Multi-Channel Synth organization?

General JUCE discussion
1

I’m working on a multi-channel synth. I’m unsure of the general design flow with regard to SynthesiserVoice usage in these channels, which are owned by juce::Synthesiser. I’m doing this currently (extremely simplified):

struct Channel
{
    Channel(int ch, juce::Synthesiser& s) : channelNumber(ch), synth( &s ) { }
     /*
     all of the noteOn/Off/etc stuff is in here
    */
    void noteOn(int midiNoteNumber, float velocity);
private:
    int channelNumber = -1;
    juce::Synthesiser* synth = nullptr;
};

struct Synth : juce::Synthesiser
{
    void noteOn(int midiChannel, int midiNoteNumber, float velocity) override;
private:
     OwnedArray<Channel> channels;
     friend Channel;
};

Channel can access the voices member of Synthesiser like this:

void Synth::noteOn(int midiChannel, int midiNoteNumber, float velocity)
{
    //after checking that midiChannel is a valid index...
    channels[midiChannel]->noteOn(midiNoteNumber, velocity);
}

void Channel::noteOn(int midiNoteNumber, float velocity)
{
    //a bunch of filtering to find the right sample and 
    //available voice leads to this:
    auto* voice = synth->findFreeVoice(sound, 
                                       channelNumber, 
                                       midiNoteNumber, 
                                       synth->isNoteStealingEnabled());

    //after checking that a voice is free
    synth->startVoice( voice, sound, channelNumber, midiNoteNumber, velocity );
}

Is this the right way to go about handling things?

Or should I be doing a vector of Synthesiser instances and have each instance handle a single channel? I haven’t found much info on the best design choice for this.

I don’t need multi-channel output, so that’s why I’m keeping all the voices in a single synth, since the juce::Synthesiser class is designed to combine all active voices into a single stereo output buffer.