LADSPA Reverb/Chorus Send: from mono to stereo

[md1872b]

Hello, I'm testing some custom LADSPA effects (both Win and Linux). CC91/CC93 work correctly and i'm happy.

However, the reverb and chorus send are mono.
I'm trying to extend this to provide LADSPA fx unit with reverb and chorus stereo send.

So I'd like to obtain something like this for LADSPA binding:
Main:L, MainR
Reverb:SendL, Reverb:SendR
Chorus:SendL, Chorus:SendR

Do you have any tips in exteding this?

I already saw I need to change the buffer for rvoice and adjusting the gain calculation for each buffer

#define FLUID_RVOICE_MAX_BUFS (6)

[derselbst]

@mawe42 is the LADSPA expert, but he seems to be busy.

So, yes the input channels used for reverb and chorus are mono currently. Even for fluidsynth's internal effects. In both cases, the fx unit is expected to create a stereo output signal. You may find more information here and some thoughts about improving this here.

Do you have any tips in exteding this?
I already saw I need to change the buffer for rvoice and adjusting the gain calculation for each buffer

You seem to be on the right track. Rather than using an ugly macro with hard-coded indexing for fluid_rvoice_buffers_set_amp and friends (as done e.g. here), one should think about defining an enum instead

typedef enum  {
DRY_L,
DRY_R,
REV_L,
REV_R,
CHO_L,
CHO_R,
RBUF_COUNT
}  fluid_rvoice_bufidx_t;

The interesting part will be to adjust the gain calculation of each buffer, here:

fluidsynth/src/synth/fluid_voice.c

Lines 805 to 817 in 5dcae73

	case GEN_REVERBSEND:
	/* The generator unit is 'tenths of a percent'. */
	voice->reverb_send = x / 1000.0f;
	fluid_clip(voice->reverb_send, 0.f, 1.f);
	UPDATE_RVOICE_BUFFERS_AMP(fluid_rvoice_buffers_set_amp, 2, fluid_voice_calculate_gain_amplitude(voice, voice->reverb_send));
	break;
	case GEN_CHORUSSEND:
	/* The generator unit is 'tenths of a percent'. */
	voice->chorus_send = x / 1000.0f;
	fluid_clip(voice->chorus_send, 0.f, 1.f);
	UPDATE_RVOICE_BUFFERS_AMP(fluid_rvoice_buffers_set_amp, 3, fluid_voice_calculate_gain_amplitude(voice, voice->chorus_send));
	break;

One might think about changing it to something like

// left reverb buffer
fluid_voice_calculate_gain_amplitude(voice, fluid_pan(voice->pan, 1) * fluid_balance(voice->balance, 1) * voice->reverb_send);

Similar for the right buffer, just like we have done for the CUSTOM_BALANCE generator just a few lines above.

More gain logic also lives here, not sure if there are other places which I'm missing...

fluidsynth/src/synth/fluid_voice.c

Lines 1848 to 1873 in 5dcae73

	int fluid_voice_set_gain(fluid_voice_t *voice, fluid_real_t gain)
	{
	fluid_real_t left, right, reverb, chorus;
	/* avoid division by zero*/
	if(gain < 0.0000001f)
	{
	gain = 0.0000001f;
	}
	voice->synth_gain = gain;
	left = fluid_voice_calculate_gain_amplitude(voice,
	fluid_pan(voice->pan, 1) * fluid_balance(voice->balance, 1));
	right = fluid_voice_calculate_gain_amplitude(voice,
	fluid_pan(voice->pan, 0) * fluid_balance(voice->balance, 0));
	reverb = fluid_voice_calculate_gain_amplitude(voice, voice->reverb_send);
	chorus = fluid_voice_calculate_gain_amplitude(voice, voice->chorus_send);
	UPDATE_RVOICE_R1(fluid_rvoice_set_synth_gain, gain);
	UPDATE_RVOICE_BUFFERS_AMP(fluid_rvoice_buffers_set_amp, 0, left);
	UPDATE_RVOICE_BUFFERS_AMP(fluid_rvoice_buffers_set_amp, 1, right);
	UPDATE_RVOICE_BUFFERS_AMP(fluid_rvoice_buffers_set_amp, 2, reverb);
	UPDATE_RVOICE_BUFFERS_AMP(fluid_rvoice_buffers_set_amp, 3, chorus);
	return FLUID_OK;
	}

[md1872b]

Thank you, I was almost close.
Now I'm facing some issues with the extension of fluid_rvoice_mixer in order to support the new buffers.
That's somewhat tricky.

[md1872b]

Got it, work as expected.

The issue was I didn't understand that I should multiply by 2 the fx_buf_count on the return of fluid_mixer_buffers_prepare, something like return offset + 2 * buffers->fx_buf_count

[derselbst]

Glad to hear! I opened a PR (#1170) and will test it myself once I find the time.