Well, sorry if i'll touch the same subject as some previous threads, but i couldn't resist: I've been reading this forum from long time ago, mainly interested on High Fi audio.
From all things i've read, (and some i've heard and tried myself), seems ESL speakers are amongst the best ones regarding distorsion figures and audio clarity (transparency is just a better term) ... But Driving them is the problem... You need either a transformer to step up the voltage output of a conventional audio amplifier to the levels required by an ESL panel (with all the problems it comes with... Does not matter how expensive it is, it can't cover the full audio spectrum, at least without sacrificing some audio quality or audio level, ore frecuency response). And you also need a powerful (100W+) amplifier that is unconditionaly stable into capacitive loads, and able to provide more current to the speaker that it would be required for a traditional dynamic loudpeaker of the same power, something that most ampls out there aren't able to do. So we will try to avoid using an out transformer ... that way there's one less source of audio signal degradation.
The other way to drive an ESL is using an specially crafted high voltage audio amplifier... But, that amp has to provide about 2Kv pp (peak to peak) or even more to each stator. I have seen several schematics of such amps on this forum. Most of them have some kinds of limitations, just because semiconductors can't handle such high voltages, at least not easily...
Nevertheless there have been several attempts to build such amplifiers, and some of them worked pretty well. Amongst all possible ways to do it, valve amplifiers are the most seen ones, just because some transmitter tubes (used as the output devices) can handle the required output voltages. But here we can see some limitations: Assume we are using an ESL panel that has a capacity of 10nF (as a big panel could have). If we try to play an audio tone of 20Khz at maximum SPL, we would need to provide 2500mA just to charge that capacity from 0v to 2000v in just 1/20000 of a second. Most high voltage valves can't provide such high current (simply due to power dissipation issues on the plate of the tube), so simply, with those amps you can run into slew rate problems easily (i seem to be hearing from some people here that real audio signals do NOT have such frecuencies with such high levels... That seems to be right, BUT it could happen that under special circunstances an amp could have to amplify such signals, and as the amp can't simply provide enough current to the panel, that amp would severely distort the audio signal in that moment. We are looking for HiFi here, so i think that it would be preferable to buld an amp that does no have that limit. Valves tend to degrade with time, so a tube amplifier requires constant monitoring and tube replacement every 10000/20000 hours (obviously , tube duration depends on how frequent the amplifier is used, and the amount of hours used by month). But if a tube degrades, so will happen with sound quality... It it is noticeable or not, is something i won't discuss here (because this post is NOt about tube amps).
But there are other problems with DirectDrive amps... Most high voltage devices (Valves, MOSFET's, BJTs) are only manufactured of a given polarity (MOSFETs are always NChannel for Hv, Valeves only exist on one polarity, similar to NChannel mosfets, and BJTs also only come on NPN types for HV). That makes (without using a transformer) nearly impossible to create a PushPull (Class AB) amplifier, so there's only left the class A kind of amplifiers, that suffer from HUGE power losses and the worst enery efficiency of all classes. Assuming we could want to build an amp that could be able to deliver 2A of current at 2000vpp in claas A, that would mean 4000W of dissipation on output devices! ... So class A and valves or mosfets or BJTs simply won't provide the performance required for a no compromise directdrive ESL amplifier...
There have been some attempts to build a DirectDrive ClassAB amp using either tubes or MOSFEts (BJT's dont provide enough gain to implement a High voltage amplifier, as most HV BTTs available were made for switching supplies that use them as switches not amplifiers and to make a BJT switch fast normally means it has low beta). Though it can be done, the VAS stage of the HV amp is just very difficult to make it work , as it also requires HV deivecs, and also has a very high dissipation... So, usually class AB HV amps simply can't be made of high quality (not to mention stability problems with HV ampls are even harder to solve)
All the above things (high power dissipation, current peaks ELS panels require, etc) made me think about class D amplification... Class D can deliver incredibly high current peaks , and inherently has a low power dissipation and high efficiency... But i wasn't able to find a HV version of such an amplifier... And i was honestly thinking it was impossible to achieve, as Class D requires to switch (in the case of HV amplification) several KVolts in a very short time. Valves are not suitable for classD, because their dissipation issues (yes, they could be used, but with big power losses). Remember we are trying to create an AMP that can deliver up to 2A of current to the panels, valves won't handle that amount of current!
On the ClassD forum, i see there exists a topology (in fact a feed back method) called UCD, that makes the amp using it to have a very low output impedance, that makes the output impedance load invariant, that feedbacks the output voltage of the filter used to demodulate the switching waveform created by the amp, and that uses a capacitor in parallel with the load... And that has very low intermodulation and distortion figures... So i asked myself if that topology could be adapted to create, perhaps, the ultimate ESL DD amplifier... That happened about a year ago.
I thought that the ESL panel could be the capacitor of the demodulator filter, and no load resistor should be needed. As the UCD feedback makes the frequency response of the LC (inductor + ESL panel) flat, no resonance would be seen at the output (i mean resonance between the panel capacity and the output inductor).... Up to that point, the amp was feasible... But...
But the problems started right there, as a switching stage using NChannel mosfets (the only ones available for such high voltages) had to be designed. I thought it was impossbiel to achieve, as no IC mosfet driver is available for such high voltage difference between the low side mosfets and the high side mosfets. And all the discrete drivers you could imagine also required very high voltage BJTs to implements the level translation ... Even worse, UCD does not warranty the duty cycle of the signal ill be 50%, so ,at first glance , coupling transformers (used to control the gates of the mosfets) would be of no use here.
Optocouplers were also discarded, as they require an auxiliary power supply to operate, and are slow compared to the UCD switching frecuency required to get High Fidelity audio...
But yesterday i found an app note (AN950) from IRF (i was looking for something totally unrelated to audio), that seemed to be the answer to the problem... I way to use a signal transformer as a gate driver with no duty cycle limitations... I added that missing part of the schematic to the "UCD" HV amp and performed some simulations on LTSpice... And seems that it works!
The amp as it is has NO bandwidth limitations, (0-20Khz) , gives 2KVpp into the ESL panel with 1vpp as the signal input...
Right now, i simply have no time to try to build it (i'm extremely busy with other projects) , but i think that posting this amp to the forum would show a new workable approach to ESL DD amplification... And perhaps, to the ultimate ESL amplifier...
Well, i'll throw the idea... I really would like to hear your thoughts about this.
On the attachment , i've included the schematics in LTSpice, the required libraries, models, the IRF app note (just in case it is not available anymore), and a PNG of thew schematic.
PLEASE, if you ever try to build it, BE CAREFULL, this amp CAN EASILY KILL YOU. Dont try it if you don't know what you are doing. I won't take any responsability for any damage you could do to yourself or anyone/anything that could happen.
Also note this amp REQUIRES shielding. The thing MUST be enclosed on a metal box GROUNDED , else , your radiofrequency regulation agency of your government could knok at your door for severely interferring radio communications when this amp is on. You were warned...
Well,let's hope the discussion will be interesting... I know there could be some trickier things to take into account to make the amps work, but at least, it seems to be a way to make it
From all things i've read, (and some i've heard and tried myself), seems ESL speakers are amongst the best ones regarding distorsion figures and audio clarity (transparency is just a better term) ... But Driving them is the problem... You need either a transformer to step up the voltage output of a conventional audio amplifier to the levels required by an ESL panel (with all the problems it comes with... Does not matter how expensive it is, it can't cover the full audio spectrum, at least without sacrificing some audio quality or audio level, ore frecuency response). And you also need a powerful (100W+) amplifier that is unconditionaly stable into capacitive loads, and able to provide more current to the speaker that it would be required for a traditional dynamic loudpeaker of the same power, something that most ampls out there aren't able to do. So we will try to avoid using an out transformer ... that way there's one less source of audio signal degradation.
The other way to drive an ESL is using an specially crafted high voltage audio amplifier... But, that amp has to provide about 2Kv pp (peak to peak) or even more to each stator. I have seen several schematics of such amps on this forum. Most of them have some kinds of limitations, just because semiconductors can't handle such high voltages, at least not easily...
Nevertheless there have been several attempts to build such amplifiers, and some of them worked pretty well. Amongst all possible ways to do it, valve amplifiers are the most seen ones, just because some transmitter tubes (used as the output devices) can handle the required output voltages. But here we can see some limitations: Assume we are using an ESL panel that has a capacity of 10nF (as a big panel could have). If we try to play an audio tone of 20Khz at maximum SPL, we would need to provide 2500mA just to charge that capacity from 0v to 2000v in just 1/20000 of a second. Most high voltage valves can't provide such high current (simply due to power dissipation issues on the plate of the tube), so simply, with those amps you can run into slew rate problems easily (i seem to be hearing from some people here that real audio signals do NOT have such frecuencies with such high levels... That seems to be right, BUT it could happen that under special circunstances an amp could have to amplify such signals, and as the amp can't simply provide enough current to the panel, that amp would severely distort the audio signal in that moment. We are looking for HiFi here, so i think that it would be preferable to buld an amp that does no have that limit. Valves tend to degrade with time, so a tube amplifier requires constant monitoring and tube replacement every 10000/20000 hours (obviously , tube duration depends on how frequent the amplifier is used, and the amount of hours used by month). But if a tube degrades, so will happen with sound quality... It it is noticeable or not, is something i won't discuss here (because this post is NOt about tube amps).
But there are other problems with DirectDrive amps... Most high voltage devices (Valves, MOSFET's, BJTs) are only manufactured of a given polarity (MOSFETs are always NChannel for Hv, Valeves only exist on one polarity, similar to NChannel mosfets, and BJTs also only come on NPN types for HV). That makes (without using a transformer) nearly impossible to create a PushPull (Class AB) amplifier, so there's only left the class A kind of amplifiers, that suffer from HUGE power losses and the worst enery efficiency of all classes. Assuming we could want to build an amp that could be able to deliver 2A of current at 2000vpp in claas A, that would mean 4000W of dissipation on output devices! ... So class A and valves or mosfets or BJTs simply won't provide the performance required for a no compromise directdrive ESL amplifier...
There have been some attempts to build a DirectDrive ClassAB amp using either tubes or MOSFEts (BJT's dont provide enough gain to implement a High voltage amplifier, as most HV BTTs available were made for switching supplies that use them as switches not amplifiers and to make a BJT switch fast normally means it has low beta). Though it can be done, the VAS stage of the HV amp is just very difficult to make it work , as it also requires HV deivecs, and also has a very high dissipation... So, usually class AB HV amps simply can't be made of high quality (not to mention stability problems with HV ampls are even harder to solve)
All the above things (high power dissipation, current peaks ELS panels require, etc) made me think about class D amplification... Class D can deliver incredibly high current peaks , and inherently has a low power dissipation and high efficiency... But i wasn't able to find a HV version of such an amplifier... And i was honestly thinking it was impossible to achieve, as Class D requires to switch (in the case of HV amplification) several KVolts in a very short time. Valves are not suitable for classD, because their dissipation issues (yes, they could be used, but with big power losses). Remember we are trying to create an AMP that can deliver up to 2A of current to the panels, valves won't handle that amount of current!
On the ClassD forum, i see there exists a topology (in fact a feed back method) called UCD, that makes the amp using it to have a very low output impedance, that makes the output impedance load invariant, that feedbacks the output voltage of the filter used to demodulate the switching waveform created by the amp, and that uses a capacitor in parallel with the load... And that has very low intermodulation and distortion figures... So i asked myself if that topology could be adapted to create, perhaps, the ultimate ESL DD amplifier... That happened about a year ago.
I thought that the ESL panel could be the capacitor of the demodulator filter, and no load resistor should be needed. As the UCD feedback makes the frequency response of the LC (inductor + ESL panel) flat, no resonance would be seen at the output (i mean resonance between the panel capacity and the output inductor).... Up to that point, the amp was feasible... But...
But the problems started right there, as a switching stage using NChannel mosfets (the only ones available for such high voltages) had to be designed. I thought it was impossbiel to achieve, as no IC mosfet driver is available for such high voltage difference between the low side mosfets and the high side mosfets. And all the discrete drivers you could imagine also required very high voltage BJTs to implements the level translation ... Even worse, UCD does not warranty the duty cycle of the signal ill be 50%, so ,at first glance , coupling transformers (used to control the gates of the mosfets) would be of no use here.
Optocouplers were also discarded, as they require an auxiliary power supply to operate, and are slow compared to the UCD switching frecuency required to get High Fidelity audio...
But yesterday i found an app note (AN950) from IRF (i was looking for something totally unrelated to audio), that seemed to be the answer to the problem... I way to use a signal transformer as a gate driver with no duty cycle limitations... I added that missing part of the schematic to the "UCD" HV amp and performed some simulations on LTSpice... And seems that it works!
The amp as it is has NO bandwidth limitations, (0-20Khz) , gives 2KVpp into the ESL panel with 1vpp as the signal input...
Right now, i simply have no time to try to build it (i'm extremely busy with other projects) , but i think that posting this amp to the forum would show a new workable approach to ESL DD amplification... And perhaps, to the ultimate ESL amplifier...
Well, i'll throw the idea... I really would like to hear your thoughts about this.
On the attachment , i've included the schematics in LTSpice, the required libraries, models, the IRF app note (just in case it is not available anymore), and a PNG of thew schematic.
PLEASE, if you ever try to build it, BE CAREFULL, this amp CAN EASILY KILL YOU. Dont try it if you don't know what you are doing. I won't take any responsability for any damage you could do to yourself or anyone/anything that could happen.
Also note this amp REQUIRES shielding. The thing MUST be enclosed on a metal box GROUNDED , else , your radiofrequency regulation agency of your government could knok at your door for severely interferring radio communications when this amp is on. You were warned...
Well,let's hope the discussion will be interesting... I know there could be some trickier things to take into account to make the amps work, but at least, it seems to be a way to make it
Hi Ejtagle,
With the voltage requirements I think you are too optimistic. 2kv tt will not do. I would say that for a non-compromise full range amp you need at least twice that, so 8 kv tt total drive. And yes, more is better in this case
On this bit I believe you are a bit pessimistic, even considering worst case scenarios. I've seen full amplitude swing in music up till around 6 kHz, and even this only with pretty extreme synthetic music (that isn't survivable anayway when listened to at such volumes
). So even worst case, I think it's save to assume that an amp capable of delivering full voltage swing in to a panel up to a khz or 4 will never run into trouble.
As for panel capacity, 10 nf is also pretty extreme. A very large wire stator esl has around 2 nf, but only a very small portion of this will have to be driven directly, the bulk is connected via high-value resistors ( (because of segmentation). For plate-stator esl's this is different as they don't use electrical segmentation, but even in that case I think 2-3 nf is more realistic. If you need even bigger panels then you probably have to reconsider the panel design...
Yes, but there's no need to go wild here. I did some extensive measurements on peak stator currents in the past and I believe that 50 mA/kV will do in all cases, even half that will probably never give any problems. Maybe some more for plate stators. Which results in around 200-500 mA for a 8 kV tt amp. Which still gives > 3000 W dissipation/channel in class A... More voltage, less current, same result
Couldn't agree more... That IS the main problem with DD.
But they will handle a few 100 mA so considering more realsitic demands might be used in a class D stage after all, using the concept as shown with mosfets in the attachment. This will work with tubes as well, assuming the filament to cathode isolation of the upper tube won't give any problems (which I doubt). Still tubes will have much higher voltage losses. Btw, with mosfets this will not work for a non-compromise class D design either because there are none capable of handling 4kV or more. Which means you have to stack several, which will make the output bridge too slow for class D.
Not always, you can 'drain' the power supply from the output current.
Most are, those that are not are incredible expensive, especially when you need 4 kV or more isolation. Still, you can buy commercial linear high voltage amplifers that use mosfets and optocouplers. The problem is the price and the fact that those amps are not very linear, so unsuited for audio.
The problem is, with HV designs things are often very different in real life. Spice models of mosfets do for instance not account for the voltage dependency of the internal capacities in the fets, nor do fets break down in spice... And there is the crosstalk issue, with signals swinging several kV even a fraction of a pf will inject significant signals in the lowlevel sections unless you're very carefull and spice won't account for that either. I have lot's of hv amp designs that work in spice but are a complete disaster in reality
Unfortunately, my knowledge of class D is very limited so I can't comment in detail on your proposal. But I think it might be an interesting approach, it might indeed be a way of driving the output stage of a high-voltage class D amp fast enough. Now, to build it and to solve all the other problems on the way...
so am I
Maybe somebody else is boring himself?
With the voltage requirements I think you are too optimistic. 2kv tt will not do. I would say that for a non-compromise full range amp you need at least twice that, so 8 kv tt total drive. And yes, more is better in this case
On this bit I believe you are a bit pessimistic, even considering worst case scenarios. I've seen full amplitude swing in music up till around 6 kHz, and even this only with pretty extreme synthetic music (that isn't survivable anayway when listened to at such volumes
). So even worst case, I think it's save to assume that an amp capable of delivering full voltage swing in to a panel up to a khz or 4 will never run into trouble.As for panel capacity, 10 nf is also pretty extreme. A very large wire stator esl has around 2 nf, but only a very small portion of this will have to be driven directly, the bulk is connected via high-value resistors ( (because of segmentation). For plate-stator esl's this is different as they don't use electrical segmentation, but even in that case I think 2-3 nf is more realistic. If you need even bigger panels then you probably have to reconsider the panel design...
Yes, but there's no need to go wild here. I did some extensive measurements on peak stator currents in the past and I believe that 50 mA/kV will do in all cases, even half that will probably never give any problems. Maybe some more for plate stators. Which results in around 200-500 mA for a 8 kV tt amp. Which still gives > 3000 W dissipation/channel in class A... More voltage, less current, same result
Couldn't agree more... That IS the main problem with DD.
But they will handle a few 100 mA so considering more realsitic demands might be used in a class D stage after all, using the concept as shown with mosfets in the attachment. This will work with tubes as well, assuming the filament to cathode isolation of the upper tube won't give any problems (which I doubt). Still tubes will have much higher voltage losses. Btw, with mosfets this will not work for a non-compromise class D design either because there are none capable of handling 4kV or more. Which means you have to stack several, which will make the output bridge too slow for class D.
Not always, you can 'drain' the power supply from the output current.
Most are, those that are not are incredible expensive, especially when you need 4 kV or more isolation. Still, you can buy commercial linear high voltage amplifers that use mosfets and optocouplers. The problem is the price and the fact that those amps are not very linear, so unsuited for audio.
The problem is, with HV designs things are often very different in real life. Spice models of mosfets do for instance not account for the voltage dependency of the internal capacities in the fets, nor do fets break down in spice... And there is the crosstalk issue, with signals swinging several kV even a fraction of a pf will inject significant signals in the lowlevel sections unless you're very carefull and spice won't account for that either. I have lot's of hv amp designs that work in spice but are a complete disaster in reality
Unfortunately, my knowledge of class D is very limited so I can't comment in detail on your proposal. But I think it might be an interesting approach, it might indeed be a way of driving the output stage of a high-voltage class D amp fast enough. Now, to build it and to solve all the other problems on the way...
so am I
Maybe somebody else is boring himself?
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.