Infineon's CO2 Sensor Monitors Indoor Air Quality
allaboutcircuits.com
allaboutcircuits.com
I just updated the blog post that I wrote last year [1] with the specs of this new Infineon sensor and also our experiences with doing CO2 measurement outdoors.
In summary we can say that under normal indoor conditions (e.g. small range of temperature, relative high range of CO2 etc.), the photo-acoustic NDIR sensors perform well.
However, outdoors they significantly underperform optical NDIR sensors to the extent that they are barely usable. See for example this chart [2].
It's hard to say why exactly they perform so poorly outdoors but I believe that they rely on quite complex internal algorithms that probably have been only developed with typical indoor conditions in mind and as soon as some parameters like temperature are out of the typical range, they significantly drop in their performance.
Furthermore, we also detected that they are sensitive to interference from low frequency noise which could for example come from ventilation systems, refrigerators etc. This is not surprising as they rely on their internal microphone to do the photo-acoustic measurements.
All of this and the rock-solid performance we see with optical NDIR sensors made us quite wary about the photo-acoustic sensors and not use them in our open-source hardware monitors. As I said, I believe they work quite well in general but because we saw also some really strange and unexpected behavior of these photo-acoustic sensors, I'd now always prefer optical NDIR sensors if possible.
[1] https://www.airgradient.com/blog/co2-sensors-photo-acoustic-...
[2] https://www.airgradient.com/blog/co2-sensors-photo-acoustic-...
Though i believe you can turn off the auto calibration which is probably a good idea in busy city areas where I imagine outdoor CO2 is slightly higher than 400. And manually calibrate them with an actual known source.
Why measure CO2 outdoors?
We are able to detect emission sources pretty well now with our outdoor monitors [2] that have the SenseAir S8 NDIR sensor built in. You can see some real data in the blog post I wrote about launching our global CO2 map [3].
So setting up a dense network of these sensors in a city would allow to measure if for example the introduction of low-emission zones or switching to electric buses etc. would work.
Another use case is to check for leakage in underground CO2 storage facilities.
All-in-all it's still experimental (to some extent) but we know the accuracy is there and we can see more and more use cases as outlined in my blog post. So we now work hard to get more and more of these sensors out there to get more data to identify additional use cases.
[1] https://www.airgradient.com/blog/performance-of-low-cost-co2...
[2] https://www.airgradient.com/outdoor/
[3] https://www.airgradient.com/blog/airgradient-global-co2-map/
I.e. put sensor in vacuum chamber, remove all gases, then introduce know amount of gases to produce a known target environment.
Given their reference instruments, I assume they’re also capable of maintaining their calibration for a long period of known time, before they need to be re-calibrated. They would never rely on something as inaccurate as “20 mins outdoors” to calibrate themselves.
If the sensors are in an environment where the CO2 is always elevated, how do you keep it properly calibrated (eg if the CO2 never goes under 700ppm, how do you stop it from recalibrating so that it returns "400" when it should actually return "700") I know you can just turn calibration off but won't the sensors' accuracy decrease over time?
Only a plant cells containing chlorophyll and exposed to sunlight will photosynthesise. Notably at night there isn’t much sunlight, so normal respiration dominates a plants gas conversion processes, and results in them producing net CO2.
Based on the graphs in the blog post, I assume GP is placing the CO2 production on plants because it the CO2 levels peak at night, then return to normal every morning.
We need to get those rural voters more oxygen, STAT!
> Carbon dioxide is not released during photosynthesis, but small amounts of that gas are emitted both day and night as a by-product of cellular respiration. It is worth noting that the majority of plants absorb carbon dioxide during the day for photosynthesis and do so in greater amounts than they release for cellular respiration.
So depending on the time plants could easily be pushing out more CO2 than they are pulling it at that time, which could result in CO2 concentrations increasing during that time.
For example, to know that it's futile to try to turn on ventilation to lower the indoor CO2 concentration.
But on a hot windless summer day in a temperature inversion, my meter will read 700 or even 800 outdoors, as the CO2 from vehicles and maybe even power plants just sits around without going anywhere.
In the winter, if my indoor reading is 800 I open the window a crack to improve my concentration. On certain days in the summer, I just accept it's going to be 1,000 indoors and there's nothing I can do about it.
However, to understand the performance of these different sensor technologies it is important to point out that they fail under certain environmental conditions, e.g. outdoors.
I think this is what is meant by "humming": There is a sound that is generated when the target molecule is present.
Additionally, contaminants on any surface the light strikes will generate spurious signals. Lock-in (synchronous) detection at the optical source modulation frequency can help get rid of a lot of noise but if the noise appears at that frequency, it will corrupt the measurement.
This classic paper from 1986 covers a lot of what can be done with the technique and how to do it correctly. Reviews of Modern Physics is behind a paywall but there's a copy on the web at the URL below.
Applications of photoacoustic sensing techniques
Andrew C. Tam
Rev. Mod. Phys. 58, 381 – Published 1 April 1986
I used that for an open-source CO2 monitor I designed:
We at AirGradient open sourced our monitors around 2 years ago and this has been the best decision for our company.
BTW Your "Advanced ESP32 development with ESP-IDF" guide [1] is the best "getting started" guide I've seen so far. Up to now no guide has been so clear and complete, and I've spent one week being very frustrated by how confusing the official ones are.
So thank you again. You made me rekindle the spark of creativeness that got buried by age and daily grind.
I do have a 'one day when I get free time' plan to make new firmware for them to also measure moisture and a bunch of other VOC's which have unique absorption spectra in the 800-2000nm range, since the hardware itself can be abused as a poor-man's spectrometer.
The hardware consists of an incandescent bulb and photodiode with amplifier and high res ADC.
The bulb and photodiode sit in a box with a silvered interior, which allows light from the bulb to reflect around inside the box a lot (and having some light absorbed by CO2) before hitting the photodiode.
So many of those actually measure humidity, temp and VOCs and try to derive some sort of CO2 reading out of those.
Chinese fabs are notorious for running secret shifts that make their customers’ chips to sell themselves. They’re the exact same chip but made without authorization, using the customer’s exact design.
That said, Sensirion has their own CMOS fabs in Switzerland so that’s very unlikely to be the case here. If they work at all, the counterfeit chips are probably some Chinese CO2 sensor IC that's small enough to fit into the same package, lightly customized to fit the pinout of the original. Or it's just a microcontroller inside faking it a la FTDI.
Comparing the datasheets for the PASCO2V01 and the new PASCO2V15 the old one actually seems a little better as far as CO2 measuring performance goes. They are the same on most things, but the old one has slightly better accuracy.
The new one is ±(50 ppm + 5%) between 400 ppm and 3000 ppm.
The old one is ±(30 ppm + 3%) between 400 ppm and 5000 ppm.
The big difference is this:
> Infineon has recently introduced the PASCO2V15, a new 5 V sensor to improve air quality monitoring in building environments.
Both of them require a dual voltage power supply. They both want 3.3 V for their digital components and a higher voltage for their IR emitter.
For the older one that higher voltage is 12 V. For the newer it is 5 V.
Small CO2 sensors have been available for years, for about $50. Compare [1].
Life of this new device is only 10 years, which is short for HVAC systems. A hotel might have a thousand of these. Older devices say "15+" years.
All these devices have a calibration problem. They drift. They try to correct by treating the lowest value they ever see as "normal" (that's about 400 ppm CO2 today, vs 300 PPM in 1950) and recalibrating. So they're not useful for observing a general increase in CO2. They're also not useful for greenhouses, where CO2 levels may drop below ambient CO2 due to photosynthesis. Manual recalibration is possible but requires feeding in pure nitrogen and a known nitrogen/CO2 mixture.[2]
Devices which don't need that re-calibration exist.[3] They're more complicated. Also don't seem to be stocked by the usual distributors.
[1] https://rmtplusstoragesenseair.blob.core.windows.net/docs/pu...
[2] https://www.co2meter.com/blogs/news/7512282-co2-sensor-calib...
[3] https://www.murata.com/en-us/products/sensor/co2/overview/te...
Murata announced this in 2019, and there's a part number (IMG-CA0011-00/ IMG-CA0012-00/IMG-CA0023-00) and a full data sheet.[1] But no distributor has it. Not DigiKey, Mouser, Arrow, or Newark. Even Octopart doesn't list it. It's on Murata's list of recommended products, not the discontinued list. Try contacting Murata.
[1] https://go.murata.com/rs/382-MEZ-125/images/HC%20CO2-sensor%...
Along these lines of air quality, can anyone recommend a similarly advanced PM2.5 / PM10 sensor under $100 / ea?
Edit: sorry missed your price guidance. They are quite a bit more so probably not what you're after!
Ikea sells a reasonable air purifier AND seperate sensors nowadays!
It also measures temp/humidity/voc, and has a built-in fan to speed up the response time.
Neither are cheap, around $25-40 each in small quantities. The infineon one has a full blown microcontroller handling the operation of the sensors.
To keep accuracy you would need to have a CO2 gas setup which isn't cheap either, but for indoor use I don't think it matters.
CO2 monitors often have little silent fans to draw in fresh air as well, for accuracy.
I have an SCD41 and I see a large spike in readings less than a minute after sitting down at my desk.
I really like the idea of using cheap (?) devices in a sort of mesh to feed back telemetry data on pollution. Pollution is everyone's concern, so visualising that would be cool.
Interested to hear if you had any more thoughts on this!
Once had this thing on a balcony of a shared flat in Heilbronn, Germany. Wondered what that was, previous tenant told me about it and it was never removed from there.
https://www.cs.dartmouth.edu/~sensorlab/pubs/BikeNet-SenSys0...
Not sure the market is big enough to invest my time.
A wearable CO2 monitor would do the opposite of this.
If you want to measure quality you need fixed location devices.
Hopefully this method doesn't have the same restriction.
What does the back to back ± mean? Is that the variance of accuracy from device to device? Or does the 5% reference the specific range of 400-3000?
No, it's stacked, and strictly applicable between 400-3000 ppm as characterized (despite 0-32000 ppm operating range by design).
At the lower 400 ppm end, accuracy is +/- 50 ppm +/- 20 ppm = +/- 70 ppm.
At the upper 3000 ppm end, accuracy is +/- 50 ppm +/- 150 ppm = +/- 170 ppm.
Also worth noting that the calibrated reference used to determine this published accuracy has an uncertainty of +/- 2%; ref datahseet[1] Table 7.
Caveat emptor: this published accuracy can't be trusted at face value when sampling more than 1 meas/min; ref ibid.[1] Table 4.
[1] https://www.infineon.com/dgdl/Infineon-PASCO2V15-DataSheet-v...
My takeaway is that it draws a lot more power than you'd expect, thanks to the incandescent light source, and unless there's quite a lot of airflow over the sensor, it'll exhibit self-heating at any poll rate under every ten minutes.