The electrostatic hybride speakers
After reading an article in the German newspaper Die Zeit on Martin Logan's electrostatic speakers (ELS) I was fascinated by the idea of having something like these speakers at home. Unfortunately, the prize of one of Martin Logan's speakers is more than the rent of my apartment for the whole year and I realized that having a pair of Logan's is not in the right proportion.
So, why not build my own set of ELSs? A quick look with google and I found The Audio Circuit, a wonderful web page on all you need to know about ESLs. Equipped with that knowledge, I started the project.
First, I looked for cheap step-up transformers which turned out to be difficult. Most of the transformers did not meet my ideas about performance or did not meet my budget. Luckily, The Audio Circuit offered the Newton NNS 0.3 electronics kit for a short time with reduced prize and I got two step-up transformers (1:150), two high voltage supplies (3.7 kV) and some other minor parts for a very good prize. Second, I got 6 micron thick Mylar foil from Martin-Jan Dijkstra for reasonable money.
Then, the construction began. As stators I used 1000 mm x 200 mm anodized perforated aluminum plates (0.8 mm thick) from Alfer available e.g. at Obi (prize is about 14 Euro). As the Al2O3 layer on the plates is a very good insulator, no coating was needed. As spacers, double sided sticky tape (1 mm thick, 19 mm wide) was used. You can get at also at e.g. Obi. It is conventionally used to fix mirrors etc. to bathroom walls. I stacked two of them to get a membrane-stator distance of 2 mm. Fixing the membrane on the first stator just needs some 10 minutes. I did not put the film under any tensile strain but laid it flat on the ground and dropped the stator with the sticky tape onto the foil. After that, I coated the foil with pyrolytic graphite solution such that the membrane becomes grey in colour. I spayed the graphite solution (solvent was just methylated spirit) with an old perfume dispenser on the foil and rubbed the graphite afterwards with a cosmetic patch. Then, the foil was gently tensilized using a hot air gun. The tension needs to be high enough that you can hear the resonance of the membrane (say some 30-50 Hz). Next, the double sided sticky tape was applied to the second stator. I placed a conductive stripe on the tape close to the inner edge. I just used 2 mm wide stripes of ordinary aluminum foil. Further I added a contact to the aluminum foil. Finally, the two stators were put together and the ESL panel was ready.
As the ESL is an acoustic dipole, the frequency response drops with 6dB per octave below a critical frequency f, which is roughly given by 1/4 of the wavelength of the shortest dimension of the panel (here 20 cm corresponing to about 500 Hz). So, I need some kind of a woofer to get reasonable sound pressures at low frequencies. As woofers, I used Alcone 6.5 HE (75 Euros each) in a bass reflex enclosure of about 30 l volume. I designed some fancy shape so the enclosure acts as a stand for the light ESL panel. The woofer front and the ESL have to be on the same plane to ensure a correct time response of the combined speakers. This plane is tilted by some 5 degree off vertical to lift the sweet spot of listening off the ground. See the Unigraphics plot of the enclosure below.
To keep things cheap, I took 16 mm MDF plates for the enclosure strengthened with some internal 22x22 mm massive wood bars to damp vibrations of the sides. The enclosures were build in two days followed by the assembly of the complete hybride speakers. The real picture of one speaker is shown below. Initial listening test of the speakers showed a fantastic clearness and directness, but a coloured sound, some problems with the intermediate frequency range and a too strong bass.
The problem of the midrange performance was due to an error in the polarity of the step-up transformer. This caused a 180 degrees phase difference at the cross over frequency between the woofer and ESL branch around 500 Hz. By changing the polarity the hybride sounded much better. The bass was reduced by about 6 dB using two resistors in the cross over circuit of the woofer. I also fine tuned the frequency of cross over to the lowest possible to allow that most of the sound is reproduced by the panel. Especially, the area of my panels is larger than that of the NSS Final 0.3 increasing the capacity of the panels. This higher capacity is in principle transformed up by the step-up transformer. So, on the primary side of the transformer, a larger capacity is seen as well and the capacitor for the cross over had to be almost trippled to get a roll off below 400 Hz. Finally, I adjusted the acoustical damping of the enclosure to get a good response at low frequencies. Below you can see the result of measurements of the hybride using speaker workshop. The frequency response (black) is nicely flat between 70 and 20 kHz. The ripples are just of the order of +-2 dB. This is responsible for the very natural sound of the speakers. Notice that the response was measured ungated and outside on my terrace to give an accurate picture of the response (like in an anechoic chamber) and to minimize room resonances and sound reflections. The plotted response is the phase correct summation of the woofer and the ESL response measured at 10 cm distance (to avoid artifacts in the frequency response due to the size of the dipole emitter). You can also see that the phase (red) is bahaving quite well. Only at the cross over frequency a significant phase shift is accumulated but it is unavoidable in analogue filters. So, the low frequency part of the spectrum accumulates a 180 degrees phase shift. Also note that the microphone was not calibrated, i.e. some of the spectral and phase imperfections are also due to the microphone (though it is a very linear microphone).
A nice frequency response is only halve the rent. For an accurate two dimensional sound picture the phase and group velocity is important. To get these, the impulse and step responses can be measured with the MLS principle. Below, the impulse response of the ESL and the woofer are depicted. Note that the pulse was a negative one. In the ideal case, the speaker should deliver a delta function (here negative). The ESL (black) is close to that (notice the break in the y-axis!), shows a nice needle like response and has only a weak overshoot. The woofer (red) has more problems. Some high frequency emission caused by ringing vibrations of the Al cone (several kHz) are visible and also resonance of the enclosure at larger times and lower frequencies.
The impulse response shown above was measured at 10 cm distance. Luckily, it changes with distance and the cone resonances of the woofer become less important. Below you see the step response for the entire ESL measured at a distance of 1 m. To suppress the echo caused by reflections of one of the walls of my terrace, it was limited to about 3 msec. You can see the almost perfect response with a short rise time and the expected gradual decay due to the lower frequency limit of the speakers. Note that for times above about 0.6 msec, the signal is negative. This comes from the 180 degrees phaseshift caused by the cross over circuit. The low frequencies emitted by the woofer have the wrong phase. This is not serious as the low frequencies play only a minor role in the 2D sound reception.
All together, the response in frequency and in the time domain are excellent and for only about 400 Euro, I got a speaker that is truly high end - thanks to The Audio Circuit.