There’s no good way to draw that in ASCII art, so I won’t try. But it’s how things like envelope feeders or other automated machinery deal with thin flat materials.

Second, the air shuffler just isn’t going to work with cards no matter what you do; it works with lottery balls because they don’t stick together or to surfaces, but cards do — and the more air you put on them, the harder they’ll be pushed against the walls and each other to form vacuum bonds. They will also fall into the separator in at a totally crazy array of angles which will jam half of them no matter how perfectly the chutes are made. (And making the chutes too perfect and smooth isn’t a solution, because that makes the vacuum bond situation worse.)

1) I claimed that the fan was underpowered. This isn’t entirely true. If the fan moved 500 cubic feet of air a minute as advertised I’d have over 30 mph wind and that would be more than enough to blow the cards around. The problem is that airflow is restricted at certain points. One of those is the is the vent. I did try taping up some of the holes and it did not help. I actually found that removing the vent entirely gave the best performance, but it was still not good enough. I think an aerospace engineer could calculate why the observed airflow is so much less the theoretical max, but I can’t. All I know is I was hoping for 10%-20% drop and I’m probably seeing 70-80%. As it stands the cards get blown around as long as they are perpendicular to the airflow they get blown around, but if they turn sideways they fall and get stuck at the bottom. One person on my website suggested using compressed air and I like that idea, but it would increase the cost. I also think replacing the fan entirely with a spinning plexiglass cube might be a good solution.

2) I completely agree with those above that a webcam could easily detect if a card is upside down. However, I was planning to go a cheaper route. I was planning to use just an Arduino and no PC. I know if I put a black dot on one side of the cards I could reliably detect that dot with IR sensor. I think I could have also probably used a specialized clear IR paint. Another option would be to find a deck of cards that allowed the IR sensor to work without modifying the cards. I am pretty sure this problem is easy to solve so I was saving it to last.

3) Cards sometimes randomly got stuck in the slide system. There is a chance that adding paint and sanding would make the slides slippery enough so this didn’t happen, but I suspect I would have needed to do some minor tweaks to the slide system. By tweaks I mean change slide positions and angles.

4) I hadn’t hooked up the servos so I don’t know if there would have been a problem with selecting a single card. My hunch is that there would be a problem here and I’m not sure of an easy way to solve this one. My best bet would be buying those slightly thicker plastic cards.