Pitch-to-MIDI Converter

Design Unknowns

We have practical knowledge of A/D audio interface, but no experience with the particular piece of hardware that will perform the A/D conversion. In fact, we don't yet know the details of the device (sample size and rate, S/N ratio, etc.) Determining these factors will be important in designing the project because it will directly effect how we implement FFT. Specifically: will we have enough processing power to deal with the samples as they come in so that we can approximate a real-time digital signal processor?

The FFT algorithm is essentially brand new material for us. At present we are researching different implementations of the algorithm. Tentatively, we will use a decimation in the time domain method with 8-sample chunks of data (8- or 16-bit -- if we use the SRAM, 8-bit might be easier to deal with); several computational steps will be performed in parallel (FFT looks well-suited to a hardware implementation due to it's inherent parallelism). Sine and cosine values will probably need to be stored in look-up tables.

We have some experience with MIDI (Musical Instrument Digital Interface), but we haven't decided on what MIDI information to transmit; the MIDI standard includes many options (such as pitch-bend, note-velocity, and a variety of "controller" variables) that we may or may not implement in our project. This will depend upon currently unknown variables, including quality of the FFT output, memory and processing power available in the microcontroller, and the amount of complexity we have time to include in the next seven weeks.