Modified sketch for improved timing precision.

[Gunghouk]

For the frequency it doesn't matter. If we mask then the frequency will move in a few large steps like a cheap videogame sound. Let's try it first like this.

That's why shifting right twice, as also suggested, to give 8bit resolution is preferred over simple masking.

[Teleno]

That's why shifting right twice, as also suggested, to give 8bit resolution is preferred over simple masking.

Well, I just say try both and find out which sound appeals to you. Test instead of speculate.

By the way, I recommend this Nano clone with LGT8F328P mcu. This chip has all the instructions and registers of the Atmega328P and much more (extra 16bit timer, 12 bit ADC, 8 bit DAC, power pins, configurable gain amplifier, 16bit arithmetic and it runs at 32MHz).

https://www.aliexpress.com/i/1005005854795426.html

Install the core libraries here https://github.com/dbuezas/lgt8fx

Then edit the file "C:\Users\....\AppData\Local\Arduino15\package s\lg t8fx\hardware\avr\2.0.7\cores\lgt8f\Arduino.h

and comment out these lines at the bottoom of the file:

 #ifndef nop
 #define  nop() __asm__ __volatile__ ("nop" ::)
 #endif​ 

This will prevent compilation errors, in particular with the Tiny4kOLED library.

[Teleno]

Latest build.

It includes all the code: PWM, FM+AM sound, OLED dislplay with current delay value

Comment out unwanted features before building & uploading. Read carefully the instructions in the file. Links to the relevant posts regarding the PCB modifications are in the file.

Arduino_PI_Teleno.zip

[SaltyDog]

By the way, I recommend this Nano clone with LGT8F328P mcu. This chip has all the instructions and registers of the Atmega328P and much more (extra 16bit timer, 12 bit ADC, 8 bit DAC, power pins, configurable gain amplifier, 16bit arithmetic and it runs at 32MHz).

Interesting, what onboard XTAL does it use 32MHz or 16MHz? Can it be a "drop in" replacement for this project, with tweaking for MHz?

[Teleno]

Interesting, what onboard XTAL does it use 32MHz or 16MHz? Can it be a "drop in" replacement for this project, with tweaking for MHz?

Mostly come with an external 16MHz crystal because they're sold as Arduino Nano's. Once the lgt8f boards are installed in Arduino IDE you can select the frequency in the menu: internal 32MHz, 16Mhz or external crystal 32Mhz or 16MHz.

If you do a clock change you have to adjust the "#define US" in the sketch to the actual number of cycles per microsecond (32 or 16 for 32MHz or 16Mz respectively).
Or better yet, replace "#define US 16" with "#define US ((byte) (F_CPU/1000000UL))" to adjust automatically.

It's more that a drop in replacement, it has an extra 16bit TImer3 wth three independent PWM channels, so all three pulses (Tx, mainSample, efeSample) can be generated in hardware.

PS: It's the one I've been using to develop this sketch

[SaltyDog]

Timer3 looks good for 3-phase motor control as well. Nice. Except, all 3 outputs are only available on the 40 pin Nano board, not the current 30 pin.

[Teleno]

Timer3 looks good for 3-phase motor control as well. Nice. Except, all 3 outputs are only available on the 40 pin Nano board, not the current 30 pin.

There's a workaround. T1 and T3 can be made to share the same prescaler, then they can be synchronized and we can get 3 PWM outputs in pace.

See registers GTCCR and PSSR.

[Teleno]

There's a workaround. T1 and T3 can be made to share the same prescaler, then they can be synchronized and we can get 3 PWM outputs in pace.

See registers GTCCR and PSSR.

Example code to sync 16 bit timers Timer1 and Timer3 in LGT8F328P.

/**
  * URL: https://dbuezas.github.io/arduino-web-timers/#mcu=LGT8F328P&timer=3&timerMode=PCPWM&clockPrescalerOrSource=256&FCPU_UI=16Mhz&OCnA_OutputPort=F1&CompareOutputModeA=clear-up%2C+set-down&OCR3A=85
  * Mode     : PCPWM
  * Period   : 31.875 us
  * Frequency: 31.37255 kHz
  * Outputs  :
  *  - F1: 33.33%, clear-up, set-down
  */
void setup(){
  DDRB =   1 << DDB1;                 // Timer1 output pin 9 (D9)
  DDRF =   1 << DDF1;                 // Timer3 output pin 1 (TX)

  GTCCR = 1 << TSM | 1 << PSRSYNC;     // Freeze Timer1 and TImer3

  TCNT1 = 0;                           // Clear timer counts
  TCNT3 = 0;

  TCCR1A = 1 << COM1A1 | 1 << WGM10;   // Timer1 configuration
  OCR1A = 85;

  TCCR3A = 1 << COM3A1 | 1 << WGM30;   // Timer3 configuration identical to Timer1
  OCR3A = 85;

  TCCR1B = 1 << CS10;                  // Start timers with prescaler 1
  TCCR3B = 1 << CS30;

  GTCCR = 0;                          // Release timers
}
void loop() {
  // put your main code here, to run repeatedly:
}
​​​ 

Without syncing




With sync (GTCCR statements commented out)




The difference is negligible taking into account that the efeSample delay is not critical. As long as the delay is constant. So we could do without the syncing.

[Teleno]

Above: Without syncing should be "With syncing", the screen shots are swapped.

[SaltyDog]

Here's the display code.

Place this snippet at the beginning of the sketch

#define USE_OLED

#ifdef USE_OLED
#include <Tiny4kOLED.h>
#include <font16x32digits.h>
const DCfont *largeFont = FONT16X32DIGITS;
const DCfont *smallFont = FONT8X16;
uint8_t width = 128;
uint8_t height = 32;
#endif​ 

This snippet at the beginning of setup()

#ifdef USE_OLED
oled.begin(width, height, sizeof(tiny4koled_init_128x32br), tiny4koled_init_128x32br);
oled.setFont(smallFont);
if (smallFont->width == 0) {
oled.setSpacing(1);
} else {
oled.setSpacing(smallFont->spacing + 1);
}
oled.on();
oled.clear();
oled.setCursor(0, 2);
oled.print("DELAY");
oled.setCursor(0, 0);
oled.print("us");
oled.setFont(largeFont);
if (smallFont->width == 0) {
oled.setSpacing(1);
} else {
oled.setSpacing(largeFont->spacing + 1);
}
#endif​ 

and this snippet in the loop()

#ifdef USE_OLED
oled.setCursor(70, 0);
oled.print(mainDelay);
oled.print(" ");
#endif​ 

This is what you'll get



If you won't use the OLED comment out the "#define USE_OLED" line and rebuild. Otherwise Nano will wait forever for a display to be ready that doesn't exist.

Install the following libraries in Arduino (using search box)

Tiny4kOLED
TinyOLED-Fonts

The display is 0.91 inch OLED 128x32 based on controller SSD1306. Like this one: https://www.aliexpress.com/item/1005006431809403.html?

Hi, I have just added the display and can now report the following findings:

1. <= 20uS no targets found.
2. 25uS all target metals found
3. 70..75 uS Gold target disappears but other metals still found, Aluminium pull-tabs faint ...

I can also report a session down at the beach scored one gold ring and "no" rubbish dug. As I hunted at 25uS, then confirmed target at 75uS.
So I am a happy man.

Different coils and dampening may give different numbers for you. But the above discrimination is possible.

[Teleno]

Now correct me if I'm wrong but 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.

[SaltyDog]

Now correct me if I'm wrong but 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.

I don't think that will work. C6/R13 is a differentiator, and is a required high pass filter(Freq >34 Hz),
It also performs the task of getting rid of the DC component for the EFE stage.

In the field the EFE compensation must work, as I am not seeing any ground effect.

[Teleno]

I don't think that will work. C6/R13 is a differentiator, and is a required high pass filter(Freq >34 Hz),
It also performs the task of getting rid of the DC component for the EFE stage.

In the field the EFE compensation must work, as I am not seeing any ground effect.

It's nor a ground effect, it's the voltage induced by the earth's magnetic field as you swing the coi.

EF is a DC offset within a period. If you pass it through an RC differentiator you get a decaying exponential. Just do the simulation.

Top: no differentiator. Bottom with differentiator.




To get rid of the EF component you wat till the target signal has vanished (that's what the long efeSample delay is for), sample it and subtract it from the (signal +EF) early component.

The differentiator is in the way. I would say the whole analog part is crappy, starting with the huge gain of the first op amp that kills any bandwidth.

[Teleno]

Hi, I have just added the display and can now report the following findings:

1. <= 20uS no targets found.
2. 25uS all target metals found
3. 70..75 uS Gold target disappears but other metals still found, Aluminium pull-tabs faint ...

I can also report a session down at the beach scored one gold ring and "no" rubbish dug. As I hunted at 25uS, then confirmed target at 75uS.
So I am a happy man.

Different coils and dampening may give different numbers for you. But the above discrimination is possible.

Glad it worked for you. I appreciate your feedback but where's my share of the bounty?

[SaltyDog]

To get rid of the EF component you wat till the target signal has vanished (that's what the long efeSample delay is for), sample it and subtract it from the (signal +EF) early component.

The differentiator is in the way. I would say the whole analog part is crappy, starting with the huge gain of the first op amp that kills any bandwidth.

Theory is one thing, where are all your results from field testing ...
The current design works very well. The op-amp has a very good gain/bandwidth product, and can easily do 1000 gain with little noise, so why not?
I have tried in the past lot's of dual op-amp designs and they have all performed worse than the single high gain solution.
Again theory is one thing, but you have to actually do the field testing ..