hermanv said:
The schematic is right, I drew it from the speaker and double-checked against the schematic I got yesterday from ML.
There should be no AC on the film(and is not), the symbol to the right is a DC source. The AC from the transformer is connected in to the stators.
I am not so sure that the ground-reference of the HV needs to be connected to a centertap. Probably it is OK to connect it at whatever connection that is available on the transformer secondary. The potential-difference between the taps is, the way I see it, unimportant in relation to the HV.
About simulating this, I know there is a pspice model for the whole package including transformer somewhere on this forum. I would gladly accept a model without transformer if someone has made one up. Must be rather complicated though, as Brian has shown.
If you want to see something weird you should look at the schematic of a Model IISW Beveridge DD-amp.
If you want to model the ESL itself, in this article you'll find some math...
Anyway, the bottom line is that everything except the panel capacitance has such small influence on the panel impedance that it can be neglected. You can verify that yourself by measuring impedance curve of a panel with and without bias, it's almost impossible to measure the difference. So it's ok to assume the panel is just a capacitor.
Modelling the transformer is much more of a challenge as parameters change with signal level. If anyone has a good basis for such a model I'd be interested!
Regards,
Martin
Anyway, the bottom line is that everything except the panel capacitance has such small influence on the panel impedance that it can be neglected. You can verify that yourself by measuring impedance curve of a panel with and without bias, it's almost impossible to measure the difference. So it's ok to assume the panel is just a capacitor.
Modelling the transformer is much more of a challenge as parameters change with signal level. If anyone has a good basis for such a model I'd be interested!
Regards,
Martin
If the AC signal is connected to the return of the DC source all of the AC signal will also be present in the output of the DC source. Only the series resistor attenuates the amount of AC now applied to the film.revintage said:
I'm surprised that the built in transformer models seem in-adequate to you. I use LTSpice (free) which allows you to specify inductance, coupling ratios (leakage inductance) series and parallel resistance and winding capacitance. In the few cases I've measured there was very good correlation between predicted and actual behavior. There is a limitation, I don't think the LTSpice model allows you to model core saturation.maudio said:
My current speakers are ML ReQuests, I'm reading the thread out of general curiousity. My speakers will soon be replaced by my home designed 3 way dynamic driver system. I'm just too lazy to do the modeling I mentioned above.
maudio said:
Well, of course, it depends on what you're trying to discern from a simulation and on how accurate you want your simulation to be. Granted, the big component is that basic capacitance. A simulation with only a capacitor load will tell you only a limited story, and may mislead as it could create instabilities in the amp that might not exist in real life. But it may be good enough to get some "ballpark" answers. A look at the impedance curves of any ESL will show that the impedance is usually far more convoluted than a simple inverse-linear relationship between reactance and frequency that would result from a pure cap. Phase plots for a cap would show -90 degrees at all frequencies, while real ESLs rarely have more than a few octaves where the phase angle exceeds -75 degrees. ESLs typically show a resonant peak at the low end, and the impedance will actually go inductive below that point. At the top end, when the mass of the diaphragm approaches the air mass, the impedance will again increase and go inductive. Furthermore, in order to make any sound, the impedance must show a lossy component (ie: a resistance). A purely reactive device would reflect all power back to the amp and could not radiate energy. Also, many designs count on stray series resistances, or even intentionally added small resistances, to prevent oscillation in the amplifier.
I've got some impedance plots of various commercial ESLs. If anyone asks, I can post some of them. True, they are "colored" by step-up transformers and, in many cases, by cross-overs. In this regard they do not reflect what a pure ESL panel will show.
That's true when measuring at the speaker terminals. But this is largely dominated by the transformer.
What I meant to say is that the impedance of the ESL element itself (measured at the stators) can be safely considered to be pure capacitive.
The problem with spice models is that they assume transformer parameters to have constant values.
But (especially core) parameters vary considerably with signal level and frequency. Try measuring Lprim at some different freqs and levels to see what I mean... has to do with eddy currents in the core.
At medium high freqs the models are ok, at low freqs and at very high freqs (above resonance) they are far from accurate
There is truth to what you say, however even pure reactances will show a variation in value when measured at different frequencies. This is due to phase error compensation problems in the test equipment more than actual changes in component values (that are non-ideal to begin with). Most of the non-linear measurement is due to stray capacitance, leakage inductance and series resistance the other transformer limitations are smaller.maudio said:
A transformer with excessive eddy current loss (or hysteresis) is not particularly suitable for use in linear audio circuits. Good transformers have distortion products (non linear behavior) of less than 1%, this is largely the same as saying that it is a component whose behavior is not very frequency dependant. As to behavior at different levels, once again good tranformers easily exceed dynamic ranges of 100 dB, this also means that they are quite linear over the design signal level range.
No, you can't just wrap wire over a chunk of iron, it's nowhere near that simple. But commercial core materials designed for linear use are pretty good. This problem has had smart people working on it for nearly 100 years.
There are certainly people who claim that tube amplifiers don't have a tube sound, they have a transformer sound. I've not seen measurements to support this, but it could be true.
In addition, you are certainly correct that worrying about small details is required for good design, but given a good transformer, it is probably pretty far down the list of problems that speakers have.
Having wound a number of trannies I agree with you... Well designed transformers are indeed not that bad throughout the audio range.
The problems I ran into with simulation are at the extreme ends, far beyond normal audio spectrum (<5 Hz and 200k-1 MHz), the models are probably not sophisticated enough for this.
I work on a system with the transformer enclosed in the feedback loop, so I need accurate simulations especially at these extremes as this is where phase margins are set.
The problems I ran into with simulation are at the extreme ends, far beyond normal audio spectrum (<5 Hz and 200k-1 MHz), the models are probably not sophisticated enough for this.
I work on a system with the transformer enclosed in the feedback loop, so I need accurate simulations especially at these extremes as this is where phase margins are set.
revintage said:, how come they work as well as other MLs? And why don´t they use the existing CT on the transformer. Must be a logical explanation to this!
This is unusual, but after some thought I see that it can (and obviously does) work. The diaphragm is not supposed to be driven by any audio voltage in any typical ESL (I think that the Final brand is the exception). The large supply resistor of many mega-ohms, along with the diaphragm resistivity, form an RC low-pass filter with the capacitance (two sided) between diaphragm and stators. This low-pass occurs at a tiny fraction of a Hz. All audio is filtered away before it "hits" the diaphragm through its DC bias supply. However, over the span of several minutes the diaphragm can receive its charge from the supply through the large resistor. If we momentarily disconnected the diaphragm wiring altogether, it would still play until the charge gradually leaked off. It doesn't need to be connected to anything for audio purposes; all it needs is a charge on its surface. You could reference the HV bias supply to the CT or to either end of the step-up transformer; for DC they're all at the same potential. There is more AC swing at either end of the transformer, so a bit more audio will leak through the slow RC filter, compared to a CT reference, but if the RC is slow enough, it really won't matter where on the step-up secondary one chooses to reference the bias supply. For ML, it just must have been a matter of some convenience to have chosen one end versus the CT; perhaps they saved $0.02 on wire.
So, back to the AM modulation transformer idea that Rcavictim brought up. After considering the ML Script bias arrangement, I'll have to correct myself - I don't think you'll need a CT on the modulation transformer secondary for anchoring the drive. You could simply reference the bias supply to either end of the secondary like the ML Script. But, instead, I would suggest placing two very large (>>1MEG) same-value resistors in series across the secondary ends to create a pseudo-CT for referencing the DC bias supply. This would be a point of zero (canceled) audio and would result in less leakage of audio at low-frequencies into the diaphragm. This would be more important for a full-range ESL than for a tweeter obviously.
BTW, for those of you unfamiliar with AM plate modulation and why we're talking about it here: The traditional way to create AM modulation was to use an audio amp to modulate the B+ going to the main RF carrier amplifier. This usually meant a large tube audio power amp whose output transformer's secondary was connected in series with the B+ supply for the RF final amp. These amps were often very powerful since you needed 50% of the desired output power in the audio amp. In other words if you needed 1000 watts of AM, you'd need a 500 watt (or slightly more) audio amp to drive the B+. In fact, the old WLW 500KW transmitter had an audio amp whose power out exceeded 300KW, and this was designed in the thirties!
The only major difference between the audio amps for plate modulation and those intended for loudspeaker use is the impedance of the secondary of the output transformer (and that the modulation transformer had to be gapped to pass the B+ current). Where a regular tube amp steps down the plate resistance to 8 ohms or so, the modulation transformer typically stepped it up. The turns ratios varied depending on the AM transmitter design, of course, but many modulation transformers had a small value step-up - just what we need to connect tubes to ESL stators. Many AM modulators did not have to be "hi-fi" as regards either low distortion or wide bandwidth, but some were actually pretty good. The WLW amp created full modulation at less than 1% distortion. So it remains to be seen if Rcavictim's modulation amps and transformers will be of any use, but I think it's a worthwhile project, and I intend to poke around in my surplus equipment heap because I know I have some modulation transformers somewhere in there.
BTW, for those of you unfamiliar with AM plate modulation and why we're talking about it here: The traditional way to create AM modulation was to use an audio amp to modulate the B+ going to the main RF carrier amplifier. This usually meant a large tube audio power amp whose output transformer's secondary was connected in series with the B+ supply for the RF final amp. These amps were often very powerful since you needed 50% of the desired output power in the audio amp. In other words if you needed 1000 watts of AM, you'd need a 500 watt (or slightly more) audio amp to drive the B+. In fact, the old WLW 500KW transmitter had an audio amp whose power out exceeded 300KW, and this was designed in the thirties!
The only major difference between the audio amps for plate modulation and those intended for loudspeaker use is the impedance of the secondary of the output transformer (and that the modulation transformer had to be gapped to pass the B+ current). Where a regular tube amp steps down the plate resistance to 8 ohms or so, the modulation transformer typically stepped it up. The turns ratios varied depending on the AM transmitter design, of course, but many modulation transformers had a small value step-up - just what we need to connect tubes to ESL stators. Many AM modulators did not have to be "hi-fi" as regards either low distortion or wide bandwidth, but some were actually pretty good. The WLW amp created full modulation at less than 1% distortion. So it remains to be seen if Rcavictim's modulation amps and transformers will be of any use, but I think it's a worthwhile project, and I intend to poke around in my surplus equipment heap because I know I have some modulation transformers somewhere in there.
Prior to securing the Hermeyer 813 amp; I connected my panels directly to the plates of a low power tube amp and tied the bias return to the ground connection on the tube amp. Worked very well, not loud however. If the modulation transformer does not have a step-up ratio, then connecting ditectly to the plates may produce a better sound. May have to put a load resistor on the secondary winding...
Good luck..
Good luck..
Who knows Dave Berning amps?
http://www.freepatentsonline.com/5612646.html
http://www.davidberning.com
As far as I understood, the basic idea is that a square wave with pulse height modulation and constant frequency is transformed by a little transformer that only has to handle this single frequency, and a rectifier is used for demodulation. But the circuitry that is actually used seems to be much more complicated. Berning amps that are currently sold do down-transformation, but I ask myself if this is possible the other way round.
http://www.freepatentsonline.com/5612646.html
http://www.davidberning.com
As far as I understood, the basic idea is that a square wave with pulse height modulation and constant frequency is transformed by a little transformer that only has to handle this single frequency, and a rectifier is used for demodulation. But the circuitry that is actually used seems to be much more complicated. Berning amps that are currently sold do down-transformation, but I ask myself if this is possible the other way round.
I looked into it long ago. I did find some very high voltage tubes (low current, though) to direct drive a diaphragm. I forgot which ones. They were in the RCA Databook.
One thing to bear in mind - maybe it has been mentioned along in this thread - is that above 20-25kV these tubes start emitting X-rays.
Serge
One thing to bear in mind - maybe it has been mentioned along in this thread - is that above 20-25kV these tubes start emitting X-rays.
Serge
From looking at various schematics and reading between the lines in my Martin Logan manual, I'm guessing that commercial ESL diaphrams are being driven at about 1000 VRMS. Does anyone have real data?
Using a push-pull driver pair, power supplies of around 1.5KV would then seem adequate, no need for 20KV which is too dangerous anyway.
So a 1 meter square panel with 1cm spacing is about 3200 pF (0.5 cm to each side stator). To charge this capacitance to a 1KV triangle wave at 20 KHz requires 64 mA. So a 200watt amp seems possible. Someone might want to check my math here, I did this fast and dirty.
Martin Logan drives the stators push pull with a ~1500 Volt DC bias on the film. This probably offers significant improvement in linearity since driving only the film results in increased attraction as it swings closer to either stator and helps problems at the zero crossing voltage.
ps. Martin Logan powder coats their stators to prevent shock and lost customers.
None of this is impossible, but a fast 200watt amplifier stable into an all capacitive load is also not trivial, maybe an AC current source is the answer.
Using a push-pull driver pair, power supplies of around 1.5KV would then seem adequate, no need for 20KV which is too dangerous anyway.
So a 1 meter square panel with 1cm spacing is about 3200 pF (0.5 cm to each side stator). To charge this capacitance to a 1KV triangle wave at 20 KHz requires 64 mA. So a 200watt amp seems possible. Someone might want to check my math here, I did this fast and dirty.
Martin Logan drives the stators push pull with a ~1500 Volt DC bias on the film. This probably offers significant improvement in linearity since driving only the film results in increased attraction as it swings closer to either stator and helps problems at the zero crossing voltage.
ps. Martin Logan powder coats their stators to prevent shock and lost customers.
None of this is impossible, but a fast 200watt amplifier stable into an all capacitive load is also not trivial, maybe an AC current source is the answer.
i think this 1cm should be 1mm.hermanv said:
But the power we all calculate doesn't seem to be right. My first esl was driven with a simple akai stereo set with an amplifier that could only deliver a 25Watts peak!!
If you calculate a little further you'll see that most of the energy will be dissipated by the step-up windings with their capacities.
In my opinion we forgetting the resistor inside the secondary windings.
I just measured this on my quad trafo's:
- 1800 for mid and high, add another 2500 Ohms for the low frequencies.
A panel with 3n2 has an impedance of 2486 Ohms(@20khz), include the 1800 Ohms trafo impedance, and a 1KV(pp) sinus requires a current of 0.165 Amp. That makes a 116 Watts.
But 3n2 is worse case scenario. Often panels are 1 or 2n.
I don't believe esl panels are inefficient, i think the step-ups are. I think 50Watts should probably be enough.
Hi,
up to a stator-stator-distance of 2mm amps with a plate voltage of as low as 1kV are able to drive the panels to quite high SPLs. Such a s-s-distance is enough for hybrid-ESLs with panels playing from ~250-300Hz on. Then its just a matter of the current capability of the amp to reach sufficient bandwidth. Even working in class A the heat and power is manageable. There are several circuit schems around for such amps, so You don´t need to reinvent the wheel.
The problems really come up when the panel shall reach down further in frequency and when the s-s-distances must therefore rise. The power needs rise quadratically with increasing s-s-distance because the driving- and polarizing voltages increase.
jau
Calvin
up to a stator-stator-distance of 2mm amps with a plate voltage of as low as 1kV are able to drive the panels to quite high SPLs. Such a s-s-distance is enough for hybrid-ESLs with panels playing from ~250-300Hz on. Then its just a matter of the current capability of the amp to reach sufficient bandwidth. Even working in class A the heat and power is manageable. There are several circuit schems around for such amps, so You don´t need to reinvent the wheel.
The problems really come up when the panel shall reach down further in frequency and when the s-s-distances must therefore rise. The power needs rise quadratically with increasing s-s-distance because the driving- and polarizing voltages increase.
jau
Calvin
The equation for sine wave current in an inductor is A = V/wL, where w equals 2 * p * F.
So it stands to reason that the current in a capacitor would be A = VwC.
If we use 1n5 for panel capacitance and pick a more reasonable frequency, say 2 KHz; to get 1KV requires 19 mA RMS or 19 Watts into the load. If we use a class A amplifier we would need about a 60 Watt amplifier per channel.
The problem is that I have no idea if 1 KV is the correct voltage and that typical amplifier ratings are full spectrum, which means 190 watts for 20 KHz into the same load and about 3 times that for amplifier power (30% efficient ??) I am cheerfully assuming that the panel is large enough so there is little gain in panel efficiency because the radiating area is increasing with respect to frequency.
So the big question is, what is the SPL rating per kilovolt of sine wave signal? And where is the efficiency turnover between panel size and source frequency?
I'll repeat myself; I think it can be done, but that it is not a trivial design problem. I'm assuming that the builder would like performance more or less similar to commercial home audio. That is to say at least 100 dB SPL probably more like 105 dB for the peaks.
So it stands to reason that the current in a capacitor would be A = VwC.
If we use 1n5 for panel capacitance and pick a more reasonable frequency, say 2 KHz; to get 1KV requires 19 mA RMS or 19 Watts into the load. If we use a class A amplifier we would need about a 60 Watt amplifier per channel.
The problem is that I have no idea if 1 KV is the correct voltage and that typical amplifier ratings are full spectrum, which means 190 watts for 20 KHz into the same load and about 3 times that for amplifier power (30% efficient ??) I am cheerfully assuming that the panel is large enough so there is little gain in panel efficiency because the radiating area is increasing with respect to frequency.
So the big question is, what is the SPL rating per kilovolt of sine wave signal? And where is the efficiency turnover between panel size and source frequency?
I'll repeat myself; I think it can be done, but that it is not a trivial design problem. I'm assuming that the builder would like performance more or less similar to commercial home audio. That is to say at least 100 dB SPL probably more like 105 dB for the peaks.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.