August 3, 2011

Simple Arduino data-collection

At this year's "Arduinos in the Physics Lab" workshop at the AAPT meeting, one of the participants asked for a simple way of using the Arduino as a tethered A/D converter for data collection direct to a computer. This is my quick & dirty demonstration solution.

Here's the code for the Arduino. It waits for a single byte 'N' to arrive on the serial port, then once that byte arrives it sends out N data pairs formatted as tab-separated millis() and analogRead() values. The readings are separated by roughly 10 milliseconds. This version of the code only reports the values of analog pin 0 (A0), but it can be easily modified to return other (or more) ports.

For the computer end, I used Python: here's the code. This was done on a Macintosh, with Pylab installed so I can use matplotlib to handle the plotting nicely. On Linux or Windows the port will be described differently, and if the program fails for you on the line 'import pylab as pl' then ... well, install pylab on your system. It's a great wrapper package for scipy, numpy, and matplotlib. The program expects two arguments: the number of points to collect and the filename where points should be saved.

Here's a sample output plot, showing relatively meaningless data from a light sensor.
One glitch I found was that there needs to be a short delay between starting the serial communications to the Arduino and sending the request for N data points. I do not know whether this is a problem with the Arduino in general, or with the Arduino Uno I was using as a testbed, or with the pyserial library, or with the Macintosh implementation of pyserial... It was a mess trying to figure out what was going on, though, because when in interactive mode everything would work perfectly but the exact same commands in a Python script would not work. The difference of course is that I would take several seconds to type commands in interactive mode, and it took me a long time to figure out what was causing the problem! The solution I used is in line 37 of the code:
The sleep value (1.5 seconds) was determined by trial and error. ser.flush() should work also, but I did not find this to be the case.

1 comment:

  1. Nice work ;)

    I'll probably re-use this. Last time, I did that logging everything from the serial communication and plotting next using GNUPlot.

    To answer (a bit) your questions about this delay that must be set before starting to read the serial buffer : it's also an issue on Linux, with pyserial and the arduino IDE. It looks like a buffer overflow...