First read after reset always 28.82 °C 81732.34 Pa

[AndreKR]

Sketch:

#include <Wire.h>
#include <BME280.h>
#include <BME280I2C.h>

BME280I2C bme;

void setup() { 
  Serial.begin(115200);

  pinMode(17, OUTPUT);
  digitalWrite(17, 1);
  delay(50);

  Serial.println(F("Starting I2C"));
  Wire.begin(2,4); // SDA = GPIO2, SCL = GPIO4

  while (!bme.begin()) {
    Serial.println(F("Could not find BMx280 sensor!"));
    delay(500);
  }

  switch(bme.chipModel())
  {
    case BME280::ChipModel_BME280:
      Serial.println(F("Found BME280 sensor, measuring temperature, humidity and pressure"));
    break;
    case BME280::ChipModel_BMP280:
      Serial.println("Found BMP280 sensor, measuring temperature and pressure only");
    break;
    default:
      Serial.println("ERROR: Found unknown sensor");
  }
}

void loop() {
  float temperature = 0;
  float humidity = 0;
  float pressure = 0;
  
  bme.read(pressure, temperature, humidity, BME280::TempUnit_Celsius, BME280::PresUnit_Pa);

  Serial.print(temperature);
  Serial.print(" °C ");
  Serial.print(pressure);
  Serial.print(" Pa");
  Serial.println("");
  
  delay(500);
}

Output:

Starting I2C
Found BMP280 sensor, measuring temperature and pressure only
28.82 °C 81732.34 Pa
25.78 °C 100944.22 Pa
25.79 °C 100947.60 Pa
25.78 °C 100943.45 Pa
25.78 °C 100944.22 Pa
25.78 °C 100944.22 Pa

[coelner]

IMHO I think it is a bug and you can solve it by make two measurements.
(and you get Pa not hPa)

@finitespace

• In BME280I2C.h you use filter_16, in all other settings Filter_Off
• do we now how long a measurement takes? I can't find a value in the datasheet
• Maybe a problem with data register shadowing?
• add in the oversampling enum a zero value for disabling this feature.
• check for register 0xF3 "status" for measuring and wait

[finitespace]

Yes, @coelner is correct, I recently changed the Filter to 16 to provide more stable measurements (I think there is a typo in his post: I think he meant change the filter not the oversampling).

The filter works by aggregating the last X samples together. So a larger filter requires more time to fill the buffer, when you reset, I would guess these values are set to their default, so it would make sense for the first read to give bogus data. I can go look at the datasheet later to see if they have any insight on this.

You can do one of two things:

1. Ignore the first read until the buffer has been populated.
2. Set the filter to off. This will provide the instantaneous value instead of the more reliable aggregated value.

[coelner]

@finitespace the part behind your name was only my thought while searching for the issue. Not directly related to this issue.

[finitespace]

Alright, I have added a routine to initialize the filter buffer. Could you pull down the filtering branch and check if it fixes the problem?

[finitespace]

I have created a pull request for this issue, I will leave it open for a little to allow review before I merge the changes. #81