| High Wattage Electrostatic Loudspeakers - Click HERE for Original Thread |
| ak_47_boy |
Say you had a 100watt amplifier, 1kw amplifier, and a panel with a 100:1 step up transformer.
The final drive voltage would be 2800v for the 100watt amp and 9000v for the 1kw. The 1kw will obviously be able to play louder but will it actually be using 100mA @ 9000v?
I don't understand where 1kw of power goes. It's not heat or sound. |
|
|
| SY |
Actually, heat and sound are exactly where it goes.
An ESL is a big capacitor. To get from zero to some voltage requires current. To get from zero to that voltage even faster takes more current. Now, when it's time to go back to zero, the higher the available current, the faster it can happen. But where does that current go? Yes, right back to the amplifier. Being reactive, to a first approximation the speaker absorbs no power, but the amp does. And that's where the heat goes; remember power factor, too.
Now, when we refine the approximation, we can see where the sound energy goes, but it's still a small proportion of the electrical energy that the amplifier needs to source and sink charging currents. It's an interesting exercise to calculate how much current it would take to drive a 1nF capacitor (a typical ESL value) at 20kHz. |
|
|
| ak_47_boy |
| Wow that actually works out to 8 ohms. |
|
|
| Eva |
| A class D amplifier would be a good alternative for driving a highly reactive load because all the reactive energy is returned back to the PSU. In case of a capacitive load (ESL), the amplifier may be designed so that the load forms part of the output LC filter (although the resonance that results when the step-up transformer is added has to be tuned to avoid trouble). |
|
|
| I_Forgot |
| quote: | Originally posted by Eva
A class D amplifier would be a good alternative for driving a highly reactive load because all the reactive energy is returned back to the PSU. In case of a capacitive load (ESL), the amplifier may be designed so that the load forms part of the output LC filter (although the resonance that results when the step-up transformer is added has to be tuned to avoid trouble). |
Alternatively, you could do class D and use an HV output bridge to drive the panel directly. You'd need some high voltage switches and they'd have to be fast. I don't know if there are such devices anywhere.
I_F |
|
|
| Calvin |
Hi,
Yeah, a HV-D-amp would be fine thing. But apart from the parts prob YouŽll probabely get into serious trouble with those aliens in the 17 quadrant near Wega-colony who simply are not into human music ;)
That thing would rather be a strong emv-sender :dead:
jauu
Calvin |
|
|
| phase_accurate |
I agree that a purpose-built class-d would be great for driving ESLs. But I am afraid that we might not be able to do a "classic" class-d amp that is capable of doing the desired voltage swing. I once consulted someone who was developing an amp intended to drive piezo actuators up to 1000 volts. The idea of an ordinary class-d half-bridge was given up quite quickly due to the huge voltage swings involved and the subsequent high snubber losses.
He then built a topology of multiple low(er)-voltage class-d amps that add up their voltages. I don't know if he even did them in some multiphase modulation scheme which would further reduce ripple (or make lower switching frequencies and therefore lower switching losses feasible).
A mutlibit (i.e. multi voltage output stage) delta-sigma topology might be quite cool as well !
The RF transmission problem could however be minimised IMO using conventional methods.
Regards
Charles |
|
|
| mzzj |
| quote: | Originally posted by I_Forgot
Alternatively, you could do class D and use an HV output bridge to drive the panel directly. You'd need some high voltage switches and they'd have to be fast. I don't know if there are such devices anywhere.
I_F |
Too bad that 6.5kV IGBT's are availlable only with insane current ratings. Probably too slow, too.
Stack of low current 600-1200v mosfets would be "intresting" for switching ~6kV . |
|
|
| phase_accurate |
I completely forgot about that. :o
It is a clever idea indeed !
Regards
Charles |
|
|
| Calvin |
Hi,
clever idea indeed, but it is rather restricted to hybrid panels, because itŽll work as a constant voltage design. A CV-electrostat is a inherent nonlinear design, creating distortions, that can only be kept sufficiently small when the movement of the membrane is very small. To allow for more movement or lower distortion YouŽll have to linearize the system with some means of feedback, creating the problem of how to create the feedback sinal and to get the system stable.
The well proven alternative is the ConstantCharge-design (with small d/s and rather lots of membrane area) with (relatively) low voltage demand, using a tranny with lower transformation factor or a more classical HV-amp running on 1kV to 2kV supply voltage. This is much easier to design, much cheaper and imo the best is, that You use a inherently linear, low-distortion transducer that doesnŽt need feedbacking in first place.
@AK47: In Your example You assumed a constant impedance of 8 Ohms. This will definitely not be the case with a tranny-coupled ESL.
The impedance varies between fractions of an Ohm up to several 100s of Ohms and the impedance curve is shaped similar to a gaussian curve.
Too You assume and calculate with rms-values. Its common and more reasonable to calculate with p-p values when You are dealing with ESLs!
Based on the assumptions of 28Vrms and 90Vrms this would give 8kVpp resp. 25.5kVpp in Your example!
jauu
Calvin |
|
|
| phase_accurate |
You need some type of synchronous rectifier to turn the voltage into DC - and you need to do it for both sides seperately (out of phase) - then it can be applied to constant-charge electrostats.
Regards
Charles |
|
|
| phase_accurate |
One might argue that you have switches again on the HV side but these would be easier to implement than a class-d stage and it can be made transformer-driven.
Regards
Charles |
|
|
| Elvee |
| quote: | Originally posted by Calvin
Hi,
clever idea indeed, but it is rather restricted to hybrid panels, because itŽll work as a constant voltage design. A CV-electrostat is a inherent nonlinear design, creating distortions, that can only be kept sufficiently small when the movement of the membrane is very small. To allow for more movement or lower distortion YouŽll have to linearize the system with some means of feedback, creating the problem of how to create the feedback sinal and to get the system stable.
|
The scheme lends itself to feedback application: the movement of the membrane can be accurately monitored by applying opposite DC voltages to the grids and picking up the displacement signal on the membrane.
A practical way of implementing this could be the insertion of a capacitor in the cold connection of the drive transformer, forming a capacitive divider with the membrane-to-grid capacitance. The superimposed HF ripple can easily be removed using cancellation techniques, with no impact on the bandwidth, thus easing feedback stability issues.
LV |
|
|
| ak_47_boy |
| Interesting! Diaphragm feedback, thats something to think about! |
|
|
| G.Kleinschmidt |
| quote: | Originally posted by phase_accurate
He then built a topology of multiple low(er)-voltage class-d amps that add up their voltages. I don't know if he even did them in some multiphase modulation scheme which would further reduce ripple (or make lower switching frequencies and therefore lower switching losses feasible). |
Actually, this is real easy to do do, as the output of each class D amp (prior to the filter) can be transformer coupled. Then you just have to series connect the secondary windings, and all your class D amps can share common supply rails.
Reminds me of a pair of 495kW 500kHz RF generators I once had to fix when employed at Flinders University. These things were run in unison for a few mS (powered 16kV plate supply provided by a room full of capacitors) for plasma fusion experiments, providing the RF excitation for the Rotomak.
Each unit had a ~10kW PWM driver, built transformer coupled as described, to provide kV's of drive to the pair of tetrodes used in the final amplifier. PWM was used to ramp up the drive voltage as the plate voltage supplied by the capacitor bank dropped off, providing a constant 990kW of RF drive for a few mS or so.
This equiptment is currently operational in some university In Houston, AFAIK (I had to repair the PWM RF drive drive unit in a hurry for an operational demonstration to the prof who buying the whole rig) That was a fun thing to work on.
Cheers,
Glen |
|
|
| phase_accurate |
| quote: | | Actually, this is real easy to do do, as the output of each class D amp (prior to the filter) can be transformer coupled. |
If there is something that I wouldn't do then it is exactly that !!:(
Regards
Charles |
|
|
| G.Kleinschmidt |
| quote: | Originally posted by phase_accurate
If there is something that I wouldn't do then it is exactly that !!:(
Regards
Charles |
Why? |
|
|
| phase_accurate |
Hint: Simply bexause you don't gain anything compared to a class-d amp driving the usual step-up transformer !
If you intend to do what you mentioned you'd need audio tranformers that have at least their lower cutoff frequency equal to the woofer/ESL crossover frequency - preferably somewhat lower.
If you are going the class-d route you'd rather use insulated stages where each one is driving the (floating) ground of the subsequent stage. Then you can implement the floating PSUs using small SMPS.
Regards
Charles |
|
|
| G.Kleinschmidt |
| quote: | Originally posted by phase_accurate
Hint: Simply bexause you don't gain anything compared to a class-d amp driving the usual step-up transformer !
If you intend to do what you mentioned you'd need audio tranformers that have at least their lower cutoff frequency equal to the woofer/ESL crossover frequency - preferably somewhat lower.
If you are going the class-d route you'd rather use insulated stages where each one is driving the (floating) ground of the subsequent stage. Then you can implement the floating PSUs using small SMPS.
Regards
Charles |
I'm sorry, but you've got it completely wrong - I said "prior to the filter". The transformers connect directly to the MOSFET outputs - ie they work at the switching frequency.
The 500kHz 10kW PWM driver I mentioned earlier consisted of 10 synchronised 1kW half bridge modules, each the size of a eurocard. The 10 transformers were toroids, about 5cm in diameter, mounted on the PCB's.
Cheers,
Glen |
|
|
| phase_accurate |
| quote: | | The transformers connect directly to the MOSFET outputs - ie they work at the switching frequency. |
I got it completely right then ! The switching signal does contain quite some spectral content of the audio signal (which is intended, isn't it ?) and the transformers must therefore be able to work from audio up to some multiples of the switching frequency.
This wouldn't be the first failed attempt in transforming a PWM signal BTW.
Regards
Charles |
|
|
| G.Kleinschmidt |
| quote: | Originally posted by phase_accurate
I got it completely right then ! The switching signal does contain quite some spectral content of the audio signal (which is intended, isn't it ?) and the transformers must therefore be able to work from audio up to some multiples of the switching frequency. |
The audio base band incurs sidebands of the carrier and its harmonics not too unlike FM of a square wave. To the best of my knowledge, with a 500kHz carrier, there is nothing of value anywhere near <20kHz, and no need for the transformers to pass it.
| quote: | Originally posted by phase_accurate
This wouldn't be the first failed attempt in transforming a PWM signal BTW. |
What on earth are you talking about? |
|
|
| phase_accurate |
| quote: | | The audio base band incurs sidebands of the carrier and its harmonics not too unlike FM of a square wave. To the best of my knowledge, with a 500kHz carrier, there is nothing of value anywhere near <20kHz, and no need for the transformers to pass it. |
The audio signal IS DEFINITELY CONTAINED in the PWM signal. How in all the world do you think a lowpass filter could extract the audio signal from a PWM signal if it weren't like that ? If you want to transform this or doing signal addition using transformers then your transformer must be able to pass frequencies from audio up to several hundred kHz.
| quote: | | What on earth are you talking about? |
There were many people discussing PWM amps chopping up the rectified mains voltage and then transforming it using a small SMPS transformer. This is a no-no !!!! There are only two solutions that come to my mind at the moment that would work. Both use a phase modulator.
Another possibility would be to use two SMPS - one doing negative voltages the other one positive voltages. Then you sum both of these before going to the output filter. But these would be SMPS and not PWM amplifiers.
Regards
Charles |
|
|
| el`Ol |
My suggestion for high voltage class D would be simulating AM with a switching device. Full sqare wave =^ full amplitude, symmetric narrow spike (see picture) =^ small amplitude. Switching rate is constant. Then the step-up-transformer, after that a rectifier with selenium diodes or thyratrons for decoding.
The transformer would only have to handle a single frequency and bandwidth wouldn`t be an issue. |
|
|
| G.Kleinschmidt |
| quote: | Originally posted by phase_accurate
How in all the world do you think a lowpass filter could extract the audio signal from a PWM signal if it weren't like that ? |
Blah blah blah.
It's magic, obviously. |
|
|
| el`Ol |
There is one thing that goes in the direction of phase-accurate`s suggestion and I had almost forgotten:
Philips had a prototype of a full-digital amp some years ago that was aimed at getting the full potential of the SACD.
It consisted of 16 1-bit switching devices with 1/16 of the full speed of the SACD. They were added up (as far as I remember some inductors were involved here). Single switching devices don`t have the necessary timing precision to do the full SACD speed and need feedback for correction, hence this solution. But with 400V switching transistors one would be able to get 6400V.
Testers who listened to the prototype were impressed. But it didn`t find its way to the market. The name of that technology was an ugly acronym, so I forgot it. |
|
|
|
|
|
