Bosch sensortec guides for bme688 gas and environmental sensor

I’m a geek for sensors and this is making me sound like a 6yo kid with a new toy.
Here’s the good stuff:

I’ve had the opportunity to talk to talk to some of their devs and even the PO of the sensor. These guys are Hyped about their stuff. And they listen to complaints and address them. If you guys are looking for a high tech sensor for your home, definitely look at the Bosch bme68x series, join their community and be active there as well.

I definitely get the feeling that this is just the beginning. They are also showing off their new air particle sensor - no moving parts, first of its kind -

Hope someone else enjoys this and that I’m not the only one :slight_smile:

The confusion matrix shown in the video looks quite impressive, but I doubt that the gas sensor is suitable for serious applications. However, it could be useful as a VOC sensor.

As with everything, for serious applications you have sensors that cost hundreds or thousands of euros.
Which for any serious application you’d be forced to buy anyway due to any number of regulations or insurance policies.

So I think you’re comparing apples to oranges.
The way I see it is, you can buy a DHT22, an air pressure sensor, a bunch of different gas sensors (each one range wildly in price - just look at the co2 sensors) code a very complex firmware to look at all of the sensors, calculate a bunch of different scenarios and output a useable result (temperature humidity air pressure altitude and air quality) including a calibration setup….

That or, you buy a bme680, get on bosch sensor tec GitHub, download their firmware examples, flash it on arduino and you’re off to the races, for something like 10 euros, and it fits inside a tin can.

Is it needed for every use case? Of course not. But one sensor like this can easily tell you if you should buy a carbon filter for an air purifier for example. That’s useful for itself. Or detect natural gas leaks. Also useful for home users who may have a gas stove.

Basically, the gas sensor seems to be sensitve to H2S and all additional gas concentrations are derived from the H2S measurement - and machine learning isn’t magic. I would be impressed if the gas sensor could be used to classify coffee blends. :slight_smile: But you’re right: you get what you pay for and the sensor offers nice features for home use - but I wouldn’t rely upon it for gas alarms.

Fun point is that the bme688 code example is precisely to recognize the smell of coffee beans :slight_smile:

In the video they classify measurements into ‘coffee’ or ‘neutral air’ …

I would be mildly impressed, if they could train a predictor (with an acceptable confusion matrix) to discern between filter coffee and espresso.

I’ve seen some gas sensors and been looking for some industrial grade sensors for other applications. Basically everything boils down to measurement method. If you have physical-chemistry sensor it is subject of aging and rusting (of various kinds) making it report false positives over unpredictable amount of time. Time of first false alarm depends on environment conditions and not time itself. Other measurement methods which are aging free are not precise or inaccurate in best case.
Yet, few months before first covid outbreak happened I saw a hope with some sensors which had fairly low energy consumption and precise measurement which could work with multiple gases - as far I remember during one of trade fairs I got a card of a high precision optic sensor which did measure humidity and size of gas molecules through small laser and shadow sizes. It could determine density of very specific subject in the air without false values. This is due to fact that each compound have a predictable characteristic if you know pressure and humidity. Obviously, measuring coffee smell is something completely different than measuring i.e. phosphine (PH3) which is easily identifiable as chemical deriverable from Mendelev table. Results coming from biologic matter will never be possible to measure (unless you track methane emmited by cows… ).
If you have a universal sensor such above optical one, you can calibrate size of stream and molecule shadow to get accurate results. Maybe not as precise as traditional physical-chemistry sensors but still good enough to deliver decent quality with no age restrictions!

1 Like

What type of sensor can one use to detect cannibis for instance?

No idea…

Given my limited knowledge in the field, I think it might be quite difficult because, there are no good solid materials which could react with cannabis or THC. A lot of sensors i.e. use various semi-passive elements will generate different resistance depending on conditions which is then used to report actual value. Most common type of such sensor is resistance temperature sensor (RTD), often referred as Pt100 or Pt1000, which gives different resistance based on temperature.
I don’t know if there are electronic THC tests. I suppose most of them will be based on chemical reaction which happens due to presence of specific chemical compound. I was looking at topic after your asked question but found nothing relevant, especially from sensor perspective.
I guess that at this moment you need to look at very specific smoke sensor. :wink: You could also try and test/train bme688 if you have enough of ganja supply. :wink:


This topic was automatically closed 41 days after the last reply. New replies are no longer allowed.