UART_RX serial receiver (from https://www.nandland.com/vhdl/modules/module-uart-serial-port-rs232.html) receives MIDI data and passes it to MIDI_IN block that decoded note on and off events, along with the note number and velocity (which is currently not used).
Note_On and Note_Off pulses that last for single clock are passed to the Voice blocks. Note_On is connected to the very first voice block only. All other voices are daisy-chained via Note_On_Fwd to Note_On links. When a voice is busy (i.e. it's been triggered and is currently playing) it forwards the Note_On signal to its Note_On_Fwd output. This way multiple voices can be triggered acheiving the polyphony.
Stochastic outputs from the voices are then mixed together via SC_Mux stochastic multiplexers/adders for left and right channels.
A single voice is composed of three sawtooth stochastic oscillators, which produde stereo output each. The ferqiencies for the oscillators are provides by the MIDI_Note2Freq block, which is a look-up table that translates MIDI note number to frequency value for current design (100MHz clock 16-bits counters). Three frequencies produced are slightly detuned for each of the three oscillators.
The Voice_Trigger block received Note_On and Note_Off signals and transforms then into Trigger output which stays at 1 when voice has been triggered and playing, and goes to 0 on note off. In case of the Busy input being active the voice is not triggered, but the Note_On gets passed to Forward output.
Triggering the voice starts the envelope generator Voice_Envelope, which is a stochastic integrator integrating up until saturation (attack phase), and then down to bipolar stochastic zero (or 0.5 unipolar) on trigger input going down.
Outputs from the three oscillators are mixed together and the results are multiplied with the envelope's output using stochastic bipolar multiplier (XNOR gate).
A single oscillator is composed to two identical chains for setero output.
The first B2SC block turns the input frequency (16 bit unsigned integer) to an equivalent stochastic signal, which is then gets integrated by SC_Integrator. Upon saturation the integrators gets reset, thus producing a sawtooth output. The binary output from the integrator is then converted to stochastic signal and shaped with the SC_Self_OR block, which perform OR operation on the signal and its slightly delayed version. This gives concave shape to the sawtooth signal produced (1 - (1 - x)^2 function).