BSEC + DevKit + BME688: Odd Gas Resistance Values

[ilcnet]

Hi dear BSTRobin
I used this code for Bme688 but when sensor sense the smoke co2 and dirty weather the values going down.at the first co2 value is 500 and iaq value is 25 a few minutes later the values are stable in 530 and 30 and this is normal but when i burn a paper or spray a ethanol in front of sensor the values start to going down! After sensing the values must be go to up not go to down, how i can solve this problem

Code:

https://paste.ofcode.org/icexEvhNqmqa657kHPgwPa

[S_Brohl]

Hi ilcnet,

From my understanding of your problem it sounds like you are describing how the sensor should actually be working. The values should go down as you expose the sensor to different gases. This is essentially a resistance operated sensor. As the sensor reacts to different substances the gas resistance values should drop, and as it is unreactive to the background air the resistance values should rise.

Have a look at my picture below, as I finally believe I have figured out why my values are saturated: 

The saturated values of 1.024e+08 represent unreactive background air at heater profile step index 0, and then as I introduce my substance the gas reacts at this step index creating drops in resistance. This in turn creates the "fingerprint" for this substance which I can then use to further train my algo.

Hope this helps you to better understand as I am also learning how this works myself. 🙂

[ilcnet]

Hi dear S_Brohl

Thanks for your response, yes i understand you say that Bme688 is a resistance sensor and for this reason that's true to values going down when sensing gases , but when i use another code such as adafruit library, the iaq value are between 7 and 25 in normal status, and after sensing gases the value going up to 380 450 and etc. So adafruit scenario is understandable for us, but with bsec library when sense gases it down to 25 and fixed.can i change the codes to working reverse? Thanks

[S_Brohl]

Ah, I read your post wrong, I thought you were referring to the gas resistance values. You are correct there.

I believe I have seen a similar problem at the early stages of my testing with the IAQ and CO2 values. Did you allow the sensor 72 hours to burn-in? Also, do you have the state saving periodically? Could just be that your sensor needs more time to calibrate.I'd output the stabilization and run-in values as Robin suggested for my problem as well.

Also, allowing the sensor to warm-up for 5 minutes after startup also seems to help a lot too.

I'm also using the Adafruit breakout board as well as the devkits, however I've built off the sample Arduino code for the BSEC2 library to fit my needs so I no longer collect the IAQ, CO2, or Breathe-VOC. It wasn't necessary for what I'm using it for, but I do recall facing a lot of issues with static values for those outputs in the beginning and they were more or less remedied by calibrating the sensor longer and ensuring the state is saved off to the EEPROM of my ESP8266.

When you flash your module I've noticed it does wipe the part of the EEPROM that saves the state, so you will have to re-calibrate each time you flash (at least that is how it has been in my case).

[S_Brohl]

Hi Robin,

It appears everything is alright after further testing. If you look at the picture I provided to ilcnet on page 3 you can see where those saturated resistance values begin to drop when my substance is added to my test setup.

I believe the heater profile used (322) is just finally "acclimated" to my environment and so there is nothing for the sensor to react to at this heater profile step. I have performed another test in a different environment to prove this hypothesis and it does appear that when in a completely different environment these saturated values no longer appear (likely due to different gas compositions in the environment, and the need for the sensor to re-stabilize).

So for now I'd say everything is working as it should, and I am back on track to collect more data. 

Should anything else come up, I'll be sure to reach out. Thanks again for your support!

Also, it would be nice if there was some kind of video or addition to the documentation that illustrates what a sensor(s) should see when properly stabilized and calibrated to an environment.