diyAudio (
-   Planars & Exotics (
-   -   angles for segmented ESL? (

Few 14th February 2005 05:43 PM

angles for segmented ESL?
I'm thinking about a multi-segment ESL that would approximate a quarter of a cylinder and fit well in the corners of a room. I'm worried about the "venitian blind effect" (variation in the high frequencies as you move your head horizontally while listening to the speakers). My quick calculations suggest the effect wouldn't be too bad if each ESL panel were 4" wide, and angled by 10 degrees relative to its neighboring panels. Can anyone tell me if I'm way off base?

I came up with 10 degrees by looking for the angle at which the output from a 4" wide panel, radiating a 15 KHz tone, would drop to one half of its on-axis output. I got something like 11 degrees, which I'm rounding down to 10 degrees. The 4" panel width comes from my experience with flat-panel ESL's using 1/16" diaphragm-to-stator spacing. I figure I'd need nine or ten such panels to form the quarter-cylinder. I'll deal with the large capacitance issue later...

I know the angled segments idea isn't new, but I don't have personal experience with it. Can anyone offer any advice or insights?

Thanks in advance.

I_Forgot 15th February 2005 03:18 AM

I built one like that
several years ago and used 6 (or was it 8?- probably 6) vertical strips per channel. If I recall correctly (I have then in a box somewhere) they were about 24" high and 4" wide. I believe I used an angle of about 170 degrees between adjacent strips. The insulators were made of 1/16" fiberglass PCB. THAT was a pain to cut!

I mounted the strips on two pieces of aluminum L stock that had been notched and bent to the desired angle. It was a real easy to mount them that way compared to trying to cut wood at several small angles. The two aluminum pieces were spaced vertically by a couple side pieces of wood.

I did not find any venetian blind effect, but even if it were there, you would not notice it unless you were moving around in front of the speakers. It will be a bigger problem near-field than far out from the speaker.

While sort-of cylindrical configuration worked for dispersing the high frequencies over a wider area (only a little wider area), I ultimately decided I liked the sound of flat panel ESLs better. The imaging of flat panels is really good. As soon as you disperse the highs around the room you have problems controlling reflections and that somewhat reduces the sharpness of the imaging.

Martin Logan builds ESLs with curved panels. I used to think it was for high frequency dispersion, but the reality is that the high frequency dispersion isn't aided much by gently curving the panels. The curve is there to make the assembly rigid. I found that out when I built a pair of 2' x 4' flat panel ESLs just to see if I could make them that size. While they worked OK, they had one big problem. 2'x4' perforated steel is heavy. When you suspend two sheets by their edges n an ESL, it acts like a drum. If you bang the frame of one of those speakers with your hand, the driver wobbles back and forth for quite a while. I consider those speakers a failure but a lesson learned.


Few 15th February 2005 01:40 PM

Thanks I_F, that helps. I like your method of mounting the angled panels.

I too have built a large flat panel ESL system. There are many features I like, but I do find I often listen off-axis when I'm doing something other than just listening--or when several people are watching a movie. Widening the listening area is one of my main motivations for thinking about the segmented approach.

Another motivation is that I'm interested in testing the idea that if you mount a quarter cylinder in a corner there won't be any early reflections from room walls (at least to a first approximation). I really like the imaging associated with mini-monitors mounted away from the walls but, at least from the (naive?) theoretical point of view, it seems true corner mounting might be an alternative solution to the early reflections problem. Trying to do this with dynamic drivers would be difficult because you can rarely get the radiating surface of the speaker close enough to the room boundaries, at least at high frequencies. It seems like it would be a cinch with ESL panels. It would also reduce the diffraction problem evident in so many conventional speakers.

I'd be interested to hear about the results of other people's efforts to corner-mount or even wall-mount drivers---ESL or otherwise. Is there a flaw in my reasoning?


Brian Beck 15th February 2005 03:21 PM


I think your plans are quite worthwhile. If you consider the placement of a quarter-cylinder radiator tight into the room corners from a reflected-image perspective, you can imagine an effective full cylinder radiator whose center line is the corner of the room. Therefore, reflections from either adjacent wall will behave the same as emissions from the rest of a virtual cylinder radiator, three quarters of which is “behind the walls”. If you make the speaker reach from floor to ceiling, you will even pick up image reflections from the floor and ceiling which will extend the line source ad infinitum. Symmetry is a wonderful tool, isn’t it?

Imaging should be excellent since you will be listening to what appears to be an infinitely small line source at the corner. This is so because the cylinder replicates the expanding wave front a millisecond a two after emission from a line source. The only concern I would have has to do with controlling the back wave, which is now trapped in the corner. I would want to fill the corner with absorbent material whose impedance increases (thicker material) as you approach the corner so as to absorb the back wave. Otherwise if the radius of the cylinder is say, 1 foot, you will have a reflected impulse coming back through the drivers at around 2 milliseconds after the first impulse. But this ought to be very controllable.

As to the “picket fence” effect, I think you should not worry too much. The effect is best described as lobing, not as a “picket fence” response anyway. There shouldn’t be any deep nulls. Any driver-to-driver angle less than 20 degrees (or stated as greater than 160 degrees) ought to show very smooth response. Only at the top of the audio range will there be some gentle lobing, but in any case this ought to be less severe that the sinx/x lobing from a planar driver!

Let us know how it works out!

SY 15th February 2005 07:08 PM

Do you not worry that the backwave reflected out from the corner will pass through the relatively transparent diaphragm and cause some severe interference effects?

Few 15th February 2005 10:52 PM

Thanks for the responses and encouragement. Brian, your description of the images forming a complete cylinder whose center is at the corner of the room is exactly what I had in mind. I also agree with all the comments about the reflections from the walls, but am (for the moment) optimistic about the possibility of absorbing most of them. I figure I'd need a two foot radius quarter cylinder, and that would provide quite a bit of room for sound absorbing material behind the panels. I'd be limiting the ESL to frequencies that are strongly absorbed by acoustic foam or wool because I'd plan on crossing over to dynamic drivers around 400-500 Hz anyway.

I remain intrigued! Thanks again for the helpful ideas.


Brian Beck 15th February 2005 11:09 PM


Well yes, that's why I mention the absorptive material. Remember that regular cone drivers in a box also suffer from this reflection effect, although the transmissibility of cones will be lower. With a large quarter cylinder ESL as “Few” intends, it should be possible to put enough absorptive material in the enclosed area to drop the intensity of the round-trip reflection to acceptable levels. I think it should be graded material, with the highest impedance at the corner due to pressure concentration there, but that’s probably not critical. With the many other advantages of this kind of scheme, it would seem a shame to abandon it due to worries about this problem. And don’t forget that the other frontal “reflections” actually build the desired wave front rather than add unwanted delayed response, so that this kind of scheme ought to suffer from far fewer room interaction effects compared to regular speakers, even with the one back wave reflection to contend with. Just MHO.

I_Forgot 15th February 2005 11:20 PM

The back wave would certainly reflect off the walls and come back through the speakers with some delay. It would be hard to predict what that would do to the sound because the distances and angles between the walls and each driver strip will vary. At high frequencies it may behave like a comb filter. At lower frequencies, maybe just some selective reinforcement/cancellation. I suspect that like almost all other ESL installations, it would be best to absorb the backwave.

The nice thing is it is pretty easy to try it out, and then we'll know for sure how it behaves. If it doesn't work out too well, the panels are still useable to make a flat speaker...


Brian Beck 16th February 2005 02:02 AM

Actually it’s not that hard to predict the results of the reflections in this case – if we assume that the drivers are arrayed in a perfect quarter circle. This assumption applies for all but the highest frequencies where chord and arc differ a bit, but then those same high frequencies are very easily absorbed anyway.

I’ve been working on a full-range ESL cylindrical speaker intended for use in a large room far away from walls, and I’ve had to consider most of this already.

Start by imagining a perfectly cylindrical speaker. There are no nearby surfaces. You’re looking down into it, so that you see a circle in cross section. Now imagine driving this speaker with a single positive-going impulse (which represents all frequencies, of course). Now you’d see a positive pressure spike growing outward from the speaker in an ever-expanding circle, concentric with the speaker, heading towards the listener who is about to hear the closest approximation to an impulse he’s probably ever heard. Inside the speaker, you’d see an ever shrinking circle moving toward the center, but with negative pressure compared to the outbound pressure front. When this circle collapses into a point in the center for an instant, the volume has been traded for pressure and we have a point source of very high (negative) pressure (really a line source in the Z-axis, but viewed on end). Now the pressure point becomes a source for an ever expanding circle which heads back toward the drivers, still in inverse polarity compared to the outbound wave. When this circle hits the driver circle, some energy will be reflected back toward the center again for another round trip, but most will pass through to follow the main wave front. If the speaker has a radius of r, then the negative pressure ring will follow the first by the time corresponding to 2r. The listener will observe a nice positive impulse followed at 2r-time by a weaker negative-going, smeared impulse (and then further reflections every 2r times). This will result in the familiar comb filter response at very predictable periodic frequency multiples.

Now if we consider the driver transmissibility (to waves passing through) as a function of frequency and consider viscosity, resistive losses and the effects of grills etc, the reflection will be corrupted, but surely it will be at least low-pass filtered in some way. I’ve done simulations of these effects and have gotten good correlation with reality. These effects will smear the reflected impulse shape in time and reduce its total energy. Now, let’s add some absorptive material inside the cylinder. If we can achieve at least 20 dB of attenuation across the 2r round trip, we would see just 1.7dB ripple instead of the severe comb filter peaks and valleys without attenuation (if driver transmissibility were as bad as to be unity). If we could achieve 40dB of 2r attenuation (at higher frequencies perhaps), the ripple would be 0.17 dB – not too bad!

Now take a quarter section and push it tight into the corner. The walls prevent flow normal to their planes, but allow flow in the shear direction, which is the same as predicted by symmetry for a whole cylinder without walls. Shades of Klipsch and Allison! Said crudely, if any one driver tries to push air “sideways” into its neighbor’s pie section, the neighbor driver pushes back equally hard and no air can flow tangentially or circumferentially – it all moves radially, away from the center (or towards it). This exactly matches the boundary condition imposed by the walls in a quarter section – or, for that matter in a half cylinder section pushed up against a long wall, far from any corner. Now fill that quarter cylinder with a pie-shaped chunk of the same absorber from the cylinder, and you’ve got the same reduction in ripple due to reflections. Ain’t symmetry grand?

I’ll say this again: If anyone thinks that these reflections don’t plague other kinds of speakers, think again. Most other speakers in boxes have all kinds of irregular and frequency-dependent wave fronts inside, much of which do pass through the cone to color the tone. And point source and planar speakers just don’t make constructive use of wall reflections as does a cylinder section. This is why Roger West sells his SALLIE back wave absorber for his magnificent speakers, but I digress. Planars are far from wavelength independent, because each frequency is treated differently in comparison with speaker dimensions – Fourier transforms of apertures and all that. And then there’s dipole cancellation. Cylinders and cylinder sections against walls can circumvent much of this wavelength dependence, and that means flat, phase coherent sound – which is our goal - if we just consume the back wave.

Sorry for the long post…

SY 16th February 2005 03:07 AM

Brian, a very nice explanation. Thanks.

Reflection from internal surfaces does indeed happen with cone drivers, but the cones are much less acoustically transparent than film diaphragms. Nonetheless, the TLs I've built have a slant board behind the woofer just for this reason. It doesn't eliminate the effect, but it does reduce it considerably.

All times are GMT. The time now is 03:36 PM.

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

Content Relevant URLs by vBSEO 3.3.2