Bosch BME sensors are crap, why people suggests them?

As title.
Bosch BME sensors are crap, why people suggests them?

I have bought three bme280 sensors from various reputable vendors and then three bme680 sensors from other reputable vendors and I have always the wrong high readings.

Adafruit, sparkfun and dfrobot are the best brand I have but they have the same problems.
This is clearly a sensor problem, a design problem.

I can feel the wrong readings, in my house the ideal temperature for me is around 21°C.

Sometimes the sensors shows 23°C and I feel cold, sometimes the sensors shows 21°C and I feel too warm even with very similar humidity.

Based on an analogic thermometer I have the readings are at least two degrees high.

I have bought those sensors for my thermostat but how can I trust them if the readings are so bad?
How Bosch can sell a similar crap and why so many people recommends those sensors?

Does your analog thermometer have a grade indication, A, B, C or any other markings to indicate its accuracy?

mikb55:
Does your analog thermometer have a grade indication, A, B, C or any other markings to indicate its accuracy?

no but my skin yes, I can assure you that every humans can feel 4°C difference

I have a board, running now, that has both a BME280 and a DS18B20 on it.

The BME280 typically reads 0.2c higher than the DS18B20.

So is the DS18B20 high by 1.8c ?

(I dont have a calibrated theremometer to hand to check)

srnet:
I have a board, running now, that has both a BME280 and a DS18B20 on it.

The BME280 typically reads 0.2c higher than the DS18B20.

So is the DS18B20 high by 1.8c ?

(I dont have a calibrated theremometer to hand to check)

you should trace a graph over an entire day and you will see that the differences are really wider in some moments of the day.

Does the temperature go down when it’s windy?
People have reported self heating problems.

mikb55:
Does the temperature go down when it's windy?
People have reported self heating problems.

temperature goes down as soon as the temperature goes down, the problem is that the readings are always too high even if the sensor is in FREE AIR with quality breakout

sblantipodi:
temperature goes down as soon as the temperature goes down, the problem is that the readings are always too high even if the sensor is in FREE AIR with quality breakout

That was an answer to a different question.
How does the air speed affect the temperature reading?

mikb55:
That was an answer to a different question.
How does the air speed affect the temperature reading?

I'm using it indoor, I haven't experienced noticeable changes with different air speed outside.

Instead of your uncalibrated analogue sensor buy a ds18b20 from a reputable supplier and mount it next to the bme280 so there is thermal contact between the two. Wrap the pair with several layers of electrical tape to act as thermal insulation and place them in a draft free environment such as a plastic food box with a lid.
Set both devices to report their value with maximum possible resolution.

The ds18b20 is factory rated at ±0.5°C accuracy from -10°C to +85°C.
The bme280 is factory rated at ±1.0°C accuracy from 0°C to +65°C.

Monitor for a day. If the two measurement ever differ by more than 1.5°C then you really do have a problem. Anything less than 1.5°C then you have nothing to complain about.

mikb55:
Instead of your uncalibrated analogue sensor buy a ds18b20 from a reputable supplier and mount it next to the bme280 so there is thermal contact between the two. Wrap the pair with several layers of electrical tape to act as thermal insulation and place them in a draft free environment such as a plastic food box with a lid.
Set both devices to report their value with maximum possible resolution.

The ds18b20 is factory rated at ±0.5°C accuracy from -10°C to +85°C.
The bme280 is factory rated at ±1.0°C accuracy from 0°C to +65°C.

Monitor for a day. If the two measurement ever differ by more than 1.5°C then you really do have a problem. Anything less than 1.5°C then you have nothing to complain about.

the problem is that those sensors are accurate in the 0.5/1.0°C range but they are not factory calibrated.
this means that you need to set an initial offset and the initial offset can vary by a wide margin.

real 0°C could be 0°C for the ds18b20 and 4°C for the bme280...
in this example
at 10°C, the ds18b20 could read between 9.5/10.5°C and the BME280 between 13/15°C due to their accuracy but you can understand that the readings are way wrong

this is the real problem of those sensors.

From the BME280 data sheet:

The integrated temperature sensor has been optimized for lowest noise and highest resolution.
Its output is used for temperature compensation of the pressure and humidity sensors and can
also be used for estimation of the ambient temperature.

Note the use of the word 'estimation' in that description.

Don

sblantipodi:
the problem is that those sensors are accurate in the 0.5/1.0°C range but they are not factory calibrated.
this means that you need to set an initial offset and the initial offset can vary by a wide margin.

real 0°C could be 0°C for the ds18b20 and 4°C for the bme280...
in this example
at 10°C, the ds18b20 could read between 9.5/10.5°C and the BME280 between 13/15°C due to their accuracy but you can understand that the readings are way wrong

this is the real problem of those sensors.

My interpretation of the ds18b20 datasheet is that no senor leaves the factory if it falls outside the quoted accuracy. I would be shocked if they don't test every sensor before shipping.

sblantipodi:
at 10°C the ds18b20 could read between 9.5/10.5°C and the BME280 between 13/15°C due to their accuracy

So at 10°C your suggesting the BME280 could indicate 15°C, an error of +5°C ?

Do you think any of the commercial users of the sensor would tolerate that sort of error ?

srnet:
So at 10°C your suggesting the BME280 could indicate 15°C, an error of +5°C ?

Do you think any of the commercial users of the sensor would tolerate that sort of error ?

Probably I made an extreme example just to let you understand that the deviation from the real temperature is not calibrated.
This is what adafruit says for example

I have bme280 and bme680 sensors that are in the range of accuracy but the offset from one and the others is up to 3°c at least

what doe it read sitting on a block of ice?

Or, and I realise its a radical suggestion, what do they read versus a proper calibrated thermometer ?

I have 2 Adafruit BME280s & 2 from Sparkfun. In a recent 45-minute indoor test at "room temperature" in essentially still air, and not in an enclosure, I found that all are within 0.5 deg F (0.3 deg C) of the dry bulb temperature measured with the three mercury thermometers in my engineering office's Taylor sling psychrometer set. (They were between 0.3 deg F and 0.5 deg F higher than the mercury thermometers.)

FWIW, the mecury thermometers are about 12 inches long, are graduated in 0.5 deg F increments, and can be read to the nearest 0.1 deg F "by eye" (if you squint :slight_smile: ). All three mercury thermometers are within 0.1 deg F of each other. There are three because the set includes a spare.

It is really tough with the BME sensors to keep self-heating under control. Either the sensor itself or other nearby components.

I love them and will use them in many more projects.

MorganS:
It is really tough with the BME sensors to keep self-heating under control.

:confused: In my test (post #17), all four were between 0.3 and 0.5 deg F higher than the reference devices, well within the +/- 1.8 deg F (+/- 1.0 deg C) mentioned in post #9 as the accuracy given in the datasheet.