Did someone use CNC as "pick and place" robot?

[Golem]

Hi!!.
This is my first post so hello everyone!!.
Well I'll try to explain what my problem is, because I'm form Spain and sorry for my english!.
OK, I have to glue a little pieces of steel in a wood sheet. Have to glue tons of this steel pieces so I think that if I could use the CNC to put the steel in place it will be great.
Did someone made someting like that?, It's possible to do it without change the CNC?, Maybe with new software?. Thank you!!

[tobyaxis]

What your looking for is a "Pick and Place Machine". Siemens and Fuji make these types of machines for printed circuit boards.

First solder is screened on the board, in your case glue.

Second a Small Vacuum Nozzle picks up electronic components and places them precisely on the board.

Then off to the oven where the solder is heated to complete the task.

These are expensive machines that are in fact CNC's of a different kind.

[Golem]

Hi everyone!
Yeah, I need one of those, but is very expensive to buy one, as you said. Because of that I want to use a DIY CNC to do that. I'm sure that all the structure of a CNC is valid, (X,Y and Z axis), but the software is not appropiate for that. This true that I could programme the points to move to in G-code, but it will be a bit hard. Any ideas?.

[tobyaxis]

It is the same concept of a VMC but your picking up and placing components instead of cutting material.

[daedalus]

Are you just trying to do an electronics pick and place on the cheap, or is your application different from that?

If you are just doing pcbs, this has been discussed on the forum many times in the past, and the consenus was that anything built out of a CNC will be far inferior to a 10 year old pick and place from ebay.

There is a massive amount of engineering in them, which you would have to replicate, involving fields as diverse as computer vision, pneumatics, control engineering, etc. The structure of a mill is all wrong for this, as you want to place lots of parts fast in order to be competitive with other shops. If you cant afford a proper pick and place, why not just outsource the lot to a production boardmaker and save yourself the grief.

If your not doing pcbs let us know, and post some info about the parts you are gluing, and the amount of time you are saving by automation, and maybe we can give you some more useful advice.

[rowbare]

You should be able to do what you describe but it will take some work to get it setup. Depending on the size of the parts you could use a vacuum or electromagnetic pickup (though if the parts are stacked, the vacuum is the way to go).

At the very least you would have to build some kind of pickup device. This can simply be a part that you attach to the Z axis when needed or you can have in move into place under machine control.

Then you need some means of ensuring that the piece is properly oriented, see the linked video for an example of this.

Then of course you need to write a bit of clever g-code to tie it all together.

The video in the link shows a pick and place machine built from a taig cnc mill. This is a small machine but the video shows all of the concepts.

YouTube- Homebrew Surface Mount Pick and Place Taig Mill Conversion

YouTube- Homebrew Surface Mount Pick and Place Taig Mill Conversion

good luck
bob

[Eclipze]

I'm building my own pick and place for electronics. It would however be far more useful if I had a proper pick and place electronics package to drive it. Can anyone suggest a suitable software package?

[daedalus]

Eclipze, check this link out, its about the most DIY machine on sale at the moment, might be worth trying to get its manual and lift ideas from:
About the cheapest machine out there

I cant really recommend software, as usually its custom written for the tool in question. You are probably going to need to write your own anyway, as you need to control the feeders / vision system.

BTW, there is a really nice $40k motorola pick and place on ebay US at the moment for like $4k B.I.N., some days i really hate being on this side of the pond.

[Eclipze]

I had a look at the SX1010 and was considering it until I saw comments about it here...
http://www.electronicskb.com/Uwe/For...-Place-Machine
<copied below>

"I got my SX1010 exactly 1 week ago, a total disapiontment, not even in
my bad dreams I could imagine this is what I will spend $13500. We
needed the machine for prototypes and small batches of very simple
parts, for this reason we also ordered the dispenser option and nozzle
changer.

The thing is, you can call and email these guys 100 times, if you are
familiar with SMT machines and you speak about a certain option, you
are sure they mean the same but in fact you are getting something
completely different. They sent me an oudated pdf file of the 3700
series, claiming it is the same software. I was supposed to get an
idea about the SX1010 from it.
So in this file you see mechanical parts feeders and full Z control,
just 2 examples. What I got is an air actuator Z spring pressure
controlled which almost for sure will damage large resistors for
example.
The feeders.... well, this is the best sad joke i ever seen. I would
laugh if it didn't costed $13.5K+shipping, now I cry.
It is a Madell's invention, a kind of slot where you place the tape,
then pull back the sticker and you attach to it a fishing weight with
a probe crocodile. You have to raise the weight every 200-300 parts
before it touches the floor. The advance mechanism is another air
actuator which presses a headless bolt on the tape and pulls it to
expose a new part.
What else, before touching the operational facts?
The machine looks like the aluminum parts are cut by hand saw, no
finish not to mention anodizing.
The dispenser control is the same stand alone box sold everywhere on
ebay for $100-130 and by Madell at over $340.
The air settings were lableled in psi, while the gauges are in MPa.
Try set a 5psi on a 0-1 MPa, 1" diameter gauge.
Then, when the Z was not moving, they said the 0-5 psi setting is a
mistake, it should be 30-50. It took us 3 hours to get this answer,
while thinking all the time we do something wrong.
The machine vs. computer dispay us up side down, if you want to move X
or Y you better sit behind the machine or if you are able to think
everything miirorwise then you are ok.
The dispalyed units are motor steps, not real dimensions. You have to
calculate everything using 1step=1/2000 inch.
In their web site and ebay store they claim the SX1010 is delivered
with 20 feeders, in fact they offer only 15.
I paid ( I don't believe I've done this) extra for another 5 feeders
and also ordered 5 "manual feeders".
As I described above, there is no difference between the manual and
the other (which ar not automatic of course).
An important thing with the feeders, if you plan to use anything more
than standard passives, you will have a problem with the height, as
the slots milled in a piece of aluminum are set at a thickness which
is for a minimal tape height. So tantalum caps, leds, and other taller
than a 0805 resistor are not feedable, you must use the manual feeder
where you have to stand like an idiot and push the parts one by one.
Right, they say there is another option, a loose parts tray.
They are not exactly loose, the tray should have specific places where
the parts are sitting, arranged as a matrix.

They claim the machine is delivered pnp, just add an air compressor
and computer and you are ready to place parts.
In fact they insisted to supply also a computer without monitor for
another $600.
I won't spend words on the computer's elemnts and quality, but the
hell, why it was supplied with 1/2 ton of viruses in it?
I plugged an USB stick in order to copy the user's manual from the
machine's computer and when I took it to my desk computer the anti-
virus went crazy, their computer infected my USB stick with a
dangerous virus.
It took several scans to clean all the junk from the machine's disk.
About the software, we had to look in all folders to find it and then
it was in 2 folders. From one location it worked and passed the homing
initialization, from the right folder (according to their instructions
it didn't). At this point we also received by email an updated manual,
once again for the 3700. Here we discovered for the first time about
the feeders, as the other manual had no mention of them.
We started to work as described, after 3 days looking to solve the air
issues, the right most feeder being unreachable by the pick up head as
it was installed by them in a wrong place and other 20 problems with
numerous emails back and forth.
Once again, we agreed to buy their computer as they claimed this is
the only method to supply the machine calibrated.
I don't know which calibration it was, but nothing was in place.
We are indeed stupid, we try to set everything up according to the
manual and nothing works. Why? because most settings in the book are
for the 3700. Usually every engineer according to a smart sentence
"opens the book only when nothing works" but here we worked according
to the book, where on every page you have to set something with the Z
height, while the machine has no Z control, simply one tiny subject
they forget to mention anywhere.
By the way, speaking about documentation, you need a lot of guessing
in installing the machine and attaching all connectors and hoses as it
is supplied without even a piece of toilet paper showing where
everything goes. We spent many hours figuring why the look-up camera
isn't working until we found a connector that plugs on it's bottom and
you have to disassemble the machine from it's table once again to have
it plugged and then it is squeezed between th camera and the table.

About operation, sophisticated windows based software and all the
wonderful stories, facto if you have a few pieces to assemble do it
with your soldering iron as you do it now, it will be much faster.
You have to become a photographer in order to use this machine and set
it up, everything goes on snap photos, absolutely everything, home
pointer, each feeder with the right part in it, the part above the
look-up camera, in every angle, fiducials, each part after assembly if
you want the inspection made, you mother in law's photo and etc.
Assuming somebody needs it indeed for prototypes, the setup of a 100K
parts/hour machine with 300 feeders takes much less.
So you might find yourself spending a full day on setup, and god saves
you if you need to change a reel with not exactly the same production
lot as the first one, you have to take all photo again.

I thought it is the best idea to buy a new machine, the price is
tempting when you compare it to other desktop types, now I deeply
regret this decision, it was much better to buy an used machine for
the same price, where the head knows where each part sits, you can
load a pick and place standard file, set the offset and you are ready
to go. I did this google search about Madell as well, without any
results before I ordered it, I thought that if nobody speaks about it
it is also a good thing, as one of the other options, Manncorp is
severely criticized everywhere.

I could add $1000 and buy a ECM97 professional used machine or even a
Juki in good condition, but somtimes stupidity strikes you. I am
considering returning this machine, as I am not in the US the shipping
will cost another $1000, they will not refund all the money, the
return is at least 15% although I have enough reasons to receive full
refund.
So it costed me 13,500 + 1000 shipping, if I spend another $1000 to
return it and $2025 for cancellation I think we might keep it here and
make it a vey high cost dot glue dispenser for securing parts, I see
no other use for it.

I share my experience with you, I think such poor product is a shame
to manufacture, I saw Chinese machines looking and working 1000 times
better."


I only need to place 0805, SOT323 and 0.5mm pitch TSSOPs. Larger production work I get done externally, but I always prefer to build the first 100 units myself and have some niche market products that I couldn't afford to get loaded due to the low volume. Given the machines are +$10k plus, and I'd probably spend more getting them up to speed, I've been designing my own for some time. Only over these holidays have I almost finished the CAD models... just working on the wiring paths now. You are probably right about the software needing to be custom.... there will be some specifics regarding tape feeding, trays and paste dispension I'd need to address. I'm hoping Mach3 has the flexibility to create a useful driver, and then have translation software for the pick and place data to required gcode segments.

Although I'm not spending much to build it... cost aside, it's actually more about the fun of designing and building it. Some people build ships in a bottle, others find endless amusement designing CNC's

[daedalus]

I did consider building one myself a while back, but gave up as i didnt have the time to dedicate to it. You really do need to do your own software, as mach cant do closed loop control, and without visual servoing this will be a massive challenge to achieve.

Visual servoing lets you avoid calibration if your clever about it (like putting optical tags on your feeders), and saves a lot of time mechanically alligning parts, but obviously theres a fair bit of dev work involved. One of my areas is computer vision, so i might be able to steer you towards some open source libraries if your interested.

Do you want to commercialise the design? If not it might be worth doing some kind of open source design model, get some other people involved, as you really need software devs, a control guy, a vision guy, and mech eng's to do a project like that well.

[Eclipze]

I've kept the design to have both flexibility for modification as well as choosing components that aren't difficult to find - nor too expensive. So it's something that could lead to commercialisation, however this path isn't really what I've been aiming for. Too much of a requirement for support hahaa.... It's ultimately a DIY model, which is being built for my own commercial use. But I've got more interest in making it an open source platform. I would be really good to get design input/expansion from others with different skill sets, particularly for the software side. But I will consider manufacturing kits or partial kits for those interested in purchasing the mechanical component/assemblies etc...

It has two heads, which the second head for my machine will be paste dispension. However any significant prototype runs I machine stencils. It has a web camera for one of the heads. Mechanical alignment is done at the point of pickup, for each individual feeder. However I have plans to incorporate a mouse sensor for optical alignment - which I've already got the basics established with the sensor. I'd like to make it a neat self contained chip module, providing X/Y offset and rotation data. It's not on the cards to start with though, as it's an optional extra that I can add after I have the core mechanics working with precision. But there is potential to add a separate upward camera for the job. The deck area is flexible, so you could have a wide PCB area and feeders on one side, or smaller PCB area with feeders/trays both sides, and/or designate an area for camera, head changing rack etc...

[scsi]

Hi there!

I'd like to respond to a couple of things discussed on this thread as well as to show off my DIY PnP

daedalus -

1st - that Motorola machine on eBay costs way over $40K. I'd say the SPG gantry system alone is about $50K. I'd buy it in a blink of an eye but I already have a bigger one sitting in my garage and another one won't fit. They used to have fourteen of these and managed to move ten somehow.

2nd - you're right on Madell SX machine being the most DIY friendly.

3rd - you can get the software/vision package for it from Madell. At least I did. Anticipating an obvious question I will tell you that the price was a little over $2K and I believe it was worth it.

Eclipze -

1st - I'd hate to put a nail in the coffin of the open source software development effort for the low cost PnP, but I'm a software developer myself with enough experience and after looking at all the stuff that's involved I decided to buy the ready-built package. Actually, Madell was the only PnP manufacturer that agreed to sell their software to a stranger and I was asking them all. Basically all low cost PnP vendors either denied to sell their stuff as it cannot be adapted to a DIY XYZ system or never responded.

2nd - the quote from the guy being disappointed about Madell tells a good story of a dude who decided to purchase the cheapest PnP machine possible and simply had wrong expectations about it. It's quite possible that he never looked at the other products in the sub $50K range or he would've found that there is no miracle there and everything has to be set up and tuned manually. Including taking pictures of your mother in law. The machines begin miraculously work when you pay over $100K or even $1M for a chip mounter. And even then it's not a plug-n-play experience with lots of phone calls and tech visits. The guy also criticizes the SX feeder functionality but forgets to mention the price per feeder. In other machines every feeder costs between $500 and $1K. With Madell's simple approach to the tape advancement it's essentially under one dollar per reel. That's the cost of the fishing sinker they are using to peel the cover tape and assuming you can make the tape guide yourself (the milling of it is shown in one of the videos below). Yet the SX mechanism is quite reliable and according to Madell, it can handle 0201 parts. I've tried it down to 0402 and it works amazingly well.

In fact, if it's not for Madell, I'd never even think of building a decent PnP at home. Luckily enough, Madell is using DeskCNC motion controller for their machines and I'm running DeskCNC software on it to get some parts done from time to time. I'm not telling which frame grabber and vision package they are using because they didn't want to give that information away until I paid. I still don't understand why - there is absolutely nothing to be shy about. The vision part is built on top of a very reputable industrial grade multi-channel grabber board and comes with a very decent image processing library all licensed for OEM use and stuff.

Well, now it's the show time. First, the machine in my garage is built on top of a very high end gantry system and it is a total overkill for the purpose. This can be done on almost any XYZ table that can show ~0.001" repeatability and at any speed. Size does not matter either. Second, the gantry is running on linear motors, but again, it's not necessary for the PnP. Just no backlash and higher speeds. And third, the machine is set up for milling plastic and aluminum and you will see it throwing some chips around in one of the videos.

1. The machine is moving 0402 resistors for the first time. Playing with a 1KW shop light to demonstrate the tolerance to the ambient lighting changes:

YouTube- Homemade SMT pick and place machine #4 Placing 0402 resistors

YouTube- Homemade SMT pick and place machine #4 Placing 0402 resistors

2. Testing the automatic nozzle changer and then milling some HDPE at ~600 IPM and 0.8mm DOC:

YouTube- Homemade SMT pick and place machine #3 Nozzle Changer and plastic chips

YouTube- Homemade SMT pick and place machine #3 Nozzle Changer and plastic chips

3. Lower quality shot. The machine performs homing and then moves some 0603 resistors at a much slower speed. This was taken before I figured out the trick on how to deal with the DeskCNC acceleration profiles.

YouTube- Homemade SMT pick and place machine #2 - Anorad gantry system

YouTube- Homemade SMT pick and place machine #2 - Anorad gantry system

Questions?

Regards,
Kyryl