At the beginning of 1999 I started two ESL projects with Kari Hakanen. Hybrid and fullrange ESL:s. Stators were made simultaneusly, but rest of the project was left aside until hybrid versions were finished. Frames were completed at summer 2001 and project delayed again until spring 2003. Few experimental panels were tested during that time.
Listening setup at the end of year 2004
Fullrange- and fullrange panels
Many commercial so called fullrange panels are based on same idea as some home theater speaker systems. Compact satellite speakers with narrowband subwoofer that creates the illusion of full range operation. Usually there is a gap or clear attenuation in summed frequency response. Frequency response of typical small fullrange ESL element shows that upper bass region is greatly attenuated and this "fullrange" operation is based on resonance peak at lower bass region. Result is soft, one tone bass with poor impact.
Frequency response of one commercial "fullrange" speaker.
Good transient reproduction requires bass resonance to be removed. Remaining response sounds very thin so it has to be equalized. Unfortunately electrostatic element has poor sensitivity compared to dynamic elements. Dipole roll-off equalization results even poorer sensitivity and is thus very rarely used. Need for equalization can be decreased by increasing panel size but result is too large for most people. Portal panel was designed with this less common way and normal size constraints were ignored.
Large panel and full range operation brings several problems that have to be taken in care and that are not necessarily problems in smaller panels.
- Large panel area will result high capacitance.
- Because of high capacitances there is no other way to lift LC-resonance than to minimize spread inductances. Transformer that is optimized for low spread inductances has also rather high capacitance
- Transform ratio cannot be very high in order to keep system impedance level reasonable. System with 1:60 ratio transformer and 3.3n total equals 12uF capacitance at primary side.
- Small reactive load demands extreme current driving capabilities from amplifier. SOAR of the ouput transistors must be carefully analyzed. 12uF capacitor has 0.66ohm impedance at 20kHz.
- Spreadinductance of the transformer has to be very low and the same applies to speaker cables. 5uH total inductance in wiring limits frequency response to 20kHz and create phase shift starting from 2kHz.
To overcame these problems full range transformer was developed, few low inductance cables were prototyped and power amplifier suitable for capacitive load was build.
Portal is same generation panel than hybrid speakers. Basic contruction is similar, but some improvements have been made. Fastening is more robust, diaphragm has more uniform and much higher resistivity coating. Polarization voltage is 5.5kV. More common diode multiplier was used as a polarization supply instead of switch mode supply that was used in first hybrid panel.
Cabling was originally build inside frames but open contact with wood at connector caused too high leakage currents. New cabling has unbroken insulation from panels to electronics box. One speaker has three identical panels that are connected in parallel and driven with one transformer.
Complete frame is 230 cm high and 120 cm wide. Outer frames are about 90 x 90 mm plywood. For maximal ridigity there is one 18 mm plywood sheet straight angled compared to others. Feet of the speaker are welded from 10 mm steelprofile and filled with bitumen. Vertical 30 cm steel bar welded to both feet ensure liable connection to frame and electronics box below speaker. Box is made of 18 mm plywood and damped and hardened with sand-resin -layer.
Steps towards better sound
Portal panels have been used in my main setup since they were finished. Panel itself has remained the same but electronics before it have had several changes.
At the beginning tube amplifier output transformer was used as a step up transformer. There was no equalization and Denon electronics (power amp, preamp and CD) were used. It took a while to get used to new sound that was very bright. Upper frequency range was not what I was used to in my earlier panels. Most recordings sounded irritating. Still there was nothing seriously wrong with frequency response. There were only few recordings that were nice to listen and I that is why didn't listen wery much.
First update was new audio transformer. The one that was used was not wery well suited for fullrange operation. I was not able to find any good commercial full range transformer so it was designed by ourselves. Difference was huge. Irritating upper range that was thought to be a directivity or power response problem was crearly distortion problem. Low frequencies take tranformer close to saturation and cause high amounts of low frequency harmonic distortion and high amounts of intermodulation distortion to upper frequencies.
Next step was new power amplifier. It was specially designed to drive capacitive low impedance load. It has massive power supply and 16 power devices that are capable to drive 0.6 ohms impedance at full power. Amplifier distortion at nominal power 70W is less than 0.001% throughout audio band. After this change upper region harchness was mostly gone and it was clear that panel itself performed excellently at mid and high frequencies.
Next update was more like a philosophical problem. Is it OK to use equalization in fullrange ESL-panel. I decided to try. Equalization filter has 4dB step corrector below 150Hz and adjustable constant phase tilt corrector. Tilt correction was originally about 6dB (20kHz was 6dB attenuated compared to 20Hz) but later changes in electronics have decreased correction to 3dB. Low frequency equalization is a compromise between sensitivity and low frequency limit.
After those changes only small modifications have been done. Low distortion DAC was designed and new low distortion version from the equalizer was build. Few low inductance cables have been tested also. It was a surprise to me how dramatic influence electronics can bring to the system. Other DIYers should remember this also while improving their panels.
Two most significant problems compared to most high prised systems I have heard have been low frequency extension and poor sensitivity and dynamics even if those are less problematic with this panel compared to several other panels. For example Audiostatic DCI was at least 10-15 dB less sensitive at same room and same listening distance. Frequency balance and integration is excellent and stereoimage such that only panel element can produce.
Aligning speakers makes huge difference to stereoimage and depth. Rather close listening distance gives best balance but extreme directivity and minimal room interaction makes them sound nearly like headphones. By increasing listening distance depth and stereoimage will improve but few desibels are lost from bass region. It must be noticed that even if panels are extremely directional they are not very difficult to listen and beaming is not a major problem. Also sound is not too dark even if listened from side of the panel like it is with most hybrid panels.