pymcu-lib-dht

DHT-family temperature & humidity sensor driver for PyMCU — DHT11, DHT22/AM2302, and DHT21/AM2301 — compiled to native machine code, with no interpreter on the chip.

pymcu install dht

Two module names, three APIs, one driver

Which module name and shape you get depends on the stdlib layer your project declares. Native and MicroPython both import from dht — that is real MicroPython's own module name; CircuitPython imports from adafruit_dht, matching that ecosystem's real name instead.

Native (stdlib unset):

from dht import DHT11, DHT22   # DHT21 is also available: an alias of DHT22

sensor = DHT22("PD2")
sensor.measure()
humidity = sensor.humidity()        # uint16, tenths of %RH: 653 = 65.3
temperature = sensor.temperature()  # int16, tenths of C: -55 = -5.5

MicroPython (stdlib = ["micropython"]) — a MicroPython script, unchanged:

from machine import Pin
from dht import DHT11

sensor = DHT11(Pin(2))
sensor.measure()
print(sensor.humidity(), sensor.temperature())

CircuitPython (stdlib = ["circuitpython"]) — an adafruit_dht script, unchanged:

import board
from adafruit_dht import DHT22

sensor = DHT22(board.D2)
print(sensor.temperature, sensor.humidity)   # float, matching upstream

None of those files contains a single match __CHIP__.arch: the layer adapters are plain Python written against each API as it is documented upstream. Everything that has to know about a chip lives in one private module, _dht.core, and the compiler folds it away; everything that has to know what a byte means for a given sensor model lives in _dht.decode, which needs no chip dispatch of its own.

DHT11 vs. DHT22 vs. DHT21

• DHT11 — integer %RH and C, no fractional resolution, no negative temperatures. Cheapest, and the smallest to compile against (see below).
• DHT22/AM2302 — one decimal digit, negative temperatures, wider and more accurate range.
• DHT21/AM2301 — the same protocol and byte layout as the DHT22. This library exposes it as DHT21 = DHT22, a plain alias, not a second implementation.

See docs/sensors.md for the full comparison and docs/protocol.md for the byte-level differences.

Cost

The whole driver is @inline: no object is allocated on the device, and a read compiles down to the bit-banging routine at the call site (or a single shared subroutine where the compiler's outliner judges that cheaper). Nothing is held in SRAM between reads.

dht.py and compat/micropython/dht.py report tenths as a plain int16/uint16 rather than a float, on both DHT11 and DHT22: two isolated probes measured on the emulator put a float-returning driver at 4.9x the flash of an int16-tenths one (680 vs. 140 bytes) for the same logic. See docs/accuracy.md for the measurement. adafruit_dht.py is the one exception — it matches adafruit_dht's real float signature on purpose, since fidelity to that upstream API is the point of the module.

Measured figures per chip are published in the library index.

Supported hardware

The whole DHT family on AVR (ATmega328P and friends), data line on PD2-PD7. Timing is measured on the AVR emulator rather than assumed — see tests/test_timing.py for the start-signal and bit-decode-threshold checks.

Other architectures raise a compile-time error rather than returning something that looks like it worked. Ports are welcome: _dht/avr.py is the whole contract, and _dht/decode.py (the per-model byte decoding) needs no changes for a new port.

Your firmware may use any name it likes at module level: the driver's own parameters are ordinary words (mask, bit, start_low_ms), and a global that collides with one of them no longer reaches inside. That took a compiler fix, so the suite measures it rather than assuming it — see tests/test_timing.py.

See docs/ for the wire protocol, wiring diagrams, accuracy limits, sensor comparison, and a porting guide for a new architecture.

License

MIT.