Mixing Audio Input with effect

[bill-g4aep]

Hello,
I'm trying out ideas for a guitar pedal. I am trying to use OutputMixer to mix an effect (eg pitch shift) with a proportion of the clean signal. I am trying MultiOutput to split the audio source stream, and directing one output to the mixer and one through the effects, then to the mixer. It compiles and runs, but the output is not cleanly mixed, but has a buzz to it. If I disable either mixer input, and make the OutputMixer no of outputs = 1, it seems clean.
I have read through the docs that I can find. Is there a simpler or better way to do this?
I have tried using a generated sinewave input, as per your example, and it works fine until I introduce effects into one side of the mixer.
I'm using ESP32 WROOM module.
I appreciate your ideas.
Bill

mixer.txt

[pschatzmann]

There is quite a lot going on and it is diffcult to say what the issue is.
My first guess might be that you are asking for too much: I suggest to reduce the channels to 1 and you maybe try to lower the sample rate?

It would also be a bit more efficient if you get rid of the VolumeStream and just rely on adjusting weights of the mixer instead.

Finally I suggest to comment out the effects.addEffect(fuzz); so that you can check if the mixing of the clean signal via the 2 paths is working at all since the current setup relys on quite a lot of magic.

[pschatzmann]

I was playing around with your sketch, and I know now what's going wrong: the OutputMixer exptects identical write sizes of the different inputs: unfortunately it is not really defined how each component submits it's samples to the output: this could be in individual samples! This can be fixed by adding a BufferedStream into the chain which makes sure that the data is written only when the buffer size is reached!

Here is the adampted example that works for a LyraT:

#include "AudioTools.h"
#include "AudioTools/AudioLibs/AudioSTK.h"
#include "AudioTools/AudioLibs/AudioBoardStream.h"

AudioInfo info(44100, 2, 16);
AudioBoardStream i2s(LyratV43);  // or use I2SStream
MultiOutput split;
OutputMixer<int16_t> mixer(i2s, 2);  // output mixer with 2 outputs mixing to AudioBoardStream
BufferedStream effects_buffer(1024, mixer);
BufferedStream direct_buffer(1024, mixer);
AudioEffectStream effects(effects_buffer);
StreamCopy copier(split, i2s, 1024);  // copy i2sStream to csvStream
Fuzz fuzz(6.5, 500);
STKChorus chorus(6000);
PitchShift pitch(0.5);
Boost boost(1.0);

// Arduino Setup
void setup(void) {
  Serial.begin(115200);
  AudioToolsLogger.begin(Serial, AudioToolsLogLevel::Warning);
  
  // setup multioutput
  split.add(effects);
  split.add(direct_buffer);
  
  // setup effects
  //effects.addEffect(fuzz);
  //effects.addEffect(chorus);
  //effects.addEffect(boost);
  //effects.addEffect(pitch);
  effects.begin(info);

  auto cfg = i2s.defaultConfig(RXTX_MODE);
  cfg.copyFrom(info);
  cfg.sd_active = false;
  cfg.input_device = ADC_INPUT_LINE2;
  cfg.buffer_size = 256;
  cfg.buffer_count = 3;
  i2s.begin(cfg);
  i2s.setVolume(1.0);

  //mixer.setWeight(0, 1.0);
  //mixer.setWeight(1, 0.0);
  mixer.begin(1024);
}

// Arduino loop - copy data
void loop() {
  copier.copy();
}

ps. I have the gut feeling that the direct_buffer is not really necessary and could be omitted. I reduced the buffer size in i2s to minimize the lag. It would also be a good idea to replace the 1024 which are scattered across the sketch with a constant and maybe reduce the value a bit.

[bill-g4aep]

Hello, and thanks for your very swift response, amazing!
I have successfully tried your suggestions, and now I think I have the basis for further experimentation. I have noticed a few things:

Latency or delay does seem rather high. I reduced buffer size to 256 and this helped, I think. But this may effect the pitchshift (see later).

If I add STKChorus to the sketch (even a simple one), I get this message:
Error: resize: Not enough memory for 8487 frames
... although it then works OK after. Do I need to consider RAM usage or memory model? Does STK require more resources

I'm trying to create an 'Octaver' effect with down and/or up octaves, using pitchshift (0.5) or (2.0) to create the octave signal. With down octave I get the warning:
[W] PitchShift.h : 376 - phase allign failed: maybe the buffer is too small
I'm using bass guitar input, so I'm guessing that with small buffers, low frequencies may not work well. Overall pitchshift seems to be a bit unstable particularly with low frequencies. I'll experiment more, but wonder if there might be a better way to do the octaver effect. Need more research.

I probably only need one channel, but I haven't quite got a single channel to work. Maybe its in my hardware I'm using the usual WM8782 ADC and PM5102 DAC. I might swap to a LyraT in the future.

This is just a fun project. In mind, I have something like the Line6 HX Stomp box, though rather simpler.

Again, thanks

[pschatzmann]

I haven't looked into the implementation of the STK effects.
If you are running out of RAM you can try to use PSRAM with this trick

The lag is most likely caused by the i2s buffers, so try to keep them low....