air quality monitor using particle counter

chris.nafis · 2012-04-18 22:27

I was interested in what the $12 Sharp sensor could do. I hooked it up to an Arduino Ethernet to post the data to Pachube (https://pachube.com/feeds/55892). I compared it to a Dylos DC1100 Pro laser particle counter which I also connected to Pachube (https://pachube.com/feeds/55522). Using data taken at the same time, I was able to "calibrate" it to read out in particles per 0.01 cubic feet. I was nicely surprised how well the two sensors agreed for high particle counts. I put documentation at: http://www.howmuchsnow.com/arduino/airquality/

Duane Degn · 2012-04-19 07:04

Wow, this is great!

I really liked your website documenting the two sensors. The data from the two sensors did seem to correlate well.

This has made me wonder about the air quality in my house. I hope to try out the Sharp sensor for myself.

Thanks for posting this.

chris.nafis · 2012-04-20 13:26

I got the Seeed Grove Shinyei Model PPD42NS Dust Sensor hooked up to
the Arduino Ethernet last night. It is posting data to Pachube. It has a digital
interface and I'm getting better results with the Arduino then I did
when I interfaced the Sharp Optical Dust sensor using the A/D.
I need to collect more low particle concentration data... but here is
the set-up and some initial data:

http://www.howmuchsnow.com/arduino/airquality/grovedust/

ElectricAye · 2012-04-22 07:09

Very interesting. Any idea what the maximum particle count per second might be for these kinds of modules, generally speaking? Thanks for posting this.

Tracy Allen · 2012-04-22 11:49

It's great to see that data on Pachube like that. How would you describe the operation of the Sharp and Shinyea sensors, that is, I'd guess they work by optical scattering using an LED and photodiode, but can you make out how the parts are arranged?

For a number of years I've been working closely with a group in environmental health to make relatively low-cost air quality sensors. We took the tack of reverse engineering and converting household smoke detectors (with a BASIC Stamp/OWL2pe controller data logger to replace the First Alert PIC). Here are a couple of papers. PM me if you'd like to look over a copy.

Aerosol Science and Technology, 38:10541062, 2004 Combined Optical and Ionization Measurement Techniques for Inexpensive Characterization of Micrometer and Submicrometer Aerosols.

J. Air & Waste Manage. Assoc. 56:789799, 2006 An Inexpensive Dual-Chamber Particle Monitor: Laboratory Characterization

The reason for two chambers is that ionization and photoelectric technologies collect complementary kinds of information on particle size and number. The papers go a lot into the noise floor, least detectable events, and also the upper limits. Basically, the signal disappears into the noise below 50 µg per m³. There are long term drift issues. But for third world applications where levels are up to mg/m³ it has proven quite effective as a tool for assessing things like the performance of alternative cook stoves.

First world applications require detection below 10µg/m³. I'm a big fan of the Dylos too. It is amazingly sensitive. We are using that, with a Prop/SD/XB interface, for monitoring second hand smoke that drifts from room to room in houses and apartments.

It's easy to tap into the signal from the ionization smoke detector, and you can buy them for $5. Most of the cheap ones use a single chip like the Allegro A5367, which has a guard voltage on pin 14 that tracks the chamber voltage. Just run that into an ADC. To wring out the best quantitative data, the power supply has to be regulated, not simply the 9V battery provided, and it has to have temperature and humidity compensation, and most of all, as anyone who has one knows, it is quite exacting about air currents. Data from the ionization detector is hard to interpret because of the interfering real world factors. Not only that, there are regulatory problems with employing the Am²⁴¹ off-book, not in a smoke alarm per se.

An interesting chip that can be used for optical smoke detection is the Silicon Labs SL1141, proximity detector with I2C interface. AN541 talks about the smoke detector application and has interesting background information. I have a couple here but haven't gotten around to trying them out. So many samples, so little time!

chris.nafis · 2012-04-22 20:46

Tracy,

You should get involved with the Air Egg project. Sounds like they could use your experience: https://groups.google.com/forum/#!forum/airqualityegg or http://airqualityegg.wikispaces.com/
There is a Kickstarter project: http://www.kickstarter.com/projects/edborden/air-quality-egg.

Chris

chris.nafis · 2012-04-23 09:38

ElectricAye wrote: »

Very interesting. Any idea what the maximum particle count per second might be for these kinds of modules, generally speaking? Thanks for posting this.

I'm still learning, but it seems to vary depending on the sensor. They react differently to particle size / shape.
The Dylos reports out every minute on the serial port. The Grove sensor gets several readings in 30 seconds, but you integrate over the 30 seconds to get particle count.
The Sharp sensor gets a reading every 10ms.

In a test where I cooked pancakes and "smoked up" the kitchen (but not enough for the smoke alarm to go off :-)
I got readings on the Dylos of 50000 per 0.01 cubic feet

Tracy Allen · 2012-04-23 15:40

Thanks for the links, Chris. I was aware of the Kickstarter page but not the behind the scenes discussion. I've been skeptical, like many people it seems. Cheap sensors fall on a continuum, with things like temperature and light at the solid end, and things like snow depth a little farther along, and most cheap gas sensors way off at the squishy end. Accounting for or eliminating confounding factors is expensive. I can appreciate what the Egg is trying to achieve, but it is going to be tough to deal with vast amounts of ambiguous data. (Been there!) I guess they know that. It's tricky for me to get involved in the discussion due to a tangle of NDAs and academic unpublished research, etc.

There are good particles, like pancakes and roasting turkey, and bad ones like too much tobacco smoke or too much wood smoke, but it is hard to sort it out without additional information or measurements. It is especially hard for outdoor air pollution where there can be multiple sources. The Dylos, depending as it does on light scattering from particles in a red laser beam, has very different responses dependent on size, shape and albedo.

Here's a graph I made a couple of years ago while carrying around a Dylos DC1100 during a July 4th fireworks display, and then the next day on a drive through Santa Cruz and up to the picnic area at Big Basin State Park. In general the readings on the right-hand red scale are far above the list printed on the back of the Dylos, that goes from excellent (<75) to very poor (>3000). Those numbers would serve as a guide for people who suffer from asthma or respiratory distress. The 0.01 cubic foot is near the volume that a child takes in in each breath. The ratio of large to small is an indication of the age of the particle distribution, because large particles settle much faster, and also particles tend to aggregate through time.

chris.nafis · 2012-04-23 16:36

Tracy, I like the graph! I got started on air quality because I wanted to put together a project with the local school districts (I've built weather stations with them in the past). It didn't look like the CO and other gas detectors would have enough "resolution" to see the polution in the US (according to Mass/NY CO levels have dropped down to 2-3ppm). I thought the dust sensors might be able to correlate to the pollen counts that the local allergist publishes. I was looking for a sensor that can show trends... it doesn't neccesarily have to be accurate. I then saw the Air Egg project, and that other people are working on similar ideas. With my weatherstation and lightning detection experiences, I can see the power of large networks of lower quality sensors. The key is getting the data organized in a useful place on the Internet. I was a fan of the Rabbit Processors becausethey were a cheap way to get data to the Internet. Now I see the Arduino filling a similar role. I've used Stamps in previous non-network applications, but haven't played with the Propeller. I would like to see a product that could quickly/inexpensively connect a sensor to the Internet (such as the Dylos and Pachube). It would be nice if we could convince WeatherUnderground to add additional sensor data. Maybe we should be trying to convince companies like Davis to add the sensor to their weatherstations?

chris.nafis · 2012-04-23 19:32

Tracy,

I added your dylos 0.5/2.5 micron ratio to my Pachube page (https://pachube.com/feeds/55522).
I'm looking forward to see if it helps explain things.

Chris

chris.nafis · 2012-05-09 10:42

Found great project/ paper from Drexel University
on using the Dylos DC1100 particle counts and calibrating them to ug/m^3
http://www.cleanair.org/sites/default/files/Drexel%20Air%20Monitoring_-_Final_Report_-_Team_19_0.pdf

Tracy Allen · 2012-05-15 22:00

Thanks for the link to that paper, Chris. Very interesting project and nicely executed.

When did Pachube become cosm? It is interesting to see the patterns in your data scroll by. I just took a look at the trailing week for the two size bins and the ratio.

You have to wonder about the long slow trends in the fines, the occasional spikes, and sometimes double peaks per day in the larger particles. The ratio has its own patterns. So many details. Wind, speed and direction, what is going on upwind would a lot to do with this.

I agree, extensive coverage of sensors could be very interesting. It's not that people haven't tried. There are formidable difficulties, funding of course, and the logistics of so much data, and the quality of the data. There are schemes afoot to attach air quality monitoring equipment to things like post office or UPS or waste management trucks, to collect data as they drive around. But that has obvious bias in the type of environment it samples.

mbaresso · 2012-12-11 08:25

Actuallay i am doing some studay about automotive partilce counting . So thos paper you mentioned helpes me a lot. ty

chris.nafis · 2013-05-10 20:29

Lately I've been playing with the Shinyei Sensors. It is much better than the Sharp. I recently hooked it up to an Electric Imp to make an inexpensive WIFI Air particle sensor that posts it data to COSM
( http://www.howmuchsnow.com/WIFIparticle/ )

Tracy Allen · 2013-05-11 09:42

Hi Chris, that's good to see. A couple of students here are using it in projects too. The Propeller interface is drop-dead simple using the counter modules. Counter A is set up to measure the total time the output stays low (which is the primary datum) and counter B is set up to totalize 1->0 transitions. At the logging interval, such as 30 seconds, the totals are captured and the counters are post-reset to zero. The time low over the 30 second interval is the official Shinyei datum, so we're not sure what to do with the transition count, or if acquisition of individual pulse widths might also carry useful information. Needs more lab characterization. No doubt the time low per interval is pretty sensitive over the long term, and it is sure is great for the price.

[SIZE=1][FONT=courier new]PRI SetupShinyei
  frqa := frqb := 1
  ctra := 100 << 26 + SHIN_PIN    ' NEGDET mode, will count how long shinyei pin is low
  ctrb := 110 << 26 + SHIN_PIN    ' NEGEDGE mode, count 1->0 transistions from Shinyei

PRI AcquireShinyei
  shins := phsb~      ' count of transitions, and reset accumulator
  shinlow := phsa~ / (clkfreq/1000)   ' time low in milliseconds and reset accumulator
[/FONT][/SIZE]

Tracy Allen · 2013-05-11 09:48

Also,
Our lab characterization paper on the Dylos particle counter finally came out.
Downloadable as:
A low-cost particle counter as a realtime fine-particle mass monitor

chris.nafis · 2013-05-13 14:54

Nice paper! Do you think there is any chance we could use the same procedures and equipment to analyze the Shinyei sensor?
I've been asking around my area and haven't found the resources.

Tracy Allen · 2013-05-13 23:07

The Shinyei is a hard one to figure. You have to wait a long time to get an average. Over the short term of seconds or minutes its output can look pretty random. Also, there is the fact that it depends on thermal convection to bring particles up through the sensor. That introduces uncertainty about the flow rates and what exactly is being convected. For those who are following this, there is a nice animation of how it works at the bottom of this Shinyei page.

I'm sure it will happen. Comparisons like the ones you are doing with the Dylos are in an of themselves revealing, because the Dylos is far more sensitive and faster than the Shinyei and has a relatively solid basis of calibration. It comes down to co-location of the Shinyei with other "professional" instruments over an extended period of time. That is pretty much how our research on the Dylos has been done, via lab or field co-locations with instruments such as the TSI dustrak. One lab I know of is doing calibrations of the Dylos against instruments like the cantankerous TEOM (tapered element oscillating microbalance), which is an instrument gives mass readings in real time, or filters, which give average mass readings over extended time intervals. Such lab experiments are difficult and time consuming.

The bottom line for health are particles that can get deep into a person's lungs, and that would add up to a mass of particles in an aerodynamic size ranges below 10 or 2.5 microns.

chris.nafis · 2013-05-17 21:01

The SEEED studio Shinyei PPD42NS dust sensor (http://www.seeedstudio.com/depot/grove-dust-sensor-p-1050.html) comes with a cable wired for +5VDC, GND, and P1. P1 shows >1um particles. The interesting thing is that the PPD42NS also has a pin P2 that shows >2.5um particles. So this allows you to determine if the detected particles are < 2.5um.

I'll update my Arduino (http://www.howmuchsnow.com/arduino/airquality/grovedust/) and Electric Imp (http://www.kickstarter.com/projects/1652961970/wifi-air-particle-sensor) code/docs to use this.

Tracy Allen · 2013-05-18 10:11

Hi Chris,

Are you doing anything to adjust the P2 threshold? That would be pin 5 on the connector...

5 : INPUT(T1) ･･･ FOR THRESHOLD FOR [P2]

The documentation I have is quite thin, only a hint of the P2 function. There are also the two on-board potentiometers, inviting for fiddling to see what they do.

chris.nafis · 2013-05-18 18:51

I contacted Shinyei. By default P2 is calibrated for > 2.5um. Don't fiddle with the pots. You can change the P2 detection size by changing the input voltage to pin 5. P1 is always >1um. If you look at P1 and P2 on a scope you can see what is happening. If P1 only, then >1um and < 2.5um. If P1 & P2 then >2.5um. Looking at the pulse size ratios of P1&P2 can tell you that you have a mix of particle sizes.

Another interesting thing is that SEEED is not an official distributor of Shinyei. Shinyei has newer versions of the PPD42Nx. There is also a knock-off version from a Korean firm: DSM501 ( http://www.alibaba.com/product-gs/897848623/Dust_sensor_DSM501_series.html )

Tracy Allen · 2013-05-18 22:48

That's interesting that SEEED is not an official distributor for Shinyei Technology. I haven't seen it sold anywhere else, do you know? There seem to be several models. They have this PPD42.. at $15 from SEEED. There is also a PPD60 or PPD62, which seems to be more sensitive, and there is a model AES-1 that uses a laser diode instead of an led, and which they bill as an "aerosol sensor" in contrast to the others as "particle sensors". The supporting data available is... practically nonexistent. That leads me to believe they want to deal factory direct with OEMs. Leave it to experimenters to fill in between the lines.

The EPA sponsored a workshop back in March this year,
EPAs Next Generation Air Monitoring Workshop Series
Air Sensors 2013: Data Quality & Applications
A student from here presented a poster and met Tim Dye, who (as you well know Chris) has been working with the Shinyei and the Dylos. I just noticed that Tim's abstract quotes a price of $160 for the PPD62PV. Tim's poster is cleverly titled,
--A Scientist with Sensors and Spare Time: Backyard Comparisons of Particulate Matter Sensors

Posters are downloadable online from
https://sites.google.com/site/airsensors2013/final-materials

A couple of posters from Berkeley that I've been involved with:
--BEACON: Berkeley Atmospheric Carbon Observation Network
-- A robust low-cost particle monitor and data platform for evaluation of cookstove performance.

chris.nafis · 2013-05-19 08:05

Actually I hadn't seen Tim's work before. Someone from the air quality egg forum just sent me a reference to the paper last week. Tim and I are trying to connect.

chris.nafis · 2013-05-22 11:19

For those interested, here is what the insides of the Shinyei PPD42NS looks like.
You can see the IR LED on the right which focuses light at the center of the air channel.
There is a resister on the bottom that heats the air and causes an updraft that brings particles into the chamber.
On the far left is a detector in a shielded case (bent out of the way in the photo). The large lens focuses any light from the center of the air channel back
on to the imager. When particles flow past the center of the air channel, they reflect light back onto the imager.
This is converted into a pulse. The pulse lengths equate to the concentration level.

air quality monitor using particle counter

Comments