Electrostatic speakers at TUT
Electrostats have long traditions at TUT (Tampere University of Technology). Development line that our panels is based on started at the early 90's. First panel I am aware of was about 5x15cm and had a candy paper as diaphragm. I was involved around generation 5.
I started to build my first pair at the end of 1998 with Kari Hakanen. Knowledge gained from other builders was naturally a great aid for me at the beginning. First speaker was done to test new stator insulation. Stators were small 40 x 35 cm plates without frames. Speaker was ready for listening at the beginning of -99. Tests proved new insulation method successfull and so we started the development of new models
Hybrid ESL was my first full ESL project. I did it together with Kari Hakanen. Earlier tests with insulation were successfull and stators were made for several panels simultaneusly. Size was selected so that panel could produce whole speech band. Frequencies from 40 to 250 are produced with 8" woofer.
Polarisation voltage is generated with self made resonance mode converter. Topology has two advantages compared to common diode multipliers. Voltage can be adjusted with potentiometer, (In this case 2-12 kV) and supply has much higher maximum output current. Those can be good features when testing but less useful in normal use.
It is easy to locate possible leakages if current is high enough to produce a hissing sound. It is also wise to test panels with voltages higher than normally used to make sure that they are not used too close to arching voltage. Higher humidity level will increase leakages and decrease maximum voltage that panel can tolerate without arching or dramatic increase of leakages. Arching is a problem only if stators are poorly insulated or has no insulation at all.
It is rather simple to produce about 6-7 octave usable frequencyband with any kind of transformer topology. In hybrid panels that will cover needed frequency region from 150 - 20kHz. Core saturation is not a great concern and panel capacitance is typically less than 1 nF. Toroidal transformer core modified from ordinary mains transformer and has transform ratio 1:45. Two transformers are used with secondaries in series, so total ratio is 1:90.
Crossover and amplifiers
I have been using my panels with Mirror Xover with passive panel equalization network. Kari encased active crossover with woofer amplifiers.
Panel Frame and Element Structure
Frame is done from steel and it has wooden cover. Steel structure is very rigid and thus frame can be made rather narrow. Outer layer is oak veneered ply-wood. Frame and element are insulated from each other. Element is connected with tight MDF-frame that is bolted to steel from behind. No bolts are at sight and parts can still be easily detached. Cloth on both sides of the element dampen bass resonance and protects the elements from dust and other dirt. Current version has identical cloth on both sides of the element so it has only little effect on elements directivity. Thicker cloth behind the element would delay back radiation and shape polar response.
It is crucial to insulate all wiring properly. Even wood may have too high conductivity for diode ladders.
Kari's panel size and LF-element almost identical. Spacers are little differently aligned and LF-box is little larger. Frame is made from MDF and has oak-board on front surface. Panel is rigidly supported from behind with two plates straight angled to panel. Dipole response needs little different equalization but difference is not significant.
Box for Low Frequency Element
Front plate is 45 mm MDF. Sides are curved from four ply-wood sheets with total thickness of 22 mm. Top and bottom are 25 mm MDF. Box is shelf-braced with several shelfs (picture below). 70 mm reflex tube is supported from both ends and rounded from outer end. Box absorbent is mainly cotton. Nodal resonances typical to aluminium cones are more excessively damped with special foam straight behind the element.
Frequency response of Kari's hybrid speakers measured in small concrete room. Measured response is closer to power than frequency response. Exact sensitivity was not measured, but it is about 87 - 90 dB. (Bias voltage 6.5kV.)
Panel construction was tested with experimental panel so it was known how panel itself would work. Insulation works fine although material used could be better and layer thinner. Most "normal" pitfalls were avoided thanks to our more experienced friends and there were no major difficulties to get them working.
When thought afterwards reflexed woofer might not be the best choice to be used with panels but perfectly satisfactory when optimally filtered. When crossover frequency is set to about 200-250Hz there is only little need for panel equalization. Panels are large enough to be used as full range especially in small rooms with poor aqoustics. If heavily equalized maximum SPL will remain low and transformer low frequency requirements may become a problem.
Despite all the general problems, listening tests have proved design rather successful. High polarization, transform ratio and sufficient size result high sensitivity and panel can be driven without high-power amplifier. Sound is something that can be expected from panel like this. Accurate stereo image, very pleasant sounding and detailed treble. Some studio recordings sound too bright with mild equalization, but by adding more high frequency attenuation, something that makes panel sound special is lost.