The Domino Effect

When taking on a task, it's generally accepted that doing it right is always better than doing it over later.  However, there are often limitations which mean that most tasks end up being done over later because you couldn't manage to do them right the first time.

Specifically, the aforementioned limitations typically stem from a previous attempt at doing a task which wasn't fully satisfactory.  When there are a number of these, it can start a whole line of problems which spiral out of control unless you set some limits.

Motor Wiring

The motor wiring on the robot was left quite short, so the wires would only reach the existing terminals.  Due to the decision to use a Mini-ITX form factor, the wires will not reach far enough to be useful.  Therefore, I decided to replace the wiring.  This also serves as a good time to colour code the connections.

Motor Capacitors

As I needed to solder new wires to the motors, it seemed a good idea to also add capacitors to reduce EMI sources on the robot.

I measured the voltage from the motor commutation, with and without a capacitor.  Without the capacitor, a spike and ringing voltage was easily visible.  With a 100nF capacitor installed, no commutation noise was visible.

Soldering

When it came to soldering the motor connections, there were some problems.  Due to a previous failure to solder to one motor's terminals appropriately, one motor barely had any terminals left.  It wasn't too much of a chore to make a good joint, but testing afterwards showed that the commutation wasn't 100% healthy.  Occasionally, the commutator seems to weld itself.  This problem doesn't seem to occur much, especially at normal loads.

Wheel Sensors

My priority for the day was to prototype the wheel sensor circuit - this was based on the circuit included in a past report, using a load resistor, potentiometer, and comparator.  A note here - check your datasheets carefully; I read a 'maximum' LED voltage drop as a 'nominal', because none was specified.  Fortunately, the current drawn was only just at the absolute maximum, so the LEDs survived.

The aim of tuning the sensors was to generate a perfect quadrature signal.  Adjusting first for a 50% duty on each waveform showed that the assembly wasn't performing terribly well.  The existing reflective surface was an etched PCB, and when testing paper (white, and laser-printed black), I found paper to be far superior.  With this in mind, I opted to replace the patterned discs.  I used Inkscape to design the new discs, and printed them on paper.  After discussing wheel removal (to mount the new discs) with one of the lab staff, I was advised to use a sticky-backed reflective plastic, printable film instead of paper, for added durability and simpler mounting.

The last part of my day was spent removing a wheel from its bearing (which took a moderate amount of time with a fairly heavy hammer) - applying the sticker was very simple, and I'll reassemble the unit tomorrow.