Electrostatic speakers?

[Willem]

I have owned Quad electrostats for nearly all my adult life. I first heard them in I think 1970, and I knew immediately that this was how real music reproduction should sound, so I started working as a student to save for them, and bought them in 1976 (using the Quad 33/303 amp that I had alrready bought in 1971). Of course, they had three limitations: they would not play very loud, they would not play very low, and the listening position was narrow. The first two did not matter too much as I was living in smallish appartments. Also, it was the time of vinyl and its inevitable low frequency limitations.
Years later, when I was earning a lot more and we had moved into a large home, I decided to upgrade to the new Quad 2805. These adressed all problems to a fairly large extent: they could play louder (but only with a rather bigger 2x140 watt Quad 606-2 because they were also less efficient), they extended lower (I measured them flat down to 37 Hz but they fall off a cliff below that), and the listening position was wider. Compared to the ELS57 the sound was even more 'in the room' rather than coming from speakers (maybe thanks to Peter Walker's idea of concentric circles with delay lines). I considered the larger 2905 but it would have blocked the view from a wonderful panorama window.
A few years later again I decided I could do with more bass extension after all, and perhaps a bit more power, so I bought a B&W PV1d subwoofer. It did not integrate that well, and the sound was a bit woolly, so I thought that maybe all those stories about integrating stats and subs were true. I then discovered that room modes could well be my problem, so I bought an Antimode 8033 dsp room eq. This cured the problem almost completely. The bass was suddenly clean and tight, and integration was seamless. So my conclusion is that the problem of integration is that of a dipole that excites far fewer room modes, and a sub that excites a lot more. Therefore, as some have indeed argued, a dipole sub would have been best, but there are limits to domestic toleration of bulky speakers (and there are virtually none on the European market). Similarly, I am now convinced that the so-called 'speed' of speakers has nothing to do with the speaker, but with room modes. My next purchase will be a second sub for an even smoother low frequency response and perhaps even better integration.

As for my old ELS57 speakers (and my Quad 33/303 amp) I still own them. They are in storage, but I recently had to get them out when my 2805s needed a repair (due to the glue used on the early production). The ELS57s were still very enjoyable, and the family resemblance was obvious, even though the modern 2805s are undeniably better in all respects. My ELS57s are still in good working order, but in Europe, I could have had them refurbished in Germany: https://www.quad-musik.de/index.php/de/ However, that is not cheap, and I am glad I decided on getting modern ones instead. So I have finally also decided to sell my 33/303/FM3 ELS57 system.

 

[andreasmaaan]

My picture is more nebulous than that. I just thought that there might be something to McKenna's intuition that a smaller-area larger-excursion (and therefore higher velocity for a given frequency) driver may produce more of something that audiophiles label a "punch" (whether this is a real thing or not is of course debatable) and associate with dynamic drivers versus other types. For a given SPL, the total acoustic energy imparted is the same, but it is denser at creation by a smaller dynamic driver than a larger electrostatic driver. I am speculating as to whether any such punch could be the perception of secondary sound waves caused by faster and larger particle displacements. I do not have a feel for the actual physics or numbers involved.

My picture for the past many years is the idealized textbook one of a vibrating flat plate flush with a wall wherein the laminarity assumption is built-in and the particle motions are too small to affect the state of the ambient fluid medium (and so linearity is built into the model). I am unaware of the research that may have been done into modeling non-idealized driver geometry and motion and more realistic fluid motion. So turbulence and interaction with solid boundaries are possibilities in my mind that my current state of ignorance of the actual physics involved prevents me from ruling out. Certainly, the classical transition Reynolds number range (based on mean velocity, diameter, viscosity and surface roughness) for flow in a pipe would be inapplicable here. The fluid velocity field in the vicinity of the driver is not likely to be in a laminated pattern, given that the fluid motion is oscillatory and the driver motion during typical music program material is not periodic and the amplitude changes continually, and that the driver displacement is not uniform across its surface area (there are edge effects) and that there are oscillating boundary layers propagating from cabinet or headphone walls towards the center of the driver. I do not know which, if any, fluid instability types may arise. All of the non-idealities may not make a whit of difference, but I am unable to estimate whether smaller-area higher-excursion drivers exaggerate any of the non-idealities to the point of making an audible difference, a sound signature of typical dynamic drivers versus typical electrostats.

Fair enough. My competence in terms of the physics isn't adequate to address these questions. However, keep in mind that the accepted model predicts particle velocities substantially below thresholds at which it could be said that fluid behaviour becomes "turbulent".

For example, just running a couple of quick sims here on some high-Xmax 6" woofers, pedicted particle velocity for a periodic sine wave at a voice coil displacement of 12mm is only around 3.5m/sec (maximum). So the model would need to be off by a factor of around 4 for the reality to approach any normal definition of turbulence, which would be equivalent to displacements of c. 50mm - utterly impossible for a small woofer. And that's already taking a small, very high-Xmax woofer and pushing it beyond its Xmax, which is not something that normally happens, at least not in terms of the listening that presumably forms the basis of commentaries on the perceived differences between dynamic driver and ESL sound...

Moreover, when we look at known turbulence problems in audio, for example with respect to port behaviour, the measured onset of distortion and compression does indeed begin to occur at SPLs at which the same model we're discussing here predicts the onset of turbulence. I can't see why the model would appear to be experimentally well supported in the case of ports, yet woefully inadequate in the case of direct radiators.

As I said though, I'm somewhat out of my depth here. It would be interesting to read a reply from someone with expertise in fluid dynamics.

 

[Wes]

...When they came out, most systems (especially in the UK) were still mono.

View attachment 104945

Is the speaker in that photo up against a porous divider or is that patterned wallpaper?

I am thinking about the back wave of course...

 

[Wes]

Let's try a Fill in the Blank test:

Everyone would own and prefer electrostatic speakers if only they were...

1. smaller
2. cheaper
3. punchier
4. had more bass

etc.

Please number your answers.

 

[AdamG]

So we’re going on a trash this speaker type group con list? I guess if it makes you feel better have at it!

 

[Duke]

I've seen Roger Sanders further reduce their already minimal early reflections by using a diagonal or semi-diagonal setup geometry.

@Duke can you expand on this, what do you mean by the geometries you describe?

By a "diagonal" setup, I mean, with a corner between and behind the speakers. This way the first reflections off the "front two"walls, which meet to form that front corner, are angled such that they miss the listening area. And in many cases the first reflections off the other two walls, which meet to form the back corner, are likewise going to miss the listening area.

Whether this is desirable or not is I suppose a matter of choice, but if imaging precision is your priority (rather than soundstage width), then imo it makes sense.

The argument for going a bit "semi-diagonal" would including smoothing out the woofer/room interaction a bit by having each of the woofers end up at different distances from the nearby room boundaries.

 

[lashto]

https://www.stereophile.com/content/quad-reference-esl-2805-loudspeaker-measurements
That distortion measurement is quite impressive

 

[mhardy6647]

Is the speaker in that photo up against a porous divider or is that patterned wallpaper?

I am thinking about the back wave of course...

You got me.
Looks like really tacky wallpaper to me.

 

[Sir Sanders Zingmore]

By a "diagonal" setup, I mean, with a corner between and behind the speakers. This way the first reflections off the "front two"walls, which meet to form that front corner, are angled such that they miss the listening area. And in many cases the first reflections off the other two walls, which meet to form the back corner, are likewise going to miss the listening area.

Whether this is desirable or not is I suppose a matter of choice, but if imaging precision is your priority (rather than soundstage width), then imo it makes sense.

The argument for going a bit "semi-diagonal" would including smoothing out the woofer/room interaction a bit by having each of the woofers end up at different distances from the nearby room boundaries.

aah yes I know exactly what you mean now and had actually spoken to Roger about it in the past, thanks.
He does it because as you say, it smooths out the woofer interactions. Because his speakers beam and because of dipole cancellation there are virtually no first reflections anyway so I don't think that is an issue with this setup

 

[Willem]

The wallpaper in that Quad advertisement is what the British of the 1950s thought of as modern.

 

[JustAnandaDourEyedDude]

Fair enough. My competence in terms of the physics isn't adequate to address these questions. However, keep in mind that the accepted model predicts particle velocities substantially below thresholds at which it could be said that fluid behaviour becomes "turbulent".

For example, just running a couple of quick sims here on some high-Xmax 6" woofers, pedicted particle velocity for a periodic sine wave at a voice coil displacement of 12mm is only around 3.5m/sec (maximum). So the model would need to be off by a factor of around 4 for the reality to approach any normal definition of turbulence, which would be equivalent to displacements of c. 50mm - utterly impossible for a small woofer. And that's already taking a small, very high-Xmax woofer and pushing it beyond its Xmax, which is not something that normally happens, at least not in terms of the listening that presumably forms the basis of commentaries on the perceived differences between dynamic driver and ESL sound...

Moreover, when we look at known turbulence problems in audio, for example with respect to port behaviour, the measured onset of distortion and compression does indeed begin to occur at SPLs at which the same model we're discussing here predicts the onset of turbulence. I can't see why the model would appear to be experimentally well supported in the case of ports, yet woefully inadequate in the case of direct radiators.

As I said though, I'm somewhat out of my depth here. It would be interesting to read a reply from someone with expertise in fluid dynamics.

Since you have run simulations, you already have a far better feel for the numbers involved than I do. 3.5m/s is indeed a pretty low speed (Mach number of less than 0.01), so the flow is within the regime where the particle motions can be considered "incompressible" as a good approximation, except for acoustic effects that they generate of course. The predictions you cite of the turbulence model you use indicate that the flow is probably indeed laminar. Laminar flow does not preclude there being unsteady vorticity in it, which would mean the driver would be acting on a non-uniform flowfield and that the resulting acoustics may be non-uniform.

Though I did not consider turbulence as the only possible departure from flow ideality that may lead to an audible difference (if any), it is interesting that you are using an accepted turbulence model presumably developed for acoustics applications. And that the model is calibrated with experimental data to even predict the transition boundary between laminar and turbulent flow is impressive. Although it boggles my mind that it could predict onset of turbulence for both ports and direct radiators, given that the geometries are quite different, though there is of course a commonality of unsteady oscillatory flow. In a Google Scholar search, I see a hit for a 2004 paper by Aktas and Farouk on acoustic streaming, where Google pulls up phrases like high streaming Reynolds numbers, turbulent streaming, turbulence model and maximum oscillatory flow velocity of 7m/s. So possibly the application-specific turbulence model you use may be based on research into acoustic streaming by oscillatory flows. I do not think the transition models used in codes in conjunction with standard RANS turbulence models would predict turbulence onset well in acoustics applications, because those transition and turbulence models were developed over 120 years of research in the context of (quasi-)steady flow over flat plates and in straight pipes of uniform circular cross-section and to some extent free shear flows. No universally valid turbulence model has ever been developed, nor is one likely to be found.

Something which I forgot to mention in my previous post is that in the case of the electrostatic driver, the plates that act as electrodes or capacitor plates ("stators" I think they are referred to as) would surely interfere with the flow caused by the diaphragm motion. I understand that in practice the stators are more lattices than plates, but there would still be some non-ideal effect. All of the preceding I am sure has been researched by at least the manufacturer's engineering teams. Edit: Also of course, the larger diameter or width of the electrostatic driver means that the ratio of driver diameter to wavelength (for a given tone) is different between typical dynamic drivers and electrostatic drivers in speakers, and also that the "nearfield" is larger for the latter than for the former. If trying to scale to equivalent listening positions relative to nearfield, the SPL would need to be raised accordingly for electrostats.

My knowledge and understanding of turbulence is woefully and shamefully inadequate, in spite of four decades of tangling with fluid dynamics . I comfort myself with the thought that some of the most brilliant physicists and mathematicians have tackled the topic over a span of more than a century, yet have failed to produce a satisfactory complete fundamental mathematical theory of turbulence, and it remains one of the great unsolved problems of physics. Though to be sure, they have made great inroads into the topic, and contributed much great insight and progress in understanding it. Fluid dynamics, like most general areas of scientific knowledge now, is a vast body of knowledge, and huge swaths of fluid dynamics remain outside my awareness. I was unaware of even the phenomenon of evanescent waves that was recently mentioned by NTK in a piano bass thread here, and understanding of that phenomenon does not even require consideration of viscous effects (be they laminar or turbulent)!

 

[MattHooper]

A few years later again I decided I could do with more bass extension after all, and perhaps a bit more power, so I bought a B&W PV1d subwoofer. It did not integrate that well, and the sound was a bit woolly, so I thought that maybe all those stories about integrating stats and subs were true. I then discovered that room modes could well be my problem, so I bought an Antimode 8033 dsp room eq. This cured the problem almost completely. The bass was suddenly clean and tight, and integration was seamless. So my conclusion is that the problem of integration is that of a dipole that excites far fewer room modes, and a sub that excites a lot more. Therefore, as some have indeed argued, a dipole sub would have been best, but there are limits to domestic toleration of bulky speakers (and there are virtually none on the European market). Similarly, I am now convinced that the so-called 'speed' of speakers has nothing to do with the speaker, but with room modes. My next purchase will be a second sub for an even smoother low frequency response and perhaps even better integration.

Very interesting! And the antimode integration does make sense.

That said, I don't feel fully on board with presuming the distinctive sound of stats - what people often think of as "speed/transparency/boxless" issue is only room nodes. I say that because I've had numerous stand mounted speakers with limited low frequency extension, which I presume are not likely exciting the room's bass nodes, and they all have that density/punch character that I haven't heard from stats. Even my tiny little Spendor S3/5s have it, even though they seem to utterly "disappear" as sound sources in terms of soundstaging and imaging, almost as fully as ESLs, but they still have that "box speaker" sound.

 

[Duke]

I then discovered that room modes could well be my problem... I am now convinced that the so-called 'speed' of speakers has nothing to do with the speaker, but with room modes.

I agree with your observation.

Many years ago, as a long-term hard-core amateur speaker builder, I embarked on a quest to build a subwoofer "fast enough" to "keep up with" Quads and Maggies, AND have good impact. So I built transmission lines of many different geometries, equalized dipoles, sealed boxes (equalized and not), low-tuned vented boxes, aperiodics, isobarics, pretty much everything except for horns. Nothing combined "fast enough" with impact. The equalized dipole sub came closest but didn't have much impact.

Then I had a conversation with Earl Geddes, in which he identified room modes as the issue and an asymmetrically distributed multi-sub system as a solution. So I tried that and imo it worked well, and I've been making such for fifteen years now.

My next purchase will be a second sub for an even smoother low frequency response and perhaps even better integration.

When you add your second sub, let me suggest introducing as much asymmetry as you reasonably can. Like if your first sub is along the front wall and closer to your left speaker, maybe put the second sub along the right-hand wall and closer to the rear of the room.

Also, if this second sub will be positioned well away from the main speakers (as I have suggested), then you want to make sure it doesn't pass upper bass energy loud enough to betray its presence. So imo you want a fairly steep lowpass filter (I use 4th order), and/or a fairly low lowpass filter frequency.

 

[Blumlein 88]

I've found if you have some box speakers with really good boxes. Either super thick or otherwise designed so nearly all the sound is from drivers and not the box, if you EQ to mimic a 'stat, then you start to get some of the same 'stat sound. It is pretty close, but no cigar. I would think the dipolar nature is the last part of it.

Most full range stats have a smiley faced response curve and truly are anemic in the bass though they might have a wide resonant peak below 100 hz that often can fool you into thinking they go lower than they do. I also would say the big Soundlabs go further toward having a real solid low end than any of the other stats I've owned or heard.

 

[andreasmaaan]

Since you have run simulations, you already have a far better feel for the numbers involved than I do. 3.5m/s is indeed a pretty low speed (Mach number of less than 0.01), so the flow is within the regime where the particle motions can be considered "incompressible" as a good approximation, except for acoustic effects that they generate of course. The predictions you cite of the turbulence model you use indicate that the flow is probably indeed laminar. Laminar flow does not preclude there being unsteady vorticity in it, which would mean the driver would be acting on a non-uniform flowfield and that the resulting acoustics may be non-uniform.

Though I did not consider turbulence as the only possible departure from flow ideality that may lead to an audible difference (if any), it is interesting that you are using an accepted turbulence model presumably developed for acoustics applications. And that the model is calibrated with experimental data to even predict the transition boundary between laminar and turbulent flow is impressive. Although it boggles my mind that it could predict onset of turbulence for both ports and direct radiators, given that the geometries are quite different, though there is of course a commonality of unsteady oscillatory flow. In a Google Scholar search, I see a hit for a 2004 paper by Aktas and Farouk on acoustic streaming, where Google pulls up phrases like high streaming Reynolds numbers, turbulent streaming, turbulence model and maximum oscillatory flow velocity of 7m/s. So possibly the application-specific turbulence model you use may be based on research into acoustic streaming by oscillatory flows. I do not think the transition models used in codes in conjunction with standard RANS turbulence models would predict turbulence onset well in acoustics applications, because those transition and turbulence models were developed over 120 years of research in the context of (quasi-)steady flow over flat plates and in straight pipes of uniform circular cross-section and to some extent free shear flows. No universally valid turbulence model has ever been developed, nor is one likely to be found.

Something which I forgot to mention in my previous post is that in the case of the electrostatic driver, the plates that act as electrodes or capacitor plates ("stators" I think they are referred to as) would surely interfere with the flow caused by the diaphragm motion. I understand that in practice the stators are more lattices than plates, but there would still be some non-ideal effect. All of the preceding I am sure has been researched by at least the manufacturer's engineering teams. Edit: Also of course, the larger diameter or width of the electrostatic driver means that the ratio of driver diameter to wavelength (for a given tone) is different between typical dynamic drivers and electrostatic drivers in speakers, and also that the "nearfield" is larger for the latter than for the former. If trying to scale to equivalent listening positions relative to nearfield, the SPL would need to be raised accordingly for electrostats.

My knowledge and understanding of turbulence is woefully and shamefully inadequate, in spite of four decades of tangling with fluid dynamics . I comfort myself with the thought that some of the most brilliant physicists and mathematicians have tackled the topic over a span of more than a century, yet have failed to produce a satisfactory complete fundamental mathematical theory of turbulence, and it remains one of the great unsolved problems of physics. Though to be sure, they have made great inroads into the topic, and contributed much great insight and progress in understanding it. Fluid dynamics, like most general areas of scientific knowledge now, is a vast body of knowledge, and huge swaths of fluid dynamics remain outside my awareness. I was unaware of even the phenomenon of evanescent waves that was recently mentioned by NTK in a piano bass thread here, and understanding of that phenomenon does not even require consideration of viscous effects (be they laminar or turbulent)!

Those are some fascinating snippets of insight into the topic of turbulence, and if your knowledge and understanding of the subject are woefully and shamefully inadequate, then I don't think mine would even make it onto the scale
When I said "It would be interesting to read a reply from someone with more expertise in the field of fluid dynamics", I didn't realise I had already received one.

Regarding the existing models used by acoustic modelling software, I should mention (if it's not clear already) that these predict air particle velocity only. To my knowledge, the secondary (but obviously more relevant) question as to the onset/extent of turbulence in a given system is not addressed (at least not by any software I am aware of). The designer is left to take the modelled air particle velocities and contemplate for themself whether turbulence is likely to be an issue. Being relatively ignorant of this field, I generally follow rules of thumb that may or may not be very accurate. As mentioned before, though, research I've come across concerning air particle velocity in ports and distortion/compression seems to track quite nicely with these rules of thumb. Interestingly, in the canonical textbooks on loudspeaker design that I've read, turbulence is dealt with in detail in respect of ports, but not even mentioned in respect of dynamic drivers. (That doesn't mean it can't possibly be an issue, of course!)

I completely agree with your other points regarding diaphragm:wavelength ratio, and suspect that this is a contributor to perceived (and of course measurable) differences between electrostat and dynamic driver sound - along with the radical differences in polar radiation.

 

[Blumlein 88]

http://www.phaselinearhistory.stereomanuals.com/andromeda.htm

Anyone ever hear these dipoles from Bob Carver? I have. They had something like the early Carver Holographic cross talk cancellation built into them. They actually sounded pretty good. The Holographic thing worked quite well if you had a big room to stay well away from the boundaries. And it was the most head in a vice speaker ever if that Holographic system was going to work. Widest imaging you'll ever hear I think.

Those funny looking drivers up top are like if you took a regular 4 inch midrange driver. Pointed it at the ceiling and mounted it in a baffle so half the cone is on the front and half on the back. Bizarre, but it worked.

 

[andreasmaaan]

My next purchase will be a second sub for an even smoother low frequency response and perhaps even better integration.

When you add your second sub, let me suggest introducing as much asymmetry as you reasonably can. Like if your first sub is along the front wall and closer to your left speaker, maybe put the second sub along the right-hand wall and closer to the rear of the room.

Or, to play devil's advocate, if your room is symmetrical, place them as symmetrically as possible, ideally in the centres of two opposing walls, to allow them to cancel (some) room modes

 

[RayDunzl]

Martin Logan - black
JBL LSR 308 - red

Speakers located with JBL adjacent to and outboard of the ML at the same height - JBL at the mid-height of the panels.

Reflections in impulse response

Room is symmetrical at the speaker end, open on the left rear corner

ML
7ms - dipole bounce off rear wall
27ms - double room length bounce

JBL
Several early reflections below 12ms
Higher level of reflections relative to direct sound


ML fires down the room and not to the sides or vertically. reflections are heard from the same direction as the direct sound

JBL sprays, reflections heard from locations "out of place". Assuming walls, floor, ceiling contributing.

ETC

JBL about 10dB and as mucch as 20dB less "Direct to Reflected Noise" ratio

My otherwise unsubstantiated opinion is the difference in the reflected energy is the main contribution to the "clear" sound often reported with panels.

The room is "quieter" with the ML playing at the same level as the JBL.

 

[T.J. McKenna]

I had the obverse view. Cones are trying to produce room filling sound at a very small spot of a cone. The sound intensity at the surface has to be extremely high. High enough you have the air in a non-linear zone. Spreading everything out over the big ole panel keeps surface sound intensity low where air remains linear.

But don't real instruments do this? Hit a snare and everything emanates from what is more or less a single point. Can you think of a single instrument where the sound is generated in similar fashion to an ESL?

 

[maverickronin]

But don't real instruments do this? Hit a snare and everything emanates from what is more or less a single point. Can you think of a single instrument where the sound is generated in similar fashion to an ESL?

Music Production =/= Music Reproduction