With this script you can read out the value of your LYWSD03MMC sensor, e.g. with Raspberry PI. Note Raspbery Pi 4 has a very limited bluetooth range. PI Zero W gives much longer range.

This sensor doesn't transmit its values in the advertisment data, like the LYWSDCGQ Bluetooth thermometer. This is more privacy friendly since no one can sniff your temperature readings. On the other side this means you have to establish a bluetooth connection with the device to get the data. When you're connected no other connection is accepted, meaning if you hold the connection no other can readout your temperature and humidity.

Once you're connected the LYWSD03MMC it advertises its values about every 6 seconds, so about 10 temperature/humidity readings per minute.

You need Python3 3.7 or above because of the dataclasses used in the Callback Function. If you don't have Python 3.7 please take the previous version from here https://raw.githubusercontent.com/JsBergbau/MiTemperature2/5d7b215d7b22d4c21d9244f8a4102513b928f2c7/LYWSD03MMC.py This version is a bit behind and connection error handling has a bug. If you really need this script, please open and issue and I'll post a new bugfree version.

For example Raspbian Stretch has only Python 3.5.3. If you like to upgrade your Distribution to current Buster release follow this Tutorial https://pimylifeup.com/upgrade-raspbian-stretch-to-raspbian-buster/ If doing so: Omit the rpi-update step.

If you like installing/compiling Python3.7 please take a look at this tutorial https://gist.github.com/SeppPenner/6a5a30ebc8f79936fa136c524417761d However it took about 5 hours to compile/run the regressiontests on a Raspberry PI3B. I use this compiled version directly without install. If you do, too, you have to change the first line in the script, pointing to your compiled Python version. For bluepy you can copy the bluepy-folder from home/pi/.local/lib/python3.7/site-packages/bluepy to Python-3.7.4/Lib and do a chmod +x bluepy-helper in Python-3.7.4/Lib/bluepy

Prequisites: python3 bluez python3-pip bluepy requests install via

sudo apt install python3 bluez python3-pip

pip3 install bluepy requests

If you use integrated MQTT client paho-mqtt is needed. Install via

pip3 install paho-mqtt

Instead of pip3 it may be just pip, depeding of your installation.

Additional requirements if you want to use the atc version. If you don't use ATC version, please ignore this section.

apt install bluetooth libbluetooth-dev
pip3 install pybluez

Bluetooth LE Scanning needs root. To run the script for AT with normal user rights, please execute

sudo setcap cap_net_raw,cap_net_admin+eip $(eval readlink -f `which python3`)

After a Python upgrade, you have to redo the above step.

---------------------------------------------
MiTemperature2 / ATC Thermometer version 3.0
---------------------------------------------
usage: LYWSD03MMC.py [-h] [--device AA:BB:CC:DD:EE:FF] [--battery ]
                     [--count N] [--interface N] [--unreachable-count N]
                     [--mqttconfigfile MQTTCONFIGFILE] [--round] [--debounce]
                     [--offset OFFSET] [--TwoPointCalibration]
                     [--calpoint1 CALPOINT1] [--offset1 OFFSET1]
                     [--calpoint2 CALPOINT2] [--offset2 OFFSET2]
                     [--callback CALLBACK] [--httpcallback URL] [--name NAME] [--skipidentical N]
                     [--influxdb N] [--atc] [--watchdogtimer X]
                     [--devicelistfile DEVICELISTFILE] [--onlydevicelist]
                     [--rssi]

optional arguments:
  -h, --help            show this help message and exit
  --device AA:BB:CC:DD:EE:FF, -d AA:BB:CC:DD:EE:FF
                        Set the device MAC-Address in format AA:BB:CC:DD:EE:FF
  --battery [], -b []   Get estimated battery level, in ATC-Mode: Get battery
                        level from device
  --count N, -c N       Read/Receive N measurements and then exit script
  --interface N, -i N   Specifiy the interface number to use, e.g. 1 for hci1
  --unreachable-count N, -urc N
                        Exit after N unsuccessful connection tries
  --mqttconfigfile MQTTCONFIGFILE, -mcf MQTTCONFIGFILE
                        specify a configurationfile for MQTT-Broker

Rounding and debouncing:
  --round, -r           Round temperature to one decimal place
  --debounce, -deb      Enable this option to get more stable temperature
                        values, requires -r option

Offset calibration mode:
  --offset OFFSET, -o OFFSET
                        Enter an offset to the reported humidity value

2 Point Calibration:
  --TwoPointCalibration, -2p
                        Use complex calibration mode. All arguments below are
                        required
  --calpoint1 CALPOINT1, -p1 CALPOINT1
                        Enter the first calibration point
  --offset1 OFFSET1, -o1 OFFSET1
                        Enter the offset for the first calibration point
  --calpoint2 CALPOINT2, -p2 CALPOINT2
                        Enter the second calibration point
  --offset2 OFFSET2, -o2 OFFSET2
                        Enter the offset for the second calibration point

Callback related arguments:
  --callback CALLBACK, -call CALLBACK
                        Pass the path to a program/script that will be called
                        on each new measurement
  --httpcallback URL, -http URL
                        Pass the URL to a program/script that will be called
                        on each new measurement
  --name NAME, -n NAME  Give this sensor a name reported to the callback
                        script
  --skipidentical N, -skip N
                        N consecutive identical measurements won't be reported
                        to callbackfunction
  --influxdb N, -infl N
                        Optimize for writing data to influxdb,1 timestamp
                        optimization, 2 integer optimization

ATC mode related arguments:
  --atc, -a             Read the data of devices with custom ATC firmware
                        flashed, use --battery to get battery level
                        additionaly in percent
  --watchdogtimer X, -wdt X
                        Re-enable scanning after not receiving any BLE packet
                        after X seconds
  --devicelistfile DEVICELISTFILE, -df DEVICELISTFILE
                        Specify a device list file giving further details to
                        devices
  --onlydevicelist, -odl
                        Only read devices which are in the device list file
  --rssi, -rs           Report RSSI via callback

Note: When using rounding option you could see 0.1 degress more in the script output than shown on the display. Obviously the LYWSD03MMC just truncates the second decimal place.

Reading the battery level with the standard Bluetooth Low Energy characteristics doesn't work. It always returns 99 % battery level. Or to be correct, sometimes 10 % when the battery is really empty, see JsBergbau#1 (comment) . But often before that device just shuts down before it can report another battery level. With every measurement the Aqara sensor also transmits the battery voltage. This voltage is transformed into a battery level 3.1V are 100%, 2.1V 0%.

The --count option is intended to save even more power. So far it is not proven, that only connecting at some interval will actually save power. See this discussion JsBergbau#3 (comment)

With the --interface option you specify the number of the bluetooth adapter to use. So --interface 1 for using hci1

With --influxdb 1 you can use a influxdb optimized output. In this mode a timestamp with the current data is sent every 10 seconds to influxdb. Or technically speaking, each received measurement is snapped to a grid of 10 seconds. Don't use this feature together with --skipidentical otherwise it won't help. To use RLE compression for timestamps influxdb requires all 1000 timestamps which are mostly in a block to have the same interval. Only one missing timestamp leads to s8b compression for timestamps. Since influxdb handles identical values very efficiently you save much more space by writing every 10 seconds instead of skipping identical values. Without RLE 1000 timestamps needed about 1129 Bytes of data in my measurement. With RLE its only 12 Byte. Of course there are now more measurements stored in influxdb, but still overall size in influxdb is still lower. Depends also environment, of course. With a very steady environment and very seldom writing identical valus then size in influxdb would be smaller not writing every 10 seconds, of course. Integer optimizsation --influxdb 2is not implemented yet.

--unreachable-count N, -urc N Use this option when you want to exit your script after collection the measurement but your sensor is somehow not reachable. Then after the specified number of failed connection tries the script will exit.

Thanks to https://github.com/pvvx/ATC_MiThermometer there is an alternative firmware which sends out the measurements as Bluetooth Low Energy Advertisments. In this mode you don't have to connect to the sensor. This saves a lot of power, especially in cases where the signal strength is low, see JsBergbau#32 I've also noticed a higher range. In addition you can have multiple receivers, see Node-RED section https://github.com/JsBergbau/MiTemperature2#node-red-flows

For longer batterylife and higher stability ATC mode is recommended. In the flasher Webpage selecting only Atc1441 as Advertising type is recommended. You can than increase the Advertising interval to save power for sensors with good reception. For sensors with weaker reception I would keep the default of 2500 ms. You could even increase it for sensors with very bad reception.

In ATC mode the script listens for BLE advertisments and filters for ATC flashed LYWSD03MMC sensors. So you start only one instance of this script and it reads out all your sensors. You can have multiple receivers and thus have a kind of cell network and your sensors are portable in a quite wide range. Use it optimally with influxdb and --influxdb 1. With this option timestamps are snapped to 10s and since influxdb only stores one value for one timestamp you won't have duplicate data in your database.

ATC firmware gives temperature only with one decimal place, so rounding and debouncing options are not available. This is no real disadavantage because accuracy is impaired because of a lacking capciator, see pvvx/ATC_MiThermometer#11 (comment)

--watchdogtimer X sometimes your device leaves BLE scanning mode, then you won't receive any data anymore. To avoid this after X seconds without receiving any BLE packet (not only from ATC sensors) BLE scanning mode is re-enabled. If you have configured your ATC LYWSD03MMC to advertise new data every 10 seconds, I recommend a setting of 5 seconds, so --watchdogtimer 5. On a Raspberry PI Zero W polling 10 sensors (2 are currently unreachable) this re-enabling BLE scan can happen a few times per minute. When polling 8 reachable sensors it happens less frequent but still at least once or twice a minute.

One note about new advertising data: The ATC LYWSD03MMC sends out the data about every 2,5 seconds. After 10 seconds (4 advertising periods) it advertises new data and to detect this, it increases the paket counter. Only the values of first paket of this series with the same paketcounter is displayed and reported by callback, since the data in one series is identical.

--devicelistfile <filename> Use this option to give your sensors a name/alias. This file can also be on a network drive. So you can keep it up to date on a single place for multiple receivers. Also in this file is space for calibration data and you can give each device it's own MQTT topic.

Note: ATC firmware shows other humidity values than the stock Xiaomi firmware, so you have to re-calibrate your sensors. The temperature value is unchanged compared to the Xiaomi sensor firmware.

An example is given in the sensors.ini file in this repository. It is quite self explaining

[info]
info1=now all available options are listet. If offset1, offset2, calpoint1 and calpoint2 are given 2Point calibration is used instead of humidityOffset.
info1= Note options are case sensitive
;info4=Use semicolon to comment out lines
sensorname=Specify an easy readable name
humidityOffset=-5
offset1 = -10
offset2 = 10
calpoint1 = 33
calpoint2 = 75
topic=This sensor data will be published with this name when using integrated MQTT
;If no topic is given, then default topic from MQTT config is used

;[A4:C1:38:DD:EE:FF]
; sensorname=Bathroom
; humidityOffset=-30
; offset1 = -10
; offset2 = -10
; calpoint1 = 33
; calpoint2 = 75
; topic = basement/Bathroom

[A4:C1:38:1F:4D:81]
sensorname=Living Room
offset1 = -2
offset2 = 2
calpoint1 = 33
calpoint2 = 75
topic=basement/Living_Room

--onlydevicelist Use this option to read only the data from sensors which are in your device list. This is quite useful if you have some spare sensors and you don't want your database get flooded with this data.

--rssi Reports the RSSI via callback

--battery is also available in ATC mode. Instead of estimating the battery level like in connection mode, the batterylevel in percent is reported exactly as on device's screen.

Hint for storing the data in influx: When you have configured an advertisment interval of 10 seconds: Ideally store one measurement every 25 seconds to use very efficient RLE compression for your measurements. With storing the data every 25s, almost every timestamp is stored. This leads to RLE compression of the timestamp thus saving a lot of space in influxdb. With an interval of 20 seconds in tests it occured quite often, that timestamp slots were not filled and thus no RLE compression can be used.

With original firmware where you connect to each sensor every 6 seconds an measurement is sent and storing every 10 seconds a measurement is a good value.

ATC mode uses passive scanning for saving battery life of the sensors. Read more about this here JsBergbau#41 (comment)

Since Version 3 of MiTemperature2 there is built in support for MQTT. Especially if you receive a lot of sensors executing a callback for each measurement is quite expensive. For a raspberry PI Zero W for example with a few sensors I have an average CPU usage of 35 %, after using built in MQTT support only about 10 % CPU is used on average and there are also other services running. Quite a lot performance measurements were made to keep CPU usage for MQTT transmit low.

With callback every single message spawns a new process, a new TCP connection is established and then closed again. With integrated MQTT support one connection is opened at startup and then maintained all the time, saving a lot of CPU cycles and network overhead.

A testscript was made and tested on a raspberry PI4 booted with opion force_turbo = 1. With that option CPU runs always at 1500 MHz so measurements are easier to compare. Because perf stat wouldn't report the used CPU cycles measurement was taken via time command. 2000 generated values were sent and the time measured. With callback to Node-RED using curl time was about 77 seconds. Using mosquitto_pub and the same JSON string time was about 42. So mosquitto_pub and MQTT is a lot faster and efficient than using HTTP via curl. Then finally 2000 values were sent via integrated MQTT support. This took about 2 seconds whereas most of that time was required to import all libraries and so on. Transmitting 10.000 values via integrated MQTT only took about 4 seconds. So there is a massive performance plus with integrated MQTT.

Data is transmitted that way (but can be changed to subtopic mode, see below):

{
  "temperature": 12.5,
  "humidity": 58,
  "voltage": 2.882,
  "calibratedHumidity": 57,
  "battery": 75,
  "timestamp": 1619805618,
  "sensor": "Living Room",
  "rssi": -88,
  "receiver": "raspberrypi"
}

or when you don't use a device list file then in sensorname there is the mac address of the sensor.

If you don't have provided any calibration data calibratedHumidity is the humidity value. So you can always use calibratedHumidity as the humidity value.

These JSON messages are all send via MQTT QOS 1. So if connection to your broker gets lost, data isn't lost. Because it contains the timestamp you won't lose one single value. Currently the implementation of the used library has a max message queue of 65555. So when having 10 thermometers and each transmitting every 10 seconds a new value this makes one message per second, so your broker can be unreachable for about 18 hours bevore you will lose messages. If you have a scenario where 65555 messages are too few and data could be lost, please open an issue and I'll try to find a solution.

Important: To also not lose any message on the MQTT receiver, please also set QOS Level 1 and a client ID and disable clean-session flag. This way your MQTT broker stores all messages until your receiver is available again. Mosquitto stores default 1000 messages maximum per client with unlimited total size. You can change this via max_queued_messages see https://mosquitto.org/man/mosquitto-conf-5.html In the Node-RED sample flow (see below) the settings are setup this way.

The example mqtt.conf file looks like

[MQTT]
broker=127.0.0.1
port=1883
topic=ATCThermometer
;username=
;password=
receivername=
;subtopics=temperature,humidity


[DEFAULT]
;Do not change here, leave it as it is
broker=127.0.0.1
username=
password=
port=1883
topic=ATCThermometer
;lastwill message
lastwill=
;lastwill topic
lwt=
clientid=
;If not specified, hostname is used as receivername
receivername=
;enable subtopics like temperature and humidity, combine with comma without spaces so "temperature,humidity"
;to disable JSON output, add "nojson"
;subtopics=nojson,temperature,humidity will send temperature and humidity
subtopics=

broker, username, password and port are self-explaining. Also lastwill and lastwill topic (lwt) can be easily understood. Topic is the default topic under which data of all thermometers is published. You can override this per sensor, see above in the sensors.ini configuration. This default topic is the same as in the Node-RED flow example to receive all thermometers at one place.

If you need seperate readings like sensors/basement/kitchen/temperature and sensors/basement/kitchen/humidity you can use the subtopics option. So if you need voltage, temperature and humidity, subtopics option would look like subtopics=temperature,voltage,humidity and it would be published under

ATCThermometer/temperature
ATCThermometer/voltage
ATCThermometer/humidity

or when you have specified a topic for your particular sensor like basement/livingroom/ then it would like

basement/livingroom/temperature
basement/livingroom/voltage
basement/livingroom/humidity

When using subtopics options you should configure a topic for each sensor otherwise values would get mixed under topic ATCThermometer/temperature, of course.

Still there is also the JSON output at topic ATCThermometer respectively basement/livingroom To disable JSON output in this case, append nojson to suptopics, so in this case subtopics=temperature,voltage,humidity,nojson

Since the values transmitted as subtopics contain no timestamp it makes no sense to cache them in case of a transmission failure. So these subtopic values are only sent as MQTT QOS level 0.

From a point of efficiency subtopics are less efficient because each subtopic needs a seperate transmission with its own overhead. So please use this feature only when really needed like with software like Homie that doesn't support JSON format.

Use sudo hcitool lescan --duplicate to get the MAC of your Sensor. This sensor only sends its measurements only via notifications. There are quite often notifications because the temperature is measured with a precision of 2 decimal places, but only one shown on the display (and this value is truncated, see above). Trying to directly read/poll the characteristics returns always zeroes.

The temperature values often change between the same values. To get cleaner temperature curves a debouncing function has been implemented. See here JsBergbau#2 for more info.

When looking at the specifications this LYWSD03MMC Sensor is specified from 0 °C to 60 °C. The LYWSDCGQ (the Bluetooth Temperatur sensor with the round display and an AAA battery) is specified from -9.9. I can confirm the LYWSD03MMC also goes down to -9.9 °C. At colder temperatures it only shows an "L". But even at lower temperatures the correct temperature is still sent! So you even could use ist to watch the temperature in your freezer which is a lot below -9.9 °C. However batterylife may be significantly reduced at those low temperatures.

There is currently no way to detect a too high battery drain except having empty batteries in less than 2 month. If you encouter lots of empty batteries, please reduce distance between LYWSD03MMC sensor and your Bluetooth receiver. With a voltage drop of 0.1 V in 1.5 month everything is perfectly fine. If it is a bit more, don't worry. Can be a button cell of lower quality. For more infos please visit JsBergbau#32

 ./LYWSD03MMC.py -d AA:BB:CC:DD:EE:FF -r -b
Trying to connect to AA:BB:CC:DD:EE:FF
Temperature: 20.6
Humidity: 54
Battery voltage: 2.944
Battery level: 84

Temperature: 20.6
Humidity: 54
Battery voltage: 2.944
Battery level: 84

Temperature: 20.6
Humidity: 54
Battery voltage: 2.944
Battery level: 84

Temperature: 20.6
Humidity: 54
Battery voltage: 2.944
Battery level: 84

Temperature: 20.6
Humidity: 54
Battery voltage: 2.944
Battery level: 84

Temperature: 20.6
Humidity: 54
Battery voltage: 2.944
Battery level: 84

Temperature: 20.6
Humidity: 54
Battery voltage: 2.944
Battery level: 84

Temperature: 20.6
Humidity: 54
Battery voltage: 2.944
Battery level: 84

Temperature: 20.6
Humidity: 54
Battery voltage: 2.944
Battery level: 84

../LYWSD03MMC.py --atc --mqttconfigfile mqtt.conf --devicelistfile MeineSensoren.ini
---------------------------------------------
MiTemperature2 / ATC Thermometer version 3.0
---------------------------------------------
Script started in ATC Mode
----------------------------
In this mode all devices within reach are read out, unless a devicelistfile and --onlydevicelist is specified.
Also --name Argument is ignored, if you require names, please use --devicelistfile.
In this mode rounding and debouncing are not available, since ATC firmware sends out only one decimal place.
ATC mode usually requires root rights. If you want to use it with normal user rights,
please execute "sudo setcap cap_net_raw,cap_net_admin+eip $(eval readlink -f `which python3`)"
You have to redo this step if you upgrade your python version.
----------------------------
Power ON bluetooth device 0
Bluetooth device 0 is already enabled
Enable LE scan
scan params: interval=1280.000ms window=1280.000ms own_bdaddr=public whitelist=no
socket filter set to ptype=HCI_EVENT_PKT event=LE_META_EVENT
Listening ...
BLE packet: A4:C1:38:AA:BB:CC 00 1110161a18a4c138aabbcc00c42914094c38 -87
Temperature:  19.6
Humidity:  41
Battery voltage: 2.38 V
RSSI: -87 dBm
Battery: 20 %
Humidity calibrated (2 points calibration):  40

BLE packet: A4:C1:38:AA:BB:CC 00 1110161a18a4c138aabbcc00b22c530b84c3 -93
Temperature:  17.8
Humidity:  44
Battery voltage: 2.948 V
RSSI: -93 dBm
Battery: 83 %
Humidity calibrated (2 points calibration):  44

MQTT connected with result code 0
MQTT published, Client: <paho.mqtt.client.Client object at 0xb60b0bf0>  Userdata: None  mid: 1
MQTT published, Client: <paho.mqtt.client.Client object at 0xb60b0bf0>  Userdata: None  mid: 2
BLE packet: A4:C1:38:AA:BB:CC 00 1110161a18a4c138aabbcc00c7343c0ac79b -62
Temperature:  19.9
Humidity:  52
Battery voltage: 2.759 V
RSSI: -62 dBm
Battery: 60 %

MQTT published, Client: <paho.mqtt.client.Client object at 0xb60b0bf0>  Userdata: None  mid: 3
BLE packet: A4:C1:38:AA:BB:CC 00 1110161a18a4c138aabbcc00b72a510b7499 -84
Temperature:  18.3
Humidity:  42
Battery voltage: 2.932 V
RSSI: -84 dBm
Battery: 81 %
Humidity calibrated (2 points calibration):  42

MQTT published, Client: <paho.mqtt.client.Client object at 0xb60b0bf0>  Userdata: None  mid: 4
BLE packet: A4:C1:38:AA:BB:CC 00 1110161a18a4c138aabbcc00c4344e0b5811 -85
Temperature:  19.6
Humidity:  52
Battery voltage: 2.904 V
RSSI: -85 dBm
Battery: 78 %
Humidity calibrated (2 points calibration):  52

MQTT published, Client: <paho.mqtt.client.Client object at 0xb60b0bf0>  Userdata: None  mid: 5
BLE packet: A4:C1:38:AA:BB:CC 00 1110161a18a4c138aabbcc00c835150957bd -50
Temperature:  20.0
Humidity:  53
Battery voltage: 2.391 V
RSSI: -50 dBm
Battery: 21 %
Humidity calibrated (2 points calibration):  53

If you like gatttool you can use it, too. However it didn't notice when BT connection was lost, while this Python-Script automatically reestablishes the connection.

gatttool -I
connect AA:BB:CC:DD:EE:FF
#enable notifications
char-write-req 0x0038 0100
#enable lower power mode, please don't do when you want to interact with the device, because often then connection gets lost
char-write-req 0x0046 f40100
#Read battery-Level, just for reference, since it is always 99 consider: note value is in Hex format
char-read-hnd 0x001b

Strictly speaking enabling notifications every time is not necessary since the device remembers it between connects. However to make it always work, the Python-Script enables them upon every connection establishment.

Notification format Notification handle = 0x0036 value: f8 07 4a d6 0b f8 07 is the temperature as signed INT16 in little endian format. Divide it by 100 to get the temperature in degree Celsius 4a is the humidity. Only integer output :( d6 and 0b are the battery voltage in Millivolts in little endian format.

Lower power mode: To safe power the connection interval is reduced. For more details, please see this Issue JsBergbau#18 (comment)

There can be done a lot more with that sensor. It stores highest and lowest values at hour level, has an integrated realtime clock and a few things more like changing the values for the comfort icon on the display. Credits go to jaggil who investigated on this and documented it well. You find the results here JsBergbau#1 Just search for jaggil.

Sometimes script fails to connect and tries to connect forever. Just exec killall bluepy-helper You can even do this while script is running. It will disconnect, but recovery automatically.

Since version 1.1 there is a watchdog-Thread checking when connection is lost for at least 60 seconds and then killing the corresponding bluepy-helper, so that other connections aren't affected. This is a workaround for an obvious bug in bluepy. This bug only occured so far when trying to (re)connect. Then this call to bluepy blocks sometimes forever.

If that doesn't help, a problem with the bluetooth stack could be the cause. To resolve:

sudo hciconfig hci0 down
sudo hciconfig hci0 up

Sometimes bluetooth gets stucks, especially if you have other software accessing/using bluetooth on your devices. After a reboot everything was fine again. If that happens a lot, you can use an additional Blueooth receiver and use it with the --interface option.

Note: If you have calibrated your sensors and flash ATC firmware, you have to calibrate them again.

Especially humidity value is often not very accurate. You get better results if you calibrate against a known humidity. This can be done very easy with common salt (NaCl). Make a saturated solution and put it together with the Xiaomi Bluetooth thermometer in an airtight box. Ensure that no (salt) water gets in contact with the device. Saltwater is very corrosive. Wait about 24 - 48 hours with a constant temperature. You should now have about 75 % relative humidity. I don't know how long it takes for the sensors to drift. So I will redo this procedure about every year.

A quite constant temperature while calibration is very important, because with temperature also humidity changes and it takes some time for the system to rebalance the humidity. In my experiments at 20 °C room temperature it took about 48 hours until humidity readings were quite stable and didn't change anymore. So give it time. If you take the sensors out of calibration humidity too early they haven't reached the final value yet.

E.g. mine shows 79 % RH when actually there is 75 %. Excecute the script with -o -4 to substract 4 from the readout value.

The offset is not linear over the whole humidity values, see https://www.rotronic.com/media/productattachments/files/c/a/capacitive_humidity_sensor_final.pdf Page 2.

Linearity Errors. The typical response of a relative humidity capacitive sensor (between 0 and 100 percent RH) is not linear. Depending on the correction made by the electronic circuits, the instrument may have a linearity error.

So you should calibrate at another point. MagnesiumChloride is recommended as giving about 33% RH at 20 °C. Please use very pure MgCl. It is also sold as bath salt and stuff like that. For high accuracy please use a purity > 99 %.

Also Calciumchloride is suitable, but the humidity depends more on temperature. Be sure to have 20 °C. https://www.salzwiki.de/index.php/Deliqueszenzfeuchte CaCl is often found in these small non-electric dehumidifiers which you can refill with refill packs.

My Xiaomi Bluetooth thermometer shows 39% RH at 33% RH. So wie here have an offset of 6. Another hygrometer show 69 % at 75% RH and 33% RH at 33% RH. So offset +6 at 75% TH and offset 0 at 33% RH. Example for the Xiaomi to use 2 point calibration: At 75% RH you have to substract 4 from the readout value, at 33% RH you have to substract 6.

-2p -p2 75 -o2 -4 -p1 33 -o1 -6

-2p: Enables 2 point calibration
-p2 75: Point 2 at 75% RH
-o2 -4: Offset -4 at Point 2
-p1 33: Point 2 at 33% RH
-o1 -6: Offset -6 at Point 2

Note the values in between are interpolated linear and the result is rounded to the nearest whole number. It makes no sense to give floatingpoint number when the input is none.

Output example:

./LYWSD03MMC.py -d AA:BB:CC:DD:EE:FF -2p -p2 75 -o2 -4 -p1 33 -o1 -6
Trying to connect to AA:BB:CC:DD:EE:FF
Temperature: 20.62
Humidity: 54
Battery voltage: 2.944
Calibrated humidity: 49

Temperature: 20.6
Humidity: 54
Battery voltage: 2.944
Calibrated humidity: 49

Temperature: 20.61
Humidity: 54
Battery voltage: 2.944
Calibrated humidity: 49

Via the --callback option a script can be passed to sent the data to. Example ./LYWSD03MMC.py -d AA:BB:CC:DD:EE:FF -2p -p2 75 -o2 -4 -p1 33 -o1 -6 --name MySensor --callback sendToFile.sh If you don't give the sensor a name, the MAC-Address is used. The callback script must be within the same folder as this script. The values outputted depends on the options like calibration or battery. So the format is printed in the first argument. Example callback

#!/bin/bash
# This is quite useful for testing
echo $@ >> data.txt
exit 0

Gives in data.txt sensorname,temperature,humidity,voltage,humidityCalibrated,timestamp MySensor 20.61 54 2.944 49 1582120122

Whereas the timestamp is in the Unix timestamp format in UTC (seconds since 1.1.1970 00:00).

Via the --httpcallback option a formatted URL can be passed to sent the data to. Example ./LYWSD03MMC.py -d AA:BB:CC:DD:EE:FF -2p -p2 75 -o2 -4 -p1 33 -o1 -6 --name MySensor --httpcallback "http://127.0.0.1:8080/myscript?name={sensorname}&temp={temperature}&hum={humidity}&bat={batteryLevel}"

This will call the script at the given URL and fill in the formatted values. Just like the built in MQTT support this is less expensive than executing a script via the --callback option every time a measurement is received. Supported values are: sensorname, temperature, humidity, voltage, humidityCalibrated, batteryLevel, rssi, timestamp.

There is an option not to report identical data to the callback. To distinguish between a failure and constantly the same values are read, the option takes the number after which identical measurements the data is reportet to the callback. Use the --skipidentical N for this. E.g. --skipidentical 1 means 1 identical measurement is skipped, so only every second identical measurement is reportet to callback. I recommend numbers between 10 and 50, giving at least every minute respectively 5 minutes a call to the callback script (With 10 and 50 the actual time is slightly higher than 1 respectively 5 minutes). It is recommended to use the --round and --debounce option, otherwise there is a lot of noise with changing the temperature. See JsBergbau#2

All data received from the sensor is stored in a list and transmitted sequentially. This means if your backend like influxdb is not reachable when a new measurement is received, it will be tried again later (currently waiting 5 seconds before the next try). Thus no data is lost when your storage engine has some trouble. There is no upper limit (the only limit should be the RAM). Keep this in mind when specifing a wrong backend.

"sendToInflux.sh" is an example script for sending the data to influxdb via http-API. Precision was set to the level of seconds. This gives better compression ratios in influxdb.

Read instruction about integartion with Prometheus Push Gateway

Finally there are flows for Node-RED. Especially if you have multiple receivers this is quite comfortable to manage the name and calibration data of your sensors at one place in Node-Red. No need to change configuration in any of your receivers.

This solution makes it also very easy to have multiple receivers and you can easily move the devices around between them. As long as one device reaches one receiver everything will work. If it reaches multiple receivers you can even reboot them without data loss.

There are two slightly different versions. Node-RED flows Callback mode.json sends directly via curl and HTTP the data to Node-RED. This version is only intended when you don't have a MQTT broker.

Node-RED flows MQTT mode.json is the more efficient version because it is intended to be used with the integraded MQTT support and obviously a MQTT broker.

To use MiTemperature2 script with Node-RED import the flows from one of these files.

With that flow you can even start the script in ATC-Mode via Node-RED. If you are user pi and in your home directory, clone this Repo via git clone https://github.com/JsBergbau/MiTemperature2, make LYWSD03MMC.py executable (also ./sendToNodeRed.sh when using callback version) and the preconfigured path to MiTemperature2 script doesn't have to be changed.

The Node-RED flow is documented with comments for easy usage. If theres missing some information, just open an issue and I'll have a look.

For easily switching between filtering in Node-Red or directly in the MiTemperature2 script, there are two scripts included.

usage: iniToJSON.py [-h] --readfile filename [--writefile filename]

optional arguments:
  -h, --help            show this help message and exit
  --readfile filename, -rf filename
                        Specify filename of ini-File to convert
  --writefile filename, -wf filename
                        Specify filename of json-File to convert

Use these to convert between ini and JSON for usage in Node-RED.

usage: jsonToIni.py [-h] --readfile filename --writefile filename

optional arguments:
  -h, --help            show this help message and exit
  --readfile filename, -rf filename
                        Specify filename of json-File to convert
  --writefile filename, -wf filename
                        Specify filename of json-File to convert

Before using this callback script, please check if it is possible to do the same with integrated MQTT support.

etpedro https://github.com/etpedro has added a sendToMQTT_HA callback script that sends the data to an MQTT server and leverages Home-Assistant MQTT Autodiscovery creating a device for each sensor under MQTT integration. Each device has 4 entities: temperature, humidity, battery percentage and battery voltage.

Replace each variable according to your environment:

mqtt.host - MQTT Host

mqtt.username - MQTT Username

mqtt.passwd- MQTT Password