Hi all,
I'm planning to build some ESL loudspeakers, and i'd like some input from the people here. There might be some flaws in my design with which you could help me.
ESLs have always interested me, and for long I've planned on building one. But I think that it's ugly to have a transformer between your amplifier and the ESL unit. So I designed an amplifier to drive them directly, with some high-voltage FETs (momentarily 10n100's, which shoud be able to handle an output of 800V, but I can easily put in transistors or FETs with higer voltages). At the moment I've ordered the first test PCB's, so now I've got time to build some good ESLs.
This low voltage-output means that I have to use a very small spacing between the stator and the membrane. Ideal would be to have a membrane-stator-distance of 0.5mm. (So I can use around 1kV on the membrane and 800Vpp audio)
At the moment I've already built a small unit with +/- 0.5mm spacing and a size of 20x40cm. This unit works, but leaks very much, it cant use more than 500V charge. But the sensitivity is quite good. At 400Vpp it is around loud enough for normal listening to speach or pop-music.
Now I plan to build a large 1mx1m full-range unit in the same way. My plan was the following:
Stator design
I planned to use something like this as a frame:
and on that I can glue a layer of 'woven wire' (how do i transloate the dutch 'gaas' to english?). I can buy sheets of woven wire, where the wires are around 1.7mm apart. It's still quite sturdy.
The spacing of the frame is around 3cm. So the woven wire is quite well supported. I hope...
spacers
I plan on using sheets of styrene as spacer. Just cut it up to 3mm broad strips running horizontally over the stator. I think it's easiest to glue them in position.
What kind of glue is good? At the moment I use bisonkit contact-glue. But this conducts a little bit, which is bad.
membrane
I've bought some mylar from the flowershop. Its around 10g/mm^2. Is this too heavy?
I've calculated that I need around 2kg of tension no the membrane in the vertical direction. I don't yet really know what's the best way to get tension on the membrane. But hanging it down with a weight, and then glueing it to the stator seems best to me...
As coating I'll use kontaktchemie 100 antistatik.
I've done some calculations on ESLs, and i get the following (theoretical) specifications.
max. sound pressure: 104dB (in practice more likely is 100dB)
Capacitance: 8.8nF
lowest frequency due to limited movement of membrane: 55Hz (more likely to be 70 Hz)
The panel won't output frequencies lower then +/-70Hz in this configuration. (I could make that 35Hz If I double the spacer thickness, but that costs me 6dB sound pressure)
What's the relation between the size of the element and the lowest frequency due to phase-cancellation? I wasn't able to include that in my simple equation.
Please comment!
I'm planning to build some ESL loudspeakers, and i'd like some input from the people here. There might be some flaws in my design with which you could help me.
ESLs have always interested me, and for long I've planned on building one. But I think that it's ugly to have a transformer between your amplifier and the ESL unit. So I designed an amplifier to drive them directly, with some high-voltage FETs (momentarily 10n100's, which shoud be able to handle an output of 800V, but I can easily put in transistors or FETs with higer voltages). At the moment I've ordered the first test PCB's, so now I've got time to build some good ESLs.
This low voltage-output means that I have to use a very small spacing between the stator and the membrane. Ideal would be to have a membrane-stator-distance of 0.5mm. (So I can use around 1kV on the membrane and 800Vpp audio)
At the moment I've already built a small unit with +/- 0.5mm spacing and a size of 20x40cm. This unit works, but leaks very much, it cant use more than 500V charge. But the sensitivity is quite good. At 400Vpp it is around loud enough for normal listening to speach or pop-music.
Now I plan to build a large 1mx1m full-range unit in the same way. My plan was the following:
Stator design
I planned to use something like this as a frame:
An externally hosted image should be here but it was not working when we last tested it.
and on that I can glue a layer of 'woven wire' (how do i transloate the dutch 'gaas' to english?). I can buy sheets of woven wire, where the wires are around 1.7mm apart. It's still quite sturdy.
The spacing of the frame is around 3cm. So the woven wire is quite well supported. I hope...
spacers
I plan on using sheets of styrene as spacer. Just cut it up to 3mm broad strips running horizontally over the stator. I think it's easiest to glue them in position.
What kind of glue is good? At the moment I use bisonkit contact-glue. But this conducts a little bit, which is bad.
membrane
I've bought some mylar from the flowershop. Its around 10g/mm^2. Is this too heavy?
I've calculated that I need around 2kg of tension no the membrane in the vertical direction. I don't yet really know what's the best way to get tension on the membrane. But hanging it down with a weight, and then glueing it to the stator seems best to me...
As coating I'll use kontaktchemie 100 antistatik.
I've done some calculations on ESLs, and i get the following (theoretical) specifications.
max. sound pressure: 104dB (in practice more likely is 100dB)
Capacitance: 8.8nF
lowest frequency due to limited movement of membrane: 55Hz (more likely to be 70 Hz)
The panel won't output frequencies lower then +/-70Hz in this configuration. (I could make that 35Hz If I double the spacer thickness, but that costs me 6dB sound pressure)
What's the relation between the size of the element and the lowest frequency due to phase-cancellation? I wasn't able to include that in my simple equation.
Please comment!
Hi,
One couldn´t give serious advice with just the few details you mentioned.
I think, that the way You start the design might not be successful, because You don´t start at the beginning of a typical design process but in the midst.
Choosing some material and hack them into something together will not result in a good ESL design, but just a ´somehow´ working prototype.
Since You already built a prototype and from the way you asked, I assume You want to build a proper design this time? But what? I strongly recommend to not start Your ESL-life with the try to build a big fullrange ESL from scratch. The end result will most definitely not be optimal.
Think of hybrid panels first. There are a lot of Pros for hybrid panels and a lot of Cons to fullrange panels.
First think about a spec-sheet which should contain the aims You want to reach in detail, for example which frequency range needs to be covered, size constraints, drive constraints etc. etc.
Then think about rough design outlines, like the dimensions, which kind of stator fits the demands, cost, build effort, sourceabilty, etc. etc. There are a lot of tips in this forum about design guidelines and material-choices. Invest some time and the search-function is Your friend
Then I´d start looking for materials that might suit the demands. Mylar from a Flowershop won´t be suitable material......if it is Mylar anyhow.
Again there are a lot of tips in this forum about sources where to get and to buy quality material from.
The better Your work in these pre-building stages is the higher will be the probability of success. The actual building process then becomes just a minor problem.
jauu
Calvin
One couldn´t give serious advice with just the few details you mentioned.
I think, that the way You start the design might not be successful, because You don´t start at the beginning of a typical design process but in the midst.
Choosing some material and hack them into something together will not result in a good ESL design, but just a ´somehow´ working prototype.
Since You already built a prototype and from the way you asked, I assume You want to build a proper design this time? But what? I strongly recommend to not start Your ESL-life with the try to build a big fullrange ESL from scratch. The end result will most definitely not be optimal.
Think of hybrid panels first. There are a lot of Pros for hybrid panels and a lot of Cons to fullrange panels.
First think about a spec-sheet which should contain the aims You want to reach in detail, for example which frequency range needs to be covered, size constraints, drive constraints etc. etc.
Then think about rough design outlines, like the dimensions, which kind of stator fits the demands, cost, build effort, sourceabilty, etc. etc. There are a lot of tips in this forum about design guidelines and material-choices. Invest some time and the search-function is Your friend
Then I´d start looking for materials that might suit the demands. Mylar from a Flowershop won´t be suitable material......if it is Mylar anyhow.
Again there are a lot of tips in this forum about sources where to get and to buy quality material from.
The better Your work in these pre-building stages is the higher will be the probability of success. The actual building process then becomes just a minor problem.
jauu
Calvin
At the moment I'm already listening to a hybrid system which I bought and repaired for very little money. (final 80i + cheap, bad subwoofer)
I want to go bigger now. So I'd like to build a full-range unit, but I want to build it suitable for my direct-drive amplifier. This gives the limitation of 800Vrms audio. (comparable with a 50W amp with a 1:40 step-up)
As far as I've read this would be enough for a full-range system.
I know this is quite difficult. I don't mind it not being perfect the first time. I just don't feel like building something I already have (a hybrid design)
Of course I can make a spec-sheet, and all that paperwork. But my experience (in electronics) learned me that doing that is only useful when you exactly know where the problems are going to be. Which is only the case if you've built some ESLs already.
Spec-sheet:
- I'm going to build a 'simple' dipole speaker.
- frequency-repsonse 70Hz-20kHz flat
- 800Vaudio for ~ 95dB @ 2m (or as good as it gets)
- cheap! (and easy to build)
Most difficult is the low-frequency-repsonse. But I can't really find much information about that. I could calculate/simulate it myself, but that would take quite a lot of time.. And there is a large chance of mistakes, so i wouldn't even trust my calculations...
A problem is that quite a lot of the theory on the internet (in my humble opinion) is not correct. So I don't really know who to trust.
For practical building tips, there's of course a lot of information. But most involve a lot(!) of work. The design I propose here is a lot less work. I might be able to completely build a unit in 1 day this way
I've already found out that I really should use a better diaphragm, and a better coating. (which didn't surprise me, calculations tell the same )
My main question is: Is it quite possible to build an ESL with only 0.5mm-0.7mm clearance between stator and diaphragm? (I don't think many people have tried that before.)
I want to go bigger now. So I'd like to build a full-range unit, but I want to build it suitable for my direct-drive amplifier. This gives the limitation of 800Vrms audio. (comparable with a 50W amp with a 1:40 step-up)
As far as I've read this would be enough for a full-range system.
I know this is quite difficult. I don't mind it not being perfect the first time. I just don't feel like building something I already have (a hybrid design)
Of course I can make a spec-sheet, and all that paperwork. But my experience (in electronics) learned me that doing that is only useful when you exactly know where the problems are going to be. Which is only the case if you've built some ESLs already.
Spec-sheet:
- I'm going to build a 'simple' dipole speaker.
- frequency-repsonse 70Hz-20kHz flat
- 800Vaudio for ~ 95dB @ 2m (or as good as it gets)
- cheap! (and easy to build)
Most difficult is the low-frequency-repsonse. But I can't really find much information about that. I could calculate/simulate it myself, but that would take quite a lot of time.. And there is a large chance of mistakes, so i wouldn't even trust my calculations...
A problem is that quite a lot of the theory on the internet (in my humble opinion) is not correct. So I don't really know who to trust.
For practical building tips, there's of course a lot of information. But most involve a lot(!) of work. The design I propose here is a lot less work. I might be able to completely build a unit in 1 day this way
I've already found out that I really should use a better diaphragm, and a better coating. (which didn't surprise me, calculations tell the same
)My main question is: Is it quite possible to build an ESL with only 0.5mm-0.7mm clearance between stator and diaphragm? (I don't think many people have tried that before.)
The "woven wire" you mentioned sounds like it will have some sections of wire closer to the diaphragm than other sections. If that's true, you are likely to run into some limitations that otherwise wouldn't trouble you. The points where the woven wires are close to the diaphragm will limit the diaphragm's displacement and the magnitude of the bias voltage you can apply, and the fact that other regions of wire are farther away from the diaphragm means that you'll sacrifice some of the panel's sensitivity--especially if you are going to use very small diaphragm-stator spacing. If the variation in wire-to-diaphragm distance were small compared to the average wire-to-diaphragm distance, it might be less problematic. In your application, it seems to me it will likely cause difficulties. On the other hand, if you can build one in a day, it won't cost much time to try it out.
Few
Few
Hi,
- simple dipole: that´s the ESL´s natural way to work, go ahead.
- 70Hz-20kHz: quite sensible choice, which allows to stay within reasonable dimensions. Rough suggestion: 0.3-0.5m² and d/s: 2-2.5mm
- 800V (I assume peak-peak, and how much current capability?) would only be sufficient to drive a small tweeter or headphone sized panel with very small d/s distance of <1mm to the suggested SPL-levels. For a d/s of 2-2.5mm rather calculate with 4000Vp-p. I wouldn´t drive a panel with a d/s of 0.5-0.7mm d/s lower than 400-500Hz.
The lower frequency limit is no matter of the d/s. The d/s only defines the dynamic range, i.e. the maximum SPL-level that can be reached at a certain frequency. Excursion demands increase with lowering frequency.
- cheap: Besides material You might get for free, a wire stator design is one of the cheapest possibilities that allows for good end results.
- woven wire or a woven mesh was tested against other stator-designs by Matthew Lattis in audioXpress. Efficiency was stated below par. The uneven surface and as such varying distances of d/s could be a reason for that. As a drawback I regard that it´s nearly impossible to insulate such a mesh structure. On the other hand could a very fine mesh not only supply for a very homogenous electrical field (needed the more, the smaller the d/s) but also for a mechanical damping.
- 0.5-0.7mm d/s is typically used with small tweeter- and headphone Panels, but hardly ever with larger panels. At one will a very high degree of mechanical precision be needed (increased costs and effort), at second will such a small d/s seriously restrict the bandwidth towards lower frequencies and at third will such a panel show a high capacitance value and as such represent a very difficult load to drive. The current needed to drive a capacitance depends on the capacitance value, the drive voltage, the frequency and the signal shape. For sinusiodial signals the formula reads: Ipeak= C*pi*f*Vpp
Example: symmetrical ESL-panel, panel diaphragm size: 0.5m², d/s: 0.7mm -> C~6.3nF, Vpp: 800Vp-p, f: 20kHz.
Ipeak= 6.3e-9*3.14*2e4*8e2 = 1.6e-1Apeak = 160mApeak.
If You fudge a bit and define 20kHz as the -3dB frequency (definition of bandwidth limit) the current demand halves to 80mApeak. Still though this is a current value most ESL-amps can´t supply for.
jauu
Calvin
- simple dipole: that´s the ESL´s natural way to work, go ahead.
- 70Hz-20kHz: quite sensible choice, which allows to stay within reasonable dimensions. Rough suggestion: 0.3-0.5m² and d/s: 2-2.5mm
- 800V (I assume peak-peak, and how much current capability?) would only be sufficient to drive a small tweeter or headphone sized panel with very small d/s distance of <1mm to the suggested SPL-levels. For a d/s of 2-2.5mm rather calculate with 4000Vp-p. I wouldn´t drive a panel with a d/s of 0.5-0.7mm d/s lower than 400-500Hz.
The lower frequency limit is no matter of the d/s. The d/s only defines the dynamic range, i.e. the maximum SPL-level that can be reached at a certain frequency. Excursion demands increase with lowering frequency.
- cheap: Besides material You might get for free, a wire stator design is one of the cheapest possibilities that allows for good end results.
- woven wire or a woven mesh was tested against other stator-designs by Matthew Lattis in audioXpress. Efficiency was stated below par. The uneven surface and as such varying distances of d/s could be a reason for that. As a drawback I regard that it´s nearly impossible to insulate such a mesh structure. On the other hand could a very fine mesh not only supply for a very homogenous electrical field (needed the more, the smaller the d/s) but also for a mechanical damping.
- 0.5-0.7mm d/s is typically used with small tweeter- and headphone Panels, but hardly ever with larger panels. At one will a very high degree of mechanical precision be needed (increased costs and effort), at second will such a small d/s seriously restrict the bandwidth towards lower frequencies and at third will such a panel show a high capacitance value and as such represent a very difficult load to drive. The current needed to drive a capacitance depends on the capacitance value, the drive voltage, the frequency and the signal shape. For sinusiodial signals the formula reads: Ipeak= C*pi*f*Vpp
Example: symmetrical ESL-panel, panel diaphragm size: 0.5m², d/s: 0.7mm -> C~6.3nF, Vpp: 800Vp-p, f: 20kHz.
Ipeak= 6.3e-9*3.14*2e4*8e2 = 1.6e-1Apeak = 160mApeak.
If You fudge a bit and define 20kHz as the -3dB frequency (definition of bandwidth limit) the current demand halves to 80mApeak. Still though this is a current value most ESL-amps can´t supply for.
jauu
Calvin
Hi,
I have not made any ESLs yet, just repaired my Quad ESL. So I am a beginner too. I recommend you to invest in a micrometer screw thickness gauge. Fold the mylar 8x or 16x so you can measure precisely. Second, mylar from the flower shop is simply not good. You need 10-12 um thickness for the bass section and 2.5 to 5 um for the treble. Calvin, correct me if I am wrong. Also you can try Johnson's Pronto Antistatic spray for diaphragm treatment. Similar to your stator structure was available over here as part of fluorescent lamp armatures (light diffusor grid) and I was also thinking about experimenting with them in an ESL. Is the more recent Quad ESL63 using something similar?
Good luck with your project,
Laszlo
I have not made any ESLs yet, just repaired my Quad ESL. So I am a beginner too. I recommend you to invest in a micrometer screw thickness gauge. Fold the mylar 8x or 16x so you can measure precisely. Second, mylar from the flower shop is simply not good. You need 10-12 um thickness for the bass section and 2.5 to 5 um for the treble. Calvin, correct me if I am wrong. Also you can try Johnson's Pronto Antistatic spray for diaphragm treatment. Similar to your stator structure was available over here as part of fluorescent lamp armatures (light diffusor grid) and I was also thinking about experimenting with them in an ESL. Is the more recent Quad ESL63 using something similar?
Good luck with your project,
Laszlo
The amplifier is at this moment capable of ~200Vrms, with ~300V supply voltage. (600Vpp) This is because it's a bridged amplifier, otherwise you'd never get the balanced input-signals an ESL requires... (wouldn't be a real problem in constant charge mode, but who cares; more decibels )
When I've got the PCB's I dare go to higher voltages, and if I change the transistors I should be able to supply 2.4kVpp -> ~800Vrms output.
The maximum current is around 200mA. Which wouldn't be enough according to your calculations, because the amplifier is in bridge mode.
But the max current is (for music) just not needed. A normal stepup transformer couldn't supply those currents either, because of wire resistance. I don't bother getting it higher, although with small modifications >1A would be easily possible with this design. But going to these values is dangerous for the transistors. A strong 20kHz note would fry the transistors with around 1kW of peak dissipation. If needed I can upgrade the current capability, but this is not needed. So let's say 'current capability is sufficient'.
I won't make 20kHz a -3dB-point. I'd rather let my amplifier have its -3dB@40kHz but just let it go into a current-clip at very high frequencies. This gives the best sound quality (as long as you don't have more than 400Vpp of 20kHz, but what hellish music would sound like that ) Because there is no transformer means things are quite different.
And now for some theory about the lowest frequency. Excursion demands are indeed the most important thing here.
- At some frequency the excursion becomes so large that tension force becomes larger than the electrical force. When both forces are equal you're at the -3dB (or -6dB?) point. (here the force to make a sound pressure wave is included in the excursion)
- The tension can be taken lower at larger d/s, this is the reason that at low d/s lower frequencies can be used. And the tension should be taken as low as possible! This makes a difference in distorsion.
- I see a lot about resonance-frequencies, these are unimportant if frequencies well below resonance are still at smal excursions. The resonance is very much overdamped.
- One BIG problem is that the in a lot of dipole designs air can go from the front to the back of the panel without a problem. This means far less weight of air being moved -> higher excursion. But this effect I'm not able to calculate easily, so I ignore it. In my calculations I still make the assumption that the panel is much bigger than wavelength.
You're right about the woven wire, it is actually not that good at 0.5mm. It would probably be optimal at 0.8mm. (well, larger is of course always possible)
It's of good quality very regularly woven, it can easily be made totally flat if glued onto something. I think less then 0.2mm accuracy can easily be achieved with this material.
It is very much like this:
)When I've got the PCB's I dare go to higher voltages, and if I change the transistors I should be able to supply 2.4kVpp -> ~800Vrms output.
The maximum current is around 200mA. Which wouldn't be enough according to your calculations, because the amplifier is in bridge mode.
But the max current is (for music) just not needed. A normal stepup transformer couldn't supply those currents either, because of wire resistance. I don't bother getting it higher, although with small modifications >1A would be easily possible with this design. But going to these values is dangerous for the transistors. A strong 20kHz note would fry the transistors with around 1kW of peak dissipation. If needed I can upgrade the current capability, but this is not needed. So let's say 'current capability is sufficient'.
I won't make 20kHz a -3dB-point. I'd rather let my amplifier have its -3dB@40kHz but just let it go into a current-clip at very high frequencies. This gives the best sound quality (as long as you don't have more than 400Vpp of 20kHz, but what hellish music would sound like that
) Because there is no transformer means things are quite different.And now for some theory about the lowest frequency. Excursion demands are indeed the most important thing here.
- At some frequency the excursion becomes so large that tension force becomes larger than the electrical force. When both forces are equal you're at the -3dB (or -6dB?) point. (here the force to make a sound pressure wave is included in the excursion)
- The tension can be taken lower at larger d/s, this is the reason that at low d/s lower frequencies can be used. And the tension should be taken as low as possible! This makes a difference in distorsion.
- I see a lot about resonance-frequencies, these are unimportant if frequencies well below resonance are still at smal excursions. The resonance is very much overdamped.
- One BIG problem is that the in a lot of dipole designs air can go from the front to the back of the panel without a problem. This means far less weight of air being moved -> higher excursion. But this effect I'm not able to calculate easily, so I ignore it. In my calculations I still make the assumption that the panel is much bigger than wavelength.
You're right about the woven wire, it is actually not that good at 0.5mm. It would probably be optimal at 0.8mm. (well, larger is of course always possible)
It's of good quality very regularly woven, it can easily be made totally flat if glued onto something. I think less then 0.2mm accuracy can easily be achieved with this material.
It is very much like this:
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
[QUOTE=zweetvoetje; I think less then 0.2mm accuracy can easily be achieved with this material.
Hello Zweetvoetje,
I have made my ESL's with metal grids like you propose, but with insulated wire glued on them (and segmentation). They were presented at the last ESL-club meeting in june in Den Haag. (I must try to make some new photo's)
In my case these grids are 22.5x177.5cm with a mesh grid of 44.4mm.
My experience tells me that your estimated accuracy is by a large amount overestimated. Even if your fine mesh could be very accurate, the metal grids are industrial products with rather large tolerances and hardly flat (at least the ones I have, but on construction sites I did not yet find others with tighter tolerances. Besides that the metal grids are hardly cheap, unless you can get them as overstock or something.
Edwin
Hello Zweetvoetje,
I have made my ESL's with metal grids like you propose, but with insulated wire glued on them (and segmentation). They were presented at the last ESL-club meeting in june in Den Haag. (I must try to make some new photo's)
In my case these grids are 22.5x177.5cm with a mesh grid of 44.4mm.
My experience tells me that your estimated accuracy is by a large amount overestimated. Even if your fine mesh could be very accurate, the metal grids are industrial products with rather large tolerances and hardly flat (at least the ones I have, but on construction sites I did not yet find others with tighter tolerances. Besides that the metal grids are hardly cheap, unless you can get them as overstock or something.
Edwin
Hi,
another problem of the wire mesh could be that it does not remain flat after gluing it to the supports. It might need to be stretched before gluing, but that could prove to be problematic too.
If you really want to go down this route (and I can't blame you, it's certainly a fast way to build a stator), you could consider using plastic mosquito mesh (used to install in window frames used to let fresh air into the room). Those would be easy to stretch a little (so they don't bulge after gluing), and are probably flatter than wire mesh. They could be sprayed with conductive paint after gluing.
Oh and as others said, mylar from the flower shop is WAY too thick, you really need the thin tensilized mylar stuff (6 or 12 micron for a full range ESL would be okay).
Best,
Kenneth
another problem of the wire mesh could be that it does not remain flat after gluing it to the supports. It might need to be stretched before gluing, but that could prove to be problematic too.
If you really want to go down this route (and I can't blame you, it's certainly a fast way to build a stator), you could consider using plastic mosquito mesh (used to install in window frames used to let fresh air into the room). Those would be easy to stretch a little (so they don't bulge after gluing), and are probably flatter than wire mesh. They could be sprayed with conductive paint after gluing.
Oh and as others said, mylar from the flower shop is WAY too thick, you really need the thin tensilized mylar stuff (6 or 12 micron for a full range ESL would be okay).
Best,
Kenneth
Hi,
I recommend You to read same ESL-theory stuff, because some of Your assumptions are false.
A Audio-tranny can deliver quite high currents to the stators. The winding resistances typically beeing in the low kOhm-range there´s enough headroom for a couple of hundreds of mA. The capabilities are more restricted by the transformer´s wattage and the driving amplifier´s wattage.
Using a bigger core with higher wattage is an easy solution.
The capabilities of a HV-amp are not that easily increased, since the heat-power-loss in the output transistors (or tubes) quickly gets excessively high. That means You have to work with stacked and/or paralleled devices and huge heatsinks or very powerful Tubes.
Since ESLs become current demanding the higher the frequency rises a bandwidth limit of 40kHz is not easy to reach. Only segmented panels which present the amp with a rather small capacitance could be driven that high. It´s nearly impossible to drive non-segmented panels that high in frequency. A bandwidth limit of 40kHz means enough drive current under all circumstances for frequencies up to 20kHz.
Btw: a 12µm thick diaphragm will start dropping in SPL mass-related around 12-15kHz. With diaphragms of 6µm and less weight-related SPL-drop is beyond 20kHz.
It´s probabely the single most failure made by ESL-newbies to think that a low mechanical tension does anything good. Always use the highest tension possible. If You want a lower bandwidth limit of 70Hz then place the resonance around 80Hz-90Hz (after applying the HV-polarizing voltage the Fs will drop to the correct value. Aging will too have a Fs-dropping effect. A resonance lower than the desired bandwidth limit, say 40Hz in this case, is off of the optimum. The panel will quickly be overdriven so dynamics are poor and efficiency drops considerably.
Your assumption of reduced distortions with lower mechanical tension is wrong. There is a relationship between the value of polarizing voltage and distortions with higher polarizing voltages lead to decreasing distortion values. Low mechanical tension means lower possible polarizing voltage, hence higher distortions. The only panel where You can and should place the Fs below the desired crossover frequency -but still as high as possible- is a hybrid-panel.
The weight of the moving system and the level of excursion are not related with each other. The excursion level needed for a given SPL-level is a function of moved air volume (diaphragm area*excursion) and frequency and other circumstances like measuring distance and distribution character, room, etc.
A precision of less than 0.2mm in such a big panel You intend to build is not an easy task. Don´t underestimate this. Plastics for spacers, especially acrylics often exhibit greater tolerances. And a cube louvre intended for lighting will hardly be within this tolerance range.
Its easy to build something cheap and quick that shares certain details with an decent ESL. If You intend to build a decent ESL, matters become much more serious.
jauu
Calvin
I recommend You to read same ESL-theory stuff, because some of Your assumptions are false.
A Audio-tranny can deliver quite high currents to the stators. The winding resistances typically beeing in the low kOhm-range there´s enough headroom for a couple of hundreds of mA. The capabilities are more restricted by the transformer´s wattage and the driving amplifier´s wattage.
Using a bigger core with higher wattage is an easy solution.
The capabilities of a HV-amp are not that easily increased, since the heat-power-loss in the output transistors (or tubes) quickly gets excessively high. That means You have to work with stacked and/or paralleled devices and huge heatsinks or very powerful Tubes.
Since ESLs become current demanding the higher the frequency rises a bandwidth limit of 40kHz is not easy to reach. Only segmented panels which present the amp with a rather small capacitance could be driven that high. It´s nearly impossible to drive non-segmented panels that high in frequency. A bandwidth limit of 40kHz means enough drive current under all circumstances for frequencies up to 20kHz.
Btw: a 12µm thick diaphragm will start dropping in SPL mass-related around 12-15kHz. With diaphragms of 6µm and less weight-related SPL-drop is beyond 20kHz.
It´s probabely the single most failure made by ESL-newbies to think that a low mechanical tension does anything good. Always use the highest tension possible. If You want a lower bandwidth limit of 70Hz then place the resonance around 80Hz-90Hz (after applying the HV-polarizing voltage the Fs will drop to the correct value. Aging will too have a Fs-dropping effect. A resonance lower than the desired bandwidth limit, say 40Hz in this case, is off of the optimum. The panel will quickly be overdriven so dynamics are poor and efficiency drops considerably.
Your assumption of reduced distortions with lower mechanical tension is wrong. There is a relationship between the value of polarizing voltage and distortions with higher polarizing voltages lead to decreasing distortion values. Low mechanical tension means lower possible polarizing voltage, hence higher distortions. The only panel where You can and should place the Fs below the desired crossover frequency -but still as high as possible- is a hybrid-panel.
The weight of the moving system and the level of excursion are not related with each other. The excursion level needed for a given SPL-level is a function of moved air volume (diaphragm area*excursion) and frequency and other circumstances like measuring distance and distribution character, room, etc.
A precision of less than 0.2mm in such a big panel You intend to build is not an easy task. Don´t underestimate this. Plastics for spacers, especially acrylics often exhibit greater tolerances. And a cube louvre intended for lighting will hardly be within this tolerance range.
Its easy to build something cheap and quick that shares certain details with an decent ESL. If You intend to build a decent ESL, matters become much more serious.
jauu
Calvin
This is how I built my small test-unit: I made a grid from PCB, somewhat resembling the large steel grids I want to use for the large unit. This had large tolerances, 1mm, or somthing like that.
Than i used some of the wiremesh, taped it on a table (it's easily stretchable), so it was very flat. Than I put a grid over it, and soldered it all to become sturdy. This way the grid has big tolerances, but the surface of the stator is quite flat. This can also be done for large units. Best way seems to me to first build one stator. Then tape an extra sheet of wiremesh on the stator, and build the other side of the stator on top of the first stator. This way both sides have the same faults, and this should work quite well. At least I think so.
@Calvin: It's not the secondary part of the transformer. It's the primary part of the transformer which is the problem. (But I must say that I've never measured the primary resistance of such a transformer) 1:100 transformers are quite normal, but using such a transformer means that you can't have more than 200-300pF of capacitance, or go below the 8 Ohm impedance. And then you're assured that wiring-resistances quickly show up in the frequency response.
I must say that I only have experience with high frequency transformers (>1MHz that is), and there 1:100 just doesn't work. 1:10 is really the limit, but even that is very uncommon. So that's why I have my doubts about these transformers, but this can very well be my fault.
Heat is only a concern when testing units, not with normal music use. My amplifier uses for music a negligible amount of current (class AB). With normal music the dissipated power is not noticeable higher than without music. So it's almost the same as the quiescent current.(3W) But when putting in a 20kHz square my (weak) testing power supply can't handle it (way over 10W).
Tubes can handle much less power than HV-transistors. (well, normal tubes that is, of course, with steam cooled tubes, with 300A filament current it's easy, but those beasts are expensive to operate
BTW, the heat problem is exactly the same for a HV amplifier, as for a normal amplifier with transformer.
Point is that the resonance could (without problems) be right within the bandwidth. For example (just a brainfart):
0.7mm clearance; 3cm spacing; 1200V charge; 10g/m^2 weight of diaphragm gives (I haven't rechecked all my calculations, i hope its correct):
minimal tension for stability around 8N/m; fres: 480Hz; theoretical lowest bandwidth: 28Hz (theoretical; not practical, because infinite size of panel is assumed)
This shows that in this case the resonance is totally overdamped. If you cutoff the frequencies below 70Hz this might work. The 'air-loading' might get considerably lower (+/-3 times) than with inifinite size, so with small side panels this should be possible.
Dropping of the tension with age is quite bad Can something be done about it? (something different than just use the largest possible tension?)
And yes, I'm gonna use real ESL-mylar.
Than i used some of the wiremesh, taped it on a table (it's easily stretchable), so it was very flat. Than I put a grid over it, and soldered it all to become sturdy. This way the grid has big tolerances, but the surface of the stator is quite flat. This can also be done for large units. Best way seems to me to first build one stator. Then tape an extra sheet of wiremesh on the stator, and build the other side of the stator on top of the first stator. This way both sides have the same faults, and this should work quite well. At least I think so.
@Calvin: It's not the secondary part of the transformer. It's the primary part of the transformer which is the problem. (But I must say that I've never measured the primary resistance of such a transformer) 1:100 transformers are quite normal, but using such a transformer means that you can't have more than 200-300pF of capacitance, or go below the 8 Ohm impedance. And then you're assured that wiring-resistances quickly show up in the frequency response.
I must say that I only have experience with high frequency transformers (>1MHz that is), and there 1:100 just doesn't work. 1:10 is really the limit, but even that is very uncommon. So that's why I have my doubts about these transformers, but this can very well be my fault.
Heat is only a concern when testing units, not with normal music use. My amplifier uses for music a negligible amount of current (class AB). With normal music the dissipated power is not noticeable higher than without music. So it's almost the same as the quiescent current.(3W) But when putting in a 20kHz square my (weak) testing power supply can't handle it (way over 10W).
Tubes can handle much less power than HV-transistors. (well, normal tubes that is, of course, with steam cooled tubes, with 300A filament current it's easy, but those beasts are expensive to operate
BTW, the heat problem is exactly the same for a HV amplifier, as for a normal amplifier with transformer.
Point is that the resonance could (without problems) be right within the bandwidth. For example (just a brainfart):
0.7mm clearance; 3cm spacing; 1200V charge; 10g/m^2 weight of diaphragm gives (I haven't rechecked all my calculations, i hope its correct):
minimal tension for stability around 8N/m; fres: 480Hz; theoretical lowest bandwidth: 28Hz (theoretical; not practical, because infinite size of panel is assumed)
This shows that in this case the resonance is totally overdamped. If you cutoff the frequencies below 70Hz this might work. The 'air-loading' might get considerably lower (+/-3 times) than with inifinite size, so with small side panels this should be possible.
Dropping of the tension with age is quite bad
Can something be done about it? (something different than just use the largest possible tension?)And yes, I'm gonna use real ESL-mylar.
Last edited:
Hello zweetvoetje,
Airload on the diaphragm doubles when going from no baffle to infinite baffle.
This lowers fundamental resonance by factor of 1.414= sqrt(2).
By what means have you determined that the diaphragm resonance is overdamped?
I'm pretty sure you will never find an ESL diaphragm that is overdamped without significant application of damping material, like felt, to cover the openings in the stators.
You might check out this thread for some measurements showing typical diaphragm resonance behavior:
http://www.diyaudio.com/forums/planars-exotics/168069-esls-have-bad-decay-plots-2.html#post2209896
(posts #11 & 12 in this thread give some details on the type of diaphragm resonance modes to expect)
Here are some measurements comparing damping material:
http://www.diyaudio.com/forums/plan...con-dots-resonance-control-3.html#post1958582
Last edited:
Hi,
You can drive in class AB only with a pushpull design. In this case the idle heat can be acceptably low because of low idle current. With music the current flowing trough the device quickly increases and so does heat. The more so the more demanding the load is. 1nF is already a difficult load. A panel with Your suggested dimensions would show a capactive value of >6nF which means load impedance of 1/6 compared to the 1nF example and meaning 6x the current. Even with power devices rated at >100W it asks for serious heat sinking to dissipate 20-30W per device or even more. So a tube capable of dissipating a continous plate power of >30W is not at all inferior to the transistor on the heatsink.
Most ESL amps are not true pushpull, but of bridge connected SingleEnded type. This requires class A idle currents. This means that the maximum current that can be supplied to the load equals the idle current when simple Plate-, Drain-, or Collector-CurrentSource is used (in form of a resistor or a active CS) and twice the idle current when a active signal controlled CS is used (which is prone to oscillation with complex loads).
Class-A means a lot of heat in idle mode.
I don´t know what You calculate on which basis, but from experience the calculated values usually vary too much from praxis to be reliable. In any case is 480Hz way too high.
With 12µm film You can reach up to ~250HZ and only lower with thinner films (4µm ~150-170Hz).
Below the resonance frequency, which always is of high Q-value (Q>2) unless dampened electronically or mechanically, the amplitude response drops sharply. Without additional damping there is no way to create an ´overdamped´ situation (Q<0.5) with a tightly stretched film.
BTW I´ve difficulties to understand Your specs.
0.7mm is probabely the d/s-value (diaphragm-stator-distance)?
3cm spacing means which dimension?
10g/m² would translate to a Film thickness of ~7µm.
This would allow for Fs of up to 200Hz initially, settling at ~180Hz after aging.
This values only for pure mechanical streching without heat treatment. With additional heat treatment the mechanical tension and such the Fs will be considerably lower (120-140Hz) and even lower with pure heat stretching (~100Hz).
Rem: estimated values guessed out of experience, under the premise of segment sizes designed after the 1:70-1:100 rule.
Aging will always lead to a lowering of Fs. Heat treatment is a form of pre-aging. A heat treated panel doesn´t alter its Fs much any more. The burn-in time is considerably reduced against a untreated panel which might take half a year to reach its final working point.
jauu
Calvin
You can drive in class AB only with a pushpull design. In this case the idle heat can be acceptably low because of low idle current. With music the current flowing trough the device quickly increases and so does heat. The more so the more demanding the load is. 1nF is already a difficult load. A panel with Your suggested dimensions would show a capactive value of >6nF which means load impedance of 1/6 compared to the 1nF example and meaning 6x the current. Even with power devices rated at >100W it asks for serious heat sinking to dissipate 20-30W per device or even more. So a tube capable of dissipating a continous plate power of >30W is not at all inferior to the transistor on the heatsink.
Most ESL amps are not true pushpull, but of bridge connected SingleEnded type. This requires class A idle currents. This means that the maximum current that can be supplied to the load equals the idle current when simple Plate-, Drain-, or Collector-CurrentSource is used (in form of a resistor or a active CS) and twice the idle current when a active signal controlled CS is used (which is prone to oscillation with complex loads).
Class-A means a lot of heat in idle mode.
I don´t know what You calculate on which basis, but from experience the calculated values usually vary too much from praxis to be reliable. In any case is 480Hz way too high.
With 12µm film You can reach up to ~250HZ and only lower with thinner films (4µm ~150-170Hz).
Below the resonance frequency, which always is of high Q-value (Q>2) unless dampened electronically or mechanically, the amplitude response drops sharply. Without additional damping there is no way to create an ´overdamped´ situation (Q<0.5) with a tightly stretched film.
BTW I´ve difficulties to understand Your specs.
0.7mm is probabely the d/s-value (diaphragm-stator-distance)?
3cm spacing means which dimension?
10g/m² would translate to a Film thickness of ~7µm.
This would allow for Fs of up to 200Hz initially, settling at ~180Hz after aging.
This values only for pure mechanical streching without heat treatment. With additional heat treatment the mechanical tension and such the Fs will be considerably lower (120-140Hz) and even lower with pure heat stretching (~100Hz).
Rem: estimated values guessed out of experience, under the premise of segment sizes designed after the 1:70-1:100 rule.
Aging will always lead to a lowering of Fs. Heat treatment is a form of pre-aging. A heat treated panel doesn´t alter its Fs much any more. The burn-in time is considerably reduced against a untreated panel which might take half a year to reach its final working point.
jauu
Calvin
@bolserst: I must say that I don't have that much knowledge about acoustics.
But the airload should really depend very much on the frequency. (for frequencies where the wavelength of the sound wave is smaller than the size of the ESL)
At low frequencies the air can easily travel from the front to the back, without much time delay. When this is the case the airload will be very much lower. At high frequencies there will be a too large phase-difference for air going from front to back. At these frequencies you'd expect the normal amount of air moved for such a wave.
@Calvin:
Class A was nice in 1950, but it's really very much outdated.
Because it's a bridged design the power is divided over 4 transistors, so even if it is 100W in total, it'll only be 25W per transistor. That's easy to keep cool... At the moment I don't even need to cooling-ribs on my transistors
Class AB normally makes the transistors hot, because they have to supply current continuously. But that's not the case, because an ESL only has capacity, no resistance. And this difference makes the design of such an amplifier quite different from a normal amplifier. Low frequencies use almost no current.
About my calculations:
0.7mm d/s indeed; 3cm is the distance between spacers;
For the lowest (infinite size) frequency I've calculated the frequency where the force from making soundwaves becomes equal to the backpulling (tension-) force of the membrane.
The resonance-frequency is calculated with the simple 'guitar string' formula. First you calculate the sound velocity of transversal waves through the diaphragm, and than you look at the frequency where a half wave fits in between the spacers. In vacuum this resonance would be clearly visible; but with air, it would be damped.
I can see that this resonance is damped because it's still much higher than the lowest frequency. When there is a resonance, the sound output should vanish rapidly below the resonance. (like here: http://www.dolmetsch.com/resonance.jpg but here it vanishes above resonance, meaning it's a different system)
As long as the resonce-frequency is very high, it won't be a problem. But at low frequencies, the air loading, and so also the damping, may get much lower. (especially without a baffle, and at the edges of the ESL)
This all is why I want a very high resonance frequency. I totally agree that a low frequency will be bad.
There might also be other resonances at frequencies where the panelsize is around 1; 2; 3; etc. times the wavelength or something like that. These are acoustic resonances, and here the remedies are totally different.
About heat treatment... What temperature do I need? (just a hairdryer?)
I assume heat treatment should be done under tension, right?
But the airload should really depend very much on the frequency. (for frequencies where the wavelength of the sound wave is smaller than the size of the ESL)
At low frequencies the air can easily travel from the front to the back, without much time delay. When this is the case the airload will be very much lower. At high frequencies there will be a too large phase-difference for air going from front to back. At these frequencies you'd expect the normal amount of air moved for such a wave.
@Calvin:
Class A was nice in 1950, but it's really very much outdated.
Because it's a bridged design the power is divided over 4 transistors, so even if it is 100W in total, it'll only be 25W per transistor. That's easy to keep cool... At the moment I don't even need to cooling-ribs on my transistors
Class AB normally makes the transistors hot, because they have to supply current continuously. But that's not the case, because an ESL only has capacity, no resistance. And this difference makes the design of such an amplifier quite different from a normal amplifier. Low frequencies use almost no current.
About my calculations:
0.7mm d/s indeed; 3cm is the distance between spacers;
For the lowest (infinite size) frequency I've calculated the frequency where the force from making soundwaves becomes equal to the backpulling (tension-) force of the membrane.
The resonance-frequency is calculated with the simple 'guitar string' formula. First you calculate the sound velocity of transversal waves through the diaphragm, and than you look at the frequency where a half wave fits in between the spacers. In vacuum this resonance would be clearly visible; but with air, it would be damped.
I can see that this resonance is damped because it's still much higher than the lowest frequency. When there is a resonance, the sound output should vanish rapidly below the resonance. (like here: http://www.dolmetsch.com/resonance.jpg but here it vanishes above resonance, meaning it's a different system)
As long as the resonce-frequency is very high, it won't be a problem. But at low frequencies, the air loading, and so also the damping, may get much lower. (especially without a baffle, and at the edges of the ESL)
This all is why I want a very high resonance frequency. I totally agree that a low frequency will be bad.
There might also be other resonances at frequencies where the panelsize is around 1; 2; 3; etc. times the wavelength or something like that. These are acoustic resonances, and here the remedies are totally different.
About heat treatment... What temperature do I need? (just a hairdryer?)
I assume heat treatment should be done under tension, right?
Last edited:
Hi,
class A is apart from efficiency still the best working class technically as well as sonically. As such its not outdated. And most amplifiers are still biased in class-A in input and driver stage and only AB-biased in the power stage.
Class A has the advantage that it allows for omittance of global feedback, while class-AB needs feedback.
Since the ESL-cap represents a complex load global feedback may be the reason for instability of the amp.
A bridged amp design doesn´t tell anything about its possibility to work in class A or class AB. It depends on the concept of each leg of the bridge as I explained earlier.
Indeed presents the cap a very high impedance at low frequencies wich means that the current demand is low and the ESL efficiency is high. Its the frequency range where basically just voltage demands count.
That is typically not the frequency range where the drive problems occur. Its the high frequency range that demands current, hence power. So the low frequency demand of the ESL dictates the voltage range the amp needs to cover while the high frequency demand of the ESL dictates
the current range. Since the amps Bias -and to a certain degree the amplitude response too- is constant a lot of power is wasted in the highfrequency range.
Regarding Fs I can only repeat my earlier thread and hint You that bolserst too has told You about the flaws and failures in Your assumptions and Thoughts about the resonance Fs.
jauu
Calvin
class A is apart from efficiency still the best working class technically as well as sonically. As such its not outdated. And most amplifiers are still biased in class-A in input and driver stage and only AB-biased in the power stage.
Class A has the advantage that it allows for omittance of global feedback, while class-AB needs feedback.
Since the ESL-cap represents a complex load global feedback may be the reason for instability of the amp.
A bridged amp design doesn´t tell anything about its possibility to work in class A or class AB. It depends on the concept of each leg of the bridge as I explained earlier.
Indeed presents the cap a very high impedance at low frequencies wich means that the current demand is low and the ESL efficiency is high. Its the frequency range where basically just voltage demands count.
That is typically not the frequency range where the drive problems occur. Its the high frequency range that demands current, hence power. So the low frequency demand of the ESL dictates the voltage range the amp needs to cover while the high frequency demand of the ESL dictates
the current range. Since the amps Bias -and to a certain degree the amplitude response too- is constant a lot of power is wasted in the highfrequency range.
Regarding Fs I can only repeat my earlier thread and hint You that bolserst too has told You about the flaws and failures in Your assumptions and Thoughts about the resonance Fs.
jauu
Calvin
I perfectly know how all amplifier classes work. That's not the problem.
I totally don't agree that an ESL is a problematic load. It is (very) problematic with a transformer in front, but without it it's very easy.
Normal amplifiers have to be able to deal with loads which can range from largely capacitive to largely inductive, and they have to remain stable over this whole range. For an ESL it's capacitive and nothing else. At the low frequencies the resistive part of the impedance is getting larger, but these frequencies are not the main problem for stability. At high-frequencies it's really >95% capacity. So there it's as easy as it gets.
But the amplifier setup should be somewhat different than that of an amplifier for Ohmic loads.
My amplifier remained really totally stable, except for some really HF oscillations of >100MHz. Those HF oscillations were not in the entire feedback loop, but just in one transistor at very large transients (>300V in 10us). (quite difficult to measure though, luckily I've recently bought a 350MHz oscilloscope, otherwise i wouldn't even have noticed it)
I really don't get how you get these low resonance frequencies. You use lots of tension. Looking at how you build you units it could easily be 100N/m in one direction.
In a link of Bolserst he gives this: F = sqrt(T/M) * sqrt( (m/a)^2 + (n/b)^2)
Which is the same as I use (however, i do the same in one dimension; but that shouldn't matter) Put in some random numbers: (1,1) mode; T=100N/m; M=10g/m^2; a=b=10cm gives: f= 1.4kHz. Taking a much lower tension of 1N/m (which probably isn't even stable) still gives 140Hz. All much higher than the resonances measured. My only conclusion is that other things are happening here.
I totally don't agree that an ESL is a problematic load. It is (very) problematic with a transformer in front, but without it it's very easy.
Normal amplifiers have to be able to deal with loads which can range from largely capacitive to largely inductive, and they have to remain stable over this whole range. For an ESL it's capacitive and nothing else. At the low frequencies the resistive part of the impedance is getting larger, but these frequencies are not the main problem for stability. At high-frequencies it's really >95% capacity. So there it's as easy as it gets.
But the amplifier setup should be somewhat different than that of an amplifier for Ohmic loads.
My amplifier remained really totally stable, except for some really HF oscillations of >100MHz. Those HF oscillations were not in the entire feedback loop, but just in one transistor at very large transients (>300V in 10us). (quite difficult to measure though, luckily I've recently bought a 350MHz oscilloscope, otherwise i wouldn't even have noticed it)
I really don't get how you get these low resonance frequencies. You use lots of tension. Looking at how you build you units it could easily be 100N/m in one direction.
In a link of Bolserst he gives this: F = sqrt(T/M) * sqrt( (m/a)^2 + (n/b)^2)
Which is the same as I use (however, i do the same in one dimension; but that shouldn't matter) Put in some random numbers: (1,1) mode; T=100N/m; M=10g/m^2; a=b=10cm gives: f= 1.4kHz. Taking a much lower tension of 1N/m (which probably isn't even stable) still gives 140Hz. All much higher than the resonances measured. My only conclusion is that other things are happening here.
Just did some measurements on the transformer of my final 80i's. I'm quite impressed how easily these things are able to ouput high frequencies. Just above 20kHz the output was decreasing, but at 50kHz it was still only -6dB.
I can't say that the total response is very flat. Putting in a square wave clearly shows a lot of bumps and dips. But this has got to be expected and I estimate that the differences stay below +/-2dB within the audible range.
I can't say that the total response is very flat. Putting in a square wave clearly shows a lot of bumps and dips. But this has got to be expected and I estimate that the differences stay below +/-2dB within the audible range.
Mass:
From Beranek, the airload on a circular ESL diaphragm with NO baffle can be estimated by:
M = 2.67 * a^3 * p0 (kg)
a = radius of diaphragm (m)
po = density of air = 1.18 Kg/m^3
In practice, even the small baffle contribution from the frame edges adds considerably to the airload. For your 10cm x 10cm case, I'd estimate the airload would be something like 0.5g. Adding a large baffle around the ESL will nearly double this.
The close proximity of the stators may also be inducing some additional airload on the diaphragm.
Tension:
Be aware that the HV bias applied to the diaphragm results in a negative stiffness term being adding to the diaphragm tension. This results in significant lowering of the fundamental resonant frequency.
http://www.diyaudio.com/forums/plan...agm-resonance-change-hv-bias.html#post1886659
Heat treating does modify the tension to a long term stable value. If film was stretched very tight before heat treating, it will loosen some. If film was applied with little tension, it will tighten up some.
For 6micron film and heat treating at 150 - 160 degC you should expect fundamental resonance of about 100Hz for a 10cm x 10cm ESL with no baffle.
Damping:
Airload does not contribute any significant damping to the diaphragm resonance modes. Expect high Q fundamental resonance with peak of +12dB to +24dB relative to midband unless resistive damping or notch filters are used. Q is dependent on ratio of tension to mass for your diaphragm. Higher tension results in higher Q.
Hi,
the easiest load for an amp is a pure resistive one, the most difficult ones are solely complex as are caps and inductors. Caps especially, because they represent a non-constant frequency dependant load and they reduce the amp´s phase margin until it starts oscillating, because of the negative feedback changing into a positive feedback.
Typical amplifier are only working well only into rather ´soft´complex loads with phase values staying below +-45°. There are some that will work perfectly well even into +-60°. The number of amps that work stable into pure capacitive loads is small. This is regardless of a transformer switched between amp and ESL or a direct driven ESL. The transformer actually eases the loading by reducing the phase shift, the more so the higher the transformers losses are.
Cap loading even makes certain amp-topologies useless or very difficult to stabilitze. For example would a SRPP-circuit be very favourable here because it produces high gain, low output impedance and beeing a pushpull-topology it could provide twice the bias-current onto the load (high-efficiency class-A). But it only works correctly into a constant load impedance- hence resistive load. It ceases beeing pushpull at all when working into complex loads. Similarities appear when a voltage controlled current source is used as active load. No problem when the load is resistive, problems to get it work stable with complex loads.
BTW: since You talked about a bridge design in Your amp would You like to share with us how You realized the 4 drive-signals for the 4 bridge-legs?
Just ask Yourself why basically each and every OP-amp datasheet covers the subject of capacitive loading and gives hints and formulas how to cope with such loads if it all were that easy as You claim?
Anyway I´ve gotten the impression, that I´m wasting my time here trying to help someone who doesn´t need help, because he thinks he knows better. I just hope that other readers can profit on a bit out of this thread.
jauu
Calvin
the easiest load for an amp is a pure resistive one, the most difficult ones are solely complex as are caps and inductors. Caps especially, because they represent a non-constant frequency dependant load and they reduce the amp´s phase margin until it starts oscillating, because of the negative feedback changing into a positive feedback.
Typical amplifier are only working well only into rather ´soft´complex loads with phase values staying below +-45°. There are some that will work perfectly well even into +-60°. The number of amps that work stable into pure capacitive loads is small. This is regardless of a transformer switched between amp and ESL or a direct driven ESL. The transformer actually eases the loading by reducing the phase shift, the more so the higher the transformers losses are.
Cap loading even makes certain amp-topologies useless or very difficult to stabilitze. For example would a SRPP-circuit be very favourable here because it produces high gain, low output impedance and beeing a pushpull-topology it could provide twice the bias-current onto the load (high-efficiency class-A). But it only works correctly into a constant load impedance- hence resistive load. It ceases beeing pushpull at all when working into complex loads. Similarities appear when a voltage controlled current source is used as active load. No problem when the load is resistive, problems to get it work stable with complex loads.
BTW: since You talked about a bridge design in Your amp would You like to share with us how You realized the 4 drive-signals for the 4 bridge-legs?
Just ask Yourself why basically each and every OP-amp datasheet covers the subject of capacitive loading and gives hints and formulas how to cope with such loads if it all were that easy as You claim?
Anyway I´ve gotten the impression, that I´m wasting my time here trying to help someone who doesn´t need help, because he thinks he knows better. I just hope that other readers can profit on a bit out of this thread.
jauu
Calvin
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.