by Andrew Radford
I first thought electrostatic headphones were a joke. It didn’t sound safe to strap high voltage transducers to your head. But after my work with electrostatic loudspeakers, I had the skill and knowledge to try building a pair and listening to them.
Basically an electrostatic headphone works just like an electrostatic loudspeaker, of course on a smaller scale. It modulates a diaphragm using electrostatic force, rather than electromagnetic force as in conventional speakers. They are usually a charged sheet of plastic film suspended between two conductive sheets called stators. The stators are ideally acoustically transparent but in reality are perforated metal.
Rather than build a pair from scratch, I decided to modify an existing pair of dynamic headphones to be electrostatic. A look around the house yielded the perfect candidates – an OLD pair of headphones, and they are of the 1970’s bulky construction, a simple circular design, like half a ‘barrel’ for each ear. Not pretty to look at, but a lot easier to retrofit with electrostatic panels than the modern curvy compact ones.
The transducers can be made any size and shape, just as long as they are flat! There needs to be at least 10mm of spacer so the glue has a nice bit of area to stick to. I would think that the minimum diaphragm size would be a circle of about 50mm in diameter. Less than that and the bass would suffer.
Unbelievably, the first prototype I built needed nothing more than mylar to get working, all other parts were adapted from the headphones. The headphones already had thin perforated metal grills crimped around the 95mm circular driver. There were also circles of card 0.5mm thick. So I used these parts along with mylar to make a monophonic earphone! Once I got that working I set to work making a proper stereo pair.
Here is an exploded view of what the panels are like; please note it is not to scale, the holes are actually only 1.5mm in diameter.
The stators I used were 1mm stainless steel perforated metal. The holes were about 1.5mm thick and about 30% open area. I found it in the junk pile in the garage. There is lots of scope for change here, any perforated metal can be used, so long as it has holes less than 2mm, and at least 25% open area. It can also be a lot thinner than the 1mm thick steel I used.
The most difficult aspect was cutting it into a 95mm circle, I used tin snips, which warped the metal, and I could not get it perfectly flat again. There is not much room for dishing when the gap between the stator and the diaphragm is only 0.5mm. Thinner metal that could be cut with ordinary scissors would be better.
Spacers need to be good insulators so I used plastic. The trick is finding plastic of the right thickness (0.4 -1mm) that also takes glue. I eventually found a cover of a ring binder, it was a kind of textured plastic which holds epoxy better than it otherwise would. It is still marginal. Polycarbonate (lexan) or acrylic (perspex) are also good. Even cardboard worked in my first prototype, although I did soak it in polyester resin diluted with acetone.
This needs to be thin film. The best is mylar. It is hard to find, and can usually only be got in rolls of a couple of kms. I can also be bought from the ESL Info Exchange, linked to in the links section. Try asking a Dupont distributor, they may sell you a single roll.
The best way I found to build these ES panels is similar to Roger Sander’s technique described in “The Electrostatic Loudspeaker Design Cookbook” which I recommend you reading if you are serious about making a useful and practical ES loudspeaker or headphones.
The metal stators must be cut to the correct shape to fit the headphone cans. In this case, a circle. There are many ways to cut metal. The first time, I used heavy duty tin snips. These work quickly, but unfortunately they bend and war the metal, which is unacceptable when the spacing between the stator and the diaphragm is only half a millimeter. Jigsaws and nibbling tools are better.
The stators should be painted. Unlike larger Electrostatic speakers, which utilise several thousand volts, the 500V can be insulated by a good coat of enamel paint, so it is a good idea. There must be a small area that is unpainted however, to provide an electrical contact. I used a piece of brass shim. You can see on the last photo above the area of bare metal that I sanded back. (The dark gunk around the edge of the metal in that photo is epoxy glue from a previous rebuild)
The four stator/spacer assemblies should then be assembled using epoxy.
It is best to use a thin film of glue rather than uneven globs. Also, it is very important that the material you use for the spacers actually sticks to the glue. Even scuffed with sandpaper, the spacers I have are marginal in that regard and as a consequence the stator/spacer assemblies are quite fragile. This can sometimes work to an advantage, because if you make a mistake it is easy to rip it apart, scrape the dried epoxy off and start again.
When the stators are glued, cover them with plastic, and lay a flat heavy object on top of them (like the heavy pane of glass) so that they set straight.
The diaphragm must be under tension to work properly in the ES headphones. So the mylar must be stretched. This is probably the hardest part of the whole construction. First, a sheet of clean glass is laid down on something white, like paper. Then the mylar is laid down on the glass and stretched. I used a stretching jig made up with a wood frame and screws, but you might get away with just flattening out the mylar with tape and tensioning it up later by heatshrinking it with a heat gun. The stretching jig I used was detailed on Sheldon Stokes site, SDS labs. This jig was originally intended for stretching mylar for the refurbishing QUAD 57 Electrostatic loudspeakers. I used half the jig.
The mylar must be stretched not so it is drum tight, but just free of wrinkles. It must also be held taut unto the glue has reached full strength and will hold it on its own. For the 5-minute epoxy glue that I used, this was a few hours to be safe.
The jig is shown as a mess of tape and wood. It actually has a number of screws witch pull the mylar taut. Once the wrinkles are just out, it only takes another half turn to get it tensioned. That’s only a few millimeters. Mylar has incredible tensile strength for its thickness.
Once the mylar is tensioned, clean it with solvent (acetone, IPA etc) and lay it on a clean flat surface (glass again). Then the graphite must be applied. The graphite makes the mylar conduct electricity just enough to hold a charge, i.e. several mega ohms per inch. Consequently, only a small amount has to be ground into the surface. You can see by the picture the tiny pile of graphite powder, that is plenty. The powder should be rubbed in hard until the mylar has a grey tinge to it, then the excess must be wiped away. Any loose graphite means clicks pops and hisses in your earphones.
In order to get electrical charge to the diaphragm, there must be another brass shim contact. I cut mine in a ‘T’ shape in order to get more contact area, while still maintaining a almost complete bead of glue for the mylar to stick to, which, BTW, it doesn’t stick to very well. Mylar doesn’t seem to stick to any glue well! You could also use copper tape or maybe aluminum foil – it would be fragile. I have also had success with just leaving a small tag of the diaphragm hanging out of the assembly and connecting to it by pressing a brass shim contact against the graphite treated side.
Lay the diaphragm contact on the film of glue, and don’t get any glue on it, so that it will make a good contact with the graphite side of the mylar. Next, lay the stator/spacer that DOESN’T have the diaphragm contact under the mylar, and gently press down on the glue to make it an even film.
Then the other stator/spacer with the diaphragm contact is placed directly on top, making sure it is aligned and that the all contacts suit the wiring on you headphone cans. Then cover them with plastic (or better still.. mylar!) and cover them with another pane of glass and a weight.
When it is all cured, cut the assembly free from the mylar, and with a sharp blade, trace around the outside of the spacer to remove all trace of the diaphragm. This step is important – you must be diligent. It is also why my stators are of slightly smaller diameter than the spacers, so that there is more than the tiny 0.5mm spacer thickness between the stator and the ragged edge of the cut diaphragm. If a small piece of the diaphragm shorts the 0.5mm distance to one of the stators, charge will be lost and the results my be audible. The mylar should be skinned tight, with absolutely no wrinkles. If there are, rip it up and do it again. You can tell when the diaphragm is well tensioned, flick it with your fingernail. If you hear a nice tone then is OK. You can also hold the panel up and look through the perforated metal at a reflection on the shiny diaphragm surface, and get an idea of the tension that way. If you can see the diaphragm is wrinkled, rip it up, start over. Also, if for some reason there is a dead short between the stators and you can’t find it, rip it up, start again. Check that the brass shims are actually making good contact with the perf. metal also.
The headphone enclosure
This must provide complete support for the ES panels for they are most probably pretty fragile.
The pictures show (what is left of) the old dynamic headphones which I butchered brutally and without compunction to accept a new pair of electrostatic transducers. The backs were cut off so the phones had only enough plastic to mount the ES panels. They resembled ‘rings’ with just a 20mm deep plastic cylinder supporting the ear cushion.
The original cord and drivers were junked. The new cord is about 1.5m of computer IDC cable (10 strands). It has an insulation rating of 300V, and I only use every second strand to further reduce loss of signal in the wire due to capacitance, and reduce the risk of insulation breakdown. The plug is a 25 pin D connector, not ideal but all I had available at the time. Again, I soldered to every second pin.
How exactly they are mounted depends on the headphones, but mine are just friction fit, held fast with the wires soldered to the brass tabs, and the odd drop of glue. 5mm foam for acoustic damping is recommended.
The above pictures were taken with the ear cushion and foam rubber removed so you can see the detail of the stators. In the leftmost picture you can see the red spacer ring behind the stator. The stator is painted black.
Driving Electrostatic Headhphones
A Step-Up Adapter for Power Amplifiers
Of course ES headphones can’t run off normal headphone sockets, they need there own special high voltage amplifier and a high voltage bias source. The bias voltage is around 450V. This was achieved with a voltage doubling rectifier running off 230VAC, which was from two little transformers back to back to provide isolation from the mains. For lower voltage mains (such as 115V for the USA), the design can be used as-is, except have a 115:24 transformer then a 12:115 transformer to get 230V. Or else use 115V and add a few more steps to the ladder rectifier, i.e. a few more capacitors and diodes to get 115 * 3 or maybe 4.
The bias voltage is not super critical. The higher the voltage, the more efficient the phones, cause there is more charge to attract. This is good because it means that a lower ratio on the transformer can be used. Go too high, and it will become unstable, and the diaphragm will ‘collapse’ into one of the stators even with no signal, i.e., the attractive force of one stator (the two stators are never exactly equally attractive) becomes enough to overcome the tension force holding the diaphragm in equilibrium. I have been meaning to try a variable bias with 500V and a pot. Since this is truly a no-load voltage, a high value pot could be used, so long as it is rated for the voltage.
The capacitors used in the voltage doubler were what I had laying around. So here is a guideline: The caps need to be non-polar, high voltage. So the cheap 630V polyester caps would be ideal for the job. The higher the capacitance, the more ‘grunty’ the bias supply, but since the only load is leakage diaphragm->stator, a VERY small value could be used. On an ESL speaker many times the area of headphones, 0.047uF capacitors are used. I believe the values could go smaller with phones. There would not be much to gain from going too small though. My prototype uses 2uF which is HUGE, bordering on unsafe! My bias supply can be turned off, and even the 2uF is enough to keep the phones going for 30 minutes, and I would consider my pair to be VERY leaky.
I used a 20Mohm resistor in line with the bias source, which is a current limiting resistor. The diaphragm needs to only be charged up to be affected by the stators, so the bias supply provides no current during normal operation (only leakage current, which is next to zero) so no current equals no voltage drop across the resistor. In a short circuit fault condition, The resistor limits the current to a minuscule amount as the full 450V is dropped across the 20Mohms. Since these headphones are around your cranium, this current limiting resistor is your friend. Don’t leave it out.
The stators need around a 300V voltage swing, and need next to no current. So a high voltage, high output impedance amplifier is ideal. These can be built easily with signal tubes such as the 12AU7 and the 6SN7. Eventually I would like to build such an amplifier. In the mean time, I created a cheap way of operating the headphones using the 100V line to speaker level impedance matching transformers used in PA applications. These can be got from any electronics store, they are around 2-5WRMS and are only a few bucks.
They usually have 5K and 2K tappings on the primary, and 16, 8, 4 and maybe 2 on the secondary. But don’t be fooled, the 2.5K tapping is NOT the center tap for the 5K tap! (Remember – turns ratio = square root of ratio of impedances). However, in this application you can get away with making a ‘virtual’ center tap by tagging a couple of 1Mohm resistors across the 5K tap, this will not load the transformer at all.
The common and 5K taps go to the stators, the negative of the bias supply goes to the virtual center tap and the positive to the diaphragm contact. Hook the other side of the transformer to a power amplifier. It will not load the amp at all. I recommend a fairly powerful amp, just because they have a greater voltage swing than smaller ones. Try different tap settings eg 8,4 ohm, the smaller the value the higher the step up ratio and the louder the headphones. It is a good idea to start on a high value eg 16ohm and work down!
I run on the 2-ohm tap with a 50W amp that has 35V supply rails. You will almost certainly get different results because factors such as bias voltage and diaphragm/stator spacing determine efficiency and maximum sound level. Also make sure the amp likes having no load, some amps especially valve amps are iffy about loads and may like a dummy resistive load.
The resistors are all 1/4W, the diodes are 1N4007 (1000V) and the capacitors are 2uF, 250V, but these should really have a higher voltage rating and don’t need to be nearly that big (0.01uf would probably be ample). The case is made out of a car battery charger. Note how the headphones have been reduced to ‘ring supports’ for the ES transducer and ear cushion. There is 4mm foam rubber on either side of the ES panel. This is acoustic damping, and also keeps fingers off the metal stator.
The upper 2 transformers in the picture are the 4W PA transformers, the grubby little ones are the bias supply transformers. The diode/cap ladder is under the D25 plug.
A Tube Electrostatic Amplifier
I have constructed a valve ESL headphone amplifier as detailed in the TubeCAD.com article Electrostatic Headphones by John Broskie. The amplifier uses 2 6SN7 tubes and 2 12AX7 tubes. Currently I have just built a prototype, and am working on reducing hum and increasing bass response. The sound is promising, though!
A lot of the part selection was dictated by what I had on hand, as you can see the power transformer is ridiculously oversized. The switch next to it is the HT switch, this is switched on once the heaters are warm, to prevent cathode stripping since there is a solid state rectifier.
I built these headphones for fun, and to replace the awful cheap ‘earbuds’ that I was previously using. But these headphones do work exceptionally well. Having never listened to a high end commercial offering like Sennheiser or Stax, I cannot tell you how they compare. However they are clearly superior to any ‘affordable’ headphone, and for price you can’t really complain (especially considering Sennheiser Orpheus ES headphones are US$14,990!!).
Bass is very refined and DEEP. Not like most cheap headphones, which either have no bass at all or augment the bass and end up muddled (“Now with Bass Boost System”).
Midrange is clean and rich, very natural and not harsh at all, and the highs are especially clear. You can listen to these headphones for long periods with far less fatigue than earbuds. Cymbals sound like Cymbals, not like hissy-fits.
Volume is acceptable, when you turn it up too loud the diaphragm hits the stator and gives a very audible snap crackle and pop which is VERY annoying. The headphones still almost make it to what I would consider maximum listening level. They are by no means quiet. They could be made to go louder by increasing spacer thickness (and hence excursion limit), but that would also mean an even greater increase in voltage swing to achieve the same output… it quickly gets out of hand.
Since these ES phones are open-backed, other people can hear what you are listening to, but it is not intrusive. It is like when someone turns up there walkman REAL LOUD. But it also means any outside sounds go straight through the diaphragm and you can hear them easily. They don’t block out the outside noise like normal headphones, so they are best listened to in the dead of night….