Rising edge interrupt triggering multiple times on STM32 Nucleo - c++

Question

I am using the STM32 NUCLEO-F401RE microcontroller board.
I have a speaker programmed to change frequency by a set amount when the joystick is pushed up/down. My issue is that sometimes (more often than not) when the joystick is pushed up/down the frequency increases/decreases multiple times, implying that the ISR is executing multiple times. Also, the InterruptIn object is set to trigger on the rising edge however sometimes it also executes on the falling edge (when the joystick is returend to neutral after pushing up/down). Any help for getting over this?
void upISR()
{
if (greenLED.getStatus())
{
myTicker.detach();
frequency+=200;
myTicker.attach(callback(&spkr, &Speaker::toggle), 0.5/frequency);
}
}
'
int main()
{
InterruptIn up(A2);
InterruptIn down(A3);
InterruptIn fire(D4);
up.rise(&upISR);
down.rise(&downISR);
fire.rise(&toggleISR);
redLED.on();
while (1){}
}

Answer 1

Mechanical switch bounce is a feature of all mechanical switches to a lesser or greater extent. It is often necessary to implement "debouncing" in software especially if the switch is directly driving an interrupt as in this case.
A quick Google search for software denounce techniques yields some rather poor techniques IMO. I seen it done poorly more times than well unfortunately.
I suggest that in the switch ISR you start (or restart in the event of a "bounce") a hardware timer for a period of say 20ms or so (longer than the switch bounce time, but shorter than the time you could possibly to genuinely release the switch). Then in the timer ISR, you test the state of the switch and change the frequency accordingly:
Pseudocode:
void upISR()
{
debounceTimerRestart() ;
}
void downISR()
{
debounceTimerRestart() ;
}
void debounceTimerISR()
{
debounceTimerStop() ;
tDirection dir = getJoystickDir() ;
swithc( dir )
{
case UP :
{
increaseFrquency() ;
}
break ;
case DN :
{
decreaseFrquency() ;
}
break ;
}
}
What this does is trigger a timer interrupt shortly ("debounce time") after the switch stops bouncing. Note the timer is "single-shot" not periodic.
Below I present an enhancement at #BenVoigt's suggestion (in comments). I am keeping it separate to make it clear it was his work. The above will generally work, but if you have a particularly poor switch the following would resolve issues, and at little cost, so you may as well:
void debounceTimerISR()
{
debounceTimerStop() ;
static tDirection previous_dir = CENTRE ;
tDirection dir = getJoystickDir() ;
// If the state changed...
if( previous_dir != dir )
{
previous_dir = dir ;
switch( dir )
{
case UP :
{
increaseFrquency() ;
}
break ;
case DN :
{
decreaseFrquency() ;
}
break ;
}
}
}

Answer 2

Simple do not use EXTI for mechanical yousticks and buttons.
Use regular interrupt (for example systick) to poll the status of the pins.

Answer 3

We clearly believe this is the normal and expected bouncing of the switch. Mechanically a switch is some piece of metal that when acted on moves that metal from one pole to another, even if they do not resemble a wiper and two poles. The metal that moves will collide and bounce, the electrical connection will show that. The bouncing is often slow enough for a processor to get multiple interrupts, although that may be an under-sampling of all the bounces possibly seen electrically. If you try to look at it on a scope the scope itself may not-intentionally be filtering some of it (but so will your chip).
One way to see the problem is as with anything, research first then write the application later. This is not a solution but a way to characterize the problem for your system
switch_isr ( void )
{
...
some_global_variable <<= 1;
some_global_variable |= (pin_state_register>>pin_number)&1;
...
}
main ( void )
{
...
some_local_variable = 0;
while(1)
{
if(some_local_variable != some_global_variable)
{
some_local_variable = some_global_variable;
primitive_hex_print(some_local_variable);
}
}
}
No reason to expect to see every state change in the shifted variable, but you should see some and get a feel for the problem. Another way is to just have a counter increment on every interrupt, print periodically in the foreground and you will see one button press may result in multiple counts. And from the time it takes for the printouts to stop changing roughly in human time the settling time.
Filtering is all about state changes per unit time though and you have to have some flavor of time, be it a loop in the foreground that polls some information set by the interrupt (up/down counters, etc), or state changes relative to a timer/clock.
I do not know what the complete rules are for your assignment, if you can only have an interrupt for each switch and not a timer, or preferably a timer instead, I do not see a clean solution that will actually work. You would have to filter in the foreground but all that is doing is polling a copy of the pin state collected by the interrupt and is that any different than not using the interrupt? You cannot use Clifford's answer if you cannot set a timer interrupt, if you could use a timer and an interrupt then you could just periodically sample the switch states with that interrupt or a copy of the pin state collected by the pin state change interrupts and filter in the timer interrupt. Not the same as Clifford's but in all cases you need state change history relative to time to see when the thing settles.
Without a time reference and states not changing with respect to time (which a pin interrupt cannot show since the state has not changed) you cannot filter out the bounces. Instead work on your dexterity and how you flick the joystick up and down.