BME680 measurement results interpretation. About repetability/representativity/accuracy ?

[wtc-peg]

Hello,

I’m running BME680 in ULP mode, using BSEC library on stm32,  on two different sensors, could someone help me to explain behaviors ?
The two sensors are at the same place in a room.
You can see joined pictures of measurements (T,H,VOC,CO2), for 24hours in an office with nobody inside.
Both sensors were running for 3 days before this day of measurements.
The global sensor is also equipped with a CO2 NDIR sensor, that show a quit stable CO2 level all along these 24hours.
At the end of period, for both sensors, you can see arrival of 3 people in the office (CO2 increasing and VOC also).

First of all, it is nice to see that the morphologies of the curves are similar.

But my mains  questions are:

1. Why Both sensors get such an increase of VOC for about 0-4 degrees of temperature increase (around 15h00)... in the worst case (250), and according to bosh IAQ table interpretation, it says to me that air is very very polluted (but nobody inside no, paints or anything like this !!??). Why is this value so big ?
2. Why the top of measurement of two BME680 sensors are so different one from each other : Max IAQ at 250 or 140 around 15h00 ?
3. VOC seems "to much" tied to temperature measurement even using BSEC that should already compensate. Anyway to improve that ?

Please help to interpret these results.

[handytech · Accepted solution]

It is very difficult to comment about the IAQ output without seeing the raw sensor signals, accuracy or ground truth. We have seen cases where an increase in the ambient temperature was also causing an increase in VOC outgassing in the environment (e.g. from carpets) in which case in the correlating IAQ increase was in fact expected.

Both BSEC and the BME680 need some time to compensate for the sensor's history and environment (self-calibration). This takes longer the less the gas sensor is operated. Therefore if you are operating in ULP, it is expected that you can already qualitatively detect changes in air quality, while accuracy and sensor-to-sensor deviation will improve over time of operation.

Note that 250IAQ doesn't mean "very very polluted" air but more generally "typical polluted" air, as BSEC will self-calibrate so that 250IAQ corresponds to the "worst air quality" it has observed during its history window.

[wtc-peg]

Hello,

I come back, as I could make some moretests with the same sensor still in ULP mode.

I took it outside (on a window) for two days, quite far from center of town near Toulon in France ;O).

Measurements ar shown below, but here are my questions:

1) Could you advice me for a domestic product that I could put close from the sensor (ie: somes tens of cm below) to get IAQ higher than 300, fror about 3 or 4 hours ?
I tried with smelty polyester resins (expecting for "embeded styren" to act), but I never get more than 200 or 250 IAQ and even it was decreasing after some hours and my resin was still very smelty...

2) On the curves below the sensor is outside fro 2 days, during last night humidity climbed a little and then the IAQ climbed up to 250 without any apparent reason (outside).
In the morning even if humidity decreases the IAQ stays at th same level.
Do you have any idea of what I could do wrong ?

[handytech]

---

@wtc-peg wrote:

1) Could you advice me for a domestic product that I could put close from the sensor (ie: somes tens of cm below) to get IAQ higher than 300, fror about 3 or 4 hours ?
I tried with smelty polyester resins (expecting for "embeded styren" to act), but I never get more than 200 or 250 IAQ and even it was decreasing after some hours and my resin was still very smelty...

---

If you are using the standard IAQ output, I believe the observed behaviour could be explained by the self-calibration process. The IAQ output learns from the air quality history, and automatically adjusts its settings so that 25IAQ corresponds to typ. "good air" observed, and 250IAQ to typ. "bad air". Therefore if you expose the sensors to bad air for several hours during your tests, you could be observing the IAQ output learning and adjusting to it. To confirm this, you could have a look at the sIAQ output, that will probably show much larger IAQ values during the same period.

---

@wtc-peg wrote:

2) On the curves below the sensor is outside fro 2 days, during last night humidity climbed a little and then the IAQ climbed up to 250 without any apparent reason (outside).
In the morning even if humidity decreases the IAQ stays at th same level.
Do you have any idea of what I could do wrong ?

---

Seeing raw sensor data (i.e. raw temp, humidity and gas values) could possibly give us a hint here, but without knowing the environmental conditions, we cannot tell from the graph alone what is causing this behavior, it could as well be due to some compounds in the outside air during the test.

[wtc-peg]

But if IAQ is automatically "coerced" between 25 and 250 how can we detect presence of nocive gazes in a room (paints or other ?) ?

Here is some more data.
I can't have output of raw data easily as sensor is already embeded, and i have to make new firmware for it.
But here is the behavior of sensor "Outside for 3 days" for IAQ.
It seems really T/H dependant as what we can expect form output fresh air is only a very few increase of CO2 by night...

BUT notice that sensor as begun its ULP life in indoor environment (as show in the original post)... Should I reset the periodically recorded calibration data ?