In an other thread someone told me that Berning probably uses such a complex topology because a simple modulator-amplifier-rectifier topology would only have 25% efficiency because a capacitive coupling is necessary to get the signal on ground. But since this is not the case for an ESL, would such a simple topology work here?
Additional info: Berning cascades low voltage rectifiers, so high voltage at the output doesn´t seem to be a problem.
Additional info: Berning cascades low voltage rectifiers, so high voltage at the output doesn´t seem to be a problem.
On the point of panef efficiency: I've once did a calc, which gave me really high efficiency of a panel.. Here it is:
Panel size: 0.5*0.5m, spacing 6+6mm.
At 20kHz that gives 43kOhm.
Assuming 2kV driving signal, it's 50mA, ~100Wt.
Now, let's calculate the pressure:
P=Isig*Eht/(2*d*c*pi*r)
d is spacing, c - speed of sound, r- distance to measurement point.
Assuming 12kV EHT - high, yet possible, especially with such a spacing, value.
That's 24Pa = 27dbPa. 1 dbPa = 94 dbspl, 27+94=121. At 100Wt!
So sensitivity is 101 db? A somewhat too high value.. That's an just excellent sensitivity. I'll do some more calcs and post em soon.
Panel size: 0.5*0.5m, spacing 6+6mm.
At 20kHz that gives 43kOhm.
Assuming 2kV driving signal, it's 50mA, ~100Wt.
Now, let's calculate the pressure:
P=Isig*Eht/(2*d*c*pi*r)
d is spacing, c - speed of sound, r- distance to measurement point.
Assuming 12kV EHT - high, yet possible, especially with such a spacing, value.
That's 24Pa = 27dbPa. 1 dbPa = 94 dbspl, 27+94=121. At 100Wt!
So sensitivity is 101 db? A somewhat too high value.. That's an just excellent sensitivity. I'll do some more calcs and post em soon.
Some more thoughts:
A switching amp is extremely efficient when driving ESL - it recuperates energy, transferring it from PS caps into panel and back. A low-voltage amp has been built by a guy, which showed recuperation.
There are two issues with switching amps: quality and voltage capabilities. Quality is a hard question, guys here at Class D say that D-class is good, but, really, how good it is? That's whai I do not know.
Voltage capabilities: you'd need a switching device both fast and able to withstand HV. Trafo is of no help, as it reduces efficiency and makes recuperation much less desired. There are no such transistors, tubes aren't switching, then what? In some future, possibly, GFETs will be able to do so.
But then, are Class D really good? In the pretty much hi-end sense good, I mean, comparable to a class A tube amp, directly driving an ESL? Is it possible without boosting switching frequency into 100 MHz range, as that range has it's own problems with switching, which won't be solved by any type of active element?
A switching amp is extremely efficient when driving ESL - it recuperates energy, transferring it from PS caps into panel and back. A low-voltage amp has been built by a guy, which showed recuperation.
There are two issues with switching amps: quality and voltage capabilities. Quality is a hard question, guys here at Class D say that D-class is good, but, really, how good it is? That's whai I do not know.
Voltage capabilities: you'd need a switching device both fast and able to withstand HV. Trafo is of no help, as it reduces efficiency and makes recuperation much less desired. There are no such transistors, tubes aren't switching, then what? In some future, possibly, GFETs will be able to do so.
But then, are Class D really good? In the pretty much hi-end sense good, I mean, comparable to a class A tube amp, directly driving an ESL? Is it possible without boosting switching frequency into 100 MHz range, as that range has it's own problems with switching, which won't be solved by any type of active element?
Yea, I know they were meant as an example. Those are also much faster at lower current densities. Smaller (i.e. less than 200 Amp) IGBTs exist, they are also faster. MOS-FETS of up to 1500V switching at up to 5 MHz are available from SANYO.
Tubes rightly or wrongly deliver a certain sonic signature when used as linear amplifiers. Some like the signature others do not. But that special effect doesn't make tubes a superior or automatic solution to other topologies such as class D.
While tubes could make a pretty good linear ESL driver by taking advantage of their nature. I would spend more effort in looking for a solid state switch if I was pursuing a high voltage class D design.
Tubes rightly or wrongly deliver a certain sonic signature when used as linear amplifiers. Some like the signature others do not. But that special effect doesn't make tubes a superior or automatic solution to other topologies such as class D.
While tubes could make a pretty good linear ESL driver by taking advantage of their nature. I would spend more effort in looking for a solid state switch if I was pursuing a high voltage class D design.
Yes, that's quite logical to choose tubes for linear amp and SS for switching. 
I'm trying to address a different point:
Can a D-amp be on par with A-amp, regarding only quality? For long time, D-amps weren't good for anything but subwoofers, lacking highs. At once, ESL is much better than dynamic head for highs reproduction, so, coupling ESL with highs-lacking amp is certainly a good idea.
I haven't been reading Class D forum for long, but from what i understood, currently the best sounding D-amps are 'analog' designs, self-oscillating, having an analog input, and analog feedback into input from the end of LP filter. That is, as i understand, essentially a 1-bit ADC with output directrly connected to a 1-bit high-power DAC Theory seems to give quite good DD and noise for even a relatively low frequency unit with simple filtering, but that seems not to be very true, actual well-performing amps have high order filtering and high sampling frequency, which results in RF-type problems. Besides, an analog input isn't very appealing if you have a digital source, but fully digital amps doesn't seem to perform well, they are said to be inferior to 'analog' ones. And we don't know for sure how well a 'analog' D-amp performs against a good A-amp at highs..
I'm trying to address a different point:
Can a D-amp be on par with A-amp, regarding only quality? For long time, D-amps weren't good for anything but subwoofers, lacking highs. At once, ESL is much better than dynamic head for highs reproduction, so, coupling ESL with highs-lacking amp is certainly a good idea.
I haven't been reading Class D forum for long, but from what i understood, currently the best sounding D-amps are 'analog' designs, self-oscillating, having an analog input, and analog feedback into input from the end of LP filter. That is, as i understand, essentially a 1-bit ADC with output directrly connected to a 1-bit high-power DAC
Theory seems to give quite good DD and noise for even a relatively low frequency unit with simple filtering, but that seems not to be very true, actual well-performing amps have high order filtering and high sampling frequency, which results in RF-type problems. Besides, an analog input isn't very appealing if you have a digital source, but fully digital amps doesn't seem to perform well, they are said to be inferior to 'analog' ones. And we don't know for sure how well a 'analog' D-amp performs against a good A-amp at highs..There is no reason that an all digital amp could not be made or sound as good as the best analog designs, but with todays technologies it is damn hard.
It's not really true that the current class D designs are really one bit DACS. Think about it; A one bit DAC would need to switch at 2^16 * 44.1 KHz or 2,890,137,600 Hz (2,890 Mhz) in order to exactly cover every possible output state. All current class D amps truncate information.
Probably all or nearly all analog amps truncate as well, a 100 watt amp needs to output 400 microvolts for the least significant bit, this is below the amps noise floor. And that's at 100 watts, at a more normal listening level of say 5 watts we now need to control down to 20 uV.
I think most class D designs use a triangle wave fed to a comparator to generate a variable output pulse width. This is really an analog design disguised with a switching output. Making a triangle wave linear to 1/(2^16) is no picnic either.
It's not really true that the current class D designs are really one bit DACS. Think about it; A one bit DAC would need to switch at 2^16 * 44.1 KHz or 2,890,137,600 Hz (2,890 Mhz) in order to exactly cover every possible output state. All current class D amps truncate information.
Probably all or nearly all analog amps truncate as well, a 100 watt amp needs to output 400 microvolts for the least significant bit, this is below the amps noise floor. And that's at 100 watts, at a more normal listening level of say 5 watts we now need to control down to 20 uV.
I think most class D designs use a triangle wave fed to a comparator to generate a variable output pulse width. This is really an analog design disguised with a switching output. Making a triangle wave linear to 1/(2^16) is no picnic either.
el`Ol
Good point, but you'd need a good headphone
Well, now i got a clue that, because D-amps use a triangle wave and a comparator, the width of impulse controlling output switches is not discrete, and as such, requred frequency isn't extreme.
There is at least one person on forum who's quite an expert in Class D, yet prefers his OTL+ESL to them any day, though i don't know anything of his system, and subjectivist/objectivist approach. I'm concerned with general linearity of Class D, not only THD+N, but (T)IMD and else.
Good point, but you'd need a good headphone
Well, now i got a clue that, because D-amps use a triangle wave and a comparator, the width of impulse controlling output switches is not discrete, and as such, requred frequency isn't extreme.
There is at least one person on forum who's quite an expert in Class D, yet prefers his OTL+ESL to them any day, though i don't know anything of his system, and subjectivist/objectivist approach. I'm concerned with general linearity of Class D, not only THD+N, but (T)IMD and else.
It has been discussed in this thread that ESL's might need current drive to get flat frequency response.
I have found parts of Baxandall's article on esl theory, with explainations for P=Eht*Isig/(2pi*c*d*r) formula, having questioned it before. It is said that, if you drive a panel with constant voltage swing, you'll emit the same total acoustic power over all frequencies applied, while on-axis SPL will rise, due to beaming, while constant current swing will reduce total power at higher frequencies, but have constant on-axis SPL.
I see a series of consequences:
1) We are limited not by higher, but, instead, by lower frequencies. To get flat frequency response, we have to have a constant maximum current, which is limited by Umax/Zlf, where Umax is maximum voltage swing before breakdown, and Zlf is impedance at lowest supposed frequency of our panel. The resulting current is small, electrical power is small, and SPL isn't so great. We get a speaker that's not loud, but is at least quite sensitive. You just can't churn in more power - it will arc at LF.
2) If we attempt to get rid of beaming effect, in any way - we'll need a constant voltage drive, a real lot more power is required at HF, but the resulting speaker won't have higher SPL,as all the power will be radiated in all directions. The speaker gets really insensitive. That seems to justify low sensitivity of "dispersing" speakers, such as Quads, some Stax's old speakers, and such.
I have found parts of Baxandall's article on esl theory, with explainations for P=Eht*Isig/(2pi*c*d*r) formula, having questioned it before. It is said that, if you drive a panel with constant voltage swing, you'll emit the same total acoustic power over all frequencies applied, while on-axis SPL will rise, due to beaming, while constant current swing will reduce total power at higher frequencies, but have constant on-axis SPL.
I see a series of consequences:
1) We are limited not by higher, but, instead, by lower frequencies. To get flat frequency response, we have to have a constant maximum current, which is limited by Umax/Zlf, where Umax is maximum voltage swing before breakdown, and Zlf is impedance at lowest supposed frequency of our panel. The resulting current is small, electrical power is small, and SPL isn't so great. We get a speaker that's not loud, but is at least quite sensitive. You just can't churn in more power - it will arc at LF.
2) If we attempt to get rid of beaming effect, in any way - we'll need a constant voltage drive, a real lot more power is required at HF, but the resulting speaker won't have higher SPL,as all the power will be radiated in all directions. The speaker gets really insensitive. That seems to justify low sensitivity of "dispersing" speakers, such as Quads, some Stax's old speakers, and such.
I'm reasonably sure that for a given panel size and a given curvature, beaming vs frequency is a physical panel attribute. Unless bandwidth is limited I don't believe any amount of circuit topology or drive scheme variations will have any effect on beaming.
I am assuming of course that the panel frequency response is kept so that the power radiated into the room is flat.
One exception is the Quad technique of time shifting the drive signal to panels that are not electrically in parallel. However, once this physical design is manifested the electronics drive will again have no effect on beaming.
I am assuming of course that the panel frequency response is kept so that the power radiated into the room is flat.
One exception is the Quad technique of time shifting the drive signal to panels that are not electrically in parallel. However, once this physical design is manifested the electronics drive will again have no effect on beaming.
I do not disagree that a current drive offers promise for ESL panels, only that a current drive of and by itself does not effect beaming.
My limited reading on the output level vs frequency of an ESL panel leads me to believe that reduction in drive voltage vs frequency operates linearly over a smaller spectrum of the panel than the panel will be asked to reproduce.
A current drive may get you initially closer to a flat acoustic response, but frequency shaping networks will most likely still be needed. If the current drive and panel are inside a feedback loop and corrective networks are also necessary, stability might be quite a challenge.
The traditional constant voltage drive amplifier and its' feedback loop followed by frequency shaping networks and then the panel are probably easier to stabilize. In a more traditional design, the panel dimensions have near zero effect on the amplifier's output voltage. With a current drive, changes in panel dimensions will probably require changes in both shaping and stability networks.
I own a pair of M-L ReQuests, for a long time I thought they represented near state of the art resolution in the midrange. I have since worked with my friend to design and build a three way conventional cone driver speaker which exceeds the M-L in clarity, resolution and dynamics. There are of course other ESL manufacturers who may have made fewer concessions to a price point design, but I now feel that for a typical sized room, a full range ESL is problematic.
My limited reading on the output level vs frequency of an ESL panel leads me to believe that reduction in drive voltage vs frequency operates linearly over a smaller spectrum of the panel than the panel will be asked to reproduce.
A current drive may get you initially closer to a flat acoustic response, but frequency shaping networks will most likely still be needed. If the current drive and panel are inside a feedback loop and corrective networks are also necessary, stability might be quite a challenge.
The traditional constant voltage drive amplifier and its' feedback loop followed by frequency shaping networks and then the panel are probably easier to stabilize. In a more traditional design, the panel dimensions have near zero effect on the amplifier's output voltage. With a current drive, changes in panel dimensions will probably require changes in both shaping and stability networks.
I own a pair of M-L ReQuests, for a long time I thought they represented near state of the art resolution in the midrange. I have since worked with my friend to design and build a three way conventional cone driver speaker which exceeds the M-L in clarity, resolution and dynamics. There are of course other ESL manufacturers who may have made fewer concessions to a price point design, but I now feel that for a typical sized room, a full range ESL is problematic.
I came across this output-transformer recently and it looks promising, methinks. Also, not too pricey.
The ATH90 @ the Wüstens
http://www.die-wuestens.de/atrafo.htm
500V directly from the plates of a pair of 6550 or KT88 to the primary results in 5-7kV on the secondary side (the secondary is multi-tapped sez the germish text).
Does someone have a schematic that'd fit?
The ATH90 @ the Wüstens
http://www.die-wuestens.de/atrafo.htm
500V directly from the plates of a pair of 6550 or KT88 to the primary results in 5-7kV on the secondary side (the secondary is multi-tapped sez the germish text).
Does someone have a schematic that'd fit?
Hi,
while this solution looks very tempting on first glance it is not so on the second.
The problem lies within the fact that stepping voltages up is a different matter than stepping them down.
The impedance of the panel (usual capacity is several 100pF to 2nF) is transformed up. This means that not only the capacitive load seen by the tubes rises but the currents will do too. Especially single ended designs could quickly run into current limitations. Pushpull designs are better off in this regard.
The interwinding capacity of the tranny´s windings is no longer shunted by the very low speaker impedance but by the much higher plate impedance. This will introduce serious bandwidth limitations. SE-amps are better in this regard.
It´s not without reason that this solution is not a popular one within the ESL community.
Since there is no datasheet or useful information about this special tranny I regard the chances of success as low. Before buying I´d ask the manufacturer to proove the usabilty of his tranny with facts and measurements.
jauu
Calvin
while this solution looks very tempting on first glance it is not so on the second.
The problem lies within the fact that stepping voltages up is a different matter than stepping them down.
The impedance of the panel (usual capacity is several 100pF to 2nF) is transformed up. This means that not only the capacitive load seen by the tubes rises but the currents will do too. Especially single ended designs could quickly run into current limitations. Pushpull designs are better off in this regard.
The interwinding capacity of the tranny´s windings is no longer shunted by the very low speaker impedance but by the much higher plate impedance. This will introduce serious bandwidth limitations. SE-amps are better in this regard.
It´s not without reason that this solution is not a popular one within the ESL community.
Since there is no datasheet or useful information about this special tranny I regard the chances of success as low. Before buying I´d ask the manufacturer to proove the usabilty of his tranny with facts and measurements.
jauu
Calvin
I think a direct drive high-voltage amplifier (AKA Sanders amps) is far, far better sound than any transformer-mediated drive. Did some A-B testing (long ago) with a local audio club in attendance. However, very dangerous. See my recent post:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=135296
http://www.diyaudio.com/forums/showthread.php?s=&threadid=135296
Hi,
it all depends on the circumstances, doesn´t it?
If we look at the points within the chain where differences can occur, we find:
The kind of panel: high-or low efficiency? what frq-range? what Impedance-range?
The kind of Transformer: which transformation-factor? Which kind of core? which kind of winding? how large is the bandwidth with a certain load? how much distortion?
The kind of Amplifier: w. or wo feedback? complexity? Output devices? OP-devices bias? Distortions? damping-factor? Voltage-range? Behaviour of amp regarding ESL-loads?
So what was A-B tested here, that would allow for such a general statement??
Rodeodave faces a completely different setup problem! It could even be regarded as a ´not direct-coupled´ amp, since he asked for a special transformer coupling a rather conventional tube-amp to the panel instead of using the same conventional tube-amp coupled via two audio trannies to the panel.
Since most audio-trannies are of rather low quality to reduce stress on the driving amp and to beef up HF-performance, I´m not really surprised that even a old acoustat-x might play better. A really good transformer (e.g. Amplimo, certain c-core-types) on the other hand outperforms most direct coupled amps. It´s now hard though to find a ´capable´ amp.
jauu
Calvin
it all depends on the circumstances, doesn´t it?
If we look at the points within the chain where differences can occur, we find:
The kind of panel: high-or low efficiency? what frq-range? what Impedance-range?
The kind of Transformer: which transformation-factor? Which kind of core? which kind of winding? how large is the bandwidth with a certain load? how much distortion?
The kind of Amplifier: w. or wo feedback? complexity? Output devices? OP-devices bias? Distortions? damping-factor? Voltage-range? Behaviour of amp regarding ESL-loads?
So what was A-B tested here, that would allow for such a general statement??
Rodeodave faces a completely different setup problem! It could even be regarded as a ´not direct-coupled´ amp, since he asked for a special transformer coupling a rather conventional tube-amp to the panel instead of using the same conventional tube-amp coupled via two audio trannies to the panel.
Since most audio-trannies are of rather low quality to reduce stress on the driving amp and to beef up HF-performance, I´m not really surprised that even a old acoustat-x might play better. A really good transformer (e.g. Amplimo, certain c-core-types) on the other hand outperforms most direct coupled amps. It´s now hard though to find a ´capable´ amp.
jauu
Calvin
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.