and 'Glolfnut' on the DIY Audio Forum.

Jazz

'Golfnut', actually.

cheers

 

[Newman]

Proper control of early reflections is crucial part of acoustic design for high performant systems.

Only if you mean, by 'early', reflections off the speaker itself, eg diffraction etc. Not if you mean room surfaces. I mean, I just showed you the experimental outcome. What didn't you understand?

 

[Jazzman53]

Yes, that's the post I was familiar with and referring to when I said above that those are not measurements, they are simulations from theory.

Which is not to say I doubt they translate into measurements, but I never see anyone publishing such measurements. It would be nice to see what is happening in practice with real designs.

Of special note, and importance to some of the discussion of Martin Logan speakers in this thread, is the middle of the three simulations, being for a curved stat panel. It shows just what a poor solution it is to curve the panel: slightly better than the flat panel's laser beam, but nowhere near as effective as the segmented simulation. @Arindal


For people who don't want to deal with ResearchGate, Rod has published his paper on AES with free access, link.


'Golfnut', actually.

cheers

HI Newman,

Thanks for the link to Dr. Rod's paper (I have a PDF copy and was unaware of that link). Also thanks for clarifying what I neglected to mention-- that those directivity sonograms are computer sims.

I'm not equipped to measure (and I have not measured) the polar response of my segmented flat panels. I can only attest to a night & day difference in dispersion patterns, in subjective listening tests of segmented versus unsegmented flat panels (I've built and owned both).

I once built a flat panel ESL with switch-selectable wide & narrow dispersion modes, which made is rather easy to do A/B comparisons. The stators were electrically segmented, and I inserted a multi-pole rotary switch to selectively leave engaged or disable the RC network. Switch OFF left the RC network engaged for WIDE mode, and switch ON disabled/jumped over it for NARROW mode. If anyone is interested, here's a link to the build page: http://jazzman-esl-page.blogspot.com/2008/03/a-segmented-wire-stator-esl-with.html

Interestingly; the panel in WIDE mode (segmentation enabled) exhibited an essentially flat frequency response from the top octave all the way down to the RC network's low-frequency cutoff at around 300Hz.

Whereas the panel in NARROW mode (segmentation disabled) exhibited a rising, peaky frequency response which is a characteristic of tall/narrow (12" x 48") unsegmented panels. I attribute the unsegmented response to the combine effects beaming highs and the onset of the dipole phase cancellation at around 700Hz, progressively rolling off frequencies below that threshold.

With the panel in NARROW mode (RC network disabled) I had to overlay EQ's to flatten the response. Whereas in WIDE mode (RC network enabled), no EQ's were needed (well, almost none-- I did pull the response down about 2db's at 3kHz, which is the old "BBC Dip").

The switch mode feature was fun for a while but the novelty quickly wore off because I not only had to power down to switch modes, lest arcing destroy the switch, but I also had to re-EQ the panel each time because the separate modes produced different response curves (as noted above).

Switching modes proved to be inconvenient and impractical, and I ended up listening in WIDE mode all the time anyway, so I omitted the switch mode feature in subsequent speaker builds.

Jazz

 

[tengiz]

I once built a flat panel ESL with switch-selectable wide & narrow dispersion modes, which made is rather easy to do A/B comparisons. The stators were electrically segmented, and I inserted a multi-pole rotary switch to selectively leave engaged or disable the RC network. Switch OFF left the RC network engaged for WIDE mode, and switch ON disabled/jumped over it for NARROW mode. If anyone is interested, here's a link to the build page: http://jazzman-esl-page.blogspot.com/2008/03/a-segmented-wire-stator-esl-with.html

This is so awesome!

My understanding is that the frequency-dependent phase differences across the segmented stators cause the diaphragm to move in a non-rigid fashion. Could this introduce any unintended modal behavior or similar effects, given how different it is from the motion of a uniformly driven, non-segmented diaphragm that simply oscillates between the stators?

Also, would it be viable to replace the discrete resistor network with a continuous resistive strip of appropriate linear resistance?