Pressure source vs volume velocity source... a bit tricky, here goes...
You could think of the electrical equivalents as a voltage source (pressure) and a current source (volume velocity), with both driving the acoustic resistance of the air.
With a conventional speaker, the motion of the cone is almost entirely determined by the force from the electromagnetic motor and the mass of the cone. This motion would be the almost the same in the absence of air. Think of the cone sweeping out a given volume per cycle (hence volume velocity). So the conventional speaker has the same displacement independent of the atmospheric pressure. As the atmospheric pressure goes up, the acoustic resistance goes up, and the acoustic pressure generated by the speaker increases.
The electrostatic motor applies a constant force per unit area to the membrane, and if the membrane is massless, 100% of the pressure is applied to the air. If the atmospheric pressure drops, acoustic resistance drops, the acoustic pressure must still be the same, so the displacement of the membrane increases.
So at high altitudes (Denver or Black Hawk say - low atmospheric pressures), the SPL from conventional speaker goes down. With an ESL, the SPL stays the same, but the membrane displacement goes up, and they cannot play as loud without the membrane hitting the stator.
Note - I've omitted the complicating factor of different frequency dependence of the acoustic resistance for the two speakers, but hopefully you get the idea.
You could think of the electrical equivalents as a voltage source (pressure) and a current source (volume velocity), with both driving the acoustic resistance of the air.
With a conventional speaker, the motion of the cone is almost entirely determined by the force from the electromagnetic motor and the mass of the cone. This motion would be the almost the same in the absence of air. Think of the cone sweeping out a given volume per cycle (hence volume velocity). So the conventional speaker has the same displacement independent of the atmospheric pressure. As the atmospheric pressure goes up, the acoustic resistance goes up, and the acoustic pressure generated by the speaker increases.
The electrostatic motor applies a constant force per unit area to the membrane, and if the membrane is massless, 100% of the pressure is applied to the air. If the atmospheric pressure drops, acoustic resistance drops, the acoustic pressure must still be the same, so the displacement of the membrane increases.
So at high altitudes (Denver or Black Hawk say - low atmospheric pressures), the SPL from conventional speaker goes down. With an ESL, the SPL stays the same, but the membrane displacement goes up, and they cannot play as loud without the membrane hitting the stator.
Note - I've omitted the complicating factor of different frequency dependence of the acoustic resistance for the two speakers, but hopefully you get the idea.
Before my copy of "Einstein at Breakfast" reaches me, I'd like to ask a physicist a question.
Forever I've been ridiculing cone-drivers air mismatch with "... shaking a heavy piece of cardboard at thin air..." as compared to light-as-air ESLs. But can you or Bolserst provide an everyday tactile analogy to impedance mismatch?
Like in the form of: riding a see-saw with a 2 lb bag of sand, an automobile going up a hill, 3 ice cubes in a bucket of water....
B.
Gears on a bike, going up a hill. The transmission of energy is optimized with things like horns. Bass ports acoustically amplify through resonance, but horns couple the transfer of volume velocity to acoustic pressure, like gears they optimize efficiency in energy transfer by matching the load at the point of transfer.
Hi B
“…provide an everyday tactile analogy to impedance mismatch?”
Hmm, another tricky one. I’ve not checked this, but the following is what I expect to happen….
Imagine you have a longish piece of rope, and you shake one end and send a pulse down the rope...(silicon rubber tubing might be better if you want to do the experiment – something with not much loss)
If the far end of the rope is waving free in the breeze, it will shake about and the pulse will return down the rope back towards you, it will have been reflected.
If the rope is anchored on a wall, the wave will again be reflected, and it will come back towards you, but the shape of the pulse will be upside down from before.
If the rope is connected to another piece of rope of the same mass per unit length, the wave will not be reflected at all but continue down the second piece of rope.
You can also imagine that when the second piece of rope has a different mass per unit length, that the wave is partially reflected, and partially transmitted – also the phase of the reflection depends on whether the second piece is heavier or lighter that the first. 100% of the energy is transferred only when the impedances are matched. The reflectance or transmittance of the interface between the two regions depends on the ratio of the characteristic impedances.
The same thing happens with all sorts of different waves when they cross boundaries into regions of different characteristic impedance. An acoustic wave in air for example is reflected off the surface of water, because of the impedance mismatch.
An interesting trick is to insert a third piece of rope between the first two, but with a carefully chosen intermediate mass per unit length. Then there is a reflection from both interfaces. If the spacing between the two interfaces is correct, the two reflections will be out of phase and cancel each other– you get no reflection and 100% transmittance - this is how antireflection coatings on glasses work. But it only works over a narrow range of frequencies – the wavelength has to be right.
Another trick is to have a graded connection between the two section of rope, say several different sections, all with the same impedance ratio to their neighbours (ie an exponential match) then there will also be very little reflection. This is how stealth technologies work, e.g on a submarine – multiple layers of rubber of different densities to provide an impedance match between the water and the steel. In this way there will be very little acoustic reflection from the surface and it will be invisible to sonar. The rubber is also lossy so the energy is dissipated in the rubber. The big rubber wedges in anechoic rooms do the same thing.
Hope this makes sense
R
“…provide an everyday tactile analogy to impedance mismatch?”
Hmm, another tricky one. I’ve not checked this, but the following is what I expect to happen….
Imagine you have a longish piece of rope, and you shake one end and send a pulse down the rope...(silicon rubber tubing might be better if you want to do the experiment – something with not much loss)
If the far end of the rope is waving free in the breeze, it will shake about and the pulse will return down the rope back towards you, it will have been reflected.
If the rope is anchored on a wall, the wave will again be reflected, and it will come back towards you, but the shape of the pulse will be upside down from before.
If the rope is connected to another piece of rope of the same mass per unit length, the wave will not be reflected at all but continue down the second piece of rope.
You can also imagine that when the second piece of rope has a different mass per unit length, that the wave is partially reflected, and partially transmitted – also the phase of the reflection depends on whether the second piece is heavier or lighter that the first. 100% of the energy is transferred only when the impedances are matched. The reflectance or transmittance of the interface between the two regions depends on the ratio of the characteristic impedances.
The same thing happens with all sorts of different waves when they cross boundaries into regions of different characteristic impedance. An acoustic wave in air for example is reflected off the surface of water, because of the impedance mismatch.
An interesting trick is to insert a third piece of rope between the first two, but with a carefully chosen intermediate mass per unit length. Then there is a reflection from both interfaces. If the spacing between the two interfaces is correct, the two reflections will be out of phase and cancel each other– you get no reflection and 100% transmittance - this is how antireflection coatings on glasses work. But it only works over a narrow range of frequencies – the wavelength has to be right.
Another trick is to have a graded connection between the two section of rope, say several different sections, all with the same impedance ratio to their neighbours (ie an exponential match) then there will also be very little reflection. This is how stealth technologies work, e.g on a submarine – multiple layers of rubber of different densities to provide an impedance match between the water and the steel. In this way there will be very little acoustic reflection from the surface and it will be invisible to sonar. The rubber is also lossy so the energy is dissipated in the rubber. The big rubber wedges in anechoic rooms do the same thing.
Hope this makes sense
R
And you can model the reflected portion, the transmitted portion and the absorbed portion, with two port anaylsis and putting the pressure equations in a matrix, and solve in matlab to plot the imaginary terms to show the phase shift. The classical way to measure absorption is with an impedance tube. It might be an interesting exercise to plot the radiation impedance of a flat panel vs a cone in matlab. Have you ever done anything like that?
This is what I consider disappointing. "Rudimentary and hidebound to a flawed grasp of room acoustics." please say something constructive, explain why you said this, or just remove it.
Thanks for elaborate metaphor. But not quite quite... visceral or intuitive.
Perhaps impedance matching to air can be pictured using a circuit that needs a transformer but aint got one (duh). Or an airplane wing lift angle. A Rice-Kellogg driver is like...
Not any kind of direct analogy, but as an example of inelegant design, I liken an R-K driver to powering a vehicle with an internal combustion engine (that has zero torque at stall speed) to an electric motor that can have max torque at zero and no stall speed.
B.
He literally described the three phenomena, transmission, reflection, and absorption, and gave simple visual analogies of real world cases.
Impedance matching is about optimizing energy transfer, doesn't matter if its electrical, acoustical, or mechanical, the simplest visual analogy is literally gears. Low gear, up hill, high gear down hill.
The reflection, absorption, and transmission was a nice touch, and the rope is an easy way to picture it.
You should just thank him.
Impedance matching is about optimizing energy transfer, doesn't matter if its electrical, acoustical, or mechanical, the simplest visual analogy is literally gears. Low gear, up hill, high gear down hill.
The reflection, absorption, and transmission was a nice touch, and the rope is an easy way to picture it.
You should just thank him.
Hi B
Ha. I thought you might want more 😊.
I’m not sure whether impedance matching is a useful concept in this case. While there is certainly a massive inefficiency in the energy transfer between the electrical input to conventional drivers and the air (of the order of 1 %) - most of the energy is dissipated in the voice coil and the speaker surrounds.
I’m not sure that you would want the mechanical impedance of the driver to be matched to the air. I mentioned earlier than conventional drivers are volume velocity sources (analogous to current sources). Like any current source, they work best when their output impedance is much greater than the impedance of the load.
Most ESLs are close to 100% efficient - over much of the pass band, the electrical power is transferred and radiated as acoustic with power close to 100% efficiency. But there is a catch.
In conventional drivers, which behave as a resistor at low frequencies, the voltage across the output transistors of the amplifier falls as the output current goes up. With ESLs, which behave as a large capacitor, the output current is maximum when output voltage is zero, so the full power supply voltage is across the output transistors. Although the ESL itself is efficient, it forces the amplifier to be very inefficient and dissipate high powers.
Ha. I thought you might want more 😊.
I’m not sure whether impedance matching is a useful concept in this case. While there is certainly a massive inefficiency in the energy transfer between the electrical input to conventional drivers and the air (of the order of 1 %) - most of the energy is dissipated in the voice coil and the speaker surrounds.
I’m not sure that you would want the mechanical impedance of the driver to be matched to the air. I mentioned earlier than conventional drivers are volume velocity sources (analogous to current sources). Like any current source, they work best when their output impedance is much greater than the impedance of the load.
Most ESLs are close to 100% efficient - over much of the pass band, the electrical power is transferred and radiated as acoustic with power close to 100% efficiency. But there is a catch.
In conventional drivers, which behave as a resistor at low frequencies, the voltage across the output transistors of the amplifier falls as the output current goes up. With ESLs, which behave as a large capacitor, the output current is maximum when output voltage is zero, so the full power supply voltage is across the output transistors. Although the ESL itself is efficient, it forces the amplifier to be very inefficient and dissipate high powers.
Yes. This was a dead end pursued by speaker 'engineers' in da 30s to the 70s. Irrelevant today except to the Public Addres mob.
The real requirement is 'good sound' with 'practical' amps. The first people to realise this were Rice & Kellog. They invented 'bass response' when everyone else just wanted 'louder'. They had to design their own amp giving a huge 1W of undistorted power.
Their original papers reward reading again & again to see how far ahead of the time they were and how many modern speaker 'gurus' still don't understand basic concepts R&K first grasped. There's good reason why their invention is pre-eminent from cheapo USB speakers to big PA to the most expensive audiofool monsters.
There are in fact only 2 transduction methods which are inherently capable of wide band flat response ie 'good sound'. One is Rice & Kellog's moving coil cone speaker.
The other is the Constant Charge Electrostatic Speaker per Walker/Baxandall. R&K is easier to realise cos even the cheapo versions sorta approach the ideal requirements. But the simplest CCELSpWB requires precise & difficult manufacture.
All other methods are just toys.
But rather than claim 100% efficiency, its more correct to say the Bass/Size/Sensitivity equation is pretty high IF implemented properly.
In theory, a CCELSpWB the size of the ESL63 should be 94dB/W @ 1m. I worked this out in Jurassic times with help from Peter Walker.
Alas, it requires special bits including transformers hand carved from Unobtainium by Huntingdon virgins. Modern implementations waste loadsa power in yucky resistors.
Last edited:
With due respect to your (and kgrlee's) legitimate explanation, I have an empiricist's reservations.
Without some match of impedance in the form of radiation resistance, the air does not "speak" back to an R-K driver. To generate faithful cone motions you must rely on the motor and bits of rubber and starched cotton to keep the cone assembly from flapping in the breeze irregularly. Nothing is controlling the cone.
While you can laser-trim the mechanical parts into a semblance of linearity, all bets are off when the assembly resonance is in the sound passband (as it always the case with low woofers), or when other pieces want to jiggle at odd frequencies, when aerodynamic irregularities arise, etc. Or when somebody asks if your speaker can do square waves.
It is inexplicable to me why today we persevere making speaker systems without feedback.
Footnote: I'm trying to remember from old pictures if R-K speakers had surrounds and air control around the spiders; if they didn't, their grasp of the work of the driver was "incomplete"
B.
Last edited:
I'm not that really interested in whether the air speaks to a speaker
[*] I'm interested in what makes the sound of a speaker 'liked'. The measuring instrument is your DBLT panel. Like any measuring instrument, you need to check & calibrate it every now & then. I've been privileged to have some of the best ears in the business on my panel.
Conducting DBLTs properly, using them to design speakers is VERY expensive ... but I could write books on the subject so won't go into that.
But doing them for nearly 2 decades shape my prejudices on speaker design. Here are some of them.
The best Moving Coil speakers surpassed ESL in midrange & HF performance around the time of the Spendor BC1.
I've tested this in DBLTs but can also provide 'Waterfall' curves as evidence. We were among the first to do Delayed Resonance Testing (which the BBC invented) Waterfalls are actually KEFplots (CDS invented by KEF).
ESL63 midrange & treble is only OK. The reason is the dustcovers. Fig 6 in Peter Walkers AES article is without dustcovers. Response is MUCH more ragged with.
I've listened to an ESL63 without dustcovers and it has very good mid & treble but you have to replace diaphragms every month.
What a well designed ESL does do better than anything else is low distortion bass. When I heard an ESL63 play 50Hz, I was astonished as I had never heard lower distortion bass at low & medium levels.
So my ultimate speaker would have moving coil mid & treble but big ESL dipole bass.
There's other good stuff on dipoles and their interaction with rooms which I won't go into but are pointers for how to position your ESL63 etc.
Where do you think the 'assembly resonances' of an ESL lie? Peter Walker can tell you a lot about that.
I've designed speakers which can 'do square waves'. But why would I want to? I have an AES paper on the subject "Is Linear Phase Worthwhile?"
B, you really owe it to yourself to have a look at R&K's papers.
[*] Actually I am. Great Guru Baxandall wrote 'Loudspeakers as Microphones' and I used some of his theory & concepts in my Powered Integrated Super Sub technology.
Last edited:
Thats fascinating thanks. Are there any possible ESL amp designs that can handle the demands with less dissipation?
No. Think about it this way....
The total power used by an amp is the power supply voltage times current
In a conventional speaker, that power dissipation is split between the amp and speaker
In an ESL, the power is (almost) dissipated entirely in the amp. Almost because there is a little bit of power radiated as sound.
The total power used by an amp is the power supply voltage times current
In a conventional speaker, that power dissipation is split between the amp and speaker
In an ESL, the power is (almost) dissipated entirely in the amp. Almost because there is a little bit of power radiated as sound.
Yes. This was a dead end pursued by speaker 'engineers' in da 30s to the 70s. Irrelevant today except to the Public Addres mob.
The real requirement is 'good sound' with 'practical' amps. The first people to realise this were Rice & Kellog. They invented 'bass response' when everyone else just wanted 'louder'. They had to design their own amp giving a huge 1W of undistorted power.
Their original papers reward reading again & again to see how far ahead of the time they were and how many modern speaker 'gurus' still don't understand basic concepts R&K first grasped. There's good reason why their invention is pre-eminent from cheapo USB speakers to big PA to the most expensive audiofool monsters.
There are in fact only 2 transduction methods which are inherently capable of wide band flat response ie 'good sound'. One is Rice & Kellog's moving coil cone speaker.
The other is the Constant Charge Electrostatic Loudspeaker per Walker/Baxandall. R&K is easier to realise cos even the cheapo versions sorta approach the ideal requirements. But the simplest CCESLpWB requires precise & difficult manufacture.
All other methods are just toys.
But rather than claim 100% efficiency, its more correct to say the Bass/Size/Sensitivity equation is pretty high IF implemented properly.
In theory, a CCESLpWB the size of the ESL63 should be 94dB/W @ 1m. I worked this out in Jurassic times with help from Peter Walker.
Alas, it requires special bits including transformers hand carved from Unobtainium by Huntingdon virgins. Modern implementations waste loadsa power in yucky resistors.
Last edited:
Thanks for making the theory so easy to grasp and then following up with some practical advice.
Can i ask your view on the best bandwidth for ESL? Is there any merit only using ESL for midrange? Most ESL seem to misbehave over 10kHz with complex resonances and plummeting impedance. Why not run them like a three way conventional design with cone woofer : ESL : ribbon? I see lots of two way woofer : ESL designs. Why not a three way?
Good to be near-full-range if you can be. While my amps complain when I try to play loud high freq test tones, not much of an issue on ordinary listening. Been a long time since I sweated about speaker performance over 12kHz. Easy for you to test if you need to sweat about that last part of an octave too... there's really not much that's detectable up there.
But also makes a lot of sense to crossover to an ESL tweeter for several good reasons, just as it does for Rice-Kellogg systems.
For a few decades, I used ESL tweeters in a tri-amp'ed system. That tweeter was a mildly curved panel holding an array of 4 circular ESLs cells (designed by Denessen who was related in business to Jansen), each about 4 inches in diameter. X-over was around 4500 Hz, leaving enough treble to give the tweeters an important role in the sound quality department but still leaving the great middle for the main panel.
Exceedingly great tweeting and the amps were happy about it too.
But really a piece of cake to create ESL tweeters with simple stators (can be 3D printed???), modest ladder bias supplies, and cheap step-up transformers. I don't know why they aren't common.
B.
Last edited: