ESL Direct Drive SS amp

Folks:

I've touched on this a few times in some other threads.

I'm interested in building some (curved) electrostatic speakers for a home theater setup. I'd like to do four but two is definite - I can arc the front two enough so that the listening couch will be in the primary arc of the speaker output. This thread isn't about that :)

The speakers, according to Wagner, would be 8Kv. Even with a 63v rails, (200watt roughly) with a 75:1 step up transformer, that's only 4725v!

I was planning on six Opti-MOS amplifiers for a home theater and I've gotten the following information from Randy Sloan:

I'm not aware of any types of high-voltage L-MOSFETs available, and going to D-MOS families would probably destroy the fidelity you are trying to achieve with electrostatic speakers. Some of the "double-die" L-MOSFETs available have Vds parameters of 200-volts, and by going with cascoded outputs (commonly called "totem poleing"), you can raise the maximum rail potentials to 400-volts. This technique would raise the maximum peak-to-peak capabilities with L-MOSFETs (not bipolars) to 800-volts. Of course, this is still a far cry from 8-Kv.

(Randy Sloan)

This is interesting though:
I still believe that L-MOSFETs are the best output devices for the application you are considering, since they are much more immune to reactive loading than bipolar devices. (Randy Sloan)

I've been thinking of a Class D or T amp, as I may be able to find a switching transistor that's capable of higher voltages. From email generated from the Class-T thread:


Driving an 8kV panel directly will require some very VERY exotic output devices. So exotic, in fact, that I haven't found any either. There are a few good candidates in the 1kV range from ST, however, which would allow for a fairly modest turns ratio from your output transformer. Take a look at the STU7NB100 as a possibility. It's Rdson is 1.5 Ohms, so efficiency goes out the window, but I'm guessing you're not all that concerned with efficiency when contemplating an 8kV power supply anyway.

I really think an output transformer is the way to go. The money you will spend on the transformers is offset somewhat by being able to use a less expensive power supply and less expensive (and far more robust) output devices.

Given that it's seemingly difficult to find high voltage output devices (yes, I know, tubes, that's another thread), I'm thinking now of just building the Opti-MOS and go for a high voltage rail, with subsequently lower power.

Rather surprisingly, the amp that this will replace, an SA-500 from Techniques (yes, ancient) was listed as a 55w amp, but I checked the voltage across the filter and it's high (something like 45 volts, 87 or so rail to rail). So the rail voltage is a lot higher than a 55w amp would normally be, so I'm going to size my replacement Opti-MOS amp for at least that.

I'm eventually going to bi-amp a woofer in the same cabinet as the ESL, so I'm going to build six Opti-MOS's now 'regular'. I'm trying to pick the output voltage of the transformer, I may just pick 90VCT to give a high drive voltage. My feeling is that if I custom wind a 16ohm output transformer for the ESL, I wouldn't be in that high of a wattage but would get the drive voltage I need. The 75:1 output transformers with a 90VCT don't get close to 8Kv anyway, which would explain why some people say ESL's require 'powerfull amps'.

It just doesn't look like a solid state amp can do 8Kv and still be 'hifi'.

Thoughts?
 

Nelson Pass

The one and only
Paid Member
2001-03-29 12:38 am
Just because you bias the panels at 8 KV doesn't mean
you want to drive them that high.

You can cascode gain devices with high voltage Mosfets
or tubes and get enough swing, say 2KV p-p.

Your other alternative is to put some feedback around the
step up transformer. This worked miracles with my old
Dayton-Wrights.
 
Hiya,

That is one thing that has interested me over the years, building a direct drive amp for my quad electrostats.

I seen a circuit in a audio amatuer book. The amp however, was a monster!

I read awhile back in a hi-fi mag that a company called EAR had made an amp for some quad 63's
and the sound quality was stunning.

Does anyone know anything about this product/circuit or know of any DD amps for quad's ?

Would be interested.
 

MRehorst

Member
2002-05-17 8:48 pm
Since you're going to build the ESLs anyway, you can dramatically reduce their drive requirements by making the insulators thinner.

Making the insulators thinner reduces max spl out, unless you raise the crossover point, and results in lowering the drive voltage requirement.

You're going to want to use bass drivers anyway (the ESLs will sound a bit too 'thin', otherwise, especially for movies which tend to emphasize very low bass for special effects), so by raising the crossover point you can maintain the max spl and ease the drive requirements at the same time.

I'm afraid I can't give you any numbers on max spl versus drive voltage- that's where you have to experiment. It is also a function of the size you make the panels. Also, curving the panels has very little effect on high frequency dispersion. Curving large panels is primarily a means of stiffening the structure.

I've been looking at Sloan's OptiMOS amp myself recently. I like the design and philosophy behind it. I especially like the protection for the amp and for the speakers, something lacking in a lot of other so-called "high-end" designs. Let us know how they work out for you.

MR
 
Mark:

I was planning on the 'Compact no compromise TL' speaker as detailed in Wagner, but I don't have the book with me (Ok, I'm at work :rolleyes: ). I completly agree, OTTOMH I believe the cross over between the Transmission line woofer and the ESL was 500Hz. I think that even with that, he still spec'd an 8Kv drive voltage, which is why I mentioned 8Kv in the thread, but I'll double check. Then again, this was a 'no compromise' design.

What I'll probably do is do a regular Opti-MOS for now, more for bi-amping existing 8 ohm speakers, then wire up a special version with say 100 or 200v rails, and wind my own transformers for the exact voltage I need.

I'll let you know how it works, I'm ordering all the parts this week. I have a couple of old Technics 3 way's with 8" woofers that I'll use once I get the Opti-MOS wired up, just to test.

It's interesting what you said about curving the panels - as I'm crossing over at 500hz, I wonder if a flat panel would work as well? I can put a beam deflector behind the speakers to bounce the sound to my wife :D but would prefer to make it wider dispersion if possible. I don't care if it's curved or not, just so the entire couch can hear the sound. This is in a large room (24x24 or so, cathedral ceiling).

For the center speaker, I was thinking of using a flat panel, as if I turn it sideways under the projector screen (12' wide), it'll hit the entire couch anyway, as long as you are sitting down. So I could build a flat panel first to test. I can build the curved stretching rig but not looking forward to that.
 
John,
I built a pair of ESLs and used a step-up transformer, so that I didn't have to worry about the cursed direct-drive. But, in order to understand it, I measured the voltage of my stators while they were playing. I'm not sure if I did this correctly but, here are the results.
Volume: moderate listening level
Bias: <2KV (any more and I would get arcing. I messed up in construction)
Spacing: 3/32 + a little for glue.
AC Voltage Reading on Fluke 23 AFTER 50:1 step-up transformer: ~40 Volts. That's right: 40 Volts. The voltage would fluctuate with music volume, as expected. Now, I don't know what frequency the meter thought it was reading. Perhaps the meter couldn't accurately measure higher frequencies, but it still shocked me.
It doesn't seem logical to me for your stators to swing 8 kilovolts. Aim for something much more reasonable and put as much bias as you can on your membrane. Good Luck.
-Dan
 
Roger Sanders suggests in the final chapter of his book (which you really should read if you haven't) that the best way to go could be an amplifier that puts out around 500v p-p combined with a transformer with a smaller than usual turns ratio say 10:1 compared to 50:1 or even 100:1. As he says we have the technology avaliable all that's required is a designer to sit down and do it.

That advantage is the a 10:1 transformer will perform much better (frequency response) than a 50:1 also this will greatly reduce the capacitance seen by the amplifier. If you enclose it in a feedback loop as Mr Pass suggested you could have a real winner.
 

MRehorst

Member
2002-05-17 8:48 pm
reading HV audio swing

jgwinner said:
Dan:

Is the Fluke a true RMS meter? If not, I'd guess that the non-sine waveform would mess up the reading, but I'm with you, 40 volts is 'shocking' (and I'll excuse the pun ;) although it was pretty funny).

I'll research the 8Kv tonight - it came righout out of Wagner's book.

The only way to get a meter to give you a correct answer (pk-pk swing) is to feed the amp a relatively low frequency, full scale, sine wave from a test CD (must be high enough F to go through the crossover, if there is one, and low enough F for the meter to get a good reading) then measure the voltage using a high voltage probe on the meter. Convert rms reading on the meter to pk-pk by multiplying by 2 pi. If you use a complex music signal, you won't know how to convert from rms to pk-pk. The other way to go is make up a HV probe for an oscilloscope and directly observe the pk-pk swing.

I believe 8kV is quite a bit higher than necessary. I think about a 3 kV swing will be adequate to drive you out of the room if the speaker is made with insulators about 1/16th inch thick.

I think the idea of using an higher voltage output amp (say a couple hundred vpp with limited current capability) and a low turns ratio transformer would be the best way to go. Low turns ratio will give really wide bandwidth. The problem here is you gotta make the transformer yourself (maybe tube interstage xfmr would be good?).

MR
 
I said Wagner before, but I was really quoting out of Sander's book (Whoops! :eek: ).

I double checked, I was pretty close it's a 7Kv voltage I'd be needed to shoot for; I'm building the 'compact no compromise Hybrid T/L' on page 173 of the Roger R. Sanders book "The electrostatic loudspeaker design cookbook'. This is for a 70m/s spacing. This design is what I'm looking for so I don't see a reason to deviate from the plans much. The 8Kv came from page 57.

The note you make Dave is basically why I was trying to see if I can up the voltage. Note 7Kv is peak voltage.

Sanders notes that a 250w amp with a rail voltage of +/-80 (160v swing) would be around 7Kv with a 1:50 step up transformer.

As pointed out in my MOSFET parts thread, Exicon has a few MOSFET's that look pretty good, the ECX10N20/ECX10P20 is a 200v, 8a, 125W version. If I have a typical output stage, I could push it as high as say 180v per rail or 360v I could get by with a 20:1 output transformer.

I'll post about the spec's of this cross on the other thread

This is assuming the total power would stay under the output parts, typically 125W. I *think* if the output transformer was 16ohms, this would work.

Thoughts?
 
Who wants a transformer for a electrostatic anyway?

I feel that amplifier distortion does not matter - as long as it's below 0.1% at full power. 0.01% is fairly straightforward to design for, so we are as well design for it.

Amplifiers were good enough in the late 50's (the williamson valve amp with NFB), distortion wise, relative to the distortion of speakers, even now adays!

I am building a 60's class a amplifier at the moment, but don't say that modern pop music is better than the beatles!

People like Douglas Self try to design the perfect amp, but it is several orders of magnitude lower than _any_ acoustic loudspeakers distortion (especially moving coil). He criticises subjectivesm regularly, but engages in it :)

I prefer looking at the overall picture. Speakers and room acoustics are much more important.

I do feel however, that acoustic distortion _definately_ does matter. This will require an SPL meter for measuring room response etc.

Very few good quality mics, for example, can get an accuracy of 0.01% distortion, let alone the much bigger mass in the voice coil of a relatively big moving coil midrange/treble unit.

Here, the electrostat wipes the slate, or has the potential too. Certainly in the midrange and treble.

I prefer technically superior solutions (especially in speakers) rather than subjectivism.

We do need somebody sharp in amplifier design, to knock up a direct drive amp. :) And then tweak it!.

Reproducing the square wave and acoustic transient response would be a good test.

The Ongaku or Gaku-On, (audionote's silver wired se class A triode amps) where great amps, but it could not drive the midrange of a speaker that could milk it!. (unless going thru' two transformers, or even one special.)

Take Care and god bless,

Kevin

PS. Had a small goblet or two of red wine tonight :)
 
I've been checking transformers - Plitron makes a 60v (120VCT) transformer, and a 175v (350VCT). The 60v gives me a 80ish rail, the 175 gives me a 240ish rail. Either much smaller than or much higher than the 200v Exicon max. :(

I thought about running the power transformer 'backwards' but that still doesn't give me what I'd need, around 150-180v rails (assuming 10% margin on the 200v output devices).

I guess I could 'Totem Pole' the outputs.

I'll look for other transformers.

(For the mean while, I'll probably build with 60v rails, and use a 50:1 transformer - when going to bi-amp this would be the subwoofer channel).
 
Power transformers turned around backwards are typically
horrible for this application because of severely limited
frequency response.

Here is a link to one of the direct drive amps.

http://www.audiocircuit.com/9041-esl-circuit/9041IMAI-CO.htm

Yes its a tube hybrid.

Lots of changes can be made to improve the 1979 design
however output tubes are still the way to go.

I have a version that replaces the tubes with p and n channel
fets similar to the koss electrostatic headphone circuit.
Requires a total of 10 p channel and 10 n channel to get
to a 5kv power supply. Frequency response starts to fall
off at about 10khz.