I think I should have used Java’s piped streams to interconnect the components of the synth, so you can go ahead and modify it to your liking.
The nice part about the synthesis part is that I optimized it by basically generating one period of subaudible frequency and then just selecting every n-th sample, depending on frequency.