Jazzing up the Arduino PI - full recipe

[Teleno]

Correction:

delayVal = round(map(analogRead(delayPin), 0, 1023, defMainDelay, defMaxDelay)); // new delay value

Do you realize you're rounding integers?

long is an integer type and a division of integers in C, C++ is also an integer, not a float.

[Gunghouk]

I think truncating the MAP output instead of rounding gives approx equal pot rotation angle per uS

delayVal = trunc(map(analogRead(delayPin), 0, 1023, defMainDelay, defMaxDelay)); // new delay value

[Gunghouk]

Do you realize you're rounding integers?

long is an integer type and a division of integers in C, C++ is also an integer, not a float.

Erm.....looking at the Arduino reference for MAP and my Excel sheet shows fractions........

My bad again

[Teleno]

Correction:

delayVal = round(map(analogRead(delayPin), 0, 1023, defMainDelay, defMaxDelay)); // new delay value

Check this out this simulation https://onlinegdb.com/fyy3Uo5KI

#include <stdlib.h>
#include <stdio.h>
#include <math.h>

#define defMainDelay 5
#define defMaxDelay 16

long delayVal;
long lastDelayVal = -1;
long noisyDelayVal;

long map(long x, long in_min, long in_max, long out_min, long out_max)
{
return (x - in_min) * (out_max - out_min) / (in_max - in_min) + out_min;
}


int main() {

    
    // swipe all  pot values
    for (long analogRead = 2; analogRead < 1024; analogRead++){
        
        // noiseless delayVal update
        delayVal = round(map(analogRead, 0, 1023, defMainDelay, defMaxDelay));
        if (delayVal > lastDelayVal){
            printf("current delayVal is %ld\n", delayVal);
            lastDelayVal = delayVal;
        }
        
        // subtract 1 bit of noise
        long noisyAnalogRead = analogRead - 1;
        noisyDelayVal = round(map(noisyAnalogRead, 0, 1023, defMainDelay, defMaxDelay));
        
        // if there's flicker the mapped value will be less than the noiseless value.
        if (noisyDelayVal < delayVal){
            printf("%ld %ld %s\n", analogRead, noisyDelayVal, "<- flicker");
        }

        // add 1 bit of noise
        noisyAnalogRead = analogRead + 1;
        noisyDelayVal = round(map(noisyAnalogRead, 0, 1023, defMainDelay, defMaxDelay));
        
        // if there's flicker the mapped value will be more than the noiseless value.
        if (noisyDelayVal > delayVal){
            printf("%ld %ld %s\n", analogRead, noisyDelayVal, "<- flicker");
        }
    }
    
    return 0;
}
​
​ 

It only adds +- 1 bit of noise, it would be much worse in real life.

output:

current delayVal is 5
92 6 <- flicker
current delayVal is 6
93 5 <- flicker
185 7 <- flicker
current delayVal is 7
186 6 <- flicker
278 8 <- flicker
current delayVal is 8
279 7 <- flicker
371 9 <- flicker
current delayVal is 9
372 8 <- flicker
464 10 <- flicker
current delayVal is 10
465 9 <- flicker
557 11 <- flicker
current delayVal is 11
558 10 <- flicker
650 12 <- flicker
current delayVal is 12
651 11 <- flicker
743 13 <- flicker
current delayVal is 13
744 12 <- flicker
836 14 <- flicker
current delayVal is 14
837 13 <- flicker
929 15 <- flicker
current delayVal is 15
930 14 <- flicker
1022 16 <- flicker
current delayVal is 16
1023 15 <- flicker​

[Teleno]

Now the same wiith +-1 bit of noise and +-1 bit hysteresis.

https://onlinegdb.com/wyqRfLz6M

#include <stdlib.h>
#include <stdio.h>
#include <math.h>

#define defMainDelay 5
#define defMaxDelay 16

long delayVal;
long lastDelayVal = -1;
long noisyDelayVal;

long threshold_up, threshold_down = 0;

long map(long x, long in_min, long in_max, long out_min, long out_max)
{
return (x - in_min) * (out_max - out_min) / (in_max - in_min) + out_min;
}


int hysteresis(long analog){
    int accept = 11;
    int reject = 0;
    
    if(analog > threshold_up || analog < threshold_down){
      threshold_up = (analog < 1024)? analog + 1 : 1024;
      threshold_down = (analog > 0)? analog - 1 : 0;
      return accept;
    }
    else return reject;
}


int main() {

    
    // swipe all  pot values
    for (long analogRead = 2; analogRead < 1024; analogRead++){
        
        // noiseless delayVal update
        delayVal = round(map(analogRead, 0, 1023, defMainDelay, defMaxDelay));
        if (delayVal > lastDelayVal){
            printf("current delayVal is %ld\n", delayVal);
            lastDelayVal = delayVal;
        }
        
        // subtract 1 bit of noise
        long noisyAnalogRead = analogRead - 1;
        if(hysteresis(noisyAnalogRead) == 1) {
            noisyDelayVal = round(map(noisyAnalogRead, 0, 1023, defMainDelay, defMaxDelay));
            
            // if there's flicker the mapped value will be less than the noiseless value.
            if (noisyDelayVal < delayVal){
                printf("%ld %ld %s\n", analogRead, noisyDelayVal, "<- flicker");
            }
        }

        // add 1 bit of noise
        noisyAnalogRead = analogRead + 1;
        if(hysteresis(noisyAnalogRead) == 1) {
            noisyDelayVal = round(map(noisyAnalogRead, 0, 1023, defMainDelay, defMaxDelay));
            
            // if there's flicker the mapped value will be more than the noiseless value.
            if (noisyDelayVal > delayVal){
                printf("%ld %ld %s\n", analogRead, noisyDelayVal, "<- flicker");
            }
        }
    }
    
    return 0;
}

​ 

Output

current delayVal is 5
current delayVal is 6
current delayVal is 7
current delayVal is 8
current delayVal is 9
current delayVal is 10
current delayVal is 11
current delayVal is 12
current delayVal is 13
current delayVal is 14
current delayVal is 15
current delayVal is 16​

[Gunghouk]

Thank you for your patience and for taking the time to point out my misconceptions

[Teleno]

Thank you for your patience and for taking the time to point out my misconceptions

No problem. I want to make sure the code is solid so I had to do the simulations.

On the bench you can notice the hysyeresis on the pot. Once you set a delay it takes an additional displacement to go back to the previous value and another to return to the current delay.

So here's the new version with stable pot.

Version 1.0.3

ArduinoPI_fullPWM_1_0_3.zip

Changes in 1.0.3:

- Pot reading validated against adaptive hysteresis thresholds before being accepted (+- 8 bit noise).

[Gunghouk]

Nice....... erm life would be so much easier with an encoder

Toggle power with the encoder button and set delay with the encoder. I've plans to show the TX power in the bottom left of the oled.

TXlo or TXhi

[Teleno]

Nice....... erm life would be so much easier with an encoder

Toggle power with the encoder button and set delay with the encoder. I've plans to show the TX power in the bottom left of the oled.

TXlo or TXhi

Version 1.0.4

Changes in 1.0.4:

- TX Boost icon added to the FontLargeDigits.h file.Code for displaying it on switch changes.


ArduinoPI_fullPWM_1_0_4.zip

[Gunghouk]

[Teleno]

For those willing to try the LGT8F328P based Nano there are two Arduino cores you can install:

https://github.com/dbuezas/lgt8fx
https://github.com/nulllaborg/arduino_nulllab

For compatibility with the OLED library, edit the "Arduino.h" of any of the above cores and delete or comment out these lines at the bottom of the file:

/* deleted for compatibility with the OLED library Tiny4kOLED
#ifndef nop
#define    nop()    __asm__ __volatile__ ("nop" ::)
#endif
*/​ 

[Teleno]

Version 1.0.5

Changes in 1.0.5:
- Added low-pass IIR filter for the pot reading, reducing noise substantially. Less hysteresis is necessary, allowing a larger delay range.
This requires a higer refresh rate for the pot, so it's now read 1000 per second.

ArduinoPI_fullPWM_1_0_5.zip

I want to point out that I believe C6 and R13 between TP6 and the input of the analog switches are totally wrong.

C6/R13 turn the EF signal (which is constant within one period) into a decaying exponential. After efeSampleDelay the signal being subtracted at the differential integrator is not the EF signal level that's riding on the main sample, but a decayed version at a lower level, breaking the EF compensation scheme.

In MOD3 (see /doc folder in the sketch distribution) I suggest removing C6 and shorting its pads, and removing R13. Then adjust the offset at TP6.

[Gunghouk]

I thought this was an elegant running average solution used in George's Voodoo Project. I've put the call in the main loop. I have averaging count set to 64.

if (readPot) {
// IIR low pass filter
//potReading = analogRead(delayPin);

getPotReading(); // running average filter
readPot = false;

// update audio frequency.
#ifdef USE_FREQ_MODULATION
word target = analogRead(targetPin);
freq = map(target, 0, 635, FREQ_MIN, FREQ_MAX);
freq = constrain(freq, FREQ_MIN, FREQ_MAX);
OCR2A = TIMER2_CLK / freq - 1;
#endif
}


unsigned int getPotReading()
{
potReading = analogRead(delayPin); // Acquire reading
averagingTotal -= averagingArray[averagingIndex]; // Remove oldest reading from the total
averagingTotal += potReading; // Add latest reading to the total
averagingArray[averagingIndex] = potReading; // Add latest reading to pot averaging array
potReading = averagingTotal/averagingCount; // Average readings

if(++averagingIndex >= averagingCount) // Increment PI index
averagingIndex = 0; // Wrap if needed

return potReading;
}

[Teleno]

I thought this was an elegant running average solution used in George's Voodoo Project. I've put the call in the main loop. I have averaging count set to 64.

if (readPot) {
// IIR low pass filter
//potReading = analogRead(delayPin);

getPotReading(); // running average filter
readPot = false;

// update audio frequency.
#ifdef USE_FREQ_MODULATION
word target = analogRead(targetPin);
freq = map(target, 0, 635, FREQ_MIN, FREQ_MAX);
freq = constrain(freq, FREQ_MIN, FREQ_MAX);
OCR2A = TIMER2_CLK / freq - 1;
#endif
}


unsigned int getPotReading()
{
potReading = analogRead(delayPin); // Acquire reading
averagingTotal -= averagingArray[averagingIndex]; // Remove oldest reading from the total
averagingTotal += potReading; // Add latest reading to the total
averagingArray[averagingIndex] = potReading; // Add latest reading to pot averaging array
potReading = averagingTotal/averagingCount; // Average readings

if(++averagingIndex >= averagingCount) // Increment PI index
averagingIndex = 0; // Wrap if needed

return potReading;
}

That's a lot more code and memory footprint than a simple low pass filter anyway.

[Gunghouk]

That's a lot more code and memory footprint than a simple low pass filter anyway.

For the array for sure but there's plenty of memory available but as an alternative or efficient replacement for the IIR I can't say. I just like its compactness and understandibilty.

All moot if the pot is replaced with an encoder.