zaterdag 13 september 2014

DIY speed measurement with Raspberry Pi

I am working on a high tech bird house for in my backyard. I am hoping next spring it will be occupied by a couple of great tits.

 Did I just write 'great tits'? Yes I did, in Dutch it is just a simple 'koolmees', but Google translator actually translates this to 'great tit'. I thought Google was being funny, but when searching images of a great tit, you actually get pictures of great tits instead of .... great tits [link removed by author].

Anyway, the plan is that a pair of little great tits start their family in my birdhouse. And I will be on the front row. The birdhouse will be equiped with a night vision camera, a bunch of sensors and a scale to weigh the bird's nest. It is all connected to a Raspberry Pi and made visible on a web site. More will follow this autumn.

One of the features will be a couple of light gates that monitor the birds flying in and out. This setup can also be used for measuring the speed of an object passing each of the light gates. In the demo video you see I can shoot my finger with a speed of 7m/s. That is more then 25km/h or 15mph. I even managed to reach 11m/s once off-camera. So you now know that you can pick your nose and shoot the content with 40km/h (25mph)!


The picture below shows the schematics. It has two 3mm IR leds (I don't know the type numbers) that are powered by the 3V3 supply of the Raspberry Pi. The leds don't need to be very powerful because the IR receivers will only be 3cm away. These receivers are BPW85 NPN phototransistors that are sensitive to visible and IR light. The base voltage is dependent of the amount of light that it receives which allows current to flow from collector to emitter. I used a 10K variable resistor to be able to tune the voltage drop to a HIGH level of 3V3 that I can feed to the RPi. So, when the lightpath is blocked it outputs 0V, and 3V3 when the lightpath is not blocked. I found out that when I increased the distance between IR transmitter and receiver to over 4cm the light wasn't strong enough to reach a 3V3 voltage drop. If so, you have to increase the current through the leds or make a buffer (like a Darlington pair) to make it work.

The software is made in Python. The standard way of checking for inputs on the Rpi is with polling (if GPIO.input(x): blablabla). In my previous blog post I got a tip to use edge triggering, which is faster, more reliable and relieves the CPU. Basically the program is always sleeping, only when one of the GPIO inputs goes from a HIGH ('lightpath is unblocked') to a LOW ('lightpath is blocked') level a dedicated function is called. This feature is explained very thorough on Adafruit, but the code below shows the basic example how to use it.

The code waits for a falling edge on both GPIO inputs 2 and 3 (pins 13 and 15). If one the two is triggered, it checks if the other input was already triggered. If this is the case it calculates the time difference in microseconds for a precise measurement. The distance between the light gates is 3cm, which is needed to calculate the speed.

For the upcoming birdhouse application it also checks if the bird entered or left the birdhouse, but for the booger shooter application I suggest you always shoot in one direction, that is AWAY from you.

The code can be found here (link):

zondag 17 augustus 2014

Pi-files: Frontdoor light switch based on local sunset/sunrise

I have a couple of front door LED lights which I would like to switch on automatically during the evening/night. The two conventional methods that are commonly available had their disadvantages:

A timer switch is the easiest and cheapest solution, but doesn't take into account day light savings. Besides that, in Einhoven, the Netherlands where I live in december the sun sets around 16:30 while in June it doesn't get dark before 22:00. A simple timer doesn't take that into account either.

Since you only want the lights on when it gets dark, instead of time you can also use a light sensor to distinguish day and night. You have these front door lights that have this built in. The problem that I had with these devices is that they start bouncing ('flickering') around sunset/sunrise. They constantly turn on/off which causes damage to the LED lights I was using. This cost me already several expensive led lights.

Since I already have a Raspberry Pi installed in the hall way (used for the doorbell alert) I might as well hook up the front door lights. You can find on the Internet all sorts of timetables for the sunset/sunrise, so my Rpi should be able to use that same information as well. A small search on the world wide web led me to a Python package called PyEphem. This package can be used to do astronomical calculations and can tell you for example where a certain object (e.g. the sun) is with respect to the observer (which is me standing in my front yard). Very conveniently it also can calculate for you the upcoming sunset and sunrise. Isn't that exactly what I need?!?

The LED lights are powered with a 12V AC-DC transformer. I put together a small electronics board to be able to use the GPIO's 3V3 of the Rpi to switch the 12V LEDs. I used a IRL3103 N-FET. This is not really a deliberate choice but like most of my transistor/FET choices based on what I have lying around in the lab. Be aware that the N-FET needs to be able to switch with a Gate-Source voltage of 3V3. For example an IRF540 (which I used for a motion detector for a 12V ledbar) doesn't switch with this voltage (Here I solved it by using a smaller BS170 N-FET and use that to switch a IRF 9540 P-FET). I added a manual switch to override the automatic switch in case I just want to turn on the lights. Parallel to the front door LEDs is a test led with a resistor onboard to be able to test the functionality without having to connect the actual front door lights.

I use GPIO pins 18 and 22 to switch the two LED's. You can switch LED's with only one output, but I thought it is a nice additional feature to switch them separately to be able to also create a spooky program for Halloween or something like that.
After including the PyEmphem package I have to indicated where the observer is located. I used itouchmap which can translate an address to a latitude and longitude coordinates. Being an observer in Eindhoven means you are located at [51.5, 5.5]. The PyEphem package calculated the next sunrise and sunset and returns these values. I do this every minute and compare the two; if the next sunset if further away in time than the next sunrise, it means it is night. The other way around means daytime. The LED lamps are switched based on the day/night status.

The code can be found here (link):

maandag 4 augustus 2014

Pi-files: Doorbell alert with pushmessage and mail with webcam footage

Actually the first 'project' I ever did with a Raspberry Pi was sending a push message to my Iphone. It was 2012, I was lying sick in bed and found a new app on my Iphone called Pushover (what else to do when you're sick?). With Pushover you can send and receive custom made push messages. On the website I found a simple Python script to send messages. I knew the Rpi was able to run Python code, so here my Rpi adventures started. Within 30 minutes I was able to receive 'hello world' on my phone (needless to say I wasn't lying in bed anymore). Seeing 'hello world' on your screen is like the software equivalent of the blinking led, THE coolest feature ever!

I decided to hook up this push message feature with my doorbell. The idea is that every time somebody rings the doorbell, I get a push message that there is somebody at the door. The wires of the doorbell were already connected to a wireless transmitter and I wanted to keep that functionality. I used a relay to combine the transmitter with a switch on the GPIO header of the Raspberry Pi.

The led and 330R resistor can be installed in the actual doorbell, so the person can find the doorbell and is triggered when the button is pressed in case he/she doesn't actual hear the bell (the led turns off when the doorbell is pressed). The 100u capacitor and flyback diode are to limit voltage peaks when using the relay. The 'test' switch is available close to the GPIO header to test the (software) functionality of the doorbell alert. The relay and test switch are connected to pin11 (GPIO0) of the Rpi.

The Python code looks as follows. First the necessary libraries are included and the GPIO header is configured:

#import libs
import RPi.GPIO as GPIO
import httplib, urllib
import time
from time import sleep, localtime, strftime

#display no warnings (on command line)
#set GPIO: use RPi board pin numbers
GPIO.setmode(GPIO.BOARD)  #alternative is GPIO.BCM (Broadcom names)
#set pin 11 as input
GPIO.setup(11, GPIO.IN)  #Input: doorbell (relay)

A seperate function is written for the push message. This is mainly copied from the Pushover FAQ section ('How do I send Pushover notifications in Python?'). The token ('push_token') and user ID ('push_user') you get when you sign in on the website. There you can also select the devices on which you want to receive the push messages.

#set up push message (pushover)
def push(text):

   push_token = 'xxx'
   push_user   = 'yyy'

   conn = httplib.HTTPSConnection("")
   conn.request("POST", "/1/messages.json",
                urllib.urlencode({"token": push_token,
                                        "user": push_user,
                                        "message": text,}), 
             {"Content-type": "application/x-www-form-urlencoded" })

The main loop checks for the GPIO input regularly. I added a tiny bit of delay (sleep) to relieve the CPU. When the doorbell is pressed, it takes the current date and time and adds that to the push message.

while 1:

    sleep(0.1) #relieves CPU load big time!
    #if doorbell is pressed...
    if GPIO.input(11):

          #determine date/time and add to message
          timestr_date = strftime("%a %d %b %Y", localtime())  
          timestr_time = strftime("%H:%M:%S", localtime())      

          mess = "Doorbell pressed on {}, {}.".format(timestr_date,timestr_time)
          print "\n"
          print mess


When this all worked it was time to extend the doorbell alert feature. Not only I want to know that there is someone at the door, I also would like to know who was at the door, especially when I am not at home. I added a Logitech C270 webcam to the setup to capture snapshots and a short movie.

A Python mail script is supposed to send it all to my mailbox. The mail functionality I didn't invent myself, I just relied on the beautiful internet community. On Kutuma's Ramblings I found almost exactly what I needed. The only difference is that I wanted to sent multiple attachments. For that I changed the mail function a little bit:

def mail(to, subject, text, attachments=[]):

   gmail_user = ''
   gmail_pwd  = 'YYY'

   #add attachments to a list
   assert type(attachments)==list

   msg = MIMEMultipart()

   msg['From']    = gmail_user
   msg['To']      = to
   msg['Subject'] = subject


   for attach in attachments:
     part = MIMEBase('application', 'octet-stream')
     part.set_payload(open(attach, 'rb').read())
           'attachment; filename="%s"' % os.path.basename(attach))

   mailServer = smtplib.SMTP("", 587)
   mailServer.login(gmail_user, gmail_pwd)
   mailServer.sendmail(gmail_user, to, msg.as_string())

You can now add the attachments as an comma separated array:


As you can see I added three snapshots and one movie as attachments. They all come from the Motion package which is a linux package to capture motion on a webcam. To be honest, I actually don't use the motion feature of that package, but I tried several other solutions (like FFMpeg or something) and came across a bunch of issues with compiling, frame rates etc. The Motion package served my purpose the best. 

NB. The Raspberry Pi that I am using is a model A, which is not capable of using the official Rpi camera module with its cool movie and snapshot features.

To install the Motion package, run the following command:
pi@raspberrypi ~ $  sudo apt-get install motion libv4l-0 uvccapture

I added some additional modules to the import section of the Python script described above and added the code below to the main loop. Starting the webcam is done right after the doorbell is pressed. After capturing the snapshots and sending the push message, the program waits 60 seconds until the webcam is stopped. After this the mail is send with the snapshots and movie.

import subprocess
import shutil #high-level file operations

    #if doorbell is pressed...
    if GPIO.input(11):

          #start webcam (motion package)
          print "\n"
          print "Start webcam (Motion)"
"sudo /etc/init.d/motion start", shell=True)
"sudo motion", shell=True)

          #copy (save) most recent snapshot of webcam
          print "\n"
          for k in range(3):
               sour_path = '/home/pi/rpi/webcam/motion/snapshot.jpg'
               dest_file    = '/home/pi/rpi/webcam/motion/screen%d.jpg' %k
               shutil.copy2(sour_path, dest_file)
               print 'Snapshot captured:  %s' %(dest_file)

          ****** Sending push message (see above) ******

          print "wait 60 seconds"
          #stop webcam
          print "Stop webcam"
"sudo /etc/init.d/motion stop", shell=True) 

          #send mail via Gmail


The motion package has a pretty big configuration file where you can configure all sorts of things. Most of the items speak for their own. The file can be edited with nano:

pi@raspberrypi ~ $  sudo nano /etc/motion/motion.conf

For the doorbell alert let me highlight the most important sections:

# File to store the process ID, also called pid file. (default: not defined)
process_id_file /var/run/motion/ 

The folder /var/run/motion needs to be present. In my latest Raspbian (Debian Wheezy) OS I encountered a bug (as far as I can tell) that this folder is gone every time I boot up the Rpi. For that reason I added a check and create the folder if necessary:

import os

folder   = "/var/run/motion"

if not os.path.exists(folder):
  print "Folder does not exist. Create folder " + folder

# Image width/height (pixels). Valid range: Camera dependent, default: 352/288
width 1280
height 720
Bigger images won't hurt!

# Maximum number of frames to be captured per second.
# Valid range: 2-100. Default: 100 (almost no limit).
framerate 2

I only use two frames per second to keep the movie size as small as possible.

# Always save images even if there was no motion (default: off)
output_all on

I don't use the motion detection features so I want all images to be saved

# Use ffmpeg to encode a timelapse movie
# Default value 0 = off - else save frame every Nth second
ffmpeg_timelapse 0.5

# Make automated snapshot every N seconds (default: 0 = disabled)
snapshot_interval 1

Save snapshots every second

# Target base directory for pictures and films
# Recommended to use absolute path. (Default: current working directory)
target_dir /home/pi/rpi/webcam/motion

Location where all files are stored

# File path for snapshots (jpeg or ppm) relative to target_dir
snapshot_filename snapshot

All snapshots are given the same name (no timestamp) to save storage space.

# File path for motion triggered ffmpeg films (mpeg) relative to target_dir
movie_filename movie

Each movie is given the same name (no timestamp) to save storage space.

That's it for now! I am already thinking of new features like face recognition, direct videostream, and a call back option to actually talk to the person at the door wherever I am. The sky is the limit! Unfortunately my spare time as well :)

Below a picture of my custom storage box which is located in the meter cupboard (again, is that really the correct translation for the dutch word 'meterkast'?). It houses a Raspberry Pi model A, an interface board (on the right) with the well famous heartbeat led. On the left two boards for the doorbell and one for the front door light which I will explain in another post soon.

zondag 27 juli 2014

Pi-files: Heartbeat

It has been a bit quite here on Deddies lab's blog. That doesn't mean that nothing came out of the lab. Actually a lot has happened, but I didn't had the opportunity (or was a bit too lazy) to post it here. Since I took a lot from the DIY community, I feel obliged to return also my findings, struggles and results back to that same community. Only then we all can grow!

Last couple of years, I invested most of my (spare) time in the Raspberry Pi platform. I don't think I have to explain anything about this versatile mini computer. Just google around and you will find tons of projects. I myself was attracted at first to the Rpi because of the General Purpose IO's (GPIO) possibilities which can be used to connect simple inputs (switches) and outputs (nothing beats a blinking led!). I was (still am) used to work with Atmel microcontrollers, where also the pins can be configured as input and output pins. So being able to use the same IO functionality but then combined with ethernet, a decent user interface, etc. etc. etc. opened a whole new world of possibilities.

Let's start easy; Sometimes it is difficult to see from the outside of the Pi is running or not. It has some status leds onboard, but when built in a box and stuffed away in a cupboard (like I have) the status leds are not so easy to see. So for that purpose I made a simple heartbeat. It is just a blinking led, but starts automatically when the Pi boots up, so it indicates when the Pi is alive and kicking!

I have a revision A of the RPi in my 'meterkast' (is it really 'meter cupboard'?).  For this heartbeat and some other projects (doorbell alert, front door lights) I made a dedicated interface board. It also has a led which is clearly visible from the outside

This led is connected to pin 16 of the GPIO header, which is called GPIO4. No, this doesn't make sense, and apparently there is a difference between the physical Raspberry Pi names and Broadcom names. I made a Python script (called where I can first configure the IO's:

import RPi.GPIO as GPIO
from time import sleep

#display no warnings (on command line)

#use RPi board pin numbers
#set pin 16 as output
GPIO.setup(16, GPIO.OUT)

The 'sleep' module is inserted to make the led blink. The while loop makes the led blink indefinite:

print "Start heartbeat\n"
while 1:


You can run this script from the command line:

pi@raspberrypi ~ $ sudo python /home/pi/gpio/scripts/ &

When you want to run the script automatically after boot up of the Rpi, you have to edit /etc/rc.local:

pi@raspberrypi ~ $ sudo nano /etc/rc.local

#!/bin/sh -e
# rc.local
# This script is executed at the end of each multiuser runlevel.
# Make sure that the script will "exit 0" on success or any other
# value on error.
# In order to enable or disable this script just change the execution
# bits.
# By default this script does nothing.

python /home/pi/gpio/scripts/ &

exit 0

Save the file with Ctrl+X. Make sure that the Python script is executable:

pi@raspberrypi ~ $ sudo chmod 755 /home/pi/gpio/scripts/

You can check if the script is executable if you see the 'x' in the permission overview:

pi@raspberrypi ~ $ ls -la

-rwxr-xr-x 1 root root   261 Jul 21 21:34

Reboot the Pi and see the led blink:

pi@raspberrypi ~ $ sudo shutdown -r now

donderdag 29 mei 2014

Motion Detection for 12V ledbar

On an electronic flea market I bought two 50cm 12V (700ma) ledbars. They were quite bright and only cost 6 euro a piece. I still needed a proper lighting solution for my work bench and thought these could perfectly serve that purpose.

With some two components conductive glue I glued them on a aluminum strip for thermal conductivity. A wooden frame prevents to look right into the leds when you enter the room. It definitely gives a nice bright even illuminated work bench.

I didn't want to use a switch to turn the lights on and off. It is just not cool! Recently I bought 10 PIR motion detection sensor modules (HC-SR501) on ebay for only 10 euro (How can they even make that for one euro?). I thought one of these can do the job for me when I approach my bench.

The PIR sensor module outputs 3V3 when motion is detected. Sensitivity and ON time can be adjusted on the module. It operates on supply voltages between 4.5 and 20V so the 12V (1.5A) adapter I need to power the ledbars can also power the PIR sensor. I wanted to use an IRF540 NFET to switch the leds, but this NFET doesn't open up completely with a Vgs of 3V3 which limits the current flowing through the leds. That is why I used a smaller BS170 NFET that can be triggered with 3V3. R1 is a 10K resistor to prevent a floating gate signal. Due to another 10K resistor (R2) at the drain of the BS170 an inverted signal of the PIR sensor is available, so a PFET (IRF9540) can invert back that signal to switch on the ledbars. For troubleshooting purposes I added a jumper to  put 12V directly on the ledbars and a test led (always add a test led!) to see the circuit works properly even if the ledbars are not connected.

 I stuffed everything into a small plastic storage box that I found in the kitchen and made it possible to rotate the whole box to be able to point the PIR sensor towards the workbench. It turned out to be a very nice local lighting solution. I doesn't turn on when I enter the room or walk around, only when I enter the work bench it turns on. Altogether it didn't cost me more then 15 euros, I don't believe you can buy something like this for so little money.