We had a 'Magnat' plasma tweeter in work for a while, which was used in scale-modelling for auditoria design (before numerical methods really took over, and instead of a spark source which was the other way to do it via impulse response and Fourier transform). It smelt funny
Scale modelling using ultrasonic signals is fun - if you want to try to scale the speed of sound (a bit) you can fill them with nitrogen. Well, I guess you can fill them with other gasses too, but sometimes not with a spark source