Resampling and Mixing Input and Output #556
-
|
I have tried the examples in resampling and mixing Inputs and Outputs from the Wiki documentation. The only examples I can fing is resampling only of single waves or mixing of waves that are not resampled. I want to be able to have 2 or more arrays generate 2 or more sounds each one resampled so that I can vary the frequency and also the amplitude or volume. I assume then I need them mixed so as to play out via the DAC pin25 and pin26 of the ESP32. I am able to do the "Input Mixer" with each wave resampled and it plays quite well, but I find there is a slight click of about 1 Hz at some frequencies, I think when multiplier is a very odd number eg 1.3333, or 1.555, or 1.7777 for example, other multipliers eg. 1.5, 2.0, 3.0 etc are OK and good sound. Below is example of what I have been trying.... xxxxxxxxxxxxxxxxxxx This gives good sound but some frequencies eg. uint16_t sample_rate=44100; GeneratorFromArray<int16_t> waveR1(arrayR1,0,false); AnalogAudioStream out; // out is DAC pin 25 and 26 analog InputMixer<int16_t> mixer; StreamCopy copier1(out, mixer); // copies mixer sound to out //OutputMixer<int16_t> mixer(out, 2); // Arduino Setup mixer.add(resample1); // add resampled sounds as input to mixer xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx This is resampling of Output Example...... gives a better clear sound. GeneratorFromArray<int16_t> waveR1(arrayR1,0,false); ResampleStream<int16_t> resample1(out); // this gives a clear sound playing copier1 only //xxxxxxOutputMixer<int16_t> mixer(out,2); StreamCopy copier1(resample1, sndR1); // copies sound to resample and out //OutputMixer<int16_t> mixer(out,2); //StreamCopy copier1(mixer, resample1); // Arduino Setup |
Beta Was this translation helpful? Give feedback.
Replies: 19 comments 87 replies
-
|
I have looked at all combinations of resampling and mixing I can find available and still cannot find a solution to the clicking background sounds when the "real resampled sounds" are playing. Using an OutputMixer to output direct to "out" and Resample the GeneratedSoundStream of each array then copy these into mixer has the clicks at some frequencies. I can play 1 sound or both sounds still get clicks. Using an InputMixer with " StreamCopy copier1(out, mixer); " to combine the 2 resampled streams and copier the mixed streams in loop means I can play 2sounds but still get clicks. The best option for sound quality is using Resampled direct "out" but I need to mix my 2 sound somewhere....?? xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx uint16_t sample_rate=44100; GeneratorFromArray<int16_t> waveR1(arrayR1,0,false); AnalogAudioStream out; //ResampleStream<int16_t> resample1(sndR1); // not send this to avoid interference ResampleStream<int16_t> resample2(sndH1); //resample sndH1 only send this to test OutputMixer<int16_t> mixer(out, 1); // define outputmixer with direct output out. had to set channels to 1 here, if set to 2 gettingcombination of 2 waveforms mixed up. StreamCopy copier1(mixer, resample2); // copier1 copies resample2 sound into mixer .... this works but has a click sound at some frequencies of resampling //StreamCopy buffer_size(128); //copier.buffer_size = 128); ..... size of buffer seems to not matter here xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx This below is resample to "out" Only. no mixer ... good sound all frequencies xxxxxxxx uint16_t sample_rate=44100; GeneratorFromArray<int16_t> waveR1(arrayR1,0,false); AnalogAudioStream out; ResampleStream<int16_t> resample1(out); // does this mean resample direct to out ...?? //AnalogAudioStream out; StreamCopy buffer_size(128); //copier.buffer_size = 128); // This works perfect (NO Clicks) when copier1 only or copier2 only sent out in loop eg. ... xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx uint16_t sample_rate=44100; GeneratorFromArray<int16_t> waveR1(arrayR1,0,false); AnalogAudioStream out; InputMixer<int16_t> mixer; StreamCopy copier1(out, mixer); StreamCopy buffer_size(128); //copier.buffer_size = 128); void setup(void) { mixer.add(resample1, 0.5); // 0.5 amplitude volume settings R1 // ................xxxx etc etc void loop() { //xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx |
Beta Was this translation helpful? Give feedback.
-
|
I think you would need to find out the root cause of your clicking sounds. Did you try to send the output to the Serial Printer to figure out where this is coming from ? |
Beta Was this translation helpful? Give feedback.
-
|
If you feed a sine wave as input, the result should also be a nice continuous sine wave. |
Beta Was this translation helpful? Give feedback.
-
|
"If you feed a sine wave as input, the result should also be a nice continuous sine wave"....... Yes, I get that, and it works. |
Beta Was this translation helpful? Give feedback.
-
|
if mixing with sine waves are working as expected the issue is with your specific input. The clicks should be visible if you print the output. For mixing on the output side there is an example in the corresponding Wiki: you just write all outputs one after each other that need to be mixed to the output mixer. |
Beta Was this translation helpful? Give feedback.
-
|
I have tried resampling and mixing with sinewave arrays and get the same thing. Some frequencies mix OK and others cause "glitches" for same waves and settings only changes to stepsize....... And I wasn't trying many freqs before so it is worse than I thought. 00:07:29.431 -> 1057 xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx Screenshot 1.jpg ... resample1.setStepSize(1.0); .... resample2.setStepSize(1.2); Screenshot 2.jpg ... resample1.setStepSize(1.0); .... resample2.setStepSize(1.0); Screenshot 3.jpg ... resample1.setStepSize(0.333); .... resample2.setStepSize(1.0); Screenshot 4.jpg ... resample1.setStepSize(0.25); .... resample2.setStepSize(1.0); Screenshot 5.jpg ... resample1.setStepSize(0.25); .... resample2.setStepSize(1.33); The 2 arrays are actually copies of same 256 point, int16_t arsineC256 and int16_t arsineD256. // Sketch to test for gliches at various freqs when Mixing Resampled waves... Output Mixer ........... uint16_t sample_rate=44100; GeneratorFromArray<int16_t> waveA1(arsineC256,0,false); CsvStream<int16_t> out(Serial); ResampleStream<int16_t> resample1(sndA1); OutputMixer<int16_t> mixer(out,2); StreamCopy copier1(mixer, resample1); // copies resampler to mixer and out // Arduino Setup Serial.begin(115200); // Define CSV Output auto rcfg1 = resample1.defaultConfig(); // define resample2 /* mixer.begin(); // resample1.setStepSize(1.0); // change stepsize to change freq waveA1 } // Arduino loop - copy sound to out resample1.setStepSize(1.0); // change stepsize to change freq waveA1 copier1.copy(); }
|
Beta Was this translation helpful? Give feedback.
-
|
I didn't think all the jpg files would have merged as they did on last posting. I thought they would have had at least my file name.or I would have put a title or description on the jpg pic itself but look like they are listed in the post, top to bottom,........ |
Beta Was this translation helpful? Give feedback.
-
|
I tried to reproduce your issue with the 1.0/1.2 combination in Jupyterlab : int channels = 1;
int sample_rate = 44100;
int frequency = 800;
// data
SineWaveGenerator<int16_t> sineWave1(32000);
SineWaveGenerator<int16_t> sineWave2(32000);
GeneratedSoundStream<int16_t> sound1(sineWave1);
GeneratedSoundStream<int16_t> sound2(sineWave2);
ResampleStream<int16_t> resample1(sound1);
ResampleStream<int16_t> resample2(sound2);
InputMixer<int16_t> mixer;
// setup mixer
mixer.add(resample1);
mixer.add(resample2);
// begin objects
sineWave1.begin(channels, sample_rate, frequency);
sineWave2.begin(channels, sample_rate, frequency);
resample1.begin(1.0);
resample1.begin(1.2);
mixer.begin(sound1.audioInfo());This gives the following result / and input;
The major difference seems to be that you load the data from an array and I use a continuous generator. Are you sure that you provide 1 full/complete waves from the beginning to the end for both inputs ? |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for reply, Yes, your reply uses the sinewave generator. I actually won't be using sinewaves but used them in my last posts to show the problem still exists with sine arrays, OR else there is something I am doing wrong in my sketch or my Audacity generating the sine array. You also use an InputMixer, my last pot was an Output Mixer. Just wondering can anyone try my sketch and see do they get same results for a sine array.? |
Beta Was this translation helpful? Give feedback.
-
|
The mixer is dividing the sample by the number of mixed channels. So you don't need to worry that the mixed output is clipped or ovefflowing. I suggest that you create a Gist with your data so that I can try to reproduce the result with your input data |
Beta Was this translation helpful? Give feedback.
-
|
This is tests using same array as before but using Input Mixer. // Sketch to test for gliches at various freqs when Mixing Resampled waves... Input Mixer ........... #include <AudioConfig.h> uint16_t sample_rate=44100; GeneratorFromArray<int16_t> waveA1(arsineC256,0,false); //CsvStream<int16_t> out(Serial); ResampleStream<int16_t> resample1(sndA1); InputMixer<int16_t> mixer; StreamCopy copier1(out, mixer); // copies resampler to mixer and out // Arduino Setup /* mixer.add(resample1); /* auto rcfg1 = resample1.defaultConfig(); // define resample2 // start I2S // Setup sine wave // mixer.begin(); resample1.setStepSize(1.0); // change stepsize to change freq waveA1 } // Arduino loop - copy sound to out //resample1.setStepSize(1.0); // change stepsize to change freq waveA1 copier1.copy(); }
|
Beta Was this translation helpful? Give feedback.
-
|
What is the difference between arsineC256 and arsineD256 ? |
Beta Was this translation helpful? Give feedback.
-
|
One array is good enough. You don't need to use 2. Yes please provide the data so that I can double check if I get the same issue with your data... |
Beta Was this translation helpful? Give feedback.
-
|
This is the SoundX.h file. I had to change it to .txt |
Beta Was this translation helpful? Give feedback.
-
|
I can reproduce the issue and after a closer look I am also getting this with the SineWaveGenerator. It is just much less frequent then in your examples.
Right now I have no idea what could be the root cause. I will have a closer look at this tomorrow... |
Beta Was this translation helpful? Give feedback.
-
|
OK, thanks for trying it out. At least we are both on the "same wavelength"..! ! |
Beta Was this translation helpful? Give feedback.
-
|
archrp256.txt |
Beta Was this translation helpful? Give feedback.
-
|
I thought this to be quite obvious: if you provide mono data the number of channels is 1 and if you provide stereo the channels is 2. See https://github.com/pschatzmann/arduino-audio-tools/wiki/Introduction You might get away with this (w/o samples being lost) but with the impact that the frequency is doubled, but only if number of samples is a multiple of the number of channels. ps. The mixer is mixing all channels from all input sources. Therefore if you provide a mono input you get the same signal on both I2S output channels. |
Beta Was this translation helpful? Give feedback.
-
|
For some strange reasons the readBytes callbed by the input mixer did not get all requested bytes from time to time. I tested both with SineWaveGenerator and GeneratorFromArray and confirm that it is working w/o unwanted artifacts now... |
Beta Was this translation helpful? Give feedback.
-
|
Yes, the glitches have gone. Thanks very much. Is there a limit now on the lower stepsize. I don't get any output if stepsize is less than 0.5. |
Beta Was this translation helpful? Give feedback.
-
|
Hmm, I don't see any reason why value < 0.5 should not work. |
Beta Was this translation helpful? Give feedback.
-
|
With the latest corrections the accuracy should be better now and it should work with values < 0.5 |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for all your work. The glitches have gone and the stepsize lt 0.5 seems sorted but problems still with some step sizes not being detected and stream continues at its last pitch until it comes to a step it can work with. See attached jpg file. There is also large variations in amplitude when some steps occur and amplitude is not always consistant...??. This sketch below is what I used to test the problems I have found. I prints what stepsize is playing to Serial and outputs Audio to DAC pins 25 and 26. Regards the earlier comments about channel numbers, even with channels=1 I always get a mixed signal of both arrays out of each 25 and 26 and I am recording both on Audacity.......That is why I am confused. Most examples indicate 2 channels and most are Csv out, So is I2S channels always 1 when sending to DAC internal and would it be 2 if sending to external I2S. Also is it possible to get stereo separate L and R with Audio tools library...? I can get it with Maximilian..??
// wiki example Input mix but with resampling added #include <AudioConfig.h>
#include <AudioLogger.h>
#include <AudioTools.h>
#include "SoundX.h"
const int pinPico[] = {27, 14, 12, 13, 15, 2, 32, 33};
// un-comment the line you want to test.... This is after glitch and 0.5 stepsize fix..... still some problems
// float stepArray[]={1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5, 5.0};
float stepArray[]={1.0, 1.125, 1.25, 1.333, 1.5, 1.667, 1.75, 2.0, 2.125}; // this line was running for jpg file attached to this post.
// float stepArray[]={1.0, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8};
// float stepArray[]={1.0, 1.1, 1.2, 1.3, 1.4, 2.0, 2.1, 2.2, 2.3};
// float stepArray[]={0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9};
float stepSamp1 = 0.4; //defaults step for wave1
float stepSamp2 = 1.0;
int listp = 0;
unsigned long a = millis();
unsigned long b = millis();
// Sketch to test for gliches at various freqs when Mixing Resampled waves... Input Mixer ...........
uint16_t sample_rate=44100;
uint8_t channels = 1; // set channels at 2 for stereo. set channels at 1 for mono
GeneratorFromArray<int16_t> waveA1(archrp256,0,false);
GeneratedSoundStream<int16_t> sndA1(waveA1); // Stream generated from array1
GeneratorFromArray<int16_t> waveB1(archrp256,0,false);
GeneratedSoundStream<int16_t> sndB1(waveB1); // Stream generated from array2
//CsvStream<int16_t> out(Serial);
AnalogAudioStream out;
ResampleStream<int16_t> resample1(sndA1);
ResampleStream<int16_t> resample2(sndB1);
InputMixer<int16_t> mixer;
StreamCopy copier1(out, mixer); // copies resampler to mixer and out
StreamCopy buffer_size(128); //copier.buffer_size = 128);
// Arduino Setup
void setup(void) {
Serial.begin(115200);
// AudioLogger::instance().begin(Serial, AudioLogger::Warning);
// AudioLogger::instance().begin(Serial, AudioLogger::Info);
mixer.add(resample1);
mixer.add(resample2);
/*
// Define CSV Output
auto config = out.defaultConfig();
config.sample_rate = sample_rate;
config.channels = channels;
out.begin(config);
*/
// define resample1
auto rcfg1 = resample1.defaultConfig();
rcfg1.step_size = stepSamp1; // wave1 config play at default rate
rcfg1.channels = channels;
resample1.begin(rcfg1);
// define resample2
auto rcfg2 = resample2.defaultConfig();
rcfg2.step_size = stepSamp2; // wave2 config play at default rate
rcfg2.channels = channels;
resample2.begin(rcfg2);
// start I2S
auto config = out.defaultConfig(TX_MODE);
config.sample_rate = sample_rate;
config.channels = channels;
config.bits_per_sample = 16;
config.buffer_count = 2;
config.buffer_size = 512;
out.begin(config);
// Setup sine wave
auto cfg1 = waveA1.defaultConfig();
auto cfg2 = waveB1.defaultConfig();
cfg1.channels = channels;
cfg1.sample_rate = sample_rate;
cfg2.channels = channels;
cfg2.sample_rate = sample_rate;
waveA1.begin(cfg1);
waveB1.begin(cfg2);
// mixer.begin();
// resample1.setStepSize(1.0); // change stepsize to change freq waveA1
// resample2.setStepSize(1.125); // change stepsize to change freq waveB1
}
// Arduino loop - copy sound to out
void loop() {
copier1.copy();
playlist();
}
void playlist(void) {
b = millis();
if (listp >= 9) listp=0;
if ((b - a) >= 1000) {
resample2.setStepSize(stepArray[listp]);
Serial.print("Playing sound ... ");
Serial.println(listp);
listp++;
a = millis();
}
} |
Beta Was this translation helpful? Give feedback.
-
|
I added a new InputMerge class which (does not mix but) interleaves the samples from the defined input streams. With this you can create a stereo output (channels=2) from 2 mono inputs (channels=1). The API is the same like the InputMixer. Please note, that I did not have the time to test the functionality yet.... |
Beta Was this translation helpful? Give feedback.
-
|
OK, It sounds like another audio processing option, would be interesting for some situations. |
Beta Was this translation helpful? Give feedback.
-
|
It would really help to draw some nice pictures to explain how the mixing is working. Here is a simple expanation: When 1 channel is sent to I2S, I automatically convert it to stereo duplicating the samples (because I2S needs 2 channels). This is a functionality of the I2SStream. 2 Channels (= Stereo) I am not sure that the impact of different frequency combinations is on the (perceived) amplitude and there is is no logic on my side to deal with this. |
Beta Was this translation helpful? Give feedback.
-
|
AudioTestx1.zip |
Beta Was this translation helpful? Give feedback.
-
|
I took another look at using Maximillian and have got it working better than before for my type of project......See sketch below...... //xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
#include "AudioTools.h"
#include "AudioLibs/MaximilianDSP.h"
#include "SoundMM.h"
float SndArray[]={385, 440, 495, 550, 586.67, 660, 733.33, 770, 880};
float freqD = 110.0; // set = 0 to switch off sound
float freqM = 440.0;
int listp = 0;
unsigned long a = millis();
unsigned long b = millis();
// Arduino output
AnalogAudioStream out;
Maximilian maximilian(out);
// Maximilian def
maxiSample maxSnd1; // sampleplayer one
maxiSample maxSnd2; // sampleplayer two
double sampleOut, sampleOut2;
void setup() {//some inits
// setup logging
Serial.begin(115200);
// AudioLogger::instance().begin(Serial, AudioLogger::Info);
// setup audio output
auto config = out.defaultConfig(TX_MODE);
config.sample_rate = 44100;
config.channels = 2;
config.bits_per_sample = 16;
config.buffer_count = 2;
config.buffer_size = 256;
out.begin(config);
maximilian.begin(config);
maxSnd1.setSample(archrp256);
maxSnd2.setSample(archrp256);
}
void play(float *output) { //this is where the magic happens. Very slow magic.
sampleOut= maxSnd1.play4(freqD, 0, 255); // play these 2 lines for separate L and R output
sampleOut2 = maxSnd2.play4(freqM, 0, 255);
// sampleOut = ((maxSnd1.play4(freqD, 0, 255)*0.5) + (maxSnd2.play4(freqM, 0, 255)*0.5)); // 0.5 is half each amplitude play this line for mixing
output[0] = sampleOut2; //left channel 26
output[1] = sampleOut; //right channel 25
// output[0] = output[1]; // make both output same
}
// Arduino loop - copy sound to out
void loop() {
maximilian.copy();
playlist();
}
void playlist(void) {
b = millis();
if (listp >= 9) listp=0;
if ((b - a) >= 1000) {
freqM = (SndArray[listp]);
Serial.print("Playing sound ... ");
Serial.println(listp);
listp++;
a = millis();
}
}
//xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx |
Beta Was this translation helpful? Give feedback.
-
|
I suggest that you keep it with Maximilian for the time being. |
Beta Was this translation helpful? Give feedback.
-
|
I am only able to test that resampling works is when playing 1 stream only. If resampling 2 streams have to use a mixer. Is there another way of sending 2 resampled streams..... 1 to L and 1 to R. That would separate the problem is it resampling or mixer problem..??? |
Beta Was this translation helpful? Give feedback.
-
|
Not sure if this is just the Interference of the 2 waves that you hear. In my tests the chain with 1 wave w/o mixer seems to be working, but with one channel thru the mixer it's not (even no mixing is done in this case) |
Beta Was this translation helpful? Give feedback.
-
|
Yes changing this line below, sending resample2 to out with streamCopy works for me, but there is only one sound. Also as you say when I comment out one of the sounds to the mixer......as below Regards the slightly off numbers 440.01 and 219.99, I assume you mean try this in the Maximilian sketch. These differences in expected freqs would certainly give beats..................... //xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx |
Beta Was this translation helpful? Give feedback.
-
|
Discovered that the beating effect may be caused by the Serial port being enabled and printing the sound number at same time as playing. List of freqs as below..... Play Freq ................ Output Freq It looks like the ESP32 is not hardware controlled and output freq depends on how busy the ESP is at doing other things at same time...??? |
Beta Was this translation helpful? Give feedback.
-
|
It is rather a question of application design: You would never measure a frequency by doing counts in an Arduino loop and do other things there as well. On the output side you should not use delays or long running operations (like serial output) in the same output loop or if you do so make sure that you can still provide enough samples (as defined by the sample rate) Are these mesurement for audio generated by Maximilan ? |
Beta Was this translation helpful? Give feedback.
-
|
Yes, the measurements were taken using the last Maximilian sketch I posted above. using the archrp256 array. 2 outputs separate pins. |
Beta Was this translation helpful? Give feedback.
-
|
I tried 2 Tasks.... still same problem. with the low freq at 110 the beats are heard most when 440 or 550 or 660 are playing. |
Beta Was this translation helpful? Give feedback.
-
|
When I read through the .h and .cpp files I see it uses F64 arrays...?? I have just been using float arrays generated in Audacity export. |
Beta Was this translation helpful? Give feedback.
-
|
I guess that F64 just indicates a 64bit float (=8 byte float) which is the same like a regular double on most processors. |
Beta Was this translation helpful? Give feedback.
-
|
So...... is the arrays I have been using for Maximilian OK........ just Sample Data Export from audacity values range -1 to 0 to +1 float Are you able to run my sketch above and not get beats etc...... OR ..... do you get beats as well.. I thought ESP32 was supposed to be a fast chip ......... has noone used it for Audio before .....??? |
Beta Was this translation helpful? Give feedback.
-
|
If you post a self sufficient sketch, I can try to play it, so I suggest that you don't put the data into a separate file... |
Beta Was this translation helpful? Give feedback.
-
|
ESPMaxiTaskWeb1.txt |
Beta Was this translation helpful? Give feedback.
-
|
I tried your sketch on an AI Thinker AudioKit (which uses I2S) and to me it sounds pretty good. Maybe this is because you use the internal DAC which usually has plenty of noise and has only a resolution of 8 bits. I usually just use this for testing... |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for trying out the sketch. I included this achrp256 waveform as test because it is a smooth sounding wave and has the sharp corners so can see the frequency response of the DAC. To me the waveform is re-produced at output exactly as the array so the frequency response is good. It is not "ringing" (as in oscillation within the wave) that I have a problem with as would be if DAC had a poor high frequency response. What I have problem with is I hear a "beat freq" as a "difference tone" or "vibrato" between the low freq (freqD) and the higher freq (freqM) for some of the sounds, some sounds are worse than others. I do think it is a bit better now when running the 2 cores but it is still there. Were you able to take any measurements of accuracy of the frequencies generated or the stability. |
Beta Was this translation helpful? Give feedback.
-
|
So far I was mainly concerned that recorded audio (e.g. as mp3) is sounding good and I never thought about measuring the exactness of any waves. So I will need to do some thinking how to deal with this. Normally you don't need to care abut interrupts this is nicely hidden by the API: You just need to make sure that the DMA buffer is not running empty by providing it enough data. If you need to cover some variability you would increase the buffer size, because the generation of samples is usually much faster then the consumption... |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for keeping testing. I extracted your wave file and do not find any change from before. For the low freq (freqD) I get 219.4 and for the high freq (freqM) I get 436.6. which gives a "beat freq" or "vibrato effect" of about 2 Hz, I can still hear that in the sound. I have discovered for some reason if I change the I2S sample rate in the sketch to 176400 it seems to help things. The sounds play at the the same pitch but closer to the 440 and 220 and I don't hear beats at this freq, but they are still there at higher sounds freqs. This is not curing the problem but it may give you some more clues what to look for. |
Beta Was this translation helpful? Give feedback.
-
|
I could imagine that there are some slight variations between the basic clock in your measureing tool and the ESP32. Did you try to set use_apll = true in the analog config ? I did some tests using my ResampleStream where I generated a sine wave and one with step size of 2.0 and then compared the result with Frequency Analysis in Audacity. With a big step size and a rect window I was getting exacly the expected values! |
Beta Was this translation helpful? Give feedback.
-
|
I seen you had made changes to Resample.h so I downloaded the zip library again. the maximilian example is still the same. |
Beta Was this translation helpful? Give feedback.
-
|
Maximilian is a self sufficient audio library and I am just providing the Maximilian maximilian(out); class so that my library can be used to output the audio.. |
Beta Was this translation helpful? Give feedback.
-
|
OK, I had downloaded the Maximillian library and thought Arduino was using it, but renamed it and sketch compiles and runs with audio tools library only. In the play function I use "play4" I don't see it in Audio tools anywhere.?? where is it called from.?? |
Beta Was this translation helpful? Give feedback.
-
|
What do you mean with you renamed it: Arduino libraries are loaded independent on their name! |
Beta Was this translation helpful? Give feedback.
-
|
I think the Panic regarding installing Maximillian is over. I eventually found a ZIP on the following website "https://github.com/pschatzmann/Maximilian" and this downloaded and installed successfully..... all maximillian sketches now compiling and running.....but the difference in pitch tones still exist. |
Beta Was this translation helpful? Give feedback.
-
|
My correction is only for the ResampleStream which has been used in your sketch at the top of this discussion and does not impact Maximilian |
Beta Was this translation helpful? Give feedback.
-
|
OK, but in your post about 5 days ago you suggested I should stay with Maximilian, so my sketches and test with the "beats" are all for Maximilian. I thought it was the best except for the "beats" and the inaccurate freqs generated. |
Beta Was this translation helpful? Give feedback.
-
|
Hmm resampling is changing the number of samples: So a step size of 2 should give double the frequency if you play at the same sample rate or give the same tone if you play at half the sample rate. So the formula should be target frequency / current frequency! I told you to stick with Maximilian because I could not figure out the error which was giving wrong frequencies for values between 1.0 and 2.0 - but this has been resolved now. I do not know Maximilan good enough to give you any advise, but you can always have a look at the source code |
Beta Was this translation helpful? Give feedback.
-
|
WebArRsamMix2.txt |
Beta Was this translation helpful? Give feedback.
-
|
I already answered the first question quite some time ago: I added a new InputMerge class which (does not mix but) interleaves the samples from the defined input streams. With this you can create a stereo output (channels=2) from 2 mono inputs (channels=1). The API is the same like the InputMixer. If you wrap the output in a VolumeStream you can manage the amplitude. I am pretty sure that the max amplitude says fixed across all combinations. You could double check by wrapping the output in a VolumePrint class. I can't reproduce your issue with the range between 0.1 and 1.0. I was testing this with a sine input. I guess the issue with the wrong frequency is because you specify that I2S receives stereo data but you provide mono data. Just remove the comments for the I2S config If you feel that the InputMixer is missing functionality, feel free to create a subclass and add your needed functionality. You just might need to overwrite the size_t readBytes(uint8_t* data, size_t len) method... |
Beta Was this translation helpful? Give feedback.
-
|
I just double checked: It is definitly working for me w/o error. You would need to analyze the stacktrace to figure out where the issue is coming from.. |
Beta Was this translation helpful? Give feedback.
-
|
I restored the last ZIP from recycle bin and it is doing same thing. I posted the messages from the serial monitor....what do they mean. |
Beta Was this translation helpful? Give feedback.
-
|
As you can see there is Backtrace: You are the only one who can analyse this! See https://github.com/me-no-dev/EspExceptionDecoder. Unfortunatly this only works with Arduino < 2.0 |
Beta Was this translation helpful? Give feedback.
-
|
Back working again. I copied the text of your sketch into a new sketch and installed the latest library again and all works. |
Beta Was this translation helpful? Give feedback.
-
|
I have been testing again using the stepsize array 1.0 to 2.125. If I have only 1 resample added to the mixer either resample1 or resample2 and run it , it plays perfectly, no glitches or discontinuity between freqs or at end repeat of playlist. also if I play continuous 1 sound it plays OK. The original problem in the mixing is still there. If I play 1 sound continuous at stepsize of 1.0 (and don't let it change in the playlist) and let the other change every minute as in playlist I do not get a proper mixed waveform out. I get different amplitudes for each sound and the amplitude of some changes as it plays, I can watch it on Audacity as it plays. That is not normal mixing. I posted a JPG about a week ago showing this. It is the JPG with the audio on it and black writing describing it. When both sounds the same frequency are added to the mixer either a constant or changing as playlist both sounds come out L and R so it sounds OK just as one sound. I will always have 2 different frequencies to mix and want a mixed output or 2 separate outpurts that I can feed to speskers and they will mix because both in same enclosure. I have only been testing the mixing sketch you posted......I tried the Merging audio but don't get much different outputs. |
Beta Was this translation helpful? Give feedback.
-
|
My mixer is just providing the weighted average of multiple input channels and if you don't provide any weights it's just the simple mathematical average. Feel free to add your own implementation if this is not satisfying your need. Here is the recording with 1.0 / float stepArray[] = { 1.0, 1.125, 1.25, 1.333, 1.5, 1.667, 1.75, 2.0, 2.125 }; using the AudioKit:
As you can see the max or min amplitudes are pretty consistent. I also added a startIndex parameter to the GeneratorFromArray because I read that it is recommended to use a different phase when mixing. If you plan to build some instruments you might also apply an ADSR effect which I think makes your issue only audible during a short period of time when both instruments are hitting together a common peak. |
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
If we use the same functionality and get different results, I guess then this is caused by the different hardware. Try to add a different phase/start index on one of the GeneratorFromArray. I was connecting the 'Aux Out' of the AudioKit with an audio cable to the 'Mic In' of my computer to record the audio. |
Beta Was this translation helpful? Give feedback.
-
|
Basically that is what I am doing but I am using the inbuilt DAC pins 25 and 26 to Line In on my computer. I also have my computer audio enhancements (eg Waves Audio etc) disabled otherwise the PC audio would be filtered etc by the PC sound card etc.. |
Beta Was this translation helpful? Give feedback.
-
|
Here we go test.mp3.zip |
Beta Was this translation helpful? Give feedback.
-
|
I will try MergeStream, if I had problems It was probably me. I will try it first and then we can compare. |
Beta Was this translation helpful? Give feedback.
-
|
The randomness most likly makes sure that the issue will not be noticable. Just start with a different offset in GeneratorFromArray. If you use timers I suggest to add a small random factor of some ms which should also help and not be noticable. |
Beta Was this translation helpful? Give feedback.
-
|
Do you mean randomness as it starts to play each cycle of the wave or once each loop of the playlist..?? |
Beta Was this translation helpful? Give feedback.
-
|
I mean your external events will most likely just start at random times, so that the probability that you end up with a disturbing phase is very small. |
Beta Was this translation helpful? Give feedback.
-
|
There will always be one sound already playing as was the case in our tests so starting random may not have much effect..? I have now tried out MergeStream. It is working as Separation and not Merge. I set I2S channels to 2 and each wave comes out separate on each pin 25 and 26. The sound is good, no glitches and good waveform at all frequencies and ALL at constant amplitude. There is definately something wrong with the Mixer we had been testing, maybe it is config settings I have got wrong, but the output was also wrong on your test.mp3.zip.. I tried setting I2S channels=2 and I2S channels=1 and made no difference on Mixer. |
Beta Was this translation helpful? Give feedback.
-
|
It's called merge because it merges the (mono) input channel streams into 1 output channel stream with the following logic I am not sure what you mean with "Is there an option to mix these outputs", but in my understanding it would really end up with the current implementation: 1 Channel (=Mono) Mixing |
Beta Was this translation helpful? Give feedback.
-
|
I commented out each merge.add line in turn then compiled and run the sketch. The uncommented sample input played out on both Left and Right. ( exact copies ). And it played at twice the frequency that it had been when both lines were adding to the mixer. |
Beta Was this translation helpful? Give feedback.
-
|
Yes OK, I see what you mean by merging. It puts the inputs into 1 stereo output L and R which in my case I was looking at it as 2 outputs doesn't matter to me about stereo, its just 2 outputs so I seen it as separating them so I could see and hear what each was doing regards resampling and volume. The Input Merge plays sinewaves and the archrp waves fairly well no distortion and I only hear slight beats on some frequencys. If I play a more harsh "real" recording it doesn't sound as well as in Maximilian and more difficult to get the low freq wave to stay in tune with the higher playing freqs. |
Beta Was this translation helpful? Give feedback.
-
|
Input Merge is working quite well. I have tried more waveforms and sound is good except for the beats. I can adjust at the stepsize for each step in the playlist that is beating and almost eliminate it, but it is very "fiddley". and sometimes very little adjustment can make alot of change. I think it needs some form of frequency control, so as be able to call by a particular frequency and it stays to it. I was reading through the documentation and I see the following....... I am finding the opposite..... a stepsize of 2.0 give me a higher pitch than a stepsize of 1.0 |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for letting me know. I corrected the documentation. I suggest that you do a code review of the ResamplingStream and the InputMerge class. Maybe you discover something I was missing. ps. I just simplified the implementation of getValue() in the ResamplingStream in order to minimize the risk of errors. I do not hear any difference though... |
Beta Was this translation helpful? Give feedback.
-
|
I was thinking of trying Pitch Shift. I tried it with single array and it seemed to be more accurate in the relationship between the "pitch_shift" factor and the actual output frequency than resampling was. But I would need to try 2 array sounds playing together....... Can I merge 2 pitchShift stream ........ .................................. merge.add(pitchShift, 0.5); Similar to merge 2 resample streams merge.add(resample1, 0.5); one of the many errors I get for the pitch shift is .................................... The merge resample shown above works, but merge pitchshift doesn't compile |
Beta Was this translation helpful? Give feedback.
-
|
The class PitchShiftStream currently extends AudioPrint, so only write is possible but no readBytes(). |
Beta Was this translation helpful? Give feedback.
-
|
I can output the PitchShift stream for each generated sound array but it outputs to both L and R pins together. Can this be separated easily or done by configuration. If this could be done I probably would not need the Merge or Mixer. |
Beta Was this translation helpful? Give feedback.
I think you would need to find out the root cause of your clicking sounds. Did you try to send the output to the Serial Printer to figure out where this is coming from ?