AVR Freaks

Edit to add: I like avrcandies solution too. It's like the simple moving average method but is just an average without the "moving". It's also not as continuous as a moving average though.

It's essentially "batch  averaging", like most simple "gated" freq counters.  So that lets you pick the time span...if you want a smoother response, just pick a longer interval to totalize.  Downside is that the result update rate slows down. 

----------

The exp avg is really hard to beat when you need a quick simple avg.  You do have to watch with the finite bit arithmetic, since the fractional part gets tossed (or rounded) each cycle.  A trick is to keep the latest result at 4x or 8x and maintain (remember) these few extra fraction bits scooted into the whole portion.  So the true output result is this value shifted a few places to get rid of the xx scaling.

Here is one I threw together  with Ynew=  (7*Yold+Sample)/8 ...you can see it stalls out and never reaches 1000, only 997, due to finite numerics.  (15Yold+sample)/16  would be worse yet.  With an 8x result scaling you see a perfect 1000 is reached (perfect due to rounding).  some rows not shown

n

the yellow 8 is the 8x scaling factor---causes all values from column A (inputs) to be 8x larger

Detect no pulse, measure H or L, that is it.

Could that lead to a race condition?  EX: All low coming in for a long time...eventually returns as "no pulse". Thus decision made to check the level, of the "no pulse" but just around then a hi arrives and gets read...so now instead of looking like 0% , it falsely looks like 100%.

I had the thought you could take the original wave, RC filter & ADC it...then convert that value into the DAC value...sort of an odd roundabout way & ADC tolerance issues to boot!

pulseIn() is implemented in arduino\hardware\arduino\avr\cores\arduino\wiring_pulse.c

unsigned long pulseIn(uint8_t pin, uint8_t state, unsigned long timeout)
{
	// cache the port and bit of the pin in order to speed up the
	// pulse width measuring loop and achieve finer resolution.  calling
	// digitalRead() instead yields much coarser resolution.
	uint8_t bit = digitalPinToBitMask(pin);
	uint8_t port = digitalPinToPort(pin);
	uint8_t stateMask = (state ? bit : 0);

	// convert the timeout from microseconds to a number of times through
	// the initial loop; it takes approximately 16 clock cycles per iteration
	unsigned long maxloops = microsecondsToClockCycles(timeout)/16;

	unsigned long width = countPulseASM(portInputRegister(port), bit, stateMask, maxloops);

	// prevent clockCyclesToMicroseconds to return bogus values if countPulseASM timed out
	if (width)
		return clockCyclesToMicroseconds(width * 16 + 16);
	else
		return 0;
}

which in turn seems mostly reliant on arduino\hardware\arduino\avr\cores\arduino\wiring_pulse.S:

/*
 * The following routine was generated by avr-gcc 4.8.3 with the following parameters
 * -gstabs -Wa,-ahlmsd=output.lst -dp -fverbose-asm -O2
 * on the original C function
 *
 * unsigned long pulseInSimpl(volatile uint8_t *port, uint8_t bit, uint8_t stateMask, unsigned long maxloops)
 * {
 *     unsigned long width = 0;
 *     // wait for any previous pulse to end
 *     while ((*port & bit) == stateMask)
 *         if (--maxloops == 0)
 *             return 0;
 *
 *     // wait for the pulse to start
 *     while ((*port & bit) != stateMask)
 *         if (--maxloops == 0)
 *             return 0;
 *
 *     // wait for the pulse to stop
 *     while ((*port & bit) == stateMask) {
 *         if (++width == maxloops)
 *             return 0;
 *     }
 *     return width;
 * }
 *
 * some compiler outputs were removed but the rest of the code is untouched
 */

#include <avr/io.h>

.section .text

.global countPulseASM

countPulseASM:
etc

(I didn't show the Asm as the C comment is perhaps more explanatory)

So the point is that "timeout" you pass is converted to "maxloops" and passed to the Asm and because no transition is detected it simply follows one of the "return 0" paths to indicate timeout rather than a width.

Presumably you can just digitalRead() the pin on receiving 0 to know whether it is fully high or fully low ?

This is 'freaks.  The topic isn't closed until all of the skin is off the dead horse!