I decided to build my first ESL with stators made from Glass Epoxy printed circuit board. I bought a sheet measuring 4' x 3' and divided it into four pieces for the four stators of two panels. They measured 2' x 18" and that is what decided the size of my panels. The diaphragms would be of polyester film of 12 microns thickness, coated with graphite on one side. I had the PCB board perforated with holes of 1/8" diameter, which proved to be a mistake, which I discovered when it came to paying for the work. If there is a next time, I will choose a larger hole size. The work of perforating is charged by the time it takes and larger holes would take much less time.
I obtained a sheet of 1/8" (3mm) Tufnol for the spacers and also used the same as stiffening and bracing to keep the copper clad PCB material as flat as possible. I used two-part epoxy to glue one of the panels and Bostik Axiom 100SA adhesive to glue the other. The Bostik joints proved to be stronger than the epoxy, but I am not sure about the insulating properties of the Bostik. I have also tried Loctite Super Glue Gel and that appears to be stronger than either the 4 minute Epoxy, or Bostik.
I had a lot of trouble with leakage on the first panel as made originally, so I abandoned the use of Tufnol for the spacers and bought some 3mm Polycarbonate sheet and cut it 1cm wide, the same as the Tufnol spacers. I also used High Voltage wire (as used to connect the EHT to television tubes) for the connection from the connection bolt to the diaphragm contact, for which I used a strip of conductive silver paint. That cured the leakage problem. The connection arrangements are shown in a drawing.
For stretching the polyester film, which was made by a firm called Plastic Products Ltd based in the UK. Their product code for this film is SM30. You will see from the picture of the stretching frame that it has two fixed clamps and two moveable clamps. The moveable clamps are connected to two pieces of timber by the sort of spring balance used by fisherman and the film was originally stretched so that each one of the balances had a reading of 20 pounds. Subsequently I increased that to the maximum tension of about 28 pounds.(12.5kilos).
In the centre of the frame is a piece of chipboard covered with white Melamine, which acts as a smooth level surface on which to rub in the graphite, which I applied according to the description on page 117 of the Loudspeaker Design Cookbook by Roger R. Sanders. It is not difficult to do. When the graphite has been rubbed into the surface of the film and surplus removed the film is removed and turned over so that the graphite faces downwards. Then when it is at maximum tension, adhesive is put on the spacers and the panel lowered onto the film and weighted until the bond is cured. I tried several glues and also double sided tape, but now I use Araldite 4 hour adhesive, which gives plenty of time. Then I leave it overnight to cure. Should any wrinkles remain after gluing they can be removed with the hot air gun.
The two stators are connected to a bolt, one on each side, for connecting the transformer to the stators.
When the audio transformers finally arrived, due to a misunderstanding, the secondaries had only one winding (that is they were single ended). I was very disappointed as I thought I would have to wait still longer before I could test my panels. Then I hit on the idea of using both transformers to drive one panel. I connected the primaries in parallel and the two secondaries in series and used the junction of the secondaries as the centre tap. This enabled me to get sound out of my first panel.
I did a near field SPL frequency run and found that the fundamental resonance of the panel was only about 20Hz and I had to use a VARIAC type transformer set at only 30% voltage, which means I had less than 2kV on the diaphragm otherwise it would become unstable. The reproduction was good, but the resonance was far too low.
I decided then to put an extra spacer halfway between the longer sides, thus dividing the panel into two 24" x 9" sections. When tensioning the new diaphragm, I set the tension on the spring scales to 12.5 kilos. This time the fundamental resonance was 46/47 Hz. By this time the transformers with centre tapped secondaries had arrived.
The SPL graph showed a 6dB per octave rise from about 2kHz upwards and of course the sharp peak of the fundamental resonance at around 46Hz. So I made up a 1st order low pass filter and that tamed the too bright high frequencies. Next I made notch filters with adjustable centre frequency, but when I used a Q of 5 and tuned it to suppress the peak, I had two peaks either side of the resonance. That presumably means the notch was too sharp. However, when I reduced the Q I found that the level between the resonance and say 100 Hz had fallen off, presumably because the skirts of the notch response were affecting it.
That is where I am at the moment. If I can suppress the resonance and keep a flat response, I will have bass down to at least 40Hz. If I cant get a flat response, I may put a high pass filter around 150Hz and use the subwoofer as originally intended. At this moment I have both panels mounted in frames made of veneered chip-board (Conti board) and so can listen to the sound in stereo. Despite the fact that the bass response is far from flat they sound better than any other speakers I have had and the stereo image is extremely good. That is, of course when the speakers are oriented so that your head is where the high frequency beams intersect. Elsewhere in the room and in other rooms, there is a loss of the high frequencies, but to me they still sound quite good. Of course the frequency response of my ancient ears is not very good anyway!
Today (19.03.99) I did a near field SPL frequency response run on both speakers at the same position in the room. The black trace shows the fundamental resonance suppressed and reasonably flat, but although the notch filter used with the panel being tested and shown by the red trace is the same as for the black trace it is not flat. That I must investigate further sometime. I believe the peaks and troughs in the trace at 6kHz and above, must be due to reflections from objects in the room. The graph is attached. Birmingham 18.03.99.
The two ESL speakers are now finished, except that I have not yet found a suitable inexpensive material to put on the back of the frame to make the electronics and their high voltages more inaccessable.
Was the effort worth it? Definitely yes. I can now enjoy the purest music and the best stereo imaging I have ever had.....by far!
However, because planar dipole speakers are so directional as far as the higher frequencies are concerned, it is necessary to make sure that the speakers are aimed correctly. If you follow the recommendations in chapter 14 of Roger Sanders' "The Electrostatic Loudspeaker Design Cookbook" you wont go far wrong. I would go even further and say that instead of making measurements from the speakers to the middle of your head, the measurements should be made from the left speaker to the left ear and from the right speaker to the right ear.
My photograph shows the two speakers with the subwoofer, disguised as a table, in the middle. I use the subwoofer to fill in the lower bass frequencies and to compensate for irregularities in the ESL response around 100Hz. They integrate very well.
Previously I have used my original ESLs as full range speakers and then supplemented by a central subwoofer.
Neither of these options proved satisfactory, partly because the diaphragm material I had used did not keep its tension and consequently the resonance peak was continually falling in frequency.
So I decided to design a closed box subwoofer with the feedback derived from a small permanent magnet attached to the speaker cone centre. The inspiration for this was an article by Russel Breden in Electronics World magazine. He used dual coil speakers and connected one of the coils to provide feedback signal, however I thought it could be done more cheaply using the small magnet working in a small coil to provide the feedback signal.
I constructed the box from 15mm medium density fibreboard. I wanted to use a Qtc of .5 which gives optimum transient performance before the application of feedback. The volume of the box works out for this particular Peerless driver to be 33 litres. The magnet I used was of a type used to operate reed relays and the coil was wound on a pot-core former with 0.1mm wire, almost full to the brim and it measured approximately 150 ohms. I used a Peerless 8" driver which has a concave dust cap, which facilitates attaching the magnet to the centre of the cone with glue.
The diagram shown below shows the signal processing circuitry and the two pictures show how the coil is attached to the speaker box and the original ESL mounted on top of the woofer box. I adjust the feedback control to increase feedback until the circuit becomes unstable and then back till it is stable and then back it off a little more to give stability margin as the circuit appears to be slightly temperature sensitive.
This arrangement gives a pretty flat response from 30-300Hz and the ESL has a filter which rolls it off from 300Hz downwards.
I am very satisfied with the way this hybrid combination works.
If anyone out there would like to construct similar subwoofers and let me have his opinion as to their quality of sound, I shall be very interested in his opinion.
For the future
Next, I would like to experiment with curved tweeters to widen the sweet spot. At the moment only one person can listen and get the benefit of the speaker's imaging capabilites.