dht11.js

var rpio = require('../lib/rpio');

/*
 * Repeatedly read a DHT11 attached to Pin 7 / GPIO 4 and print out the current
 * temperature and relative humidity if the read was successful.
 */
var pin = 7;

/*
 * Print debug messages if enabled.
 */
var debug = false;

function dbg()
{
	if (debug) {
		console.error.apply(console, arguments);
	}
}

/*
 * The DHT11 transmission is made up of 40 "bits", where the length of each
 * high output determines whether it is a 0 or 1.  Per the datasheet, "0" is
 * 26-28us in length, and "1" 70us.  However, in a JS environment we can't be
 * so precise, and so we determine what is high and low based on a sampling of
 * the values we received.
 *
 * These devices are notoriously unreliable, so this program will run forever
 * attempting to retrieve and print valid data until the user kills the process.
 *
 * For the most portable implementation we read all bits after initialisation
 * into a big buffer as fast as possible, and then parse afterwards.  This is
 * necessary on e.g. the original Raspberry Pi 1 as the CPU is not fast enough
 * to do something like:
 *
 *	while ((curstate = rpio.read(...)) == laststate)
 *		count++;
 *
 * for each read, which would be the normal way of doing something like this.
 */
function read_dht11(vals)
{
	/*
	 * Our read buffer of all the bits sent.  Should be plenty.  On a
	 * Raspberry Pi 4 a successful read uses up about half of this.
	 */
	var buf = new Buffer(50000);

	/*
	 * Array storing the data bits received from the DHT11 after parsing
	 * the input buffer.
	 */
	var data = new Array();

	/*
	 * The initialisation sequence as per the datasheet is to start high,
	 * pull low for at least 18ms, then back high and read.  The JavaScript
	 * function call overhead ensures that we'll actually be sleeping
	 * longer than that, but it doesn't appear to be a problem.
	 */
	rpio.open(pin, rpio.OUTPUT, rpio.HIGH);
	rpio.write(pin, rpio.LOW);
	rpio.msleep(18);

	/*
	 * Enable the pin pull-up then use the mode-setting variant of readbuf()
	 * to read the pin value into the buffer as fast as possible.  We only
	 * have around 100us before the DHT11 will start transmitting data so
	 * there needs to be no delay between configuring the pin as an input and
	 * reading the data.
	 *
	 * Configuring the pin for input, enabling the pull-up, then reading the
	 * data would take way too long.  Apart from the JavaScript function call
	 * overhead, just enabling the pull-up can require a 150us delay on some
	 * hardware.
	 */
	rpio.pud(pin, rpio.PULL_UP);
	rpio.readbuf(pin, buf, buf.length, true);
	rpio.close(pin);

	/*
	 * The data has been received, split the buffer into groups of "1"s.  By
	 * measuring the length of each group we can determine whether it is a
	 * low (short) or high (long) bit sent by the DHT11.
	 */
	buf.join('').replace(/0+/g, '0').split('0').forEach(function(bits, n) {
		data.push(bits.length);
	});

	/*
	 * In normal operation we'd expect to see 43 values: the initial high of
	 * the pull-up, a ready signal, 40 bits of data, then a continuous high
	 * to signal end of transmission.  Use shift() and pop() to remove the
	 * control bits, leaving just the data.
	 *
	 * In certain circumstances the first bit can be missing, if we didn't
	 * switch to read mode fast enough.  However as that's just the high
	 * value configured by the pull-up and not data from the DHT11 we allow
	 * it to be missing.  Everything else, even with a correct checksum (can
	 * happen!) we treat as invalid.
	 */
	if (data.length < 42 || data.length > 43) {
		dbg("Bad data read: length=%d", data.length);
		return false;
	}

	/* Remove extra first bit generated by the pull-up */
	if (data.length == 43) {
		data.shift();
	}

	/* Remove data ready bit */
	data.shift();

	/* Remove end of transmission bit */
	data.pop();

	/*
	 * Calculate the low and high water marks.  As each model of Raspberry
	 * Pi will run at different speeds, the length of each high bit will
	 * vary, so calculate the average and use that to determine what is
	 * "high" and "low".
	 *
	 * The longest "low" seen on a Raspberry Pi 4 is around 135, so the
	 * default low here should be more than sufficient.
	 */
	var low = 10000;
	var high = 0;
	for (var i = 0; i < data.length; i++) {
		if (data[i] < low)
			low = data[i];
		if (data[i] > high)
			high = data[i];
	}
	var avg = (low + high) / 2;

	/*
	 * The data received from the DHT11 is in 5 groups of 8-bits:
	 *
	 *	[0:7] integral relative humidity
	 *	[8:15] decimal relative humidity
	 *	[16:23] integral temperature
	 *	[24:31] decimal temperature
	 *	[32:39] checksum
	 *
	 * Parse the bitstream into the supplied "vals" buffer.
	 */
	vals.fill(0);
	for (var i = 0; i < data.length; i++) {
		var group = parseInt(i/8)

		/* The data is in big-endian format, shift it in. */
		vals[group] <<= 1;

		/* This should be a high bit, based on the average. */
		if (data[i] >= avg)
			vals[group] |= 1;
	}

	/*
	 * Occasionally the final checksum test can be valid even when the
	 * values are clearly bogus.  Perform additional sanity checks to
	 * ensure they are within the limitations of the DHT11.
	 */

	/* Relative humidity range is 20 - 90% */
	if (vals[0] < 20 || vals[0] > 90) {
		dbg("Bad humidity: %d%%", vals[0]);
		return false;
	}

	/* Temperature range is 0 - 50C */
	if (vals[2] > 50) {
		dbg("Bad temperature: %d", vals[0]);
		return false;
	}

	/*
	 * Validate the checksum and return whether successful or not.  The
	 * checksum is simply the value of the other 4 groups combined.
	 *
	 * In theory the total should be masked off to 8-bits, but in testing
	 * this has occasionally resulted in obviously bogus data passing the
	 * test.  It's unlikely that valid data will total >255 anyway given
	 * the limitations of the DHT11 and the lack of decimal precision in
	 * many implementations.
	 */
	if ((vals[0] + vals[1] + vals[2] + vals[3]) != vals[4]) {
		dbg("Bad checksum: %d:%d:%d:%d %d", v[0], v[1], v[2], v[3], v[4]);
		return false;
	}

	return true;
}

/*
 * Run until the user kills the process, skipping any bad reads.
 */
var v = Buffer(5);

while (true) {
	/*
	 * On some DHT11 the fractional parts are always 0, but print them out
	 * anyway as some models appear to at least support temperatures.
	 *
	 * On a successful read wait 5 seconds before the next sample.  On a
	 * failed read, wait 1 second, which is the minimum time recommended
	 * by the datasheet between samples.
	 */
	if (read_dht11(v)) {
		console.log("Temperature = %d.%dC, Humidity = %d.%d%%",
			    v[2], v[3], v[0], v[1]);
		rpio.sleep(5);
	} else {
		rpio.sleep(1);
	}
}