Better Pick And Place Machine DIY

[Sionyn]

based on what looks like 3D printer

[Spikee]

It has potential , just needs some development.

[Thor-Arne]

Promising.

Looking forward to see some actual pick-and-place, especially the component alignment and the speed.

[mikeselectricstuff]

Mechanics and speed look better than most of the lame efforts seen so far, but consequently probably not cheap. Still no sign of feeders, without which it is of minmal usefulness.
There is way more to p&p than a good x/y/z mechanism. I suspec he'll either get bored and give up, or spend not much less than a used commercial machine  before it is any real use.

[george graves]

I suspec he'll either get bored and give up, or spend not much less than a used commercial machine  before it is any real use.

You might want to re-think that.    This project is from the same guys who built this(a wall sized voice controlled automated parts storage system):

or spend not much less than a used commercial machine

Used machine?  So like 10k?  15k?  I doubt it.  I think if you used a 3d printer to help make some of the feeders and the head, I think you could make a DIY PNP, with top and bottom vision, and a rotating head for under $2-3k.  I think the vision system and software is the real issue.

[mikeselectricstuff]

I suspec he'll either get bored and give up, or spend not much less than a used commercial machine  before it is any real use.

You might want to re-think that.    This project is from the same guys who built this(a wall sized voice controlled automated parts storage system):

Classic example of solving a non-existent problem with a technological sledgehammer, and not nearly as complex as making a useable P&P. It's just voice->database->XY location.

or spend not much less than a used commercial machine

Used machine?  So like 10k?  15k?  I doubt it.  I think if you used a 3d printer to help make some of the feeders and the head, I think you could make a DIY PNP, with top and bottom vision, and a rotating head for under $2-3k.  I think the vision system and software is the real issue.

Vision is easy (as vision problems go) - all you need to do is find the centre and rotation of something in a very controllable environment at a fixed focal distance with optimal lighting. For bonus points make it find parts in trays & feeders to reduce manual setup time. For super-bonus make it find loose chip parts and shake the tray til it finds a good one the right way uo.
 
Software is a major task - you only realise how major after you've used a real P&P and notice all the little details that need taking care of to make it useable.

Even a good commercial P&P takes significant time to set up data and load parts.  Any corner-cutting on a DIY machine that makes setup take longer eats further into the range of jobs for which a machine is useful.

But useable tape feeders are still the problem that someone needs to solve to make it worthwhile.
Any P&P that can't run a reel of parts without manual intervention is nothing more than a toy.

 Most would-be DIYers have never used a real P&P machine, and have not realised that a P&P that is, say 25% as good as the cheapest real one is a lot less than 25% as useful.

An analogy that comes to mind is OCR software - 99.999% accuracy is pretty useful, but as that accuracy reduces below, say, 99%   there comes a point where it is of just no use as it requires too much manual correction.   

I'm not saying it's not doable, just that it's an order of magnitude more work than it first appears.
To anyone considering it, I'd say before you do anything, go take a good look at any commercial pick/place in action, and look at the whole process from setup to finished boards. At least you will get an idea of the complexity of the task.

[george graves]

Classic example of solving a non-existent problem with a technological sledgehammer, and not nearly as complex as making a useable P&P. It's just voice->database->XY location.

I was just repling to your comment about him not finishing the project - I agree - overkill.

And...I hear where you are coming from - but I'm sure people were saying the same thing about a DIY 3d printer several years ago.

DIY feeder:



And for software - the openpnp project just about has all the complexities that you mention worked out.

[pickle9000]

I've never seen a feeder, is it normal for it to have it's own motor?

[8086]

I've never seen a feeder, is it normal for it to have it's own motor?

It  needs one in order to advance the tape. Otherwise you would be  able to place only one component. 

[pickle9000]

What I mean is is it normal for each feeder cartridge to have a motor, or would there be a single motor to power multiple cartridges. Meaning does a machine typically use more of a mechanical solution versus one that uses electronics and say steppers as in the vid above.

[poorchava]

Some modern feeders I've seen use two approaches: either a flat stepper motor (u know, the feeder for passives is like 10-15 mm wide) or some kind of miniature motor driving a screw gear which advances the tape. So yea, usually every cartridge has it's own motor (or motors)

[mikeselectricstuff]

And...I hear where you are coming from - but I'm sure people were saying the same thing about a DIY 3d printer several years ago.

The difference  is a slow, crappy 3D printer is still useful as it will do things  you can't  do by hand. The same does not apply to a P&P

DIY feeder:

That's not a viable feeder - just a stepper with a sprocket. Too wide to get a reasonable number, no cover stripping and unlikley to work well with both paper and plastic tapes.

I quite like the multi-lane approach to feeders - my Versatronics P&P uses 10-lane feeder banks, which have one stepper and 10 solenoids, and gives good density. Bit of a PITA to reload the middle reels though.

[george graves]

That's not a viable feeder - just a stepper with a sprocket. Too wide to get a reasonable number, no cover stripping and unlikley to work well with both paper and plastic tapes.

I believe that was just a proof-of-concept someone whipped up in an afternoon.  It's not like "this is the only feeder design possible."  Come one - let's be a bit more forward thinking, eh?   

I still don't understand why you think that something that does 90% of what a $30k machine does, for a 10% of the price isn't worthwhile venture.  Enlighten me - I don't say that flippantly - I'm interested.

And, I have to disagree -  I do think it(a DIY PNP) is analogous to the DIY 3d printers.  As a side note: Have you seen the quality of prints the current round of of repraps are putting out?  "Slow" and "Crappy" aren't words I'd use to describe the prints now-a-days.  At a recent 3d printer met up at a local hackerspace, I couldn't tell the reprap prints apart from the makerbot prints - no one could.  Not to mention a DIY approach has the *huge* advantage of you being able to fix the thing when it breaks.  To me that's a huge plus.

Anyways....point is, that none of the DIY PNP machines to date are useable - I'll give you that.  But, just like with 3d printing, when a community comes together, a lot can happen.  I looked at DIY 3d printers a few years ago - the output was laughable.  Now, I'm really, really impressed.

I guess I don't see any issues that can't be over come.  Feeder width might be an issue, same with software.  But it's all just a bunch of parts and some code.  Nothing magical/mythical with a PNP.  (Unless there is some baby dolphin tears used to lube the machine that I don't know about.) - And - it's even more true if a open source group got some momentum with it - just like 3d printing.

At a $30k entry price for a new PNP, that I can't work on, is close source, and I can't expand upon - yea.... I think 100 hours of my time would be a worth it to be able to populate 50-100 boards a day while I do other work.  I kinda value my time.  Just my 2 cents.

BTW - love your videos - keep up the good work.

[JuKu]

Mechanics and speed look better than most of the lame efforts seen so far, but consequently probably not cheap.

Speed is indeed good, but speed is easy. The mechanics of this machine cost $225.

Still no sign of feeders, without which it is of minmal usefulness.

Depends on the intended use. For any real production, I agree. But for prototypes and (very) small production, feeders are definitely not required. As I've written before, I am looking for this kind of machine, or just maybe, to build one. I've posted the picture of the board that got me convinced that I need a machine like this: https://www.eevblog.com/forum/general-chat/homemade-pick-and-place-awesomeness/msg158657/#msg158657. This board has 80 small bypass caps, the biggest nunber of single component. That is only 320mm of tape. The machine only needs longish slots to where pre-peeled tapes are placed. Those might be a problem too, but that is much easier to solve.

Setup time is not an issue for one-off boards either. I can see a process where machine prompts components ("Need 16 pieces of 470pF, 0805"), the operator takes those out from parts storage, puts the tape in a slot and clicks the button labeled "slot2". While the machine places those, the user puts back components from slot1 and gets ready for the next part. Spend some thought of the work procedures, and you could save that job. If you have more slots than types of components, operator is needed only when the slots run out. And in principle, the machine can be big. Round rails, aluminium profile or whatever is used to build the contraption are not that expensive. Add a vision assisted loose component pickup capability and we'll have an extremely usable machine. With an upward looking camera, even BGAs become usable for hobbyists.

With this kind of machine, I imagine that the board would have been done in less than an hour, "setup" included. Still not good for production but huge improvement in development. A prototype house typically takes several days and charges more than half of the machine cost each time.

There is way more to p&p than a good x/y/z mechanism. I suspec he'll either get bored and give up, or spend not much less than a used commercial machine  before it is any real use.

Right, the software is the crux of the matter. Still, that would be fun, and if done as a hobby project, spent time does not count. For any commercial firm I doubt the investment in time (= salaries) would pay off for own use. Still, I think there is a market for a sub $1000 pick and place kit. But there is no reason why the part costs would be more than a few hundred.

[JuKu]

And to add to the fixed slots scheme: Imagine a board with parallel slots (say, 1mm wide), 4mm apart (the SMD tapes come in 4mm increments). You have a pile of special aluminium profile pieces, and you press them to the board at required distances. You now have configurable part tape holder system.

Now we only need a company putting out about 2k, 1k for the nozzle and 1k for the first mile of the profile. I would buy several meters with the prototype p&p kit. 

[mikeselectricstuff]

Setup time is not an issue for one-off boards either. I can see a process where machine prompts components ("Need 16 pieces of 470pF, 0805"), the operator takes those out from parts storage, puts the tape in a slot and clicks the button labeled "slot2". While the machine places those, the user puts back components from slot1 and gets ready for the next part.

I think in practice it won't be that simple, e.g.peeling the right amount of cover, the uncertainty on qty doe to mis-picks and vision fails. Not saying it's not possible, just probably not quite that easy in practice

I really don't think that would be suffiecntly faster than hand-placement for a big enough range of jobs to be worth the effort. I have a P&P but still hand-assemble small quantities, and hand-place parts on the P&P joibs where there are only one or two parts of the type.

You need to step back and look at the overall picture. Requirements range from one simple board to zillions of complicated ones.
At one end, hand assembly is the quickest and easiest option, and at the other it is a no-brainer to go to an assembly house.

So the market for a cheap or DIY P&P is where the placement count is too high for hand assembly to be viable, but low enough ( or urgent enough) that outsourcing it doesn't make sense. Any compromise in spec/funcitonality narrows that market further.
If you take out the urgency/flexibility requirement, the upper end limit can be in the tens to hundreds depending on the attitude and workoad of available assembly houses. if you have space, you can pick up an old TP9 or similar for 5-10K, which will pay for itself pretty soon.

I'm sure there is some market for a simple/DIY majchine but IMO it is very small, and any compromise in functionality, feeders in particular, makes it smaller.

With an upward looking camera, even BGAs become usable for hobbyists.

BGA viability is more about quality of solderpaste and post-inspection than placement accuracy. The need for multilayer boards for many BGAs is at least as high a deterrent than any placement difficulties.

[amyk]

I think speed might not be too high on the list of things a DIY P&P for the hobbyist would need. Even if slower than manual placement, if the accuracy is fine then the advantage is that it becomes a "set-and-forget" type of thing to automate a menial task. While it takes its time I can concentrate on doing other work. As long as the amount of time it needs attention is less than it would take to place the parts manually. So concentrate on getting the feeders and other bits working well, speed can come later.

[mikeselectricstuff]

I think speed might not be too high on the list of things a DIY P&P for the hobbyist would need. Even if slower than manual placement, if the accuracy is fine then the advantage is that it becomes a "set-and-forget" type of thing to automate a menial task. While it takes its time I can concentrate on doing other work. As long as the amount of time it needs attention is less than it would take to place the parts manually. So concentrate on getting the feeders and other bits working well, speed can come later.

To some extent yes, but there is only so far speed can go down before it becomes useless. One fundamental limit is how long the paste stays tacky.
The other is just the effective use of money/space - how long does it take to pay off even a cheap machine that does only a couple of boards a day?
If it is slow it is even more important to be reliable and not need attention, which brings us back to needing reel feeders.

Even the apparently simple method of using fixed tape lengths has potential problems. One is you need to tell the machine where they are ( vision may be useable to help speed this up, but probably not with black parts in black tape), because different jobs are going to want a different balance of feeder vs. board areas on the bed
Plastic tape has its own issues - if you need to peel the cover off a length, sometimes small parts stick to the cover tape, and plastic tape is very springy, so once the cover is off, the slightest touch ( or vibration from the machine) can easily send all the parts flying.   

[george graves]

Hard to see how a hobbyist would have sufficient use to justify the development effort. Unless the mere learning experience and development was sufficient incentive. Perhaps even that IS the hobby.

The assumption here is that there are hobbyist that building one of something.  Compared to a company that builds 100,000 of something.  And there is no middle ground.  You either make 2-3 of something - or 2-3 hundred thousand.  That's not the case at all.  Is it?

You're right - for a hobbyist it would be an "academic exercise" - as they say.  But there is a large number of people that are in the gray area between hobbyist and a corporation.  It's a niche, granted.  But that's where open source seems to thrive.  I could site examples of that - but I'm guessing I wouldn't need to really - if you have been following the trends of open source hard/software over the years.

And for mike - here is yet another 3d printed feeder design.   

[mikeselectricstuff]

And for mike - here is yet another 3d printed feeder design.   

..well at least someone has realised that feeders are  necessary, which is a start.
I can't see this design working well with plastic tapes though.
I think the wheel would need to be metal (or at least use metal pins) to get a smooth and robust enough pin action, and I don't see the point of the sideways pushing - you just need a track with a stepped ledge to guide the tape.
And he's got it back to front on his model - the index is on the wrong side - corrected on the multi-bank design.
At least it's a start.

[JuKu]

Hard to see how a hobbyist would have sufficient use to justify the development effort.

At some point, machine vs. hand placement is not a question about ease of use or convenience, it is about can I do this at all. Cases:
- There is a finite time after printing the paste that it is still tacky enough. In my experience, that 450 component board was really pushing that limit. I don't want to do that again.
- The older we get, the smaller the components become, both in reality and in perception. I can do only a few 0603 in a row by hand, and I do need a microscope for that. A board with 30 or so of those, and I need a machine. I can't place any 0402s. Ok, a machine would need to be pretty good for those - but possible for a hobbyist machine, I would think.
- BGAs.

Unless the mere learning experience and development was sufficient incentive. Perhaps even that IS the hobby.

Bingo!

[lewis]

Mechanics and speed look better than most of the lame efforts seen so far, but consequently probably not cheap. Still no sign of feeders, without which it is of minmal usefulness.
There is way more to p&p than a good x/y/z mechanism. I suspec he'll either get bored and give up, or spend not much less than a used commercial machine  before it is any real use.

You miserable old cynic you! If people want to have a go at making a P+P machine themselves, let them! If they want to try and sell a low-end one to a few customers, let them! We should be enthusiastically encouraging these people. Britain (and Hewlett Packard) was built by eccentrics in garden sheds.

[mikeselectricstuff]

Mechanics and speed look better than most of the lame efforts seen so far, but consequently probably not cheap. Still no sign of feeders, without which it is of minmal usefulness.
There is way more to p&p than a good x/y/z mechanism. I suspec he'll either get bored and give up, or spend not much less than a used commercial machine  before it is any real use.

You miserable old cynic you!

Cynical, me..?

If people want to have a go at making a P+P machine themselves, let them!

Nobody is stopping them

If they want to try and sell a low-end one to a few customers, let them!
We should be enthusiastically encouraging these people.

Yes but encouraging them to think about how to make something that is actually useful and not just diving in with a half-arsed approach that will never be more than a toy. My fear is that some over-enthuisiastic DIYer will lauch a half-arsed attempt on kickstarter, get funded by fanboyz/girlz who don't understand the issues until it's too late. That then takes a way a lot of potential market for something that's been done properly.

Britain (and Hewlett Packard) was built by eccentrics in garden sheds.

Yes - totally agree IMO the UK designed Versatronics machine is probably the best effort anyone has ever made at making a small, cheap P&P - I think they just were a bit ahead of their time in the 1990s and ran out of cash. There are still plenty of these machines still going, and if you can find a good used one (typically £3-8K with feeders) it is definitely the smallest and probably the cheapest real P&P available.

Potential P&P builders could learn a lot by using one of these machines.
 
Bottom line is I'm just trying to share my experience and help people avoid wasting their time persuing dead ends.
I did look seriously at doing a DIY P&P a while ago based on a plotter before realising it was hopeless, and that, combined with experience of having a real one gives me insight into what is and is not important.
Seemingly small issues like parts bouncing out of plastic tapes are only apparent once you start using feeders, but could easily make some approaches to feeder design unviable.   There are many other small issues like this that you only see with experience.

All I'm doing is trying to help & guide people but sometimes the truth hurts!
Nothing has dissuaded me from the opinion that a P&P without reasonable speed and a viable design for proper feeders is a dead end.

I believe that there is, now more than ever,  a good opportunity to make a cheap (say $3-5K) very useable, and scaleble machine with economic tape feeders, but it will only be viable if it can perform well enough in terms of speed, setup time and component variety to cover a sufficiently wide range of jobs from 1-offs to small runs,to be genuinely useful to enough people. 
 
It will need a lot of work and some clever design to minimise costs. I've yet to see anything that comes close to it. I hope this will change but am not holding my breath.

If people want serious suggestions on approaches that I think are worth persuing - I'm more than happy to contribute constructive ideas and share experience for free whan I have a little more time.

[mikeselectricstuff]

..and if someone wants a suggestion on something to design to make a real contribution to low-end SMD, the market is crying out for a cheap (<£200)  stencil printer that will handle random sized unframed metal stencils, with proper tensioning.

Good quality paste printing is the number 1 factor in good reflow quality.

I'm convinced it could be mostly done using thick (3.2/4.8mm) PCB material as the main structural material.  Some thought needs to be put into alignment and adjustment, although I've found that the fixed pinning method used in Eurocircuits' rather expensive printer works surprisingly well without the need for adjustments.

[gregariz]

That's not a viable feeder - just a stepper with a sprocket. Too wide to get a reasonable number, no cover stripping and unlikley to work well with both paper and plastic tapes.

I quite like the multi-lane approach to feeders - my Versatronics P&P uses 10-lane feeder banks, which have one stepper and 10 solenoids, and gives good density. Bit of a PITA to reload the middle reels though.

The older & common Juki's are all pneumatically driven. There is simply a compressed air line going into a momentary switch which shoots air out to both the tape advance sprocket and the tape cover winder. The machine head comes across to pick from the tape and after it has done so it hits the momentary switch and the tape advances before it leaves. Very simple, no motors and relatively cheap feeders. Its really nice because when you set up the tapes by hand you can apply the compressed air to the feeder bank and hit the tape advance switch by hand to make sure your aligned and ready before you set the machine up for a job.

WRT to feeders and diy pp machines. I've thought about it too and personally I think the whole feeder caper is insurmountable for the DIY'ers. I think the best idea if someone is intent on doing a machine is simply to limit themselves to cut tapes and stick vib feeders. So build a really large xy table with a really fast screw and get a crap load of 8mm and 12mm cut tape linear guides machined out of aluminium and mount them down at one end of the table. Doing it that way they may get a couple hundred parts per strip of cut tape laying down on the table which may be good for a run of 100 or so boards before you need to replace the tapes. pulling back the tape covers will be up to the operator to do before running the job. ie no feeders, no fishing sinkers, limited runs of about 100 or so boards.