What do I need to know to start programming a synth in c++/ other language?

Question

I'm still a beginner, so this will be a pretty advanced project for me, but I want to start it now and have it as an ongoing thing (I don't intend to crank out a finished product in 2 weeks, nor even to build something I will use, just a learning experience).

I want to build a synth to be used as a vst3 plugin for ableton live 9. I was thinking that the first step would be to just make something that outputs an arbitrary waveform and responds to midi input (the frequency changes with regard to the note) and a basic GUI template. Then I would try to introduce different waveforms, followed by more oscillators and detuning, followed by voices, followed by filters, followed by multiple voices etc.

How would I start doing this? I know basic c++, but I don't for instance know how to create a waveform and output it as sound. Is c++ a good language to use for this type of application?

Answer 1

Though your question is quite broad I can give some insight into where to start.

First, you will need a copy of the VST3 SDK from Steinberg

From my experience, there is quite a large learning curve for getting started with the VST SDK. So I would suggest finding a library that provides a wrapper on top of it. For example JUCE

Using a wrapper will help you get past a lot of the boilerplate programming that needs to be done to make a VST work.

As far as waveform generation, filtering, and other DSP related concepts... here is so much to the subject that I couldn't even begin to describe it.

Take a look at musicdsp and The STK for some basic concepts.

Get a book on the subject. A classic would be The Audio Programming Book

Additionally you need to make sure that you have a strong grasp of audio theory. Take a look at Introduction to Computer Music: Volume One.

And of course, Google is your friend.

EDIT:

To answer your question more completely. Yes C++ (or C) would be the language of choice for this kind of application (though not the only possible choice)

I would look into using an Audio API before delving into VST development, this will allow you to work on your skills without the trouble of VST development.

As for Audio Libraries, ehese are two are popular options:

• PortAudio (written in C)

• RtAudio (written in C++, but not c++11 or higher)

  Or you could take a look at libzaudio Which is an Audio API that I am currently working on. (It currently depends on PortAudio, but provides a more modern C++11 style way of working with audio)

Assuming you have installed PortAudio & libzaudio the following will generate a sin 440hz wave for one second:

#include <iostream>
#include <cmath>
#include <libzaudio/zaudio.hpp>

int main(int argc, char** argv)
{
    try
    {
        //bring the needed zaudio components into scope
        using zaudio::no_error;
        using zaudio::sample;
        using zaudio::sample_format;
        using zaudio::stream_params;
        using zaudio::time_point;
        using zaudio::make_stream_context;
        using zaudio::make_stream_params;
        using zaudio::make_audio_stream;
        using zaudio::start_stream;
        using zaudio::stop_stream;
        using zaudio::thread_sleep;

        //create an alias for a 32 bit float sample
        using sample_type = sample<sample_format::f32>;

        //create a stream context with the default api (portaudio currently)
        auto&& context = make_stream_context<sample_type>();

        //create a stream params object
        auto&& params = make_stream_params<sample_type>(44100,512,0,2);

        //setup to generate a sine wave
        constexpr sample_type _2pi = M_PI * 2.0;
        float hz = 440.0;
        sample_type phs = 0;
        sample_type stp = hz / params.sample_rate() * _2pi;

        //create a zaudio::stream_callback compliant lambda that generates a sine wave
        auto&& callback = [&](const sample_type* input,
                              sample_type* output,
                              time_point stream_time,
                              stream_params<sample_type>& params) noexcept
        {
            for(std::size_t i = 0; i < params.frame_count(); ++i)
            {
                for(std::size_t j = 0; j < params.output_frame_width(); ++j)
                {
                    *(output++) = std::sin(phs);
                }
                phs += stp;
                if(phs > _2pi)
                {
                    phs -= _2pi;
                }
            }
            return no_error;
        };

        //create an audio stream using the params, context, and callback created above. Uses the default error callback
        auto&& stream = make_audio_stream<sample_type>(params,context,callback);

        //start the stream
        start_stream(stream);
        //run for 1 second
        thread_sleep(std::chrono::seconds(1));
        //stop the stream
        stop_stream(stream);
    }
    catch (std::exception& e)
    {
        std::cout<<e.what()<<std::endl;
    }
    return 0;
}

(Pulled from one of my example files)

I would be happy to explain more in detail if you would like to contact me elsewhere. (A long discussion on the topic would not be permitted on Stack Overflow)