Particle Powered Air Quality Sensor Logging to Google Docs

[jaredwolff]

@fotofieber after assembling and testing a few more boards. I've noticed that the fresh sensors, do need some time to "burn-in." I'm seeing wild values in the 1000's whereas the one I've been using for a few weeks is rock solid.

Here's a capture of one I'm testing right now:

The blue is the one that's been running for a while. The red is a fresh unused sensor.

I haven't had a chance to play with the MH-Z14A. I'm currently doing an eval of all sensors that I can (easily) get my hands on to compare side by side.

All the latest and greatest code for my project is located under the Beam me Up section in my original post above. #3 I think. Here's the link again though.

[FotoFieber]

@jaredwolff
Thx for the link to your sourcecode.
How long is the time interval on the graph in your post?

I tried to use drive mode 1 (measurement every second) with an average calculation for every minute. But it didn't get better... :(

I added now a MH-Z19B to the testbed.

Awaiting still from aliexpress:
SGP30
MICS-VZ-89TE
GY-MCU680V1 BME680
Senseair S8-0053
and another CCS811.

[jaredwolff]

@fotofieber said in Particle Powered Air Quality Sensor Logging to Google Docs:

How long is the time interval on the graph in your post?

The interval was either 20 seconds or 2 minutes. My code defaults to 2 minutes. There's just no reason to be checking it that often.

The CCS811 on the other hand takes a measurement every second. (I left it in the default mode) I just take the latest measurement when it's time to send the update to the server.

Curious to see what conclusions you come to with your tests. I think we'll get there right around the same time!

[FotoFieber]

@jaredwolff
The MHZ19B is working out of the box without tuning. NDIR sensors seem to be easier to handle but are more power hungry.

In addition to the CO2 ppm, it has a temperature reading, which is not to bad:

The temperature of the BMP280 and ST7012 are quite high. They are on the same small pcb as the CCS811.

[jaredwolff]

@fotofieber said in Particle Powered Air Quality Sensor Logging to Google Docs:

The MHZ19B is working out of the box without tuning. NDIR sensors seem to be easier to handle but are more power hungry.

Nice! Glad to see something working for you. The MHZ19B seems like a good option when you have the space. I wonder how it is long term.

How many hours have you continuously run your CCS811? I'm seeing the fresh one I had normalize to zero over almost 24h of continuous running. (See the blue trace below for TVOC & C02)

I'm also not surprised by the readings you're getting on temp. Unless you get a sensors that is specifically more precise you could get some significant deviations. I would expect ±1 °C from most sensors though..

[FotoFieber]

@jaredwolff said in Particle Powered Air Quality Sensor Logging to Google Docs:

Nice! Glad to see something working for you. The MHZ19B seems like a good option when you have the space.

The MH-Z19B does use less space than the MH-Z14B. I thinkt the footprint is similar to my CCS811 PCB from aliexpress. But it is taller. I think the main argument against the MH-Z19B could be power consumption.

I wonder how it is long term.

I have some MH-Z14B working fine for more than two years. Accoriding to the datasheet, the MH-Z19B should last more than 15 years.

How many hours have you continuously run your CCS811?

More than two days. It didn't get better.

I'm seeing the fresh one I had normalize to zero over almost 24h of continuous running. (See the blue trace below for TVOC & C02)

I can't see the time on your graph. I only see 2019-0... would it be possible to show the time?

The manufacturer has improved accuracy with the last firmware of the CCS811. Which version of firmware do you have on your CCS811?

[jaredwolff]

@fotofieber said in Particle Powered Air Quality Sensor Logging to Google Docs:

The manufacturer has improved accuracy with the last firmware of the CCS811. Which version of firmware do you have on your CCS811?

Good question. According to my code: 1.1.0

@fotofieber said in Particle Powered Air Quality Sensor Logging to Google Docs:

I can't see the time on your graph. I only see 2019-0... would it be possible to show the time?

Hmm. Yea they're full timestamps. Pretty useless huh? What are you using to create your graphs? That way we're looking at the same thing.

On another note: I did get SPG30 and BME680 up and running. I ran the SPG30 while I was tinkering with the BME680. It's about 12 hours of data. (Sorry the x axis is not quite working out.)

My initial impressions from the integration:

SPG30 is about as easy as the CCS811. As you can see it has its own demons. It does require another power supply (1.8V typical for this part) And it can get power hungry if you run it tons (48mA when the heater is on).

The BME680 is not so easy to integrate into the Particle platform. Mostly due to the added archive file (i.e. static library, i.e..a file). After tinkering with what they had I have it taking in and processing data. The only thing I really care about is the IAQ calculation which is not open source. It does some other calculations on the temperature and humidity if you choose to integrated it with a space heater. (or HVAC unit for that matter)

[FotoFieber]

@jaredwolff said in Particle Powered Air Quality Sensor Logging to Google Docs:

Hmm. Yea they're full timestamps. Pretty useless huh? What are you using to create your graphs? That way we're looking at the same thing.

I am using grafana with influxdb (both in docker).

@jaredwolff

SPG30 is about as easy as the CCS811. As you can see it has its own demons. It does require another power supply (1.8V typical for this part) And it can get power hungry if you run it tons (48mA when the heater is on).

Awaiting mine impatiently. :) Power consumption is not a real problem in my usage scenario.

@jaredwolff

The BME680 is not so easy to integrate into the Particle platform. Mostly due to the added archive file (i.e. static library, i.e..a file).

Do you build with platform.io? I only found instructions for the arduino ide...

Temp and hum are measured on the heated sensor and must be calibrated to be useful according to some discussions I read.

[jaredwolff]

@fotofieber said in Particle Powered Air Quality Sensor Logging to Google Docs:

I am using grafana with influxdb (both in docker).

I like it. It may make sense to have my own infrastructure in the future. But then again, I don't feel like dealing with it all. :laughing:

@fotofieber said in Particle Powered Air Quality Sensor Logging to Google Docs:

Do you build with platform.io? I only found instructions for the arduino ide...

I'm building using the Particle SDK on my machine. They can build Arduino libraries but can't quite include external/static libraries yet (As far as I can tell) out of the box. The Particle platform, in general, has been for convenience and getting to proof of concept (really) fast. Plus they're an awesome company.

As for the data, I'm just pushing the data from the Particle platform into Google docs (as my original post described). It's great for short tests but longer data streams you definitely need a time-based database and some nice front end graphing capabilities. :smile:

Another update on the TVOC eval. Looks like the BME680 outputs on the same scale as the SPG30 and CCS811. (i.e. 500ppb == a reading of 500 IAQ on the BME680)

(The timestamps are there if you click on the image itself. )

@fotofieber said in Particle Powered Air Quality Sensor Logging to Google Docs:

Temp and hum are measured on the heated sensor and must be calibrated to be useful according to some discussions I read.

I was thinking about this more considering the discrepancy I'm seeing in temperatures and humidity. If you're serious about the results and the accuracy of the temp & humidity it may require a 3rd temperature sensor with no heater inside.

Supposedly the BME680 has a calibrated output but it hasn't fired up. I'll have to figure that out..

[FotoFieber]

Firmware upgrade time!

My CCS811 has bootloader version: 1000

and had application version: 1100

I upgraded to the newest firmware (2.0.1). Now waiting for another two days.... :) Hope then I get reasonable results like you have....

[jaredwolff]

@fotofieber yeehaw. I hope that fixes the craziness you were seeing. I'm curious if that will fix the early unreliable data I've been seeing too..

[jaredwolff]

@FotoFieber if you check out my tvoc_test branch, I implemented the CCS811 update routine. I'm interested to see how it compares to the SPG30 and BME680..

[FotoFieber]

@jaredwolff
The update has brought a big improvement.
I think one could derive airquality from the measurements, but the CO2 ppm value is still way off.

[jaredwolff]

@fotofieber that's great news! Pretty much what I've seen. I've stopped eC02 all together because it correlates 1:1 with TVOC on that part. Plus I don't know how actionable the C02 reading is. Does it indicate stale air? C02 is a byproduct of the VOCs? Maybe it's more useful for more industrial environments..

Are you logging anything other than C02?

[FotoFieber]

@jaredwolff
CO2 is an indicator of air quality. I use it since years mainly with consumer devices from Netatmo. In my graphs you can see, when people are in the rooms consume the air, when they leave, when I open the windows...

The CO2 measurement with NDIR-sensors seems much more reliable than what I see here from CCS811. They usually show quite good results just plugged in without calibration. They are easier to handle, because you don''t need to bother with handling basline (save and restore). And they don't need a temp/hum measurement to correct their values.

Yesterday arrived another NDIR-Sensor (Senseair S8) and a SGP30. I am curious, how they perform. Hope I have the first results in two days.

Are you logging anything other than C02?

In general I have much more values I log (radioactivity, luminosity, window open/close, motion...)

In this setup I log what I can get from the sensors.

[FotoFieber]

This night was a complete fail for the CCS811 I have here:

Fresh air at 20:00
Sleep till 6:15 with closed windows
-> NDIR sensors show, how air quality gets worse
-> CCS811 shows, how the air quality gets better during sleep with closed window!
Fresh air at 06:15

WTF.... :)

[FotoFieber]

S8 readings seem quite goot. (NDIR)
SGP30 :( Will try with the adafruit library instead of the sparktech.

[jaredwolff]

@fotofieber said in Particle Powered Air Quality Sensor Logging to Google Docs:

SGP30 Will try with the adafruit library instead of the sparktech.

I have that one in the same branch I mentioned before if you need a working reference.

I followed your lead on the Grafana + InfluxDB stuff. It was dead simple to setup.

I just want to get HTTPS/SSL working. Pretty nifty little piece of software. :)

[FotoFieber]

@jaredwolff said in Particle Powered Air Quality Sensor Logging to Google Docs:

I followed your lead on the Grafana + InfluxDB stuff. It was dead simple to setup.

Yes, really simple. I use docker for grafana and influxdb.

Your SGP30 CO2 ppm graph looks like mine... :(

I just want to get HTTPS/SSL working.

I use an nginx-setup here for https.

Pretty nifty little piece of software.

Yepp! They are real software gems.

[jaredwolff]

@fotofieber also, I compiled my conclusions about some of the smaller form factor TVOC/C02 sensors. It was inspired by everything we talked about here. Our back and forth conversation has been awesome. So I wanted to say thanks for the inspiration :)

The above is a screenshot from the side-by-side data of the BME680, CCS811 and SGP30.

The BME680 seemed impressive at first but I'm seeing a big difference in temperature and humidity. My thinking it's likely due to the heating elements inside keeping it a little less humid and a little more hot. (These are the compensated values by the way. The non-compensated values are even lower humidity and a hotter.)

The SGP30 is a bit of a power hog but the TVOC readings seemed to align well with the response from the BME680.

The CCS811, like we've seen throughout this thread, had some wacky responses using the 1.0.0 firmware. I haven't seen that since all of my CCS811 are up to date.

It's a toss up between all three but I'd shy away from the CCS811. If anyone here plans on building a production device, the stock is not reliable enough. (Digikey and Mouser have been out of stock lately) I'm not sure if you'd have the same problem in the EU or how things are in Asia.

Anyway my TLDR conclusions are here if anyone is interested. They include software implementation factors, hardware factors, availability factors and data output factors all rolled into one.