|
After developing the PCB based internal wheel encoders sometime ago (See Part 1 and Part 3), I finally
got back to debugging them. The robot's behavior was very erratic, I was
getting better results using timers for my turns. With the aid of an
oscilloscope I started looking at what was coming out of the encoders. The
black trace represents the left encoder, while the green trace is the right
encoder. The left encoder wasn't too bad, the front end needs de-bouncing,
which could be done either in hardware or software. The right channel looked
awful, with glitches and random pulses. I disassembled the servo and
discovered that the foil tape on the gear had started to peel and paint chips
had fallen off the tape and the gear. |
 |
| My mistake was using a cheap paint that was brushed onto the
gear and apparently didn't stick. So I stripped the gears down and
started over. This time I sprayed the gears flat black with spray
paint and applied 4 small strips of white tap ( I couldn't find my silver pinstripe
tape I used before). I also applied epoxy each edge of the tape so
it wouldn't peel off.
|
 |
| Once the gears were completed and back in the servos, I
hooked it up and got the following outputs. At first glance things
look pretty clean, but then I realized the left channel was only putting
out 1 pulse for every 4 pulses on the right side. Thinking the tape
fell off I took the servo apart again, but everything looked ok.
Then I realized the PCB with the reflector sensor was sitting higher up on
the servo, then the right one was.
|
 |
| The remedy was easy. Since the whole sensor was on a
PCB, I popped it loose and oversized the hole in the servo case a little
more, so I could move the PCB around.
|
 |
| With the servo running, I moved the PCB around till I got a good signal,
then glued the PCB in place.
|
 |
| Here's what the final signal looks like on the scope.like this one
Now I have a pair of nice clean pulse trains coming
from my encoders and can continue debugging the firmware.
|
 |
