2012-09-08

Raspberry Pi distance measuring sensor with LCD output

Measure distances from the Sharp GP2Y0A02YK0F sensor using an MCP3008 ADC and hardware SPI.


The Sharp GP2Y0A02YK0F can be powered from the 5V supply on the Raspberry Pi. The Analog output is less than 3V and so can easily work with the logic level circuit. If you buy one of these look for the cable that goes with it to save you some bother.

This project builds on two previous projects in this blog. For the Analog to Digital SPI electronics and Python code first go here: Raspberry Pi hardware SPI analog inputs using the MCP3008. For the TextStar LCD first read this: Raspberry Pi with TextStar Serial LCD Display.

In the first video below the sensor is held in a clamp at the very top left of the picture. You can just about make out the display showing the distance to my hand. At the other end of the table there's a chair so when I lift my hand up it shows the distance to that instead - about 118 cm. The second video is a close-up of the screen while I move my hand forward and backwards in front of the sensor. Again the jumps to 118 cm are when I raise a lower my hand. 



All the code is available on github. You'll need mcp3008.py and distance-screen.py to run this project. The code assumes that you have the sensor output connected to CH1 on the MCP3008. The main routine to look at in the code is the translation from the sensor output to a distance. In the datasheet there's a graph which plots distance against voltage output:
I compared this against my sensor by laying a tape measure on the table and looking at the voltage output - it matched perfectly. I was about to start working out a formula to fit this curve when I found a good set of comments on the Sparkfun page including a magic formula!

Here's my Python routine which is called 10 times a second using the on_tick handler from the screen driver:

def write_distance():
    display.position_cursor(1, 1)
    r = []
    for i in range (0,10):
        r.append(mcp3008.readadc(1))
    a = sum(r)/10.0
    v = (a/1023.0)*3.3
    d = 16.2537 * v**4 - 129.893 * v**3 + 382.268 * v**2 - 512.611 * v + 306.439
    cm = int(round(d))
    val = '%d cm' % cm
    percent = int(cm/1.5)
    display.ser.write(str(val).ljust(16))
    display.capped_bar(16, percent)


  • Firstly, take ten readings and find the average (a) - this smooths things out a little.
  • Convert this to a voltage (v)
  • Use the magic formula to get the distance (d)
  • Round this to the nearest centimetre (cm)
  • Then write this to the display along with a 16 character capped bar graph
So the code happily reads from the sensor 100 times a second and does the calculation and screen update 10 times a second. The CPU usage hovers around 2% in top.

6 comments:

Michael Horne said...

Fantastic sensor experiment - thanks for sharing :-) I'll have to add that to my Pitricorder project :-)

Anonymous said...

great idea!

do you know if it's possible to use a sensor that can deal with more then 150cm?

Anonymous said...

Hi
My Dad is blind and I'm looking into a project to make some kind of wearable device (hat, glasses, armband, kneepad etc) which detects upcoming obstructions and feeds back via audio or vibration to alert him. He keeps banging his head on things because his stick is usually sweeping low. This looks like a good start for the distance detector part. Would this be the right approach? Do you know of any projects which have done the whole thing? I haven't bought a Pi yet, just done 15 minutes surfing.
Chris (Ex-developer, now pre-sales)

jerbly said...

The infra-red sensor has a very focused point and so you may be better off using an ultrasonic sensor that has wider detection area. Otherwise you may suffer from false readings and areas where the sensor cannot detect the obstacle ahead. Perhaps a combination of the two would give you the best of both worlds.

Anonymous said...

Thanks for the tip. I've modified my thinking to have 2xHY-SRF05 ultrasonic devices for peripheral 'vision' (pointing NE & NW) and one GP2Y0A02YK0F in the centre pointing N. I prefer C to Python. Is it going to be any kind of problem driving 3 detectors?

seregus said...

Hi!
I'm really inspired by your work and using some of your ideas for my project.
But wanted to share something about using Sharp IR sensor/Sonar - I've found it's quite easier to use I2C protocol and actually both sonars and sharp sensors are available on the marked in I2C version (simply google sharp i2c or sonar i2c), because you don't need any chip in between just connect some cables, enable I2C support for Raspberry PI (uncomment 2 lines of config) and ... just a simple command return you all necessary values!