The Wikiplacer project

[stewi]

Until we have a host and website for the Wikiplacer we continue here.

In our first on-line brainstorming session, we got already quite some details accomplished.

We’ll need your and customer feedback on the requirement for max-min board size as well component range, upon which we make a decision on required placement accuracy and travel area.

Likely we’ll decide on linear motors (custom or off the shelve) for the X and Y positioning system.
We’ll strive for high accuracy for the lowest possible price. Also, the machine should run unattended at a 4 Sigma certainty, which we can only achieve with automatic and reliable component feeding method and closed loop positioning system.
In order to keep the masses low and cost for such linear motors, we’ll likely decide on a light-weight (carbon) fiberglass gantry, laminated over a foam core.
Unless we get convinced otherwise, we’ll inspect the component on the fly with a camera vision system. We briefly discussed the option to view the component from the top and through the nozzle, but discarded this concept due to development time while dealing with light reflections and background while the component is moving from pick up to placement.
With the component inspection at the placement head while traveling in X and Y, the Z axis either has to be raised in order to get a camera or mirror underneath, or the component has to be rotated or moved into the camera optics.
With many years experience in rotary heads, I conclude a small size 4 nozzle rotary head, would be best for this purpose.
A second camera will be used for board fiducial recognition and a third, stationary camera will be used for larger, odd shaped or micro BGA components.
The feeders, already proposed in the open source thread will further be simplified and size reduced. We’ll likely remove the sprocket wheel out of the design and only sense the perforation holes with sensors located at the feeder table. As a trade off, we will not be able to use clear plastic tapes.
The empty carrier and cover tape will need to be disposed properly, not requiring operators with scissors or clearing an empty tape mess.
We will also consider intelligent feeders with RF ID tag in order to store component information as well as remaining number of components in the tape. The feeder component information should minimize set up time, but should also provide early warning messages of feeder tracks going to run empty.

All we know, we are setting our goals very high. We will require your support and input but I’m convinced we get there.

[stewi]

Before I go into weekend relaxation, I would like to discuss, what I got so far.
I was briefly searching for off the shelve encoder discs, without luck so far. Also knowing, that glass tends to shatter, I'm either looking for plastic or metal discs. Finally, I think they could be custom made, with photo etched lines.
Just for looks I made 800 lines on a 50 mm disc, but I think 900 lines and quadruple sensor would be best for 0.1 resolution.
I also found in my stash 4 mm shafts and 2 linear ball bearings (NB SM4G). Bearing OD is 8 mm length 12 mm.
If I place 4 discs in a circular pattern, I come up with a head of slightly over 50 mm diameter.
I placed 4 linear shafts, 2 each in a 8.1 mm offset in order to clear the other bearings and came up with a stroke of 14 mm.
For nozzle rotation and vacuum seal I like to use tapered discs, pretty much like piston engine valves. Here we would need to experiment with different materials and find how long they last.
Also, I will have to get my hand on a Siplace nozzle and take dimensions.

So far, I have no motors yet. Small DC servos come to my mind. However, if we want to be affordable, we need to look elsewhere than Maxxon, Escap and Faulhaber. I may look into the Canon catalog, any other suggestion?

[Nails]

>>I was briefly searching for off the shelve encoder discs, without luck so far.

Check out:

US Digital | Products » Encoders » Incremental » Rotary » Disks

Maximum of 1250 lines on a 1 inch disk, 2500 on a 2 inch disk.

[daedalus]

My suggestion is to avoid making encoders if at all possible, have done so before and its not worth the effort involved.

Avago technologies have two things that spring to mind. They do stick on reflective code strips, which could be wrapped around the perimeter of your disc, giving you 200+ line per inch encoding (albeit with a small discontinuity where the cut ends meet). Last time I built with this stuff the reader chips were 2-3$ and the strip was about 1$/inch. The read heads can be purchased from farnell/digi, but the strip required making a few enquires.

You can x4 decode quadrature chips, so 200 line per inch is actually 800 count per inch.

The cheaper/easier option would be avagos 16 bit magnetic encoders, which give you an absolute orientation reading with resolution of 0.08 degree (so no homing needed). These are like $5 a chip.

[stewi]

Very well. I'll look at all options. Since we also want to rotate the disc at the turning station, we may require a two stage drum. One, where the turning motor engages and one for the encoder strip.
A gap where the strip meets on the perimeter should not effect accuracy.
We will have 2 turning stations on the 4 station rotary head. One which rotates the component before the camera 90, 180, -90 and one fine rotation after the camera.
Multiplying the encoder signal may be only required for the turning station after the camera.

One of our main focus has turned recently into the X and y gantry. If we go the mainstream way. than a X and Y gantry would come in mind. Likely with one motor in X and dual in Y.
The placement head would require dual turning stations, one z-axis and a rotary axis revolving the head from one segment to the other with 90 degree increments.

With likely less mechanical effort we could use a delta robot, which does X, Y and Z out of 3 axis arms.

http://www.youtube.com/watch?v=FgnxBaFFDi8]Linear delta-robot test - YouTube

Not going the mainstream would also add more food for thoughts:
Z-axis travel is much higher than on a standard gantry machines
There could be a second level of non tape component feeding (tray bulk feeder)
We will open a new niche market of placement machines with added robotics functions and likely get more attention

We keep on posting here, but www.wikiplacer.com will come up soon!

[stewi]

The encoder discs Nails suggested are actually made out of Mylar not glass and pretty inexpensive as well. I'll put these in my drawing.

[stewi]

I got motors and encoder discs placed in space. Will need a smaller encoder for Z-axis as well and will keep on going on WikiPlacer

[Larken]

stewi, did you use to work at a pick and place company ? That machine looks familiar