Collecting Haze Data with Arduino, Raspberry Pi, and Amazon DynamoDB

I finally decided to put those Sharp airborne dust sensors which I bought during the last haze epidemic (in 2013) to good use, along with the small mountain of parts I have at home.

I used this very simple Arduino circuit and code snippet to read the Sharp sensor.  I ended up using a 100uF capacitor and 220-ohm resistor because those are what I had on hand; performance seems unimpaired. I used an Arduino Uno, because the dust sensor needs 5V. The sensor is very noisy and jumpy, so I used an interquartile mean and 200 measurements (discarding the bottom 25% and top 25%) to get a more robust reading. The raw value still jumps around by 1 LSB.


I also modified the slope function so that a full-range reading is 250 ug/m^3 (the original formula is here). Here is the modified Arduino code:


 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
// program to read a Sharp GP2Y1010AU0F dust sensor
// this sensor produces a dust value with a maximum value of 250 ug/m3

#define NUMSAMPLES 200
#define dustPin 0
#define ledPower 2
#define delayTime 280
#define delayTime2 40
#define offTime 9680

void setup() {
  Serial.begin(9600);
  pinMode(ledPower, OUTPUT);
  pinMode(4, OUTPUT);
}

// swap sort algorithm
void swapsort(int *sorted, int num) {
  boolean done = false;    // flag to know when we're done sorting              
  int j = 0;
  int temp = 0;

  while(!done) {           // simple swap sort, sorts numbers from lowest to highest
    done = true;
    for (j = 0; j < (num - 1); j++) {
      if (sorted[j] > sorted[j + 1]){     // numbers are out of order - swap
        temp = sorted[j + 1];
        sorted [j+1] =  sorted[j] ;
        sorted [j] = temp;
        done = false;
      }
    }
  }
}


// read the dust sensor, implementing an interquartile mean
// thanks to STMicro Application Note 3964 "How to design a simple temperature measurement application using the STM32L-DISCOVERY"
int readRawDustValue() {
  int i = 0;
  int rawVal[NUMSAMPLES];

  for (i = 0; i < NUMSAMPLES; i++) {
    // ledPower is any digital pin on the arduino connected to Pin 3 on the sensor
    digitalWrite(ledPower, LOW); // power on the LED
    delayMicroseconds(delayTime);

    rawVal[i] = analogRead(dustPin); // read the dust value via pin 5 on the sensor
    delayMicroseconds(delayTime2);

    digitalWrite(ledPower, HIGH); // turn the LED off
    delayMicroseconds(offTime);
  }

  // now we have the raw values, sort them
  swapsort(rawVal, NUMSAMPLES);
  
  // drop the lowest 25% and highest 25% of the readings
  long dustVal = 0;
  for (i = NUMSAMPLES / 4; i < (NUMSAMPLES * 3 / 4); i++) {
    dustVal += rawVal[i];
  }
  dustVal /= (NUMSAMPLES / 2);
  return (dustVal);  
}


// convert the raw count to a dust value
// the full-range signal is 771 counts = 3.76V
// based on http://www.howmuchsnow.com/arduino/airquality/
//
// dust density (mg/m3) = 0.172 * V - 0.1

float calcDustDensity (int rawVal) {
  float calcVoltage = rawVal * (5.0 / 1024);
  
  // I use a different figure so that 771 counts = 248 ug/m3
  float dustDensity = ((calcVoltage * 0.688) - 0.1) * 100;
  
  if (dustDensity < 0) dustDensity = 0;
  return (dustDensity);
}


void loop() {
  int dustValue = readRawDustValue();
  float dustDensity = calcDustDensity(dustValue);
  
  Serial.print("Raw Dust Value = ");
  Serial.println(dustValue);
  Serial.print("Dust Density = ");
  Serial.println(dustDensity);
}

Unfortunately the Arduino Uno only has 2K of RAM and mine did not have an Ethernet or Wi-Fi shield, so I decided to use a Raspberry Pi (Model B, the old single-core one) to read the Arduino Uno and talk to DynamoDB.

Basically, the Arduino appears as a serial port to the RasPi and I read the values using pySerial and uploaded them using boto.

Surprisingly the RasPi behaves exactly like a "real" Linux box, installing the AWS CLI and Python SDK (boto3 and boto) is exactly the same as on a large machine. There were no hiccups during the installation (albeit the installation took a long time).


The Python script is as follows (my first Python program!) note the hard-coded AWS credentials which is really a terrible practice.


  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
#!/usr/bin/python

import boto
from boto import dynamodb2
from boto.dynamodb2.table import Table

import datetime
import time
import serial
import re
import sys

# monkey hacking to work around "inexact numeric" issue in boto
import decimal
from boto.dynamodb.types import DYNAMODB_CONTEXT
# Inhibit Inexact Exceptions
DYNAMODB_CONTEXT.traps[decimal.Inexact] = 0
# Inhibit Rounded Exceptions
DYNAMODB_CONTEXT.traps[decimal.Rounded] = 0


conn = dynamodb2.connect_to_region(
 'ap-southeast-1',
 aws_access_key_id='AKIAxxxx',
 aws_secret_access_key='R0KIxxxx',
)
table = Table(
 'dustValues',
 connection=conn
)

# open the serial port (we need to be setuid root for this)
serialport = serial.Serial("/dev/ttyACM0", 9600, timeout=0.5)

rawDustValue = 0
dustDensity = 0

oldDustValue = 0
oldDustDensity = 0

# to put a blank line..
print "\n"

while True:
 command = serialport.readline()
 matchObj = re.match( '^(.*) = (.*)$', command, re.M | re.I)
 if (matchObj):
  # hash is unixTimestamp
  unixTimestamp = int(time.time())
  timestamp = time.strftime("%Y%m%d%H%M%S")

  # Var =  Raw Dust Value
  # Val =  137
  # Var =  Dust Density
  # Val =  81.51
  if (matchObj.group(1) == 'Raw Dust Value'):
   rawDustValue = int(matchObj.group(2))
   # print "raw dust value = ", rawDustValue

  if (matchObj.group(1) == 'Dust Density'):
   dustDensity = int(float(matchObj.group(2)))
   # print "dust density = ", dustDensity

 if ((rawDustValue != oldDustValue) and (dustDensity != oldDustDensity)):
  oldDustValue = rawDustValue
  oldDustDensity = dustDensity

  # print "hash = ", str(unixTimestamp)

  # calculate the PSI (this is a very approximate value)
  # based on dustDensity (in ug/m3) and this
  # http://www.haze.gov.sg/docs/default-source/faq/computation-of-the-pollutant-standards-index-(psi).pdf
  # we only use the 24-hour PM2.5
  # note that the Sharp sensor can't distinguish particle size
  # so PM10 particles are also falsely measured

  psi = 0
  if (dustDensity <= 12):
   psi = dustDensity * 4.17
  elif (dustDensity > 12 and dustDensity <= 55):
   psi = 51 + ((dustDensity - 13) * 1.17)
  elif (dustDensity > 55 and dustDensity <= 150):
   psi = 101 + ((dustDensity - 56) * 1.05)
  elif (dustDensity > 150):
   psi = 201 + (dustDensity - 105)

  psi = int(psi)

  # shorten timestamp so it fits on the tiny PiTFT screen
  ts = int(unixTimestamp) - 1444116000

  print ts, ":",
  print "raw=", rawDustValue,
  print " ug/m3=", dustDensity,
  print " PSI=", psi, "   \r",
  sys.stdout.flush()

  # write to the table
  try:
   table.put_item(data={
    'unixTimestamp': int(unixTimestamp),
    'timestamp': timestamp,
    'rawDustValue' : int(rawDustValue),
    'dustDensity' : int(dustDensity),
    'psi' : int(psi)
   })
  except:
   # do nothing
   pass

I then placed the Python script in /etc/rc.local (making sure to append an ampersand so that booting would complete).

And voila: