Any Direct Drive ESL Amp projects someone could share?

There have been a few Direct Drive ESL Amp forays on this forum, but no completed projects posted. I am not interested in headphone ESL Amps.

Has anyone completed Neil S. McKean's ESL Amplifier?

If so, what power supply was used? Or where did you source one?

Were there any boards made for it?

Anyone update it?

Any other designs?

I understand that I am the beggar here, but if someone has made one, or another one, your experience posted here would be very helpful. If this should be in the amplifier section I am happy to move there.
Last edited:
I am absolutely sure it provides the best sound you can have. I built a Sanders-like amp maybe 25 years ago and used it very happily for maybe 15 yrs with hardly a hiccough all that time (all rough memories). Drove both Dayton-Wright XG-10s and DIY DW panels with XG8 cells. Sold it when downsizing.

Helps a lot to have a big, dusty, surplus electronics souk in your town.

Dangerous... part of the thrill, eh.

2400v B+ wired direct to panels (and negative voltage bias, of course), inductance filtered and big oil capacitors. 8068 tubes???. Ran a large bank of resistors as the effective load so the panels were hardly noticed by the amp circuit.

Keep the thread here.

Last edited:
regarding the "danger." I was planning on marrying the amplifier (mono channels) into my Hybrid ESL design in order to prevent some of that danger. Of course that won't prevent any problems during the build (and test) of the amp itself, but at least the hookup from the pre-amp to the amp should be safe (right?).

Are you familiar with McKean's amp?

For as widely a distributed design as it is, I am surprised there hasn't been project to follow on it. I'd put up some money for someone to seriously detail (and maybe improve it some) a build.

I assume most have been put off of such a project the moment they figured how much wattage be needed to drive a panel properly.
Also the amplifier would rather be a a specific design for a specific panel.
For non-segmented panels the voltage requirements may be lower while the current requirements are higher than for segmented panels.
If You wanted to design for a full power bandwidth of only 20kHz, a amp for a non-segmented (hybrid) panel would already waste more than 1kW of power.
Reducing the full power bandwidth requirement also reduces the wattage requirement, but shares the risk of current clipping ... the risk of not beeing able to drive the panel under all circumstances.
Segmented panels though requiring more voltage are easier to deal with.
The amp would probabely rather run into voltage clipping, but the lower current requirement and the panel's lower impedance variation let a amp project look more doable.
Still though such an amp would be ridicolously inefficient.

Last edited:
I looked for solid state outputs but none available at the time.

With tubes with plate caps on top, the high voltage wires are patently obvious and floating in the air. Also, I suspect tubes have a more benign failure mode than a solid-state device and fairly robust.

Not a lot of extra fuss using tubes (if you know where to buy tube sockets), just heater power and a few little things.

Maybe those are not strong enough reasons to go to old fashioned 8068 tubes, but just saying'.

I am able to understand the value of those FETs, but designing an amp is really beyond my ability.

I noticed that even Nelson Pass has mused on such an idea in the past. I don't know if a donation toward a build could entice him (or someone else).

Even if I never build such an amp (though I really want to), I absolutely love learning about audio. I have enjoyed reading all the content here. It has led to a line array, and two small projects for my teens. An ESL build is in the planning now.

Thanks for the comments. I will just keep watching for a build.
I'm still in the "considering different approaches" phase of an ESL/Amp design. I've noticed that there are two camps, the most prolific topology uses a standard audio amp driving a 1:50 to 1:100 step up transformer. And then there is the pure direct drive that uses tubes.

Both of those approaches have serious design challenges. I'm concluding that something in between might be the best choice. I think it is possible to design a class d amp using 1200 volt silicon carbide FETs. They are very fast with a low gate charge. Running them at around 900v should provide enough margin. This means that one only needs a 1:3 to 1:6 transformer ratio to reach the same ESL drive levels.

A lower ratio gets the amp "closer" to the ESL load and should facilitate tighter control of the membrane. I'm no transformer design guru, but I'm pretty sure that a 1:6 transformer is a lot easier to design than a 1:100 with equivalent power-bandwidth. Of course, the class d has to be done right with feedback, but D's can be very low noise, low distortion, highly efficient, and they don't mind driving a complex load.

I really hope I follow through on this project and not lose interest once I figure it out. I'll be sure to post some details, but it will be a while. I haven't searched this forum for a similar configuration, so it wouldn't surprise me a bit if someone has already done it.
This means that one only needs a 1:3 to 1:6 transformer ratio to reach the same ESL drive levels.
Around 1970, Mike Wright designed a system with a tube amp with that kind of logic. It drove a large wood-frame speaker, clearly a forerunner of his later designs.

I'm sure a lot of the people who heard is at an audio show in Toronto were as smitten as i was the moment I heard it.

I might be the happy compromise design but others who know much more about transformers are needed to comment.

If DIY builders are using pretty creative toroidal power transformers as 1:100 step-ups, what kind of easily available transformer would be recruited for this application?

Alexburg is exacty correct, I low ratio high voltage transformer is much more costly to construct than the current DIY methods.

Such a transformer will require many times more turns on the primary itself and the secondary turns times the ratio more.
This would require a very large core in order to keep the transformers self capacitance at a reasonable level (Maybe if that!!).

I have been pushing the idea of a direct drive amp for many many years and it wasn't until about two years ago I had finally got some FET's that would be suitable.
But I need to come up with a suitable supply for the amp.

Funds were lacking for me for my DIY ESL projects, but that will have been solved this year for me.....Finally!!
Thus the reason I have been slow at getting onto this phase of my designs.

In 2008 I had started palying with FET designs that I had started drawing up in circuitmaker since the late 90's and had proven the concept to work.

I built my First ESL in 2003 but it was not in working order when I made the amp, I didn't get my ESL's working again until 2010.

Since then I had been testing my panels and pushing them to find out exactly how much voltage they could take and find out exactly how much I needed to make them work to there limits, and, how much was enough and/or too much, to find out what range to use and keep them realible.

Of across this was the demise of my fine little panels and I started building a new model to suit my findings, That panel has been made but I still haven't put a diaphragm on it yet and test it, Sadly it has already been three years since I made it.

You can find pictures of my new build here,

Because I make smaller Desktop sized panels they require at least twice the voltage than larger panels require to maintain a high efficiency.
I use and have found that up to +8Kv of bias was good. although have used as much as 12Kv but this was the limit of my original design and stator coating breakdown became a major issue at that level until thier final demise.

My new stators have no problem holding back my test voltage of 13.8kv with the probe touching the stator.

Running such a high Bias voltage increases the efficiency of an ESL, But from what I understand you are still limited in your peak SPL due to the laws of physic's and the dielectric breakdown of the air in the gap, nevertheless at peak levels it is extremely loud even for my smaller sized panels.
I had no problems getting to +105db at 1 meter under reliable conditions.

I was able to achive an efficiency of 90db +/-1db with just a 4v peak signal into my step up transformers (1:128), this is equivelent to a 2.83Vrms signal or one watt.

But ESL's are voltage devices so the actuall impedance and power level is depicted by the frequency and the reactive components in the system.

I had a ruler flat frequency response and I still have the calibrated graphs to show.

Now on to the basic requirements,
Most ESL's may never need more than 4Kv peak across the stators under louder than normal conditions, Double this and they will be 6db louder.

Under more than normal listening conditions for my setup anything under 2Kv peak on the stators was more than enough, in fact 500-600Vpeak was all that was required to reach 90db!

Although I did have about 10-13Kv peak on the stators for about 30minutes straight many times before, They Finally broke down (2010) for the first time, I didn't have a working SPL meter at the time but I assure you that it was way too loud for even me and I am a seasoned rock guitarist and I do like it loud!!!

Therefore listening to my little panels at a comfortable 85db was a pleasure and Oooooh Soooo Very Clean this required no more than 500v peak on an average with an ocassional 1000v peak for transients.

By the way I monitored all of the voltages using my O-Scope in realtime.

A suitlable amp can be made with just a 1Kv supply as this would give you 2Kv peak on the stators.
My goal is to get at least 4Kv peak so I would have to start with a 2Kv or so power supply and then someday I will go even bigger for that full out performance.

Such as using those 4500V FET's, such an amp would be the equivelent of driving a transformer with a very large 200-300watt amplifier in order to achive my ultimate goal for when I do like it loud for 16Kv p-p or so.

My amp design is no different that any other Class A P-P design, I have already posted my designs in several of the threads already.
They can be found here,

The main thing I had to learn was how to stack the FET's in order to use a higher voltage and this was challanging but quite easy to do.
My model amp used a power resistor for the current source just like the sanders tube amp and my supply voltage was only 200v.

But I was using IRF510's (from RadioShack) and I had to stack 4 or 5 of those in order to be able to handle the supply voltage, But...It worked now all I needed was my ESL's working again as I didn't have a Bias supply anymore.

I had a few IRF740's and 820's as well.
I tried every type of FET I had on hand including some IRFZ types and they all worked once the bias was properly adjusted, all I had to do was swap them out and adjust them.

So, I set out to build a very nice regulated variable HV Bias supply and that can be found here,

That post is for the schematic but the whole build is in the thread.

Anyhow as I was saying I finally got some IRFGB30's (1000v) and various other FET's from STM ranging from 900v to 1500v, I have a 900v supply that I have to repair that is suitable for a First step in this type of design.

I also have a couple of 300-500watt tube linears I could steal the supplies out of since the tubes are shot, But I would rather get new tubes for them oneday.

I got lucky the other day and found Two microwave ovens of equivelent size in a dumpster so hopefully there will be two good MOV transformers I can use instead for a large amplifier design.

Just let me say one thing,
I have been playing with Very High voltages all of my whole life and many times I have been lucky to even be here typing this post, But HV is NO JOKE and Must Be RESPECTED!!!

But I believe that even a modest Direct Drive amp can be built very easily and enjoyed and not cost anywhere what a set of tubes would cost these days, not to mention the power supply for such a beast.

Ohh Ya, That first amp...I built two circuits and ran them P-P and used it to drive a little Piezeo element, I thought the element was going to explode!!
But I did run sine/square waves and music through and it was the cleanest time I had ever heard a piezeo element sing, I knew then that I was finally on to something.

It is just a hobby for me and I have so much going on at once that it is hard to get things done sometimes, let alone burning myself out and setting it on the back burner every once in a while.

But a D-D ESL Amp is very close to the top of my things to do next.
As I have mentioned other times I have been working with Microcontrollers and IC's lately to build a preamp for my Desktop ESL system so I don't exactly want a bunch of HV flying around on my desk right now !!! He,he,he,he ;)


jer :D

P.S. here is a link to some more links of the threads I mentioned on such design we are discussing here,

Last edited:
Using 500W transistors it's reasonable to assume 250W of dissipation per device thus yielding 300mA at 6kV p-p at 3kV DC bus.
Anything that produce more than 10..20 mA is extremely dangerous. Although AC current is not better or worse than DC current: remember Tesla (Westinghouse) vs Edison rumble. Amount of charge even with conservatively small DC bypass capacitor is enough to be used in defibrillator: read lethal. Caps are prone to fail short circuit: you have to use really reliable capacitor in series with the amp of yours.

SRPP is usually used: no need to drive high side tube/FET. The latter can be be done though. Nevertheless working with high DC voltage requires high quality workmanship including materials, layout and construction.
High potential will collect dust as crazy vacuum cleaner so it is not that easy to remove heat.
Good luck
Good Point on the output caps!

Large HV caps can be found sometimes very costly and sometimes reasonable surplus types.

In order to have a large enough value inorder to lessin the amount of voltage drop across them I was thinking about a DIY cap made out of a stack of glass plates or some polycarbonate sheets or something, which ever material that will make the least amount of audible noise.

Reminds me of years ago when I had stacks glass plate for my tesla coil experiments....I wish I still had them. ;)

SSRP setup would require the use of Depletion mode FET's, I am mainly looking at the Enhancement types due to cost and higher voltage ratings.

If you are building a segemented ESL design then your Bias current ratings won't have to be as high they would be for a large single panel.

In my case for my little panels of 35pf to 100pf the FET's will be dissipating at least half of the produced power with their 80pf Coss.

This won't be so bad of a ratio for a larger panel that has much more capcitance.

Hoping to assume that by stacking two FET's in series this will actually reduce the Coss in half to aprox. 40pf and lower if I added more.

jer :)
Last edited:
Seems inescapable, that unless you are just aiming for ESL upper tweeters, good quality "direct" drive means direct B+ and risk.

Mistakes aside, under ordinary circumstances, you just turn off the amp when you are anywhere near it. But when testing occasionally with power on, you ought to understand how to keep out of trouble.

For a dozen years or so, I ran my DD in a rack high above the floor in two different house basements with wires coming up through the floor to the speakers. The speakers being DW and the cable being their usual mil-spec quality, no real danger to the family. from any conceivable exposure. But then I've never had a cat.

Of crouse when using such voltages all the caps should have a proper discharge path especially the output caps if they are used.

In my own model I probably won't use them at all. but if I were to present a commercial moel or something intended to be safe for an Advanced DIY'er I would present them.

I know that for my own safety and setup, I know how to deal with such issues, for commercial and DIY I have spent many sleepless nights in trying to figure out such safeguards for others.

I have not found in my very First low voltage set up that I was limited to just higher frequencys, my circuit worked from DC too well up into the +150khz range, in fact that is what I set my bandwidth goals to.

It may be overkill but this would assure a very good transient response at the higher end and square wave that would be square for purity sake whether or not you may or may not be able to hear the difference.

jer :)
Last edited:
Thanks for the feedback. It sure seemed intuitive that a lower ratio transformer would be easier to build, but like I said, it's not an area that I have design experience in. I'm new to this forum and am impressed by the technical prowess that a large number of it members possess.

I have a gazillion questions regarding ESLs and there drive requirements, but I still have some due diligence to exercise before diving in to any of them.
It's not that hard to bias enhancement mode FET. Ladder structure basically drives the next step by the previous one so single HV device is better. Coss and Crss are defined at 25 VDC. Consequently at 1500 VDC (midpoint) you'll get about sq root 1500/25 - 8 times smaller. Cascode driven stage is pretty good current source, although voltage feedback is feasible but capacitive impedance of load will require compensation tuning.
One more thing that wanted to point out about having HVdc on the outputs,
Is that if you use a +/- DC supply you can still very well set the same circuit up to swing +/- and have a 0Vdc output with no signal.
No need for complimentary devices.

Tricky, But it can be done! ;)

jer :)
Last edited: