g: Slant Plate Acoustic Lenses formulaalexberg> If your findings are relevant, please share.
Has anyone Diyed a curved esl panel?
I have Acoustats that are flat, An MLogans that are all curved.
The Acoustats have only 25% open output of the panel, thats one way to get bass out of a esl,the ML have 50% open output of the panel so it hard to say what your hearing
But what i think is flat is BANG you get it all ,that is it, what ever the panel can putout
the ear get it, good bad all,you move it moves.
The curved more forgiving, sounds more like others panels, Maggys,rebbons,an some good pistons.
But just one mans O-pine
I have Acoustats that are flat, An MLogans that are all curved.
The Acoustats have only 25% open output of the panel, thats one way to get bass out of a esl,the ML have 50% open output of the panel so it hard to say what your hearing
But what i think is flat is BANG you get it all ,that is it, what ever the panel can putout
the ear get it, good bad all,you move it moves.
The curved more forgiving, sounds more like others panels, Maggys,rebbons,an some good pistons.
But just one mans O-pine
There are two problems with curved ESL membranes.
Vertical tension (along the cylinder axis) tends to cause wrinkles; the circumferential tension required to remove the wrinkles makes the membrane want to "saddle" closer to the back panel at the center between the horizontal supports. Martin Logan solves this with relatively close spaced horizontal supports, and I think also they tension the membrane with the back panel over-curved, then release the panel to final curvature, relieving some of the circumferential tension.
The second problem is that the change in membrane tension with displacement is higher outward than inward. The vertical tension creates a restoring force proportional to displacement either direction, but the horizontal tension (circumferential around the section of cylinder) aids displacement towards the cylinder axis but opposes motion away from it. This can cause waveform and intermodulation distortion proportional to displacement at lower frequencies.
A friend and I built an ESL with twelve 2"(50mm) wide segments angled at 5 degrees per segment, 72"(1.8m) tall. (Actually 4 panels 36" tall, stacked into stereo pairs). Even at 20kc, the segment patterns overlap to give a smooth polar pattern.
Their response extended down to 70 Hz in an infinite baffle (the doorway to the front porch, with plywood fill around the speaker). All we had was a swept sine wave generator and a RadioShack SPL meter, so room reflections were a problem for measuring it, but it sounds damn fine. As a dipole in the 12'x20' room, there is a hole in the response - it starts rolling off about 500 Hz, and then comes back up around the 70Hz resonance of the segments.
Vertical tension (along the cylinder axis) tends to cause wrinkles; the circumferential tension required to remove the wrinkles makes the membrane want to "saddle" closer to the back panel at the center between the horizontal supports. Martin Logan solves this with relatively close spaced horizontal supports, and I think also they tension the membrane with the back panel over-curved, then release the panel to final curvature, relieving some of the circumferential tension.
The second problem is that the change in membrane tension with displacement is higher outward than inward. The vertical tension creates a restoring force proportional to displacement either direction, but the horizontal tension (circumferential around the section of cylinder) aids displacement towards the cylinder axis but opposes motion away from it. This can cause waveform and intermodulation distortion proportional to displacement at lower frequencies.
A friend and I built an ESL with twelve 2"(50mm) wide segments angled at 5 degrees per segment, 72"(1.8m) tall. (Actually 4 panels 36" tall, stacked into stereo pairs). Even at 20kc, the segment patterns overlap to give a smooth polar pattern.
Their response extended down to 70 Hz in an infinite baffle (the doorway to the front porch, with plywood fill around the speaker). All we had was a swept sine wave generator and a RadioShack SPL meter, so room reflections were a problem for measuring it, but it sounds damn fine. As a dipole in the 12'x20' room, there is a hole in the response - it starts rolling off about 500 Hz, and then comes back up around the 70Hz resonance of the segments.
Hi technophile
I once made a setup with comparable small sized panels angled at around 7 degrees from each other and there wasn't a smooth dispersion from it (clearly audible with noise signals). I must say however that there was some 3 cm separation between the active areas of the different panels which might contribute to some interference effects. So I am curious about the way you build it; maybe you put the panels more closely together? Putting so many panels side to side to achieve a good dispersion would make the speaker quite big so I moved to electrical segmentation instead.
There might be some distortion at low frequencies which should be no deal for hybrid esl
I once made a setup with comparable small sized panels angled at around 7 degrees from each other and there wasn't a smooth dispersion from it (clearly audible with noise signals). I must say however that there was some 3 cm separation between the active areas of the different panels which might contribute to some interference effects. So I am curious about the way you build it; maybe you put the panels more closely together? Putting so many panels side to side to achieve a good dispersion would make the speaker quite big so I moved to electrical segmentation instead.
There might be some distortion at low frequencies which should be no deal for hybrid esl
We started with a continuous flat perforated metal back panel, and used a metal bender to make 5degree bends every 2 inches. We placed foam tape strips ~1/8 in/3mm wide and 1/16in/1.5mm thick on the bends. We stretched and heat tensioned a 7 micron mylar film that had been surface treated to make it slightly conductive; on the top of the membrane over foam tape spacer around the edge, we placed a copper tape with conductive adhesive to provide a path for charging the membrane, and an equipotential boundary of low impedance to minimize the effect of leakage currents. We put another set of foam spacer strips on top of the first, and then individual 2"/50mm wide screens on the spacers. Because the effective arc length is longer on this face, there were gaps on the foam spacers between screen segments. This allowed us to implement electrical segmentation first described by Kellog in US Patent 1983377. This effectively resulted in segments slightly less than 2"/50mm wide, physically separated by ~1/8"/3mm.
One important consideration we discovered was the effect of Transparency Index - see IPA Acoustics Handbook - Applications.
One important consideration we discovered was the effect of Transparency Index - see IPA Acoustics Handbook - Applications.
Did you use inductors for your electrical segmentation as taught in the Kellog Patent which describes an LC transmission line arrangment to feed a delayed full bandwidth signal to successive segments? Or, did you use the more common DIY approach of using resistors between segments which rolls off the high frequencies progressively more and more as the signal works it's way from segment to segment.
If you used inductors, can you share how you determined the apropriate values and where you sourced them from?
I find it amusing that Kellog, who is best known for his work with dynamic loudspeakers, taught concepts in his 1934 ESL patent that were later re-patented and brought to the fore-front by two well known companies.
Walker's US patent 3773984 covers the use of LC transmission lines for segmentation which was put to good use in QUAD's ESL63.
Final's US patent 7054456 covers the concept of driving a push-pull ESL by sending the stepped up audio signal to the diaphragm rather the usual(and argueably better) approach of sending differential audio voltages to that stators. Final launched a whole new line of ESLs with this "NEW" idea.
Inductors. Probably Digikey or Mouser, and I don't remember the particular values. they were matched to the segment capacitance so that we could get a reasonable reflected impedance to the amps (Carver cube) through the transformer. About 47k, and 150:1 turns ratio, so ~2 ohms, if memory serves. The file with actual values is at my friend's house.
Very interesting. So your use of the LC lumped transmission line concept was not to improve dispersion like QUAD, but to improve impedance match with the driving amplifier.
From the panel dimensions you posted: 12 sections of 2" x 72" with 1/16" D/S spacing:
Each section would have a capacitance of roughly 275pF.
Total panel capacitance would have been 3300pF which would be very difficult to drive full bandwidth(ie up to 20kHz) with a step up transformer unless it's leakage inductance was exeedingly low(<1uH for the 150:1 ratio you mentioned)
A few lumped transmission line formulas:
Zo = sqrt(L/C)
Ts = sqrt(L*C)
Fc = 1/(pi*sqrt(L*C)
Where
Zo = characteristic impedance
Ts = time delay per section
Fc = cut off frequency (above which impedance and response is no longer uniform)
L = inductance per section
C = capacitance per section
For your characteristic impedance of 47Kohm:
L = 600mH (inductance needed between each section
Fc = 24.8Khz (just above the audio band. Perhaps this might account for some of the HF behavior you attributed to the transparence index of your stators)
Ts = 12.8uS (roughly 0.17" time delay per section)
If the line is terminated with a 47Kohm resistance, the resulting impedance reflected thru the 150:1 step up transformer would be a nearly constant resistive 2 ohm for the whole of the audio band. A tough load, but evidently it could be driven full bandwidth by your Carver amplifier, unlike the 3300pF unsegmented load.
A couple follow up questions:
1) Did you feed the LC line from the inside edge? or outside edge of your panels.
The total time delay thru 12 sections would result in about 2" delay for the last last section relative to the first. Since your panels were already physically curved thru a 60 degree arc, perhaps the 2" electrical delay was not really an issue when compared to the physicall curvature of your panel.
2) I would be interested to know what inductors you used, if you can track down the part number. I know the voltage across the inductors is pretty small for frequencies < 1kHz, but by 10Khz the inductors would be seeing nearly the full voltage magnitude output from the step-up transformer secondary. Perhaps you used multi-section RF type inductors that could handle large voltages?
I have been tempted to experiment with LC transmission line ESL, but the lack of easily obtainable high inductance/voltage rated inductors has discouraged me.
Congrats on your successfull implementation.
Last edited:
Fed from one edge - it just changes the toe in a little. And the panels electrically were 2"X36", stacked vertically.
We did have a few inductors fail (arc) at high power, and never found a perfect cheap inductor.
We've experimented with smaller segments, but the construction gets complicated (it feels like it goes as n^2, for n segments), and the tolerances get more demanding.
We did have a few inductors fail (arc) at high power, and never found a perfect cheap inductor.
We've experimented with smaller segments, but the construction gets complicated (it feels like it goes as n^2, for n segments), and the tolerances get more demanding.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.