Below you will find a description of the developments of a large 'full range' electrostatic loudspeaker. The outline won't contain fine drawings and accurate measurements of components for building an ESL; what you will find are some important issues that need attention during construction and design of your ESL. For a fundamental description of the basics of an ESL I refer to the official Quad site, this site will give you a tutorial description of the true physical basics of ESL's. Reading other articles from designers on this web page is highly recommended for gathering information and idea's about ESL's, but you probably already did :-)
If you have never built an ESL before, it is wise to start with a small unit for the mid/high region instead of heading directly for the large panel. You will find during the construction that several issues are not that evident as expected: coatings, behaviour of high voltage terminals, problems with construction etc. Once this seems under control a large panel can be built.
Some electrical basics
ESL's need fairly high drive voltages. The electrical withstand fieldstrength (arcing or flash-over threshold) of air is roughly 3kV/mm. The bias applied at the ESL should be such that a static fieldstrength is obtained in the range of 1200 to 2000V/mm. 1500V/mm will do fine. This means that if the stator-to-Mylar or diaphragm distance is 2mm, you will need a bias voltage of say 2mm x 1.5kV/mm = 3kV.
With the example from above, this means that there's 3kV/mm - 1.5kV/mm = 1.5kV/mm electrical fieldstrength left to drive the unit. As the diaphragm-stator distance is 2mm, the required maximum drive voltage for this ESL will be 1.5kV/mm x 2mm = 3kVpeak for the front panel and 3kVpeak for the rear panel, ending in two times 3kV, equals 6kVpeak drive. It is up to the constructor if he wants to drive the ESL to the limit of arcing, but it this voltage that will give you the maximum obtainable acoustical output.
A bass panel with stator-diaphragm spacing of 4mm requires a bias supply of 4mm x 1.5kV/mm = 6kV or 12kVpeak drive voltage in total. If the latter cannot be achieved, the bias supply can be increased to say 2kV/mm to get slightly higher output levels.
The capacitance of an ESL is very important with respect to the performance type of step-up transformer (if used) and the required drive power of the amplifier. The capacitance of the ESL is calculated by:
C = 8854 x A / d
- C = capacitance of ESL in pF (pico Farad)
- A = surface of ESL in m2 (square meters)
- d = distance between stators in mm
Add about 25% total to the capacitance due to unknown stray capacitances.
The unwelcome trick about the capacity of an ESL is the following: Suppose we have designed a single element full-range panel. High frequency response of the ESL has been taken care of by dividing the large panel into one or two slim units. By doing so, the amplifier sees the full capacitance of the ESL at its terminals. A stator-stator distance of 8mm is required for a full range panel operating from 30 or 40Hz and onwards and a large surface to get proper output, say 1m2 in total. The capacitance of such an ESL will be roughly 8854 x 1m2 / 8mm = 1100pF plus 25%, is approximately 1380pF.
Now, the impedance, Z, or 'electrical resistive behaviour' of a capacitor is calculated by:
Z = 159 x 1,000,000 / (C x f)
- Z = impedance ('resistance') of ESL in kOhm.
- C = capacitance of ESL in pF
- f = frequency in Hz.
Suppose the available drive voltage is 6kVpeak with a certain step-up transformer and amplifier. The impedance of the ESL at 50Hz is approximately: 159 x 1,000,000 / (1380 x 50) = 2300kOhm. The power or amount of watts across a certain impedance or resistance is calculated by:
P = 500 x (U x U) / Z
- P = required drive power in Watts for ESL
- U = voltage across ESL in peak kV
- Z = impedance of ESL in kOhm.
At 50Hz the required drive power for full output power is 500 x (6 x 6) / 2300 = 8Watts, or 16Watts peak drive power.
The impedance of an ESL decreases rapidly with increasing frequency, so more drive power is needed for the higher frequencies. At 20kHz the impedance will be:
Z = 159 x 1,000,000 / (1380 x 20000) = 5.76kOhm
Meaning a drive power at 6kV of: 500 x (6 x 6) / 5.76 = 3125Watts or 6,25kW peak drive power. In other words, you will need an amplifier capable of delivering 6kW peak power to drive the ESL at its full output power at the highest frequencies. This is not recommended. In practise, it won't go that far because of the following:
- Music does not show its maximum amplitudes at the highest frequencies
- The step-up transformer used shows resistive losses, which increases the impedance at the higher frequencies and in return, the power fed by the amplifier.
It is however hopefully clear to the reader that a large full-range panel is a very tricky thing. It is NOT recommended to construct a full range panel from one panel covering the low and high frequencies. The impedance at higher frequencies will become too low, putting severe requirements on the amplifier namely the capability of delivering high output powers with low distortion at high frequencies. Very few amplifiers can do this work well. Secondly, constructing a step-up transformer capable of good high frequency response at high step-up ratio's is a very difficult task. Therefore, it is far better with the focus on good transient response of the ESL and low capacitance, to split the full-range panel into two separate units. The unit for the high frequencies shall show a capacitance of say 150pF. With 3kV drive (smaller stator-stator distance) the required drive power at full output power and 20kHz will be 85Watts or 170Wpeak. This is: much better for the amplifier, much better to construct a step-up transformer with this small capacitance for the higher frequencies and much better for the transient response due to the light weight of the diaphragm and less drive required from the amplifier.
The diaphragm and the coating
I used Mylar for the diaphragm. For best transient response of the loudspeaker (mid/high section and low section as well) take the thinnest Mylar you can get. For the mid/high section Mylar in the range of 0...3um (micron) is highly advised, and it can be used for the low frequency panel as well. This will result in best transient response of the system and in return a nice tight and firm bass. I could not get my hands on 2 or 3um Mylar, but I intend to use this for my next pair of ESL's.
Coating the diaphragm was not evident and easy!!! I read that the type of coating can be any thing like handsoap, graphite, wall-paper paste etc., but nothing performed clearly to my satisfaction however.
The issue with soaps
For a bass ESL you will need large surfaces and a high resistive coating in the range of 1Gohm to 5Gohms (refer to the Quad web site and other articles on this web site). It is fairly difficult to get a coating of 1 or 2 Gohm with soap. I finally got there and tried to 'age' the diaphragm by applying heat-cold cycles to the diaphragm with the hair dryer. Heat, cool down, heat, cool down etc. I did this several times with considerable heat applied and found out that after few times the resistive coating had disappeared and didnot come back again. I tried also with wall- paper paste. No soaps were able to get a reliable high resistance, nor would they stay good on the Mylar. Ofcourse, the quality of the coating depends upon the type of soap and things you use. Experimenting with the stuff you can get is recommended, this is just an overview of my findings.
Graphite / carbon
Dirty job and far more difficult, if not, impossible to get a high resistive coating in the Gohm range.
This worked out very well. That is to say, for the first 5 hours. The anti static sprays on silicon bases will give you a fine reliable and reproductive high resistive coating that stays on the Mylar as well after the heat-cool down cycles. This promised a lot of good. However, after having constructed a bass ESL with the anti static coating it seemed that the high voltage terminal on the Mylar was nibbling itself away due to corona effects around the HV terminal. I heard same experience from somebody later on. I had to take the ESL apart to apply a new diaphragm and coating as the HV terminal had left due to self destruction. Taking the ESL apart was not a nice job as I glued the entire construction firm and proudly together with polyurethane glue. Things got destroyed when I took the ESL apart. Golden advice: Design your ESL so that it can be dismantled easily.
The coating that worked
The type of coating that worked out fine (at least to me) was hobby glue, the child-friendly type, from TESA (fluid type of glue, not the glue stick like "Pritt stift"). The "Pritt-stift" stick glue worked out fine as well, but applying the glue to the Mylar was not that easy as the TESA fluid glue. I do not know the main ingredient of this TESA glue.
Maybe some one does?? It is possible with this glue to get a fairly high resistive coating in the range of few Gohm/square. This is how it works:
Apply few cc of glue in a desert bowl and mix it with 5 parts methanol. Mix the stuff until a fine solution is obtained. Take a paper hand tissue or kitchen paper cloth, put little glue solution on the paper and smear it out over the Mylar. Don't use a lot of glue, the Mylar just need to get wetted, that's all. Next, take a new paper tissue, wet it with methanol or water and softly take some glue from the Mylar by whipping few times softly across the surface. Let dry with the hair dryer. You should have a resistive coating in the range of few Gohms/square now. This range of resistance can be measured with some digital Ohm/voltage universal meters (not all); check the capabilities of the universal meter out at the electrical component shop in your village before buy. The resistance shall be measured by putting two coins on the Mylar about 2cm apart although not critical; You will find that the resistance is constant across large distances between the coins. If you cannot measure high resistance, there's no reason to worry, the above method guarantees a high resistance in the Gohm range. Performance can be checked by the following test (advised before starting a large panel!!):
Test drive a small piece of Mylar of 20cm x20cm. Apply the glue as described above. Put the Mylar with coating upwards on a small piece of stator and put a coin on the Mylar. Connect coin and stator to a bias supply of 5kV or so. The Mylar should bond very well to the stator if the bias is switched on. Take care about the high voltages of course as the stator is under high tension. Small traces of corona and sizzling sounds can be heard around the edges and coin, though this should not be loud. No severe corona effects shall be noticed!! If so, the coating has too low resistivity and more glue shall be wept off with methanol. Switch of the bias supply and de-charge to supply by first applying a wire to the ground terminal of the supply (stator) and then by applying the wire to the HT terminal (coin) of the supply. Be prepared for sparks. If the coating has been applied correctly you will discover that charges will stay for a looooong time on the Mylar. This is good, meaning a high resistance coating! I was able to measure for long time small voltages on the surface with a sensitive digital voltmeter and coin.
Once you understand the trick of getting the high resistance coating you will be ready for building the large bass ESL panel.
Picture of coating the Mylar with anti static spray. NOT recommended: see text!!
The isolators or spacers
I used plexiglas spacers. Although many more materials can be used like pertinax (paper) and glass fibre from PC-boards, plexiglas has a large dielectrical withstand voltage (doesnot leak and arc or flash over), and, is not dangerous to work with while fibre glass is during drilling and cutting. Plexiglas can be delivered in 1mm, 2mm, 3mm and 4mm thickness. Search the yellow pages for a supplier, it may be little difficult to find. Let the supplier cut the plexiglas to your required dimensions.
Plexiglas is not cheap. Check the plexiglas at its thickness across its length. I found out later after having glued the plexiglas to the frame that the plexiglas was indeed 4mm thick at the ends, but increased to more than 5mm in the middle!!! Check the thickness carefully and ask for a warranty if it does not meet your specifications, say +/- 0.1mm. As I already glued the plexiglas to the frame, I had reached the point of no return! I had no payment bill of the plexiglas, and had to sandpaper the plexiglas with carborundum sandpaper (sillicium carbide) to get the 4mm thickness. This type of sandpaper is not easy for sale and it had to be ordered specially for me. It is not easy to get a constant thickness of the stator by sanding. In fact, it did not work out well enough. If you need this type of sandpaper to sand the plexiglas go ask at a grave stone manufacturer; they use it as well for sanding granite and the like. Sharp edges on the spacers shall be sanded or scrapped away as these may cut into the Mylar later on.
Result after loooong hours of sanding with sillicium carbide sandpaper. Dusty and hardwork. Check the thickness of the plexiglas carefully before you accept your order.
I used perforated sheets of metal for the stators, measuring 205 x 45 cm and 2mm thick. This results in a heavy ESL. Wire stators are used as well in some designs, although I find the construction of a wired stator somewhat difficult. The stators I plan to use for my next pair of ESL will definitely be wire grid stators for the mid-high section. This type of stator has good acoustical characteristics due to its very fine mazes. The surface won't be broken up by separated high frequency sound sources like the perforated steel. Remember that the wavelength of sound at 20kHz is only 17mm so drills in the sheet of 5mm diameter and 12mm centre-centre distance is a construction close in range of the wavelength. If upper performance is required, the grid shall be made very fine. The wire grid suits OK to me for this high frequency purpose. I also used this type of stator for my small proto type ESL, see picture below.
My first proto type ESL. It is small and easy to modify. It has acoustical output from 300Hz and upwards. The high frequencies are of very, very good quality, I have never heard before. The stator is of the 'fine chicken wire' grid type.
The stators shall be painted well to avoid early arcing or flash over to the Mylar. Take all your time to paint the stators properly. It is important to get a good insulative coating on the stators.Painting the stators
The perforated metal sheet has a soft side and a rough side. The soft side shall point towards the Mylar to prevent arcing. I painted the stators with acryl paint that is obtained form a car surplus store (spray cans). You will need a lot. I used about 10 cans of paint in total measuring 600ml content (25% free additional paint)! The following layers were put on the stators:
Mylar side: Two layers of primer, 4 thick layers of paint. Opposite side: Two layers of primer, thick filler primer (4 cans of 400ml in total) to get rid of the rough surface for the better view and 2 layers of paint. Advice: First start with the primer and filler primer. The entire environment will get 'dusty' from the paint, and you will find traces of the spray everywhere around, also on the other plates close by. The paint even got under the plates lying on blocks on the ground resulting in light greyish film on the already (beautifully) painted Mylar side of the stator. I had to redo this side again. Also take care of a clean environment for the spraying activity. I discovered that there were spiders housed in the garage, that made webs under the plates within one day. They didnot give me a chance to paint the plates without any thin web strings here and there. Too bad, but it's OK, I don't really mind.
After I painted the stators with the acryl paint, I applied few thick layers of polyurethane varnish on both sides. Polyurethane varnish and acryl paint does not seem to like each other very well, so make sure that the acryl paint has been dried completely (leave it for 2 days). First apply a thin layer of polyurethane with a brush and let it dry as well for 2 days. Then apply a thick layer of polyurethane varnish. Don't apply this thick layer first time as the varnish will 'peel' together from the acryl.
The ESL Constructio
The bass ESL measures 200cm x40cm effectively and stator Mylar distance is 4mm. The section for mid and high is planned to have slightly less height (150cm or so) and shall have 5cm width. The stator for the mid/high section will be definitely a wire grid system (see above).
The frame is made of some exotic solid wood. First, I glued some 2mm plexiglas strips to the wooden frame on which the stator will be mounted. I used polyurethane glue which provides a good solid connection. It took about 24 hours before the glue reached its full strength according to the instruction. It takes quite a long time before the glue starts to harden, more than one and a half hour. Such a long time is comfortable to work with instead of the rapid type of glues. There's plenty of time to put the glue on the units and provides for easy corrections. The picture below shows the stator to the wooden frame with spacers already glued to the stator.
Wooden frame with stator. Ready for glueing the Mylar to the spacers.
Stretching the Mylar
I found stretching the Mylar a nice job to do, much better than sanding the plexiglas. The question came in mind if I should to apply a firm or weak tension in the Mylar. I pointed out earlier that a small experimental ESL at hand is very handy so I experimented with the tension in the Mylar with the small ESL. The droplet shaped ESL in one of the pictures above showed the following results at different tensions in the Mylar. (I publish the result because some might be interesting in it):
Normal tension, hand tight (first concept)
The main resonance frequency was 350Hz, and few more down below (180Hz). The frequency response rolled off at 340Hz, and the ESL showed slightly increased output around 180Hz.
Weak tension almost no force applied
Horrible frequency response with lots of resonance frequencies. I even tried to dampen them with soft sticky tape (tape for mounting mirrors against the wall). Nothing helped. I hoped to lower the resonance frequency from the ESL by the lower tension, however it is not recommended! The frequency response rolled little off at 340Hz and started again around 180Hz. I noticed that an ESL couples tight with the environment around its resonance frequency! I held my hand in front of the unit (10cm) and the speaker acted immediately! Incredible. This also proves for being a good acoustical transducer, at least it helps a lot!
High tension, perfectly stretched (pull tight)
This gave best result. The unit showed a resonance frequency of 200Hz and fine! Output started to roll off from 200Hz. No or very, very small minor resonance frequencies were detected in the higher bands, but I might be mistaken, could be some standing waves in the living as well. My conclusion for this ESL is to tighten the Mylar as good as possible. This lowered the resonance frequency with respect the first unit and gave a more flat response.
So I decided to stretch the Mylar for the bass ESL as well tight!! Stretching the Mylar is not that difficult. First make sure you have a clean and flat work bench. I used a sheet of pressed MDF or so. Cut the Mylar into a larger piece than the stator. Take a piece of tape and stick the Mylar at it's longest end to the MDF panel. Then take a piece of tape again and put it at the opposite side of the sheet of Mylar. Pull tight. There will be lots of ripples in the Mylar, but these will disappear gradually as you go cross wise along the Mylar. First do the longest sides (top and bottom of the large slim ESL), then take the shortest sides (left and right side of the ESL). Start in the middle of each side. If all is done, go around the pieces of tape again and pull again. Take a look at the Mylar at low angles across the surface and adjust here and there. You will find that stretching won't have an effect anymore in the end. This is the point that the Mylar has been stretched perfectly.
Stretching the Mylar. Use small pieces of sticky tape and stretch the Mylar cross-wise. Pull tight and go around twice.
Clean the stator with methanol before glueing things together!!. Now it is time to glue to Mylar to the stator. In the first design I used polyurethane glue, but replaced it by thin two-sided sticky tape from TESA. I discovered that this tape holds the Mylar even better than the polyurethane glue and it is less messy ofcourse. Take the stator and lay it on the Mylar in one smooth swing.
Update November 2, 2000
After few days this article was released, I got two emails from ESL constructors who claimed that the TESA tape would not hold at raised room temperatures of 30 degrees C and more. The sticky surface will get "stretchy" and as a result, decreases the mechanical tension in the Mylar. I must convince that I didnot test the TESA at high temperatures. Hair dryer tests are really good tests for reliability checks. I always did that at the laboratory when I was designing RF electronic circuits few years ago as a first indication for checking the reliablity of the circuit I designed. Should have done this now as well for checking the mechanical strength of the sticky tape.... At least, room temperature doesn't raise above 30 degrees here in Belgium, BUT the message from the emails proove that the trick with TESA tape is not reliable. Within few years the tension in the Mylar has probably left a great deal. I don't know a better substitute for now except the messy polyurethane glue. Maybe silicon kit works out fine as well. TEST-DRIVE THE PROPERTIES OF SILICON KIT BY YOURSELF WITH THE HAIR DRYER!!!
Glueing the Mylar to the stator, using thin two-sided sticky tape from TESA.
Next, the Mylar has to be cut away from the stator and the coating shall be applied (refer above in this article for the coating). Apply a low resistance loop just along the stator on the Mylar with a small wire brush and the TESA hobby glue solution. The high-voltage terminal is made form silver paint (see picture below). To my opinion this is extremely expensive material. A bottle of same shape with eau de Cologne from a shop on the Champs Elysees in Paris is cheaper. No kidding! The difference is that you will buy real silver in this case that holds its value so your money is not wasted in theory. Silver paint can be bought at a car surplus store for window defrost heater repair. A piece of sticky aluminium foil is put upside down on the silver paint. (don't take away the protective paper on the aluminium foil, as you don't need to glue the aluminium strip. This will only make the mating of the front stator more difficult as the strip will glue itself to the front stator).
Close up of the high-voltage terminal. Aluminium pointing towards the silver paint, protective paper is left on the aluminium.
Work clean and appropriate. High voltages and high impedances are involved so leakage paths are formed easily. Before glueing the front stator together, clean the edge of the Mylar with methanol and the spacer of the front stator as well. Don't economise on paper cloth. If this job is not done properly, you will end up with a sizzling and low performing ESL as the bias supply won't be capable of charging up the diaphragm due to leakage problems.
Finally the front stator is glued together. This was not done with two-sided tape, as the stator needs to be aligned with the rear stator: the holes in the plates need to be in line with both plates. Two-sided tape sticks immediately leaving no time for aligning the front plate. I used some kind of repair silicon kit that worked fine. Finally I drilled holes and tapped threads in the front stator for nylon bolts that holds the front stator firmly together to the construction. This design enables me to take to ESL apart without destroying the unit.
Final assembly of the ESL. The upper stator is glued together with repair silicon kit to allow for mating the holes in the front and rear plate.
OK, how does it sound?? A bias voltage of 8.7kV is used for the bass ESL and the panel has a resonance frequency of 16Hz. Output starts at 35Hz and upwards. The bass ESL (with the TESA hobby glue coating) has excellent bass production. To be honest, I have never heard such a bass before. It sounds firm and tight. At first instance you think there's no bass in the speaker but there's definitely a bass. It sounds deep and straight without any banging and resonances. Just clear. A pleasure to listen to. At the moment I use the large panel for the entire frequency range. The high tones are OK, but the small ESL sounded better as expected but has low output due to its small aperture.
Due to the large surface of the 'bass ESL', the sweet spot is small. I don't care about that as I listen to music from my comfortable chair in the acoustical centre. As soon as you're in there, you will experience beautiful waves of sound. The orchestra plays in between the loudspeakers; you won't 'see' the loudspeakers in the music. The instruments are clear, the ESL 'reveals all details from the recording material'. It's just great! I use a home built direct drive amplifier for the ESL's that can deliver high peak output powers into the kilowatt range.
Positioning the ESL's seemed critical. I placed them about 1m from the wall, with heavy curtains at the back of the ESL's. Placing the ESL close to the wall results in bad low frequency response. Finding the sweet spot needs accurate positioning as well with this large radiating surface but it can be done within 10 minutes. Good luck with your designs and happy listening!!
The high section needs to be separated form the low section, but this was the first initial idea. A slim mid-high unit will be build but I need to get my hands on thin Mylar for a high performance mid-high ESL. Fine wire grid will at least be used for the mid-high section and perhaps for the bass ESL as well (not sure about the mechanical stiffness). The bass ESL will be made little wider (50 or 60cm) and smaller in height. This is also due to aesthetic reasons. The 4mm spacing is maintained to assure static stability. A radiating surface of at least 0.8m2 for the bass section will be maintained for the second design.