[RomanLini]

I think you could come close to placing those 0402 caps blind. There is plenty of room around them.

Assuming you can move the nozzle over the right place on the PCB, careful choice of the nozzle size (with a little vacuum leak) should cause some self-centering of the component.

And that leads to the thought that maybe you could have the nozzle shaped in some way that assists in the self centering. I have a nozzle that will be used just for those round glass MELF diodes, and I plan to grind it with a slightly curved end so it will fit the round body of the MELF and also allow better rotation, so the MELF rotation is guaranteed fixed to nozzle rotation.

I wonder if you machined a small rectangular "recess" in the end of a special nozzle the part would auto-locate into the recess when the vacuum acts on it?

[Eclipze]

Perhaps another option would be to have an alignment nozzle for positioning/rotation of the PCB. Reducing the error with placement and/or need for a fidiucial (for component sizes we are talking about). One that has a simple tapered point, like a plum bob, which can be dropped into opposing alignment holes in the PCB. Perhaps not a full conical section or point, but one that facilitates the positioning with minimal chance of sticking.

[stewi]

Regarding nozzles, there are many after market sources for nozzles, that is why OEM are so picky giving out any info.
Here is a US source of various nozzle styles SMT Pick and Place Nozzles and Consumables
and my buddies in Austria AdoptSMT Europe GmbH - Die Gebrauchtmaschinen-Boerse - Index

Yes, melf (mini and maxi) require a special nozzle, else components have too much vacuum leakage.

[stewi]

Nails, I was reading through the OpenCV manual and it appears that even with my limited C knowledge the software does everything needed.
Coming more from the practical end of vision systems, there are a few more things to be considered.
If you have done a lot of pictures with your SLR camera, you'll remember most of these issues.
If the light source is too direct, than light is reflected back to the camera, creating shiny areas.
PCB Prototyping, MITS Fiducial camera
Consider using a larger diameter LED ring, which shines in a flat angle onto the board object.
Also, the round fiducial mark is commonly pre-tinned creating a dome shape mirror. However, the surface may be shiny in some areas and dull in others. The camera image will rather display a potato shape than a circle.
If you have three fiducials on your board, they all may appear completely different in the camera than to our eyes.

A wide angle lens distorts the image. It can be software corrected, but usually only if the distortion is linear and not pillow shape.
If you are using the camera for a component camera in 2-3 inch distance than a lot of reflected light rays will get into the camera. Some of the BGA have a gold pad, which will create a large shiny spot if the light source is too direct.

[stewi]

Nozzle sizes for Siplace revolver head. Let me know, if you need translations.

[stewi]

Perhaps another option would be to have an alignment nozzle for positioning/rotation of the PCB. Reducing the error with placement and/or need for a fidiucial (for component sizes we are talking about). One that has a simple tapered point, like a plum bob, which can be dropped into opposing alignment holes in the PCB. Perhaps not a full conical section or point, but one that facilitates the positioning with minimal chance of sticking.

If the nozzle pushes the board in X and Y from the first alignment hole, than it would require, that the board can rotate around the first alignment hole, when dipping into the next hole.
However, a tapered pin could be used to leave a pin mark in the fiducial pad at the programmed x and y position. Then you move the board to the center of the pin mark, fiducial pad and then you start your placement program.

[stewi]

I tend to agree with the majority, that vision system may be an overkill.
Most of the commercial machines use board conveyors, which either require mechanical board alignment (troublesome) or optical (fiducial) board recognition.
If we insert the board manually in the machine, we’ll likely use pins in the board mounting holes and get the boards pretty well pre-aligned.
Larger components, except fine pitch, can be aligned with mechanical jaws at the head or in an external alignment station.
Smaller than nozzle components could use a special shaped nozzle, assuming that we get component pick up fairly accurate.
What I would recommend:
Mount a USB microscope to the placement head.
Determine camera to nozzle offset by placing a small blob of silly putty in the placement area.
Lower z axis with nozzle into the silly putty to leave a print.
Zero out X and Y.
Jog the camera over to the nozzle print in the silly putty.
Write down X and Y offset.
If your entire placement program can run with an offset, than you can watch on the monitor pick up and placement. Else, you can modify your program in Excel, adding or subtracting the offset value in pick up and placement.
Make corrections, as required for pick and placements and run your batch of boards.
For inspection after placement, you can run your board program with the camera offset again.

[RomanLini]

Perhaps another option would be to have an alignment nozzle for positioning/rotation of the PCB. Reducing the error with placement and/or need for a fidiucial (for component sizes we are talking about). One that has a simple tapered point, like a plum bob, which can be dropped into opposing alignment holes in the PCB. Perhaps not a full conical section or point, but one that facilitates the positioning with minimal chance of sticking.

That's a clever idea! I had originally envisioned using some pins probably under the PCB to hold it in aligment.

Your idea allows the machine to align by itself if it could be made to work. Since both our builds have a sturdy enough 1.5mm feed pin, why not use the feed pin in any hole on the PCB that is > 1.5mm?

You could use a "back-forth" procedure like people use in their component alignment pits, so the 1.5mm pin in say a 3mm hole would not be a problem once the programming is setup.

It would need some sprung clamps or similar to allow the PCB to be moved with lowish force, but still hold it exactly in place after that...

[Eclipze]

I could certainly see the board rotating around as stewi mentioned. Aligning one hole, only to move the position of the previous in the process. Can't think of an alternative.

The method that's been of interest is to use a peg board. So I can load some pegs and push the PCB into a corner of pegs. Most boards I've had have been within 0.2mm on the routing. But I'd rather have a more accurate method for board positioning given a blind placement strategy.

[stewi]

Assuming, you can correct placement data upon a known offset in X,Y and theta, than you could use the vacuum sensor and the nozzle to detect hole's positions. You would have to come from the left and right, top and bottom to find each hole center. The vacuum sensor should detect if you hit the board or the hole.

[Eclipze]

I started thinking about all kinds of options, like placing magnets as posts for board alignment. Or small suction cups.

[stewi]

You could make yourself a vacuum box, if you have enough clearance in Z. The top plate has several small holes, which are covered in the area where you don't have a PCB. The box contains a larger hole for your vacuum cleaner (if you don't mind the noise).
You could wiggle your board straight, then turn on the vacuum.

I hate magnets anywhere close to electronics.

Yeah, my 100's post!

[Nails]

I got my nozzle today. And it is tiny. But so far, I like what I see. Very precisely made. Here's some pics. I also have a CAD pic of what I think the the pickup head should look like. It's pretty simple.

You can see on the top of it, it has prongs with a circular portion at the top. They appear to be 1/16" diameter. The prongs have some give, so they spread apart. I suspect that is how it is locked onto the mating head. Since the OD is .379", I also suspect there is a 3/8" o-ring that would seal it at the bottom. You can see it has a nice chamfer at the top which would help to center itself as the pickup head's o-ring makes contact with it.

Also note the two half-circles at the bottom near the nozzle, which would allow it to keep it's orientation in the tool rack.

I'm going to try to run and design with this, as they are inexpensive.

[Eclipze]

Hey Nails... those look really good! Those the HSC Lighting tips.

Which nozzle size do you have pictured?

[Nails]

Yes. I got them here:

Universal (UIC) Lighting Nozzles on Count On Tools

If you notice, on the page, the ID specs don't make any sense.

The nozzle pictured is for 0402.

Yeah, I really like it and am convinced that this is the way to go. It doesn't make sense to me to try to make a nozzle when they are so cheap. This one was $32. 0603 nozzle is $28

I'm in the process right now of trying to work out the design of the mechanism with rotation, approximately 3/4" downstroke, with vacuum. And I'm trying to get the spacing at approximately 0.700 apart so I can have two of them on my machine.

The trick for virtually frictionless rotary motion, frictionless up and down motion, and have a vacuum seal is to use bronze sleeve bearings at the ends of a block with a vacuum port in the middle. A rod has to be drilled out and have a small portion on the side drilled or grinded out for the vacuum to reach the tip. I happened to stumble across a rotary union with vacuum ports in my basement that I didn't even know I had. I took it apart.. The rotation and up and down motion was really smooth, and there seemed to be no air leakage, nor any side slop at all. I was pretty surprised when I took it apart that it is this simple.

[stewi]

Nails, you may ask these people from COT how the nozzle is seated, but I think Universal does not use an O-ring.
Also, I do prefer rectangular nozzles, since the components are rectangular.
I don't want to advertise Siplace too much else you may think, I'm still working with them, but even the small 0402 nozzle has more than one rectangular hole, acting like a screen and prevent, that the small components are swallowed.
Again, ask the people from COT. They don't sell machines, they sell precision made (Swiss tool machine) components.

[Nails]

I'm certain there must be a seal. However, I don't believe it is a round profile o-ring that I show in my picture. That wouldn't work.

A square profile one will, such as found on this page.

McMaster-Carr

Regarding the nozzle shape not having a rectangular hole; I'm not worried about that. These obviously have been proven to work well with a simple circular hole.

I've been radically changing my design for the head since getting this nozzle. I can't believe how beneficial having this part in hand has been. I'll post a pic when I am happy with it.

[Eclipze]

I like the UIC Lighting nozzles. Struggled to find info on the Siplace nozzles, although having a non-round pickup adds a requirement for nozzle alignment (which I don't have). The frustration is finding a supplier that ships outside the US :-(

[phomann]

I'm certain there must be a seal. However, I don't believe it is a round profile o-ring that I show in my picture. That wouldn't work.

A square profile one will, such as found on this page.

McMaster-Carr

Regarding the nozzle shape not having a rectangular hole; I'm not worried about that. These obviously have been proven to work well with a simple circular hole.

I've been radically changing my design for the head since getting this nozzle. I can't believe how beneficial having this part in hand has been. I'll post a pic when I am happy with it.

I don't believe there is a need for a seal. Remember that you have a continuous vacuum supply, so leakage is not an issue.

My Yamaha YM series uses YMH nozzles. The nozzle is spring loaded and slides inside the Z axis tube on the head. (The spring loading accommodates different component heights.) There is no seal apart from the sliding fit.

Also each head measures the vacuum and provides a lo, med, hi discrete output. When you set up a feeder you indicate the vacuum level check needed for a successful pickup, none, lo, med or hi.

Cheers,

Peter.

[scsi]

Ditto on vacuum sealing. When vacuum supply is constant and strong enough, like in case if you use a proper vacuum ejector, there is absolutely no need to seal anything. On my machine I'm using parts from Madell and there are at least three places where the vacuum can "escape". Still, even relatively small nozzle with 1.5mm ID can hold a large QFP's with no issues.

Speaking of Madell, it's very hard to beat their prices for nozzles. They do charge an arm and a leg for their hollow shaft stepper, but the rest of the nozzle components are very affordable. I posted this video already, buy given the size of the thread, here we go again
[ame=http://www.youtube.com/watch?v=bYfD0UUKyek]Homemade SMT pick and place machine #3 Nozzle Changer and plastic chips - YouTube[/ame]

I think, if you are looking to build just one machine for yourself, the $300 for a motor can be justified. For mass production, there is always a way to order motors straight from China with the shaft machined to your spec. When I priced custom steppers for my other project that required special shaft machining, the price I was given was $8.50 e.a. after a one-time $1,000 setup/tooling fee.

Now, about the board alignment via holes without fiducials. While it may work reasonably well for small components, I doubt it will be sufficient for fine pitch packages where 0.1mm alignment is critical. We need alignment for copper layer, not the holes, becasue they are typically not aligned well when the board is made at the factory.