Electrostatic speakers?

[Keith_W]

I don't think Amir had any problems measuring it. He just needed more time than he was willing to put in to get super-accurate in the top two octaves.

You don't think that having to take 2000 measurements qualifies a speaker as "difficult to measure"? And he has a robot, imagine if you had to do it manually!

 

[Duke]

I don't think Amir had any problems measuring it. He just needed more time than he was willing to put in to get super-accurate in the top two octaves.

Agreed.

I've had several conversations with Amir about the feasibility of him measuring one of my designs which uses a rear-firing horn. Briefly, he would have to run enough measurements to describe the wave expansion from the speaker, and in his words the rear-firing energy would "create an extremely complex soundfield". He estimated it would take about 8-12 hours to run the measurements alone, which he described as putting "a ton" of wear-and-tear on his Klippel unit.

In other words, it's just not practical.

 

[misterdog]

Your friends must have really crappy speakers.
Keith

Well one of my friends had a demonstration of the Kii 3 that you sell and was underwhelmed.

Though two other friends have stated that the latest Purifi woofers are the best in the world.

Feel free to call up to Sheffield though and hear a properly integrated pair of ESL's.

You could bring your latest boxes and we could compare and contrast.

Will your taste in fresh coffee match mine?, who knows, everyone has different taste preferences.
Kenyan, Columbian 2nd wave, 3rd wave it's all a variation of the same - coffee.
Taste buds vary as do hearing preferences.

Always good though that Purifi are chasing the same goal that Peter Walker did back in the 1950/60's of the lowest possible distortion in audio reproduction.

As did Kellogg and Rice back in the 1920's when they used a pig intestine covered in gold leaf as the first electrostatic loudspeaker.

I have an auditorium available locally with a 3M high ceiling and space for 80 or so people, maybe we should arrange a bake off?
We could then conduct a survey of whether the latest and best dynamic drivers are considered better, equal, or worse than my Quad 989 ESL's.

 

[misterdog]

(better, more natural, plausible ambience) in the midrange.

Precisely.

Though musical taste also comes into play. Those who enjoy the visceral extremes of rock and electronica might miss the lowest octave that ESL's can't provide.
I provide this in my system with 2x12" dipole subs, though in the full knowledge that these drivers have some 10% distortion.
With their MiniDsp plate amps I can switch this off and on at will, there is little that my ESL's miss out on.

 

[Jazzman53]

Precisely.

Though musical taste also comes into play. Those who enjoy the visceral extremes of rock and electronica might miss the lowest octave that ESL's can't provide.
I provide this in my system with 2x12" dipole subs, though in the full knowledge that these drivers have some 10% distortion.
With their MiniDsp plate amps I can switch this off and on at will, there is little that my ESL's miss out on.

ESLs, and even most conventional speakers can benefit from subs. I have and recommend dipole subs too... a pair of RiPols worked for me.

 

[Newman]

Guys, anything works for anybody. That aspect of the hobby is all about the sighted listening effect, and doesn't say a darned thing about sound reproduction.

This thread keeps lapsing into Anecdote Disguised As Sound Comment, Attribution Error, and Myth Multiplication.

cheers

 

[Newman]

You don't think that having to take 2000 measurements qualifies a speaker as "difficult to measure"?

Only if you have "difficulty" with waiting while an unattended machine completes its task. Remember, you were musing that it might be incapable of properly measuring panel speakers, that it generically "fails".

And he has a robot, imagine if you had to do it manually!

Now I think you are arguing my case more than yours! Herr Klippel loves a good day's march.

cheers

 

[Purité Audio]

Well one of my friends had a demonstration of the Kii 3 that you sell and was underwhelmed.

Though two other friends have stated that the latest Purifi woofers are the best in the world.

Feel free to call up to Sheffield though and hear a properly integrated pair of ESL's.

You could bring your latest boxes and we could compare and contrast.

Will your taste in fresh coffee match mine?, who knows, everyone has different taste preferences.
Kenyan, Columbian 2nd wave, 3rd wave it's all a variation of the same - coffee.
Taste buds vary as do hearing preferences.

Always good though that Purifi are chasing the same goal that Peter Walker did back in the 1950/60's of the lowest possible distortion in audio reproduction.

As did Kellogg and Rice back in the 1920's when they used a pig intestine covered in gold leaf as the first electrostatic loudspeaker.

I have an auditorium available locally with a 3M high ceiling and space for 80 or so people, maybe we should arrange a bake off?
We could then conduct a survey of whether the latest and best dynamic drivers are considered better, equal, or worse than my Quad 989 ESL's.

I have heard many Quads/electrostats always underwhelmed, probably something back in the 50’s.
Keith

 

[gnarly]

The limitation with the Klippel NFS is that the choice of using spherical harmonics as the expansion functions means that the math is not efficient for sound sources that aren't compact (compact means small when compared to the wavelength of sound being produced). To "fit" the radiation pattern of a non-compact source like a panel speakers (or a line array) will require an enormous number of expansion functions, and therefore an enormous number of measurement points. (The number goes up fast once the driver size to wavelength ratio exceeds a certain limit, but I am not sure how fast. Power of 2? 3? Exponential?)

Thx for the explanation. If I remember correctly, doesn't the NFS system also need an additional math module (beyond what Amir currently has) for planars and lines?

 

[NTK]

Thx for the explanation. If I remember correctly, doesn't the NFS system also need an additional math module (beyond what Amir currently has) for planars and lines?

What Klippel's "line source" option does is to measure the line array one driver at a time, then sum (vector sum) the responses of all the drivers to get the final combined response. By measuring one driver at a time, the sound source becomes "compact" and the radiation pattern is much less complicated. Therefore it does not require a lot of measurement points for each of the measurements. (You do require to measure each driver because the diffractions from each driver will be different. These diffractions are dependent on the locations of the drivers relative to the entire speaker, which will be unique to each driver.)

This trick will not work for a continuous source like a panel since you can't activate only a small portion of the panel as a time. Also, if you have multi-driver passives, this method may not work well either because you'll have to perform surgery to the passive cross-over to have access to drive the individual drivers. (See below for the link to the source of this slide.)

You don't think that having to take 2000 measurements qualifies a speaker as "difficult to measure"? And he has a robot, imagine if you had to do it manually!

Found this slide from Klippel. Looks like the number of measurement points goes up to the square of frequency (square of N+1).

Using the provided formulas, if the source is 1 m long, f = 16 kHz, c = 343, N = (1/2)×(2π 16000/343) = 146.5. Rounding up N to 147, M is therefore (147 + 1)² ≈ 22000 or more measurement points!
https://www.klippel.de/fileadmin/klippel/Files/Know_How/Literature/Transparancies/Holographic directivity measurement of line sources and sound panels.pdf

 

[gnarly]

Thank you NTK. Yikes! Planars can really chew up a bunch of measurement points, huh?!

 

[hex168]

Agreed.

I've had several conversations with Amir about the feasibility of him measuring one of my designs which uses a rear-firing horn. Briefly, he would have to run enough measurements to describe the wave expansion from the speaker, and in his words the rear-firing energy would "create an extremely complex soundfield". He estimated it would take about 8-12 hours to run the measurements alone, which he described as putting "a ton" of wear-and-tear on his Klippel unit.

In other words, it's just not practical.

Can the rear-facing horn be measured separately on the Klippel from the forward facing speaker, and then combined with magic Klippel math? I'm reasoning based on the method to measure line array drivers separately for a less-complex sound field.

 

[bmc0]

Can the rear-facing horn be measured separately on the Klippel from the forward facing speaker, and then combined with magic Klippel math?

It's certainly mathematically possible. I'm not sure what the Klippel software offers natively, but combining multiple sources could be done fairly easily in an external script assuming one can export complex pressure (or magnitude and phase) vs frequency at defined points.

 

[Arindal]

Planars can really chew up a bunch of measurement points, huh?!

Planars and very large ribbons are not just big in terms of radiating diaphragm area (like line arrays). Some also show a tendency of diaphragm breakup, approximating the behavior of a bending-wave transducer, which means that at higher frequencies parts of the diaphragm move in opposite direction at the same time, creating an almost chaotic diffuse field. For anyone being familiar with NXT panels, the behavior is not dissimilar.

I doubt that mathematical models original created for a limited number of more of less pistonic transducers or parts of a diaphragm, are meant to represent such chaotic diffuse soundfield with mathematical precision. And even if this is possible with a sufficient number of measuring points and operations, I doubt the result will help us understanding how human ears surrounded by a human head, will perceive this soundfield.

The aforementioned hypothesis is IMHO applicable mainly to large foil diaphragms which are not subject to damping or stiff suspension. Segmented ESL are not affected.

Can the rear-facing horn be measured separately on the Klippel from the forward facing speaker, and then combined with magic Klippel math?

It should be possible, if delay and phase relations as well as relative geometrical arrangement can be represented by the underlying mathematical model.

 

[Duke]

Can the rear-facing horn be measured separately on the Klippel from the forward facing speaker, and then combined with magic Klippel math? I'm reasoning based on the method to measure line array drivers separately for a less-complex sound field.

I have zero hands-on experience with the Klippel, but what you described is how I think it would work. And I trust Amir's assessment of the complexity involved.

My understanding is that the interaction of the front-firing and rear-firing drivers inevitably results in a highly complex soundfield, exacerbated in a real-world application by the fact that the rear-firing drivers significantly interact with the front-firing drivers only AFTER their output bounces off at least one wall.

 

[dlaloum]

Planars and very large ribbons are not just big in terms of radiating diaphragm area (like line arrays). Some also show a tendency of diaphragm breakup, approximating the behavior of a bending-wave transducer, which means that at higher frequencies parts of the diaphragm move in opposite direction at the same time, creating an almost chaotic diffuse field. For anyone being familiar with NXT panels, the behavior is not dissimilar.

I doubt that mathematical models original created for a limited number of more of less pistonic transducers or parts of a diaphragm, are meant to represent such chaotic diffuse soundfield with mathematical precision. And even if this is possible with a sufficient number of measuring points and operations, I doubt the result will help us understanding how human ears surrounded by a human head, will perceive this soundfield.

The aforementioned hypothesis is IMHO applicable mainly to large foil diaphragms which are not subject to damping or stiff suspension. Segmented ESL are not affected.



It should be possible, if delay and phase relations as well as relative geometrical arrangement can be represented by the underlying mathematical model.

Many planars and ribbons overcome these issues by seperating - ie: running multiple different drivers (of differing shapes and sizes) for different parts of the frequency range, or using other mechanism (like the virtual point source used in the Quad ESL63 and later - with delay lines) to shape the performance, and avoid many of the issues that can be caused by a single full range, large radiating panel.

Panels just like dynamic drivers suffer from different breakup modes when run outside of their optimal frequency range...
The original Quad ESL (57) was a twin woofer, single tweeter setup - as one example.

 

[Arindal]

@dlaloum we should separate two things here: frequency-selective segmentation of the motor (which in case of ESL almost necessarily means segmentation of the stator, not the diaphragm, you named Quad 63 as an example) and physical segmentation, i.e. segmented suspension, of the diaphragm (Soundfield, Martin Logan).

The motor segmentation might address some directivity issues by preventing the active zone from becoming overly huge for short wavelengths.

As i understand it, it does not have much impact on the diaphragm behavior related to breakup, unwanted bending wave behavior, resonances and alike. Imagine the center of a Quad 2805 or alike: We have a circular are of roughly 10cm in diameter, exciting solely the center of the foil for frequencies of 10K and above. As the foil is flexible, with very little inner damping and no suspension anywhere near, this will cause de facto bending wave behavior, breakup and resonances beyond the active zone, and even within that one, as the motor force per diaphragm area is not strong enough to force the diaphragm to pistonic behavior in case of a resonance or breakup.

Panels just like dynamic drivers suffer from different breakup modes when run outside of their optimal frequency range...

You cannot compare these cases, as the vast majority of pistonic dynamic tends to have a comparably strong suspension (surround or spider), stronger motor per diaphragm area, are smaller compared to the emanated wavelengths, and in most cases their diaphragms offer much higher stiffness or inner damping compared to a foil of a planar ESL.

There are similarities with some bending wave transducers, as I understand it, for example NXT and Manger MSW (The latter offering immense inner damping and circular motor structure, so very different diaphragm behavior), as well as tall, non-suspended ribbons if you look at one dimension solely (Magnepan, Apogee). Very large, very light fullrange cones, like Lowther, might also be filed under this category, but the majority of dynamic drivers cannot.

Defining ´Optimal frequency range´ as the one ensuring pistonic behavior and minimum diaphragm area relative to the wavelength, single-diaphragm panels are mostly operating outside their optimal frequency range, if I understand them correctly.

As mentioned, all of this is not applicable to planars with the foil itself being segmented, for example by a grid, like ML and Soundfield are doing it.

This said, I still have sympathy for certain aspects of sound quality that very good ESL can deliver, which are difficult to achieve with conventional dynamic concepts. For treble bands, I see AMTs as the single most promising solution to achieve both at the same time, but this is a completely personal judgement.

 

[MarkS]

I liked electrostatics ever since I saw them on Friends as a kid View attachment 519144

And the Emmy for "worst possible placement of electrostatic speakers on a sitcom set" goes to ... Friends!

 

[gnarly]

Planars and very large ribbons are not just big in terms of radiating diaphragm area (like line arrays). Some also show a tendency of diaphragm breakup, approximating the behavior of a bending-wave transducer, which means that at higher frequencies parts of the diaphragm move in opposite direction at the same time, creating an almost chaotic diffuse field. For anyone being familiar with NXT panels, the behavior is not dissimilar.

I doubt that mathematical models original created for a limited number of more of less pistonic transducers or parts of a diaphragm, are meant to represent such chaotic diffuse soundfield with mathematical precision. And even if this is possible with a sufficient number of measuring points and operations, I doubt the result will help us understanding how human ears surrounded by a human head, will perceive this soundfield.

The aforementioned hypothesis is IMHO applicable mainly to large foil diaphragms which are not subject to damping or stiff suspension. Segmented ESL are not affected.

I have to admit I don't put much credence into the more micro type discussions about transducers' properties/behaviors, and seldom give attention to them.
As a DIY speaker designer/builder of about every type speaker (except electrostats), I've learned it's easy enough to measure transducers in test jigs and prototypes, to find where pistonic ends and breakups begin, as well as get a grip on polars. (And not care so much about why.. ..just me...I'm curious enough, but simply more interested in find best sound I can, even it through excessive trial and error.)

I'm all about high-level pragmatic engineering.
From that vantage point, here is how I see electrostats offering a unique sound, or at least a sound different from conventional baffled & boxed speakers.
First, the well recognized dipole radiation.
Second, and at least as important imo, is that being planars they behave as line arrays in both the vertical and horizontal dimensions.

Vertical line array behavior is generally a given. I view it as a big in-room positive, owing to lines reduction in the lobing from floor and ceiling bounce. Highly desirable.

My positive experiences with DIY floor-to-ceiling and tall CBT line arrays, led me to see that is exactly what electrostats/planars are doing vertically, subject to the same line array principles regarding length of line and center-to-center spacings of line elements.

The implied c2c spacing is part of electrostat's different sound I think. The tighter and continuous c2c spacing (especially for non segmented diaphragm), let's the line behave as line for all frequencies down to where length on line becomes controlling. This is a real boon for VHF, where most all more conventional lines can't achieve necessary c2c tightness.

I think horizontal line array behavior is generally viewed as problematic, with attempts to either curve the diaphragm horizontally or segment it into vertical stripes. I still have the original CLS, which can sound awesome on certain material, but are too colored for most tracks. (I really need to break them out of the closet and EQ them.) I also have a pair of Acoustat-X with the high voltage direct drive amps, that use three identical panel segments for horizontal dispersion. Still use these for comparisons, still love them.
No EQ really needed as they have a variable HF adjustment.

Here's close range transfer, no processing /EQs, other than the amps' HF gain control.
Red is HF gain full up. Green full down.

Phase is the last part of the unique sound a think.
And not just a flat phase trace from not having a crossover(s). It's flat no matter where you measure off the diaphragm.
Flat phase can be done for any speaker, at least on-axis. Heck, I've had flat phase on everything I've built, typically 5-ways including sub as a way.
But good ole separated drivers' geometry changes phase off-axis. ...even on my syn/MEHs which have considerably tighter c2c's than conventional designs.

Anyway, my 2c on why I think stats sound a bit unique. My 2$ comment is my ears hear it

And let me go ahead and preempt any "sighted listening bias", horse-crap replies.
I've owned and used the Acoustat-X for nearly 50 years. (CLS, although less used, for 30 yr) I'm quite used to their looks and size...no romance there folks.
Heck, they are kinda small compared to a number/most of my DIYs. And double heck, I don't give a rat's ass what a speaker looks like.

I experiment so much, with many different DIY designs, and many different processing techniques...FIR and IIR...DIY's all built strictly off transfer functions and impulse Reponses.
I audibly compare DIYs to DIYs, both within same type to across types (syns vs lines vs coax vs conventional, etc), ... and compare to the stats, a couple of Meyer rigs, various old speakers I have...even a pair of 901 series VI hanging in the garage lol.
Yes, all sighted. I say so what. It's is a rare event when I hear something that stands out as a real improvement. When horses for courses is taken into account, even with a real improvement in a particular design, I know it is only good for the particular course the speaker was designed for.

In the end I care about one thing and one thing only...HOW DOES IT SOUND? (They all wiggle a little different....And I'm a measurement freak!)

 

[hex168]

I have to admit I don't put much credence into the more micro type discussions about transducers' properties/behaviors, and seldom give attention to them.
As a DIY speaker designer/builder of about every type speaker (except electrostats), I've learned it's easy enough to measure transducers in test jigs and prototypes, to find where pistonic ends and breakups begin, as well as get a grip on polars. (And not care so much about why.. ..just me...I'm curious enough, but simply more interested in find best sound I can, even it through excessive trial and error.)

I'm all about high-level pragmatic engineering.
From that vantage point, here is how I see electrostats offering a unique sound, or at least a sound different from conventional baffled & boxed speakers.
First, the well recognized dipole radiation.
Second, and at least as important imo, is that being planars they behave as line arrays in both the vertical and horizontal dimensions.

Vertical line array behavior is generally a given. I view it as a big in-room positive, owing to lines reduction in the lobing from floor and ceiling bounce. Highly desirable.

My positive experiences with DIY floor-to-ceiling and tall CBT line arrays, led me to see that is exactly what electrostats/planars are doing vertically, subject to the same line array principles regarding length of line and center-to-center spacings of line elements.

The implied c2c spacing is part of electrostat's different sound I think. The tighter and continuous c2c spacing (especially for non segmented diaphragm), let's the line behave as line for all frequencies down to where length on line becomes controlling. This is a real boon for VHF, where most all more conventional lines can't achieve necessary c2c tightness.

I think horizontal line array behavior is generally viewed as problematic, with attempts to either curve the diaphragm horizontally or segment it into vertical stripes. I still have the original CLS, which can sound awesome on certain material, but are too colored for most tracks. (I really need to break them out of the closet and EQ them.) I also have a pair of Acoustat-X with the high voltage direct drive amps, that use three identical panel segments for horizontal dispersion. Still use these for comparisons, still love them.
No EQ really needed as they have a variable HF adjustment.

View attachment 520024

Here's close range transfer, no processing /EQs, other than the amps' HF gain control.
Red is HF gain full up. Green full down.

View attachment 520025

Phase is the last part of the unique sound a think.
And not just a flat phase trace from not having a crossover(s). It's flat no matter where you measure off the diaphragm.
Flat phase can be done for any speaker, at least on-axis. Heck, I've had flat phase on everything I've built, typically 5-ways including sub as a way.
But good ole separated drivers' geometry changes phase off-axis. ...even on my syn/MEHs which have considerably tighter c2c's than conventional designs.

Anyway, my 2c on why I think stats sound a bit unique. My 2$ comment is my ears hear it

And let me go ahead and preempt any "sighted listening bias", horse-crap replies.
I've owned and used the Acoustat-X for nearly 50 years. (CLS, although less used, for 30 yr) I'm quite used to their looks and size...no romance there folks.
Heck, they are kinda small compared to a number/most of my DIYs. And double heck, I don't give a rat's ass what a speaker looks like.

I experiment so much, with many different DIY designs, and many different processing techniques...FIR and IIR...DIY's all built strictly off transfer functions and impulse Reponses.
I audibly compare DIYs to DIYs, both within same type to across types (syns vs lines vs coax vs conventional, etc), ... and compare to the stats, a couple of Meyer rigs, various old speakers I have...even a pair of 901 series VI hanging in the garage lol.
Yes, all sighted. I say so what. It's is a rare event when I hear something that stands out as a real improvement. When horses for courses is taken into account, even with a real improvement in a particular design, I know it is only good for the particular course the speaker was designed for.

In the end I care about one thing and one thing only...HOW DOES IT SOUND? (They all wiggle a little different....And I'm a measurement freak!)

"They all wiggle a little different" - I love that line and am filing the serial number off it right now.