Faster Pulses with Spare Timers?

Hi All,
Working on something that needs to produce accurate-ish (+/-20%) pulses down to 20us every 10ms for 1s. I'm currently using a combination of setInterval() and digitalPulse() (see snippet below if interested).

With a bit of dialling-in, the 1s duration is reliable enough, setInterval() is working great at 100Hz, and digitalPulse() gives good results at 200us and passable ones at around 140us.

digitalPulse can do 32us, but there's about +40/-0 us of variation in it so it's not ideal. For stuff around 100us things could be fudged so that the +40us sits as +/-20% around 100us, but it doesn't work once you get faster than that.

So the question is, what timers are free to use that don't interfere with Espruino? What timers (if any) does digitalPulse() use? Was thinking of using a one-shot timer triggered over and over by setInterval().

// Timeouts of OFFSET_MS used to account for speed of js execution. Obtain OFFSET_MS empirically.
// Timeout to start pulses.
setTimeout( () =>   { tmp1 = setInterval( () => { 
                digitalPulse( PULSE_PIN, true, pulseDuration_ms );
            }, CNST_PERIOD_MS )
        },  OFFSET_MS
    );
// Timeout to stop pulses after some time.
setTimeout( function(){ PULSE_PIN.reset(); clearInterval(tmp1); }, Duration1_ms+OFFSET_MS);

Hi - did you consider using digitalPulse with an array? That can be pretty accurate to get the repetition.

Also which firmware are you on? In 2v13 the accuracy improved a lot.

But... yes, there is free hardware. What's used is shown at the top of https://github.com/espruino/Espruino/blo­b/master/targets/nrf5x/jshardware.c and there's actually an example of doing PWM using hardware timers (as part of a capacitive sense example) at:

https://www.espruino.com/NRF52LL#hardwar­e-capacitive-sense-on-two-pins

Thanks Gordon, all useful advice.

I was on v2.12, but have updated to v2.14 and seen an improvement in timings, although there'll still be one or two long pulses in a second's worth and it still doesn't quite go down to 20us. But I agree it's much improved and if I get fed up with fiddling with timers it's probably good enough to work with.

Thank you for the links re. hardware and examples, I'll have a read up on them.

Finally got around to this, and results are excellent. Getting <0.3us jitter.
The examples definitely came in handy, although there's a bit of guesswork involved in sending the right values to the right places.

A working example is included below though that uses a button on D17 to start / stop a train of pulses.

var ll = require("NRF52LL");
// 100Hz pulses of 20us width, but triggered by button press.
// Set up D20 as output
digitalWrite(D20, 0);
// Create a task for the pin
var t0 = ll.gpiote(7, {type:"task",pin:D20});
// Create a timer using timer 2 that counts to 10000
// and has capture-compares at 20 and 10000 counts.
var tmr = ll.timer(2, {cc:[20, 10000],cc1clear:1});
// User a PPI to trigger the events.
ll.ppiEnable(0, tmr.eCompare[0], t0.tClr);
ll.ppiEnable(1, tmr.eCompare[1], t0.tSet);
//Button stuff to start/stop pulses.
let btn = false;
function checkBtn(){
    if (digitalRead(D17) == 1) {
        if (btn==false){
            // Manually start timer
            poke32(tmr.tStart,1);
            while(digitalRead(D17)){}
            btn=true;
        }
        else{
            // stop the timer.
            poke32(tmr.tStop,1);
            while(digitalRead(D17)){}
            digitalWrite(D20, 0);
            btn=false;
        }
    }
}
setInterval(checkBtn, 10);

edit - added 'var ll = require("NRF52LL");' at top of example