Diverter:

I've nearly completed the diverter build now, and have got the stepper motor running under Arduino control and tuned to go as fast as possible without mis-stepping. With eight possible positions, the diverter needs to move - at most - halfway round the ring, which takes 160 milliseconds. (Moving to the neighbouring position takes only 62ms.) A Lego brick dropped into the top of the diverter takes at least 132 milliseconds to drop out of the other end (t = sqrt(2d/g), i.e. assuming in free-fall - in reality it will mostly likely take a bit longer whilst the piece bounces around inside). This means we need to have finished the colour measurement and classification process about 30ms before the piece enters the top of the diverter tube to be sure of it arriving at the other end in the right position. It also means the machine will never process pieces faster than about 6 per second, however clever/fast it is, but y'know that's fast enough.

Feeder:

Some interesting experiments with a 500mm length of vee-shaped plastic channel. I've fitted a small motor under one end, and attached an off-centre mass (a plastic hub with a few screws in) so that it vibrates everything. I've been playing with varying the position and stiffness of the springy mounts which support the channel, plus the motor speed and incline of the channel, and observed some interesting effects:

• (obviously) changing the incline changes the speed at which the bricks move along the channel
• increasing the amplitude of the vibrations 'jiggles' the bricks and helps them to separate from each other; only quite small vibrations are required just to move the bricks along.
• choosing the right motor speed causes a resonant mode to be set up in the channel, such that one section vibrates more strongly than others. Bricks in these regions tend to separate from each other and move along more quickly.
• So, by carefully tuning everything, I hope to vibrate the channel most strongly at the 'exit' end, so that a pile of bricks dropped at the 'entrance' end move slowly at first, increasing in speed and - critically - separating from each other before they reach the exit, appearing one at a time past the colour sensor and into the diverter. 

Well, that's the plan anyway...