This is fantastic!

• I love the idea of using CO2 and/or RH as a metric for sleep quality / breathing patterns. I suppose you're getting a much stronger signal in your tent than you would in an open room? But seems worth a shot!

• Interesting to see the apparent correlations in temp, CO2, and RH, and what that might say about e.g. Craig health / mushroom health / the nature of the enclosure. I wonder if the cross-correlation time of each of these variables against the other could be used as a metric for e.g. how much air is circulating around the sensor, or how 'closed' a 'grow box system' is ...

• Could we use the CO2 sensor, in a closed container, plus infragram, as a way of correlating plant respiration w/ photosynthetic activity?

• How much are the parts in all of this, again? It's looking like it'd make for a neat little kit!

• Can't wait to see this applied to mushrooms!

I'm trying to imagine how to calibrate the CO2 sensor-- Do you think we could fill up a series standard-sized balloons of some sort with a known air mixture? Could we do it with SodaStream cans?

I volunteer to make airtight plastic balloons of a known size for the purpose.

@donblair

• Yes, it wouldn't be so bad to set up a tent in one's domicile for the purpose of science.
• It seems like a lot of correlations can be made in order to build a model for the sensor response and vice versa; this would be an interesting subject for a Public Lab sub-community to investigate using open data methods.
• Putting the sensor in a terrarium would be an interesting test case, but I'm not sure how much an infragram image would vary over a fews hours or even days.
• The CO2 sensor is the big-ticket item at $80 (see link in post for source); the Riffle is how ever much you are charging for this app (I got mine for free!); the SHT21 is a few bucks, the optional box is $10, maybe another $10 for cables and weather proof bushings (not on this prototype yet). I don't think a kit would fair well - I'm just about providing the resources for interested hobbyists to do-it-themselves. Maybe custom assembled units could be made for researchers with deeper pockets, a warm feeling about open source, and a tollerance for early stage prototypes.
• Maybe the next test should be done in an enclosed mushroom.

@mathew

This particular optical sensor is supposed to be factory calibrated, but there is no harm in checking the numbers! One of the reasons that I did not go with a cheaper MG811 MOX electrochemical gas sensor module, was that it provided only an analog output and was difficult for me to try to calibrate. While testing the MG811, I sealed it in a container with a known amount of sodium bicarbonate (baking soda) solution and injected an excess of ascorbic acid (could also use vinegar) to cause it to give off CO2. The problem that I had was that the sensor voltage was not only directly correlated to CO2 but also to water vapor to a high enough degree that the results did not make much sense. The SodaStream sounds like it might work better than what I tried. However, this inherent water vapor sensitivity would be even more of an issue in a growroom. Furthermore, the MOX sensor needs to be held at a high enough temperature to function at all so it drew over 1 watt of power, whereas the optical sensor took more like 0.1 watt. At the end of the day, Occam's razor prevailed over Uncle Pennybags' monocle.

Please explain what is meant by: "During the period when I am sleeping peacefully in my tent (after ~25 mins), the RH and CO2 readings seem to fluctuate in a correlated way. " I did a correlation for RH and CO2 from 2:30 to 3 and got a value of 0.05. This doesn't indicate any correlation. I must be misunderstanding?

@jefffalk

Well, I hadn't done any correlation analysis yet, but I tend not to trust algorithms that yield a single number value judgement. To better qualitatively show what patterns that I think I see, I have plotted the part of the data record from around 25 mins up to 60 mins since the start of the experiment (segment 1). (Note that the datetime stamps in the raw data are likely not absolutely correct since we had forgotten to synchronize the RTC.) The data sets have been rescaled by subtracting their means and dividing by the standard deviation.

The remainder of the data (segment 2) - which was not plotted in the original post - understandably looks to be not as correlated. I don't know what happened here, but I will say that I've been tossing and turning a bit more now that I've taken to sleeping on my side within the last year.

The code for these plots can be found at this link.

Thanks for making the effort to explain and for doing the experiment. Yes, there is an interval of high correlation.

Hey, I also got myself the SainSmart MH-Z14 and I must say, I am not sure if I really trust the values. Mine (using Arduino and the UART interface) reads around 700ppm indoors, rising steadily to over 1000ppm when put outdoors. These values look suspicious to me - they're twice as high as I'd have expected, and I don't really believe that outdoor CO2 levels (on my window sill on a winter day) are higher than indoors. What's more, I get a rather suspiciously good (negative) correlation between the temperature and the CO2 concentration (r-square = 0.95). Do you have any more insights into how reliable these measurements actually are? Yes, my sensor does spike when I breathe at it, so it seems to react to CO2 changes, but beyond that I haven't seen evidence that this sensor actually does what the datasheet promises.

Judging by the few websites and Youtube movies that exist about this sensor, it seems that most people get measurements around 1000, and nobody seems to question that, even though, according to [1] "NIOSH considers that indoor air concentrations of carbon dioxide that exceed 1,000 ppm are a marker suggesting inadequate ventilation." So, perhaps this sensor isn't actually working, or at least not with anywhere near the accuracy promised by the datasheet?

[1] http://en.wikipedia.org/wiki/Indoor_air_quality

@tobi_k Hi, I am glad that there are others testing out these sensors with a skeptical attitude. Very often, the abosulute baseline readings from many cheap sensors should not be trusted. Expensive instruments have all sorts of compensating analog signal conditioning circuitry and calibration trimmers that have to be tuned in a lab - thus driving costs up. Discount sensor companies cut all sorts of corners in design and manufacturing and often lie about the real world performance. I prefer to look at relative values and time change since they are usually more trustworthy. Also, simultaneous recording temperature and humidty can give you extra clues to understand the response of the CO2 sensor. And, believe me, the situation with the MOX CO2 sensors is much worse - since the sensor itself is the major source of error.

The SainSmart board is a drop-in module for many types of control systems and is unecessarily complicated, containing its own MCU. I do think that the underlying NDIR unit (metallic module) is sound, but it needs a better interface. I am making plans to see if the particular optical sensor can be obtained cheaply and designed into an open source module which is easier to calibrate and integrate into extensible sensor platforms. Again, it is important to lower one's expectations for the veracity of DIY sensor systems and to put things to the test!