For those who haven't read the threads in which it has been mentioned, I have been playing with a rudimentary plasma speaker design. I've recently decided to shelve the project due to some real safety concerns as well as the fact these don't sound very good without a LOT of work and custom components (namely the high voltage transformer). I've decided to document what I did here.
The system I used was a very simple Pulse Width Modulation amplifier that drove a high voltage transformer ripped from an electronic neon sign power supply. This is not one of the huge 60Hz high voltage sign transformers of yesteryear but a more modern switching device. It specifically was removed from an Evertron 2610 ordered from Neon Central.
This transformer has a single primary wound out of enameled copper wire driven by a half-bridge N channel FET system. The FET's ran off of line voltage through a voltage doubler rectifier to send a +338VDC square wave to the primary. The transformer also had two secondaries - the high voltage secondary, and a much lower voltage feedback secondary. The HV secondary was rated for 5.5kVrms @ 27mA but higher voltage was easily achieved once removed from its original power supply.
The pulse width modulator I used was an extremely simple analog design. Everything ran from from a regulated +15V supply. A triangle wave generator (schematic available in National Instruments application note AN-31.pdf in the signal generator section) was made using OPA2132 op amps and compared using another OPA2132 as a comparator to an analog input signal AC coupled into the single supply system to generate the PWM waveform. A true comparator would have been better, but due to the low speeds it didn't really matter.
This low-level PWM signal was send to an L6384 half bridge driver which fed two N-channel MOSFETs rated at 250V. I supplied the half bridge with rectified and filtered line voltage (169VDC). A DC blocking capacitor was put on the output of the half-bridge to prevent saturating the transformer.
The transformer in question had a strong resonance at approximately 33kHz. The PWM frequency was chosen to drive the transformer at this frequency with the PWM amplifier. This method was expected to be very very nonlinear for two reasons I won't go into here, and indeed it was. Tones fed into it with a tone generator elicited quite distorted sound, but the fundamental frequency was correct. I could run it to 650Hz before the arc would fail.
This method will never work for high quality sound. It was a simple test to see if I could really do this. My second idea for driving this is likely to be much more successful in terms of linearity, and if anyone is still reading and wants to play with this stuff I think it's a novel idea.
Since the transformer exhibited a strong high-Q resonance @ 33kHz, it could be operated at a bias point just beneath that (say 30kHz) while still maintaining a good thermal plasma arc. Using FM modulation, one could shift the drive frequency of the half-bridge switching amp according to an input audio signal. This would sweep the drive frequency from 27 to 33kHz, and generate your max 'peak-to-peak' output amplitude. This is bound to be better than the first method, but it still will be quite limited in the high frequency bandwidth. Some may remember my interest in the XR2206 function generator chip a few weeks back. One of the reasons I was looking into it was its built-in FM modulation capability. Some old-timers may recognize that this is essentially how an old 'super re-gen' tuner functions, but in reverse and with much higher power.
There is further capability here available with the feedback available from the second low voltage secondary winding. I never characterized its operation, but I expect that some form of feedback could be applied using it to further linearize the system.
Below are images of the triangle wave, a resulting low level PWM signal, and the output of the (totally unloaded and undamped) half-bridge.
The system I used was a very simple Pulse Width Modulation amplifier that drove a high voltage transformer ripped from an electronic neon sign power supply. This is not one of the huge 60Hz high voltage sign transformers of yesteryear but a more modern switching device. It specifically was removed from an Evertron 2610 ordered from Neon Central.
This transformer has a single primary wound out of enameled copper wire driven by a half-bridge N channel FET system. The FET's ran off of line voltage through a voltage doubler rectifier to send a +338VDC square wave to the primary. The transformer also had two secondaries - the high voltage secondary, and a much lower voltage feedback secondary. The HV secondary was rated for 5.5kVrms @ 27mA but higher voltage was easily achieved once removed from its original power supply.
The pulse width modulator I used was an extremely simple analog design. Everything ran from from a regulated +15V supply. A triangle wave generator (schematic available in National Instruments application note AN-31.pdf in the signal generator section) was made using OPA2132 op amps and compared using another OPA2132 as a comparator to an analog input signal AC coupled into the single supply system to generate the PWM waveform. A true comparator would have been better, but due to the low speeds it didn't really matter.
This low-level PWM signal was send to an L6384 half bridge driver which fed two N-channel MOSFETs rated at 250V. I supplied the half bridge with rectified and filtered line voltage (169VDC). A DC blocking capacitor was put on the output of the half-bridge to prevent saturating the transformer.
The transformer in question had a strong resonance at approximately 33kHz. The PWM frequency was chosen to drive the transformer at this frequency with the PWM amplifier. This method was expected to be very very nonlinear for two reasons I won't go into here, and indeed it was. Tones fed into it with a tone generator elicited quite distorted sound, but the fundamental frequency was correct. I could run it to 650Hz before the arc would fail.
This method will never work for high quality sound. It was a simple test to see if I could really do this. My second idea for driving this is likely to be much more successful in terms of linearity, and if anyone is still reading and wants to play with this stuff I think it's a novel idea.
Since the transformer exhibited a strong high-Q resonance @ 33kHz, it could be operated at a bias point just beneath that (say 30kHz) while still maintaining a good thermal plasma arc. Using FM modulation, one could shift the drive frequency of the half-bridge switching amp according to an input audio signal. This would sweep the drive frequency from 27 to 33kHz, and generate your max 'peak-to-peak' output amplitude. This is bound to be better than the first method, but it still will be quite limited in the high frequency bandwidth. Some may remember my interest in the XR2206 function generator chip a few weeks back. One of the reasons I was looking into it was its built-in FM modulation capability. Some old-timers may recognize that this is essentially how an old 'super re-gen' tuner functions, but in reverse and with much higher power.
There is further capability here available with the feedback available from the second low voltage secondary winding. I never characterized its operation, but I expect that some form of feedback could be applied using it to further linearize the system.
Below are images of the triangle wave, a resulting low level PWM signal, and the output of the (totally unloaded and undamped) half-bridge.
Comment