I have a bunch of sensors from Plantower and Honeywell, I will post some examples in September (I'm away on holidays now) using those sensors and displaying AQI on LCD.

The point I was trying to make in my previous message is you cannot really make this sensor work. With no fan and no decent factory calibration you cannot have reliable data. Look at the data sheet, the curves for min/max value are somewhere around +/-60% from real value. As air flow is generated from heating resistor results will also depend a lot on the temperature.

Hello, I think you would waste much less time buying a Plantower PMS5003 or PMS7003.
This sensor has no fan so it will never be properly calibrated even if you spend hundreds of hours tuning it.

For 12-15$ Plantower sensors will give you relatively accurate results out of the box.

Hello @franz-unix, can't you just bend the output pin of the booster at 90° and solder a wire from it to the raw pin ?

@alexsh1 air quality is more accurate than other sensors but you need to use software from Bosch. It uses temperature, humidity and pressure to adjust measures gas value with some calibration data.
As it needs a lot of memory you will need an SAMD, ESP or NRF5 processor to run it.

The best if you want to run on batteries is to run at 3.3V (or less) and avoid sensors that need 5V.
Everything is much more power hungry at 5V and sensors optimized for low power consumption run at low voltage.

I made a working version of my test board with PCF85176 LCD driver, finally reaching some decently low current consumption for an always on display. There are still some possible gains to make with lower driver voltage, and with a more recent driver that should allow to go as low as 5uA.

So finally I made a new board with my spare PCF85176, still with a 5V uno and it's much better. This is the measurement for the full board: driver, LCD and MCP1703 that generates the 3V VLCD.
I2C lines of the driver are 5V tolerant so I rewired my board to power the driver through the MCP1703. But it seems it's leaching power from the (still at 5V) I2C lines as measurement falls to 1uA
Anyway the contrast is very good as you can see on the picture, and I'm also happy with the design of my "sensor test board", not perfect but flexible enough to test different combinations of voltage sources.

So I made a first test with the embryo of my library, and I realize:

• the soldering was not so good in the end, ic is not perfectly aligned and there are 2 segments connected together. And much worse CLK pin was somehow shorted to VDD meaning the clock was "locked" and display not refreshed (it was only varying voltage once when I sent new data, pretty bad...)
• I have to be more careful when applying rules from datasheet, as pullups on the I2C have too high values and it makes the driver use much more current than expected. I made test quickly with a uno so at 5V supply which is obviously not great, with 10K pullups on the board + 50k pullups on the uno power consumption is 120uA. Without the 10K pullups on the board it goes up to 180uA. And if I disconnect the SDA/SCL lines it goes up a bit over 200uA. So it seems driver is wasting energy driving those pins high. I will solder lower values pullups tomorrow (respecting rules from datasheet this time ), use a 3.3V arduino and and see if it goes better.

@dbemowsk said in Regulators needed:

LP2950

Did you try them in the end or did you chose another one ?

I didn't know about those, they seem very nice if your node needs low current and you have a high input voltage, the max 30V is really nice for nodes like a led dimmer up to 24V.

Nice job, efficient and clean !
I'm tempted to buy this microscope now, so I can make your led light