ticks to ms calculation

RE: ticks to ms calculation

Eliot Stock — Thu, 28 May 2015 13:35:52 GMT

I was already doing a lot with floats, so it's no big deal. I agree this is not ideal if you're not already using floats, good point.

RE: ticks to ms calculation

RK — Thu, 28 May 2015 13:30:06 GMT

you imported the floating point emulation library just to do that? if you were using it anyway I guess that's ok. With a couple of ifs you can decide whether to do the multiplication or division first, or indeed use the uint64_t emulation which is quite fast and won't overflow. Either works.

RE: ticks to ms calculation

Eliot Stock — Thu, 28 May 2015 13:26:18 GMT

The risk with just rearranging the formula in that macro is that you a) end up trying to use a uint64_t beyond the preprocessor and/or b) end up overflowing a uint32_t at a low number of ticks and getting a bogus result back. I settled on this function.

uint32_t app_timer_ms(uint32_t ticks)
{
	// eg. (7 + 1) * 1000 / 32768
	//   = 8000 / 32768
	//   = 0.24414062
	float numerator = ((float)APP_TIMER_PRESCALER + 1.0f) * 1000.0f;
	float denominator = (float)APP_TIMER_CLOCK_FREQ;
	float ms_per_tick = numerator / denominator;

	uint32_t ms = ms_per_tick * ticks;

	return ms;
}

RE: ticks to ms calculation

RK — Thu, 28 May 2015 01:38:17 GMT

well just re-arrange the formula, ignoring the ROUNDED_DIV() for simplicity

APP_TIMER_TICKS = ( MS * APP_TIMER_CLOCK_FREQ ) / ( ( PRESCALAR + 1 ) * 1000 )

MS = APP_TIMER_TICKS * ( ( PRESCALAR + 1 ) * 1000 ) / APP_TIMER_CLOCK_FREQ

I'm sure you can turn that into a define, not that it really helps you that much as the number of ticks will be in a variable so the compiler can't evaluate it and compile it out into a constant which is what happens most of the time with the current macro as usually everything on the right hand side is a compile-time constant.