building a perler or any art craft bead sorter

justin weber

i had a TCS230-DB Color Sensor but never even opened it, but now it looks like ColorPAL is out and way cheaper. I originally got the tcs230 to make a m&m's sorter, but that project never got off the ground. I am quiet a noob in electronics, I did play with the propeller demo board to accomplish some things without external components.
My question is how hard would it be to build a sorter with a conveyor-belt and sorting tray that would automatically feed beads and sort them?

In my head I imagine some type of bottle with a restricted opening that the bs2 could open to let a bead out and fall onto a short stepper motor driven belt. it would advance to under the colorPAL, detect the color, and advance a motorized circular tray to the correct color and then drop the bead off the belt and then repeat the process. The primary purpose would be to sort perler beads(melty beads, hamma beads, fuze art). I don't have the skills to design it electrically, but if anybody has come across something like this before and would be willing to give me schematics or any other advise please reply to my topic.

As a side note: I do realize that you can buy presorted or bags of a certain color, so I'm looking for a very cheap sorter you can get tons of color shades cheap buying the assorted jars and this may eventually turn into a small online business so i willing to invest a little bit to make a device to sort them. The bs2 as a kit is about $80 and the colorPal is $15. I was aiming at $100, but even right there it is above that without any mounting hardware. I have no idea on what else is required. probably two stepper motors. One light duty for the belt and one fairly heavy duty for a sorting tray. maybe a couple photoeyes to detect position and if a bead even came out or the bottle/bag feeding the belt is empty. I still have the propeller demo board, but i'm sure i fried some I/O ports messing with stuff, so just assume that i want to start fresh with a bs2 and the colorPAL as a sensor.

justin weber

i have some experience programming. Ti-89 basic, C# and i understand the basic concepts of the propeller chip, so i can probably manage the code without much or any help. I just can't think of how i would start a project like this.

Post Edited (justin weber) : 12/12/2009 8:13:14 AM GMT

Mike Green

The code and the wiring is the easy part. The heart of a bead sorter is the conveyor that has to separate the beads and move them one at a time past the color sensor, then have some way to move the bead off the conveyor to a container (like your circular tray). I assume that your beads can come in all sorts of sizes and all sorts of shapes. That makes it tougher. Unless you have a design already in mind for the mechanical stuff, I suggest you start with a sorter for beads of a uniform shape and size (like M&Ms) so you can get some experience with the design issues.

One difference that comes to mind with beads as opposed to M&Ms is the beads' tendency to roll. On a conveyor belt, I'd probably use something with a dimple or groove to hold the bead in place. If the beads are large enough so the sensor could "look" at each one individually (and they'd roll), they could sit in a v-shaped channel that's tilted. You'd have a gravity feed. You'd need a "gate" at the far end with two solenoid driven "blades" that would block off the channel. The nearer one would block off the channel to keep the other beads from rolling down and the one further along would hold the subject bead while its color was being examined, then while the circular tray underneath was being turned. When the right position was reached in the tray, the further gate would be pulled back and the bead allowed to drop into the tray.

justin weber

great thoughts on the conveyor. i was thinking after reading your comment, maybe something from an RC car. like the high speed cars that have a belt between the front and back wheels on the left and right. they are mainly ribbed(or v shaped as you say) for traction and would likely let the bead say put. The beads are all about the same size with some imperfections. maybe even the feeder mechanism would channel them further into a single row. To me this part is just as hard as the conveyor mechanism. I also am wondering about any pre made small circular divider trays that have lift out tabs that would let you put a bag or jar under to release the sorted colors. From this point this project looks like it could easily go over $200. If the circle tray is used the dividers could only be so small to make sure the bead makes it in more often than not. So there are many many shades of colors in these 22,000 assorted jars and maybe you would "set" certain colors to sort and then the belt would have to advance back and reject them by pushing them back off the belt into a secondary jar for the second sort. when the bottle is empty you would empty the trays and sort the additional colors, unless a sorting tray could be made in a double stack method where one of the top dividers is a big hole to allow additional colors to make it into the bottom tray. Again I don't even know if the project is worth it -- educationally it would be awesome. I'd have to sort a couple million beads to justify the cost of building this over the extra it costs to get single 1000 pack beads at $2-3 of a single shade of color.

jameco.com has a tank wheel set: http://www.jameco.com/webapp/wcs/stores/servlet/ProductDisplay?langId=-1&storeId=10001&catalogId=10001&productId=358299. I imagine needing motor mounts and probably small hex barstock sections to hold all the rollers and place and a frame of some sort. And all the rc servo units i'm seeing don't look like they would have the torque to advance a divider tray i'm imagining. It would most likely weight under a pound and the beads weight nothing -- they are basically hardened colored glue. maybe a dish of some sort from the kitchen. like an dish made for veggies for a snack tray. the one i have has cups the pull out for cleaning and would be perfect. but those would be pretty heavy. Jameco also has a Gripper servo as well but it has huge fingers. your idea of the "blades" is probably the way to go.
I did venture to the ColorPal datasheet and looks easy and well suited to this task. It looks like the DB color sensor needs all kind of lighting and the accuracy is determined by the time given to sample the color whereas the colorpal automatically takes lighting into account and looks to have a already predetermined sample time.

Post Edited (justin weber) : 12/12/2009 6:45:26 PM GMT

Mike Green

There are all sorts of ways to do the mechanical part. I've seen sorters where there's a long line of little bins off the side of a conveyor belt paralleling the belt and there are little solenoid controlled "pushers" on the other side of the belt. The controller categorizes the "beads" and remembers where they are along the belt as the belt advances a step at a time. When the bead(s) reach the proper point, the pusher pushes the bead off the belt into the bin and the belt advances one step. You could use a servo motor to swing a little hinged gate as well ... probably be easier to use than a solenoid.

Even the belt driver could be a continuous motion servo. That way you'd be able to easily control the belt's speed. You'd need to know the position of the belt. You could do that be attaching little "flags" to the belt that would interrupt a beam of light. You'd have one flag at each position where a bead could sit and another one on the other side to mark the end of the row of little bins so the Stamp could easily keep track of the belt's position.

You could do multiple "levels" of sorting. First you'd sort all of the beads into general color groupings, then you'd take each bottle of sorted beads and sort them again into finer categories of color. If you had only 8 output bins and did a total of 9 runs, you'd be able to sort the beads into 8 x 8 or 64 color categories.

Post Edited (Mike Green) : 12/12/2009 6:48:28 PM GMT

justin weber

oh so basically like i was saying about the limited dividers. sorting for certain color and pushing colors that don't have bins off into a secondary container for a second sort, but instead of that have the "secondary" bin use that system as the sorting bins. place them parallel to the belt and advance the belt forward past the sensor to push them off. That would work better(easier) than advancing a circular tray and you could just pick up the jars and dump them into a bag. Sorting the colors by groups also sounds good. that would minimize the length of the belt. I could have jars on both sides of the sensor and it could advance the belt back towards the feeder, but I'd probably need all the room to get the beads to line up one by one. The tank wheel might still work but I'd need a bigger belt. Even if the feeder comes from overhead instead of one one side of the belt; the belt at the url doesn't look big enough for this application.

actually, this sounds like something a kid would do for a science fair project. It definately is seeming more feasible to build now, but with my limited design skills the cost of prototyping would be excessive. It's not like i know exactly what i need, buy it slap it together tweak a little and i have a working sorter. the conveyor belt and bins would be something that with more research i could manage the first time, but the feeding mechanism would probably require tons of different designs because the beads can stand up on end or roll on thier side when dropped onto the belt. There is a big hole in the top so to measure the color they would HAVE to be on thier side. Not only because of measuring area, but because sometimes the cutter at the factory gets dirty and leaves marks on the top and bottom of the beads.

Post Edited (justin weber) : 12/12/2009 7:09:34 PM GMT

Phil Pilgrim (PhiPi)

One way to perform singulation (the term used for separating items on a conveyor) is to use two belts end-to-end and to run the downstream belt slightly faster than the upstream belt. As product transfers from one to the other, gaps are created automatically. In order to feed the first belt, you could build a small shuffalo. This could be embedded in a hopper to lift beads from it, one single-file row at a time (wide shuffalo), or just one bead at a time (narrow shuffalo, as shown in the link), and drop them onto the conveyor. (I first saw a shuffalo in a pickle plant. They're really fun to watch in operation. I even built one from Legos once to feed wooden pegs onto a conveyor to sort them by size.)

As to the ColorPAL vs. TCS230-DB/TCS3200-DB question, please read the comparison chart in the ColorPAL docs. The two significant aspects to note are working distance (ColorPAL works best in contact with the subject) and field of view (TCS230's is smaller). With either sensor, you will have to stop the conveyor momentarily to take a reading, since each RGB color component is measured separately, one after the other. (If you use a single-bead shuffalo, you could take a reading at the top "step" before it's fed to the conveyor.)

-Phil

Addendum: A narrow, double-sided timing belt (toothed belt with teeth on both sides: one side for the drive sprocket; the other, to pocket the beads) would make an excellent conveyor for beads if it's accompanied by sloping rails on each side.

Post Edited (Phil Pilgrim (PhiPi)) : 12/12/2009 7:19:46 PM GMT

justin weber

phil that is very helpful. i never thought about two belts. i did read the comparison chart in the colorpal doc. i assumed that i would mount the sensor over the belt and have the bead advance under it. I would think that doing this would narrow the field of view to a single bead.Singulation would allow beads to drop out of the feeder in groups. By groups i mean blades could open for a time and a bunch would out and then shut. But even with the blade method a bead could get stuck as the blades closed, which might not matter. The two belt system would allow seperation, but I was thinking this whole time the feeding mechanism would be designed to only allow one bead to be on the conveyor at any one time. Maybe that is impratical and somehow getting one between each tooth would not only be faster but may be the only way to get them lined up for the sensor.