Pick & Place Adapter for 3D Printers

[RoGeorge]

Where would it be a good place to ask about building/upgrading 3D printers? Is there any forum similar with EEvblog, but for 3D printers instead of electronics and test equipment?

I am trying to design and build a custom delta printer. The goal would be to use it not only as a 3D printer, but as a pick and place machine, too. Since I have no previous experience with 3D printing nor pick and place, before starting the build I would like to ask about some of the design choices I made.

Which forum would you recommend for that, please?


Later edit, goal changes: An universal SMD feeder + vacuum head to transform existing 3D printers (FDM) into automated Pick and Place machines able to assemble a full PCB or a whole panel without manual intervention.

The adapter consists of an unconventional SMD feeder and a small vacuum head. The same feeder can handle all types of SMD parts, and it is placed near the existing printer. The vacuum head is attached to the existing extruder of the 3D printer, by simply using a clamp.

Pros
- The same feeder will handle all the required SMDs, regardless of their type
- Continuous feeding for identical or panelized boards, no feeder setup required in between
- Easy to attach/detach the whole adapter (in a minute or so).
- No modding or dismantle required for the existing 3D printer.
- Fully automated PCB assembly
- Should work with almost any 3D printer
- Much smaller footprint than an industrial machine
- Cheap
- Free and open source

Cons
- Much slower than industrial grade Pick and Place machines

[donotdespisethesnake]

Since I have no previous experience with 3D printing nor pick and place, before starting the build I would like to ask about some of the design choices I made.

Sounds like you should start simple and work up, unless you have previous experience building CNC machines. Multifunction devices are notoriously hard to implement well, so I would start with building a 3DP, and a dedicated PnP, then when you understand the issues of each you can think about how to integrate the two functions.

There are several forums for 3DP, DIY PnP not so much. You are rather on your own there.

[RoGeorge]

Indeed, it will be just a 3D printer at the beginning, and see how it goes. If all OK print some feeders, then switch the printing head with a vacuum needle.

For the 3D printing forum, http://forums.reprap.org/ seems a pretty active one, I'll start from there.

What would be the desired resolution of angles and position for an average P&P machine?

[hagster]

As Rogeorge says there is the reprap forum. There is also a couple of google groups.
Delta bots
https://groups.google.com/forum/m/#!forum/deltabot
Hbot/coreXY
https://groups.google.com/forum/m/#!forum/h-bot-and-corexy-3d-printers

[Rerouter]

I can give you some pointers, Deltoid printers benefit when the moving head mass is as small as possible, at least if you want the motion speed high.

For vacuum pickup heads, and rotation I honestly would not know where to begin, likely you would have a vaccuum line run down the bowden, and a micro solenoid right next to the pickup head, and possibly an LED COB and a linear CCD to play the smallest shadow game for rotation,

[donotdespisethesnake]

Indeed, it will be just a 3D printer at the beginning, and see how it goes. If all OK print some feeders, then switch the printing head with a vacuum needle.

For the 3D printing forum, http://forums.reprap.org/ seems a pretty active one, I'll start from there.

What would be the desired resolution of angles and position for an average P&P machine?

I'm not sure that a delta is a good idea for PnP. Have a close look at FirePick Delta. After promising a lot, they failed to deliver, to be fair that is partly because they tried to develop too many new features from scratch. OTOH hand successful PnP machines have all been cartesian.

A big part will be the host software, for 3DP there is a fairly well established set of mature software to choose. With PnP there is nothing similar except perhaps OpenPnP which is in development.

What is an "average" P&P machine? Since there is only one DIY machine I am aware of, there is not a large sample set. Commercial machines tend to have capabilities way beyond what you could achieve with a DIY machine. There are cheap Chinese machines at the low end which would be a more realistic target. However, if you aim for 0603 and 0.5mm pitch LQFP that would meet most hobby needs. Further improvements can be made by iteration.

Creating a cheap, reliable feeder system is the Holy Grail of low cost PnP. I saw a recent design using the PCB itself as the body, although personally I think creating discrete feeders is the wrong approach.

Your approach will depend on your experience in hardware design, electronics and software. If you are skilled in all 3 areas, then it doesn't matter too much, but most people tend to be skilled in 1 or 2, so it would be best to try to re-use existing stuff, at least initially.

[RoGeorge]

Most of the parts for a delta printer were already ordered before asking here, so the parts are coming now. It will be a custom design based on 70 cm long linear rails (already have the 3 rails), which will impose a total height of about 1 meter (3 feet).

The first question is about the belt, please.
It has 2 mm teeth (GT2 like, no-name from AliExpress, 10 meters of belt for $8.25 and free shipping) with metal wires reinforcement. My concern is that the reinforcing wires will come in touch with the Aluminum pulleys and damage their teeth.







Is it normal for the reinforcing wires to be visible between the belt's teeth?
Wouldn't those wires damage the pulleys?

[RoGeorge]

The ball-joints are Traxxas-like, cheap and no-name. They have some backlash. I don't know if that amount of backlash is normal, or if it is because of the low quality of the ball-joints. The backlash is very big, and clearly visible on the direction perpendicular to the rod. On the longitudinal direction (relative to the rod), the backlash can only be felt at hand, but it is almost invisible to the eye. There will be two springs for each pair of rods, one spring near each rods' ends, so that the perpendicular backlash will not appear during printing.

Anyway, the backlash seems pretty big to me. In the hope that any plastic deformation happened during the ball assembly will revert to the original dimensions, I already boiled the joints, and applied thermal shocks by repeatedly diving the ball-joints into hot and cold water, but with no visible reduction of the backlash. Please see the end result:



Are these ball-joints usable, or the backlash is too big?

[ar__systems]

Delta PnP is a bad idea. Delta CNC is just crazy....

[SeanB]

Would not use that belt with steel reinforcement, the best is a polyamide reinforcement, which does flex easily and does not fatigue really fast. That belting is probably barely good enough to use as a belt where you rarely flex it, for something that has to move many millions of cycles it is in no way suitable, you will find breaks in the belt after only a few dozen cycles, plus the metal will get a permanent set making it very non repeatable. Best to bin it and buy new of the right pitch and with a fibre reinforcement.

About the best use for that belting is as plant ties.

[RUWO]

hello! first of all, I would recommend joining the 3D printing discord, it is really active and diverse: https://discord.gg/b9B8Nj
second of all, I would really like to work with you on the project! I have a lot of experience with 3D printers and own 1 atm, and I'm building A second, completely custom one. I'm looking into making the second one into a PnP, so we are in the same boat. if you have discord, would you mind PM-ing me your name and discriminator? I'm working on setting up a discord server for my project. let me know if you are interested. I have a bunch of ideas to share and discuss for the machine.

[RoGeorge]

This project is more of a loose end exploration for me, working on it only here and there, so I can not commit to any kind of team work right now, but thank you for your collaboration offer, and for the Discord invitation, too. I'll keep that in mind. 

[RUWO]

I have full-time school and a lot of other projects running as well, so I can't really commit a lot of time to it either. anyway, for the PnPpart, I was planning on using a hollow-shaft nema8 stepper for component rotation, with a vacuum hose attached to one side and a ring magnet for holding nozzles at the other side. the hose would then run to a box with a y-splitter, with on one side a solenoid valve, leading to a small (think half a liter) vacuum tank, and a small vacuum pump behind it. the other side would also have a solenoid valve, but this time with something like a fish-tank pump to compensate for the vacuum and blow the part onto the PCB with a bit of force. as for the actual hard part, the feeders, I was considering using a servo/stepper to roll of the plastic protection film on top, pulling in more of the tape as it does.

[RoGeorge]

Sounds like you already have a good plan, go for it! 

I am still waiting for some parts for the 3D printer phase. After the delta printer will be ready, will see about the P&P part. So far I have only fuzzy plans about the P&P, without any details.

My plan is for going open source in both HW and SW, and to aim for a hobby maker's tool, not an industrial P&P machine, not planning to making a commercial product here.

As many people pointed out before, the feeders seems to be the difficult part. I don't understand why the feeders are such a difficult part, but I will take their word for granted. The best way to deal with a difficulty is to avoid it, in the first place, if possible. So, no feeders.

I suspect the necessity of feeders came from the era when machine vision was very slow and expensive, and the parts were placed on reels to be easily identified and picked by an almost blind machine. Now, this is not needed any more. There are plenty of cheap camera, and computing power is ridiculously cheap.

Instead of feeding the machine with reels, I plan to just peel all the parts from the reel into a box, then use vision to pick one part only from that box, then rotate and flip the part as necessary before placing it on the PCB.

[RUWO]

the advantage of reels is that you can have hundreds of parts without taking up a lot of build volume. also, it takes a lot of labor to put them all in a box, defeating the point of the PnP as you could have just placed the components on the PCB. you might want to look into cut tape if you want to avoid feeders

[RoGeorge]

- I assume making a reliable universal (and automated) reel peeler will be much easier than making a reliable universal feeder, but of course, I might be wrong.
- The idea is to put the parts in very small boxes (way smaller than a reel) by using some sort of funnel.
- Many small boxes with parts can be placed around the 3D printing surface.
- Eventually, many small boxes with various parts can be placed on a rotating platform (outside the printing area, like a carousel, similar with a CD changer mechanism) and present to the P&P head only the next required box/part.

I've seen the cut tape method before, but it seems to me a much more laborious way than a feeder setup. Another disadvantage for the cut tape will be the limited number of cut tape locations available around the printing (P&P) area.

With boxes, it's easy to present the right box near the printing are, then let the P&P head and the vision system (thinking OpenCV here) to do the rest.

Again, easier to say than to implement all these, but that's where all the fun is. The speed will not be important for non industrial use, so if that idea will manage to _reliably_ P&P at least one part/minute, I'll be as happy as a pig in mud. 

[RoGeorge]

Scribbled drafts of the above message.

On the left half, the automated peeler for SMD parts. On the right, a carousel with boxes full of peeled SMD parts.





Universal feeder - it should select, flip and angle the next SMD to be placed (based on OpenCV vision) while the current SMD is being placed on the PCB (the PCB not shown in this drawing). Detail for the path of the next SMD to be placed: from the peeled SMDs box -> SMD's flipper -> SMD's angle adjusting platform -> PCB

[donotdespisethesnake]

Your belt is faulty, send it back. If you are going to try doing new things with no experience, at least buy decent components.

Creating cheap, reliable feeders is difficult. Commercial feeders are expensive because they are very reliable. If you trade some reliability for cost, you can make them cheaper.

Your bulk tray feeder is the sort of thing seen in very high end machines. It's easy to say cameras are fast and cheap, but actually getting fast and reliable performance is not easy. Again, you might get cheaper with less performance. I would say your proposal will actually be much harder to implement than a reel feeder.

Assuming this is a research project, and you are not really expecting to ever create a working system, then just the bulk feeder would be an interesting and challenging project. If you want to create a working machine, start with an existing design and build it. Then when you have experience under your belt, you can start on modifications or creating a new design.

[RoGeorge]

Thank you for the advice. The white belt was from AliExpress, $10 total for 10 meters (30 feet) and free shipping, it doesn't worth the effort to return it. I have another one, much flexible, made from a black rubber-like material, with fiberglass insertion instead of steel wires, except the black one is GT2.5 instead of GT2.

I hope the rest of the parts won't be so bad.

Casted Al corners and linear rail




Carbon fiber rods 30cm/6mm/4mm and Traxxas-like ball-joints




2x2cm/1.6mm square hollow bar - most probably will be replaced by 20x20 Al profile




About the universal SMD feeder, the trade I was planning for is to loose in speed, in order to keep the price low. So, low price slow, but still very reliable. Since a 3D printer user usually doesn't mind to wait many hours (if not days) for a 3D print to finish, I bet the same user won't mind if the feeder is slow. TBH, I wouldn't mind if it would be even slower than a manual pick and place, as long as the whole process is automated. The only time limitation will be in the range of a couple of hours per board, so the soldering paste doesn't dry. Usually, the hobby/maker PCBs are small enough, so I hope the machine will be able to populate a full PCB or a small panel before the soldering paste starts to dry.

About the vision system, the plan is to use $5 cameras in high speed mode (they can do up to 900 FPS  ), and use a $10 RPi zero with OpenCV per each camera for image processing.


In the meantime, I decided to change the plans a little: no more dedicated P&P machine, better aim for an adapter that can be attached to an existed 3D printer. This is the new plan:

Pick & Place Adapter for 3D Printers

An adapter to transform existing 3D printers (FDM) into automated Pick and Place machines able to assemble a full PCB or a whole panel without manual intervention.

The adapter consists of an unconventional SMD feeder and a small vacuum head. The same feeder can handle all types of SMD parts, and it is placed near the existing printer. The vacuum head is attached to the existing extruder of the 3D printer, by simply using a clamp.

Pros
- The same feeder will handle all the required SMDs, regardless of their type
- Continuous feeding for identical or panelized boards, no feeder setup required in between
- Easy to attach/detach the whole adapter (in a minute or so).
- No modding or dismantle required for the existing 3D printer.
- Fully automated PCB assembly
- Should work with almost any 3D printer
- Much smaller footprint than an industrial machine
- Cheap
- Free and open source

Cons
- Much slower than industrial grade Pick and Place machines