Author Topic: O-M-G cheap pick and place machines coming!  (Read 107785 times)

0 Members and 1 Guest are viewing this topic.

Offline JuKu

  • Frequent Contributor
  • **
  • Posts: 566
  • Country: fi
    • LitePlacer - The Low Cost DIY Pick and Place Machine
Re: O-M-G cheap pick and place machines coming!
« Reply #50 on: July 31, 2014, 01:26:20 pm »
BTW, as regards to the contest entry, I was surprised to see it entered, because it's not really a "connected" device?  :-//
Perhaps I'm missing something?
There has to be a motor controller hardware. So, there is a PC connected to a motor control board connected to stepper motors. And a camera connected to something connected to the software. Would you clarify why this would not be a connected device?
http://www.liteplacer.com - The Low Cost DIY Pick and Place Machine
 

Offline Neil Jansen

  • Newbie
  • Posts: 9
Re: O-M-G cheap pick and place machines coming!
« Reply #51 on: July 31, 2014, 02:00:09 pm »
Of course.
Just to clarify, what I meant by "Destined for failure" does not mean I think the project will actually fail, nor do I want it to fail. Just in my my experience when a project tried to do too many thing it always leads to compromises, or takes away form the core focus, or sucks development time on an non-core feature etc
It's something we talk about a lot on the Amp Hour.
We're going to stay focused and vigilant towards the project at hand: making it easy to place super tiny SMT parts reliably.  I've not even tried the hotend extruder out yet on the first prototype machine, it's just hanging there like a vestigial leftover.  I have been focused, and will stay focused, on getting the SMT stuff working before I touch it.  But I'm a RepRap veteran, and 3D printing design and operation comes very naturally to me, and I realize how powerful it can be on a machine like this, for printing component feeders and fixturing in ESD-safe plastic, not to mention enclosures and brackets.  Some people still think 3D printing is nothing more than a toy, I would disagree with that.  I use it to make jigs, pogo pin fixtures, and other awesome stuff, and want to share that sort of workflow with others.

Quote
I have no doubt you'll sell machines, but that is separate issue about how viable small cheap DIY PnP machines are actually are when you look at them unbiasedly. Just because people buy them and use them doesn't actually make them viable. Some people just like to work inefficiently and charge zero for their time.

So once again, I'm not against them, I think they are cool, heck, I want one, but I maintain my view that there is a very narrow window of real value useability on these things.
You're a young strapping gent, your eyesight is good, and you have a steady hand, I assume.  The retired NASA engineers that I've talked to, that have lost the wherewithal to do the small stuff (or lived before the small stuff existed) believe that this machine is viable.  Those with vision problems, which includes my main software developer, Karl Lew, believe this machine is viable (he's slowly going blind, and can't drive any more). The guys down at my local makerspace, wanting to solder about 10,000 WS2812B LEDs into a giant tetris array think that this project is viable.   Those without SMT soldering skills, and those with mediocre soldering skills think that this project is viable.  I am the kind of guy to check my privilege before deciding what's viable and what isn't.  I can solder 0402 chip parts and 1/2mm pitch IC's, even QFN's, DFN's, and BGA's, all by hand, but so what?  Can everyone?  Does everyone want to?  Outside of a very small prideful engineering community, this is a very viable product, that has real usability.

Like I said, we're working on some really clever stuff, including a new feeder design that doesn't require futzing around getting each part inserted correctly.  This will be an open standard that hopefully middle-men companies can jump in and provide a useful service to those that'd rather not deal with the time spent loading feeders.  Also, not sure if I've covered it, but our system automatically recognizes feeders and boards, which can be placed anywhere.  This is possible thanks to our incredibly flexible high-level vision API that we've spent the last year writing.  We're also writing an intuitive GUI to replace the one in OpenPnP, which is quite busy and not intuitive.  This machine will be quite easy to use, and we've got at least 100 beta testers signed up (most are serious engineers with a proven track record of SMT assembly in one form or another), so assuming a fraction of them actually build/buy one and provide us feedback, we're on our way to making a very useful system.


Quote
BTW, as regards to the contest entry, I was surprised to see it entered, because it's not really a "connected" device?  :-//
Perhaps I'm missing something?
You control the machine via any HTML5 browser, via  node.js/Express webserver.  That could be a laptop, tablet, even iphone (we use the Twitter Bootstrap framework, so it's responsive to different devices).  That's enough to consider it a connected device, according to my interpretation of the rules.  It's not a static machine with its own UI like a NeoDen TM-220A.  You have to connect to it via Ethernet or Wi-Fi for it to be useful.
 

Offline free_electron

  • Super Contributor
  • ***
  • Posts: 8515
  • Country: us
    • SiliconValleyGarage
Re: O-M-G cheap pick and place machines coming!
« Reply #52 on: July 31, 2014, 02:05:04 pm »
Hi, I'm the lead developer of FirePick Delta, I saw this post in my referrer logs and figured I'd chime in.
great. Now we can talk instead of rant-n-rave

About the jack of all trades master of none. I too get that feeling judging from the webpage.
It even mentions its usage as a rework machine placing and replacing BGA parts. Pardon me if i inject a mandatory eyeroll here. And i quote :
Quote
Rework: Our robotic machine will also function as a rework station, capable of replacing BGA's, QFN's, TQFP's, and any other common part available.  This is a great thing for those for the electronics repair business, the video game modding community, and the smarter-than-average consumer with a broken consumer device that needs fixing.
You are creating great expectations... That aren't possible as it is only the placement step of a rework cycle , even the removal can't be done.

First of all rework implies desoldering and resoldering. Bring heat anywhere near that placement head and you end up with a puddle of goo..

Second. Bga ? With a dispenser needle as pickup and an aquarium pump that kicks in and out and creates a pulsed airflow ? I think not. Time to bring in compressed air and a venturi to create vacuum.

Third bga and tqfp need an upward looking camera to check pin alignment. These packages also require a matrix feeder (these parts come in standardized trays. So do a lot of packages like qfn son and others) i dont see that capability in the machine. The only way for those packages to identify pin 1 is using a bottum-up camera. Any part with occluded pins can not be handled unless you have this kind of camera.

Judging from you pictures you use a downward looking camera but i dont see an illuminator anywhere.

You mention paste dispensing. Are tou familiar with the concept of pullback ? To correctly dispense you need both compressed air and vacuum.


Some random thoughts : the machine is square. Can we have feeders on all sides ?


Quote
Tray feeders: Yes, although not the big JEDEC sizes.
Define big. Can you handle a tqfp100 ? What about a 144 ? Or a 208 ?  To put it in arduio terms : i want the thing to be able to place the atmega2560. You do need the jedec trays as that is the standard. See my comment above. Lots of parts in qfn , son and other packages come in these trays.

Quote
For forks sake ..

Again, lots of misinformation here.  Yes we forked OpenPnP, but we use Git, and can send pull requests back to Jason Von Nieda, who runs OpenPnP.  What is the problem here?  That's how software gets done these days.  I have good communication with Jason, we're not doing a "hard fork", we've not come to philosophical differences on OpenPnP. 
the. You are one of the few doing it right and i stand corrected. Too much stuff gets forked on philisophical differences and in the end both go nowhere. That seems not to be the case here. Good !



[/quote]Anyway, I could keep going, but not sure it'll really do much.  This machine will work great for many.  [/quote]
I will buy it if it can handle 0603 upward . Forget the big packages for now. Would be nice but i am good at handsoldering the big tqfps. It's the tedious placing of resistors and caps and small fry that bores me. But, i do want an up looking camera to handle qfn and son and real jedec trays. Too many parts these days use these packages (accelerometers, compass chips , bluetooth modules)

Another very important thing is the teach-in of the machine. Taking centroid data is not enough. Take an led for example. That is a polarised component. Having centroid data and rotational data is not enough. You need to know how the board sits in the machine and you need to know how the part sits in the machine.

What about flipped parts. Once in a while parts flip over as the tape progresses. This leads to a mispick. Mispick detection needs multiple things : vacuum integrity check to verify we did get a part on the nozzle. Sideview to verify the part didn't flip. I would at least expect vacuum check so the machine can do an in position retry , then progress the tape and try again.


I'll be keeping an eye on this ...i would even want to buy one , provided it is reliable and doesn't require constant futzing with the machine.

What i mean with that is : i was at techshop a few weeks ago where there was a gathering of 3d printing enthousiasts. I talked to a whole bunch of different machine owners. I asked them : are these machines turn-key ? The answer was unanimous : no. Everybody that has 3d printing hobby is more tinkering with the machine than actually creating parts with it. Half of them were printing 'calibration jigs' so they could make their life easier when they change materials... They all still struggled with peeling of parts during builds, sagging material, parts being too large or too small due to wrongly comle sated overprint and or material shrinkage.
If i send a file to a 3d systems machine i don't need to futz with anything. Push the button and it comes out perfect. Especially the SLS machines printing nylon.

3d printing in my hobby means : making things, not tinkering with the machine itself.
Likewise with a pick and place : shoot parts on a board. Not constantly futzing with the machine itself.
Pick and place machines can be moody little buggers. I had a juki for years in the lab to run prototypes. It ran great but it needed too much 'caring love' to my liking. ok, it was 10 years old already back in 2000...
But i will expect a machine that is at least on par with a 15 year old pick and place. Shove in reels, assign position and hit 'go' , not 'take out wrench and hammer and apply percussive maintenance to force it to behave'
« Last Edit: July 31, 2014, 02:33:40 pm by free_electron »
Professional Electron Wrangler.
Any comments, or points of view expressed, are my own and not endorsed , induced or compensated by my employer(s).
 

Offline Neil Jansen

  • Newbie
  • Posts: 9
Re: O-M-G cheap pick and place machines coming!
« Reply #53 on: July 31, 2014, 02:05:21 pm »
This is what the contest entry says.

Quote
It must transmit data to and/or from another device (computer, phone, duplicate/variation of your device, etc). This could be over the Internet, or using any other method of your choosing.

If it is transmitting control sequences from a PC to the device then it would seems to be "connected" assuming a USB cable (or serial) is "any other method of your choosing".

If you want to delve any deeper into the semantics of "connected" then it's lawyers at 10 paces.
Precisely.  I believe that we meet that requirement easily, with our html5 browser interface via wifi / Ethernet.

Here's our youtube video that we were required to make for the contest, describing how we're meeting the various requirements of the rules:

 

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13677
  • Country: gb
    • Mike's Electric Stuff
Re: O-M-G cheap pick and place machines coming!
« Reply #54 on: July 31, 2014, 02:35:22 pm »

Second. Bga ? With a dispenser needle as pickup and an aquarium pump that kicks in and out and creates a pulsed airflow ? I think not. Time to bring in compressed air and a venturi to create vacuum.
A simple cheap diaphragm vacuum pump and a solenoid valve to get clean release can work just fine. May help to have a reservoir to filter any bumpiness in the pump, though this would need a 2-way valve.
Quote
Third bga and tqfp need an upward looking camera to check pin alignment.
All parts need an upward looking camera. Ideally in the head (at least as an option) so it can image on the journey from feeder to board.   
Quote
These packages also require a matrix feeder (these parts come in standardized trays. So do a lot of packages like qfn son and others)
All you need is the ability to pick parts from trays - you don't  need auto tray-change on a machine at this level.
Quote
The only way for those packages to identify pin 1 is using a bottum-up camera. Any part with occluded pins can not be handled unless you have this kind of camera.
You don't need to auto-detect orientation - just load it right. 
Quote
You mention paste dispensing. Are tou familiar with the concept of pullback ? To correctly dispense you need both compressed air and vacuum.
Or even better, an auger  - I'm sure an auger dispenser could be made using a woodscrew or other standard part as the auger

Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13677
  • Country: gb
    • Mike's Electric Stuff
Re: O-M-G cheap pick and place machines coming!
« Reply #55 on: July 31, 2014, 02:46:55 pm »

Hi, I'm the lead developer of FirePick Delta, I saw this post in my referrer logs and figured I'd chime in.

Just wondering, have you ever used, or talked people who use real pick/place machines all day.
If not I'd very highly reccommend it, as there are lots of subtle things you only realise once you start doing this stuff in anger.

A few examples I can think of offhand  :
Never ever use any kind of oil near a feeder, as it will soak up dust from paper tape and gum up
Set the pick position of plastic tape slightly high so that the nozzle sucks the part out without bouncing parts out pf the next pocket.
It is extremely useful to be able to stop a job, tweak vision parameters for the part being picked and carry on.
small SO parts can come in tape or tube, and you may not  know which at design time, so you need to be able to override different feeder orientations seperate from the part definition. Similarly sometimes the same part (same to the extent of using the same part definition) may come on tape spaced at one to two indexes, so this also needs to be overrideable.
There are many, many more....


 


 
Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Offline free_electron

  • Super Contributor
  • ***
  • Posts: 8515
  • Country: us
    • SiliconValleyGarage
Re: O-M-G cheap pick and place machines coming!
« Reply #56 on: July 31, 2014, 02:53:23 pm »

Quote
All parts need an upward looking camera. Ideally in the head (at least as an option) so it can image on the journey from feeder to board.   
If you know poition at time of picking and use a feedback mechanism like an encoder in the rotating head then passives can be done reliably. Which leads to another question : does this machine have feedback on its steppers ? Afaik none of the 3d printers have that...

 These packages also require a matrix feeder (these parts come in standardized trays. So do a lot of packages like qfn son and others)
Quote
All you need is the ability to pick parts from trays - you don't  need auto tray-change on a machine at this level.
Agreed, but i do want to be able to use full trays and have room for at least 2 trays. I'll do a manual swap. I also want the software to have a hold-for-operator step for trays so i can swap the tray when the machine prompts me.


Quote
You don't need to auto-detect orientation - just load it right. 
There is a vision system :use it. I don't agree with the load it right filosophy. What if i put the stick in reverse ? The vision system needs to be able to detect that. That's what the damn thing is for !
On the juki the teach-in camera would go to the part and you clicked where pin 1 sits. For example sod323 packages are very hard for the machine vision. The teach in camera would show the part and you told it pin one sits left, right top or bottom. (The machine had feeders on four sides). And that was it. Very easy. Same for tantalums.

Quote
Or even better, an auger  - I'm sure an auger dispenser could be made using a woodscrew or other standard part as the auger
Mmm no. I want it to use simple 10cc syringes for dispensing.
Professional Electron Wrangler.
Any comments, or points of view expressed, are my own and not endorsed , induced or compensated by my employer(s).
 

Offline tridentsx

  • Regular Contributor
  • *
  • Posts: 97
  • Country: us
Re: O-M-G cheap pick and place machines coming!
« Reply #57 on: July 31, 2014, 02:57:33 pm »

I think the mirror is used to see the bottom of a part. No need for an extra camera,some commercial system does it the same way, but the mirror is attached to the head and flipped in place instead.
 

Offline Neil Jansen

  • Newbie
  • Posts: 9
Re: O-M-G cheap pick and place machines coming!
« Reply #58 on: July 31, 2014, 03:00:04 pm »
A simple cheap diaphragm vacuum pump and a solenoid valve to get clean release can work just fine. May help to have a reservoir to filter any bumpiness in the pump, though this would need a 2-way valve.
Our project is an open-source one, so there's nothing stopping anyone from using a Venturi and solenoid instead of a diaphragm pump.  We started with a diaphragm pump because they're cheap and available.  Will they be able to do BGAs and crazy small parts?  I don't know right now, but we will be evaluating them and making our decisions from there.  We don't want to jump right to the super expensive parts, as that is not our philosophy.  We want to offer a cheap system (one many here would easily scoff at), but also offer a proper system to anyone that needs it, and can afford it.

Quote
All parts need an upward looking camera. Ideally in the head (at least as an option) so it can image on the journey from feeder to board.   
For our cheapest version, we use a cheap mirror and some computer vision parallax-fixing code to do cheap up-looking vision.  We're currently trying to get the camera closer to the nozzle than we can get with a stock camera module pcb assembly.  We will also offer a second up-looking camera for those that have the cash and want the performance.  We're using cellphone camera modules, they're high-resolution, easy to multiplex, and allow us to access all of the hardware functions like shutter speed, ISO, and hardware shutter, unlike USB pen cams and some webcams.

Quote
All you need is the ability to pick parts from trays - you don't  need auto tray-change on a machine at this level.
Agreed.

Quote
You mention paste dispensing. Are tou familiar with the concept of pullback ? To correctly dispense you need both compressed air and vacuum.
I'm very familiar with pullback, I own a really nice shot-dispense system that does it correctly.  Currently, our FPD paste dispense prototype uses a 10cc syringe with small stepper motor on a screw drive.  We actually use modified RepRap filament calculations and GCode to control the extrude and retract.  And as mentioned earlier, the paste dispense will be closed-loop via our vision system, so it will do some calibration dots on a little tray that sits on the edge of the machine, and determine correct steps-per-dot-size from there.  This whole system is actually completely compatible with your normal air-compressor-powered shot dispense system, most have a foot control, you can wire this up to the mosfet output of the controller, as long as you interface it correctly.  Obviously tweaking the air pressure and vacuum must be done manually, but the timing can be done via the controller software. 

And of course, if someone still wants to use stencils, that's totally cool too.  If you've got 'em, use 'em.  But having paste dispense for small quick stuff is still something many will find value in.
« Last Edit: July 31, 2014, 03:05:42 pm by Neil Jansen »
 

Offline Karl Lew

  • Newbie
  • Posts: 9
Re: O-M-G cheap pick and place machines coming!
« Reply #59 on: July 31, 2014, 03:02:00 pm »
Hi!

I'm Karl. I work on FirePick software and somewhere in my past is an EE degree, although my entire career has been in software engineering from punched cards all the way through to AngularJS via Java, C# and C++.

Neil pointed out this very amazing thread of massive feedback which took me by surprise. Thank you all for laying it out there. I learned a lot just reading the thread. To all the skeptics, I would say...

"Yes we'll have it done tomorrow"

OK. I was kidding. We won't have it all done tomorrow or the day after. This is a TOUGH problem to solve. It's a problem that takes skills in mechanical, electrical and software engineering. Assembling a team to tackle these problems together is also difficult. Thanks in very large part to Hack-a-Day and the ensuing publicity, we've brought together a global team of engineers who do span those skills. We still won't have it done tomorrow.

What we will have is an iterable design. Several folks have questioned the swiss-army tool lack of focus of the design. I would agree that the fatter the swiss-army knife the more useless it is, and I only use mine for cooking when I backpack. But the reason we're paying attention to the "all-in-one use-less case" is that it aligns the common design problems very nicely so that we can tackle the common elements together and specialize later. Although a 3-in-1 machine (3D, PnP, SolderPaste) would attract notice, we expect that three separate machines evolving from the common base would be more pragmatic. It's early for that specialized divergence, so we're doing 3-in-1 to get things like the delta mechanics and software rock solid before branching out.

Once again, thanks for all your feedback. We're quite happy to chat with you folks and learn from you and work with you.

:) Karl
 

Offline Neil Jansen

  • Newbie
  • Posts: 9
Re: O-M-G cheap pick and place machines coming!
« Reply #60 on: July 31, 2014, 03:03:27 pm »
Just wondering, have you ever used, or talked people who use real pick/place machines all day.
If not I'd very highly reccommend it, as there are lots of subtle things you only realise once you start doing this stuff in anger.

Yes, I use to run an SMT line, about six years back. Many of my friends run real pick and place machines, and I talk to many of them at conferences and shows that I bring my machine to, and on the internet.

Thank you for your advice though, I certainty don't know everything, and am very interested in getting feedback from people.  EEVblog seems to be a great place to get that level of feedback, lots of people here that know their stuff.
 

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13677
  • Country: gb
    • Mike's Electric Stuff
Re: O-M-G cheap pick and place machines coming!
« Reply #61 on: July 31, 2014, 03:08:36 pm »

All parts need an upward looking camera. Ideally in the head (at least as an option) so it can image on the journey from feeder to board.   
If you know poition at time of picking and use a feedback mechanism like an encoder in the rotating head then passives can be done reliably.
Depends on how well taped they are, I've seen SOT-23s with quite a lot of variation, and even some 0603s with a fair amount of wobble space in the tape.
Quote
Which leads to another question : does this machine have feedback on its steppers ? Afaik none of the 3d printers have that...
Not really necessary as long as your speed curves are safe - may be useful if you're pushing for max speed.  It would be useful to be able to detect a head crash though
Quote
Agreed, but i do want to be able to use full trays and have room for at least 2 trays.
Another possible option is to have room for half a tray on the bed and half hanging over the side, and turn round half way (with auto part rotation, obviously)

Quote
There is a vision system :use it. I don't agree with the load it right filosophy. What if i put the stick in reverse ? The vision system needs to be able to detect that. That's what the damn thing is for !
But there are so many variants in how parts ID pin 1, it may be too much hassle to program in.
Maybe something as simple as a manual verify when picking the first part would be a good compromise


Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Offline EEVblog

  • Administrator
  • *****
  • Posts: 37626
  • Country: au
    • EEVblog
Re: O-M-G cheap pick and place machines coming!
« Reply #62 on: July 31, 2014, 03:17:14 pm »
There has to be a motor controller hardware. So, there is a PC connected to a motor control board connected to stepper motors. And a camera connected to something connected to the software. Would you clarify why this would not be a connected device?

IMO "connected" means internet/cloud/whatever connected, not just a PC to control something.
Otherwise you can join any two components together and say the design is "connected".

But technically the entry requirements state:
Quote
It must transmit data to and/or from another device (computer, phone, duplicate/variation of your device, etc). This could be over the Internet, or using any other method of your choosing.

So yeah, it meets that. But as a judge a 3D printer/CNC machine/whatever connected to a PC USB port to control it doesn't rank nearly as high on the "connected" scale as say someone who rolls their 10,000 node mesh network. It's a given that such a device is connected to the PC to control it, so there is nothing new or innovative in that aspect of it at all.
Given that the only two main things are "connected" and "open", well, you'd really want to blow our socks of with the open/documentation part.

Although we have not been given official judging guidelines yet, so this is just they way I see it.

As I mentioned in my interview video for the contest, I personally want to see something so clever that I slap my forehead and go "why didn't I think of that?".
And I suspect it's more likely than not that the winner will be such a device/idea.
 

Offline free_electron

  • Super Contributor
  • ***
  • Posts: 8515
  • Country: us
    • SiliconValleyGarage
Re: O-M-G cheap pick and place machines coming!
« Reply #63 on: July 31, 2014, 04:02:49 pm »
Depends on how well taped they are, I've seen SOT-23s with quite a lot of variation, and even some 0603s with a fair amount of wobble space in the tape.
that is a problem. true. another thing i ran into is tapes not consistently feeding the right amount. so the part sits offset on the nozzle leading to placement errors. for 1206 not a big problem. the moment you reflow it self adjusts. 0402 ? disaster ..
vision needs to compensate for that.
Quote
Not really necessary as long as your speed curves are safe - may be useful if you're pushing for max speed.  It would be useful to be able to detect a head crash though
that is one problem. which leads me to think about another potential problem. tall components like SMD electrolytics. when that head moves it better be up high as i don't want it to run into already placed components
also component drop during movement. i would strongly suggest there is a vacuum integrity check from pick to place. venturi + pressure sensor. a venturi doesn't cost anything. a venturi + pressure sensor will add max 10$ to the bom. well worth the effort. it can be put in as an optio , but provide an analog channel that is beeing monitored during travel and can have min and max set.



Quote
Another possible option is to have room for half a tray on the bed and half hanging over the side, and turn round half way (with auto part rotation, obviously)

good idea. you can actually make a terrace like holder. two sideways , one on top. the bottom ones would need rotating halfway through. another idea would be a 3dprinted custom tray holding 20 pieces per package. the tqfp's can all be lumped in one cavity kinda like an inverted pyramid so the tqfp32 falls deeper than the tqfp100. the center to center pitch qould be identical. you could build a tray holding for example 40 universal tqfp and some other often used packages. ok woul would have ot hand load the tray from the master trays , but ti would allow for many different parts in one build. the software would need to have the capability to be told U1 sits in columns 1 to 4 , u2 in colums 5 to 8 , u3 in columns 9 to 12 and so on. i'm just thinking about a project i am working on now. there are two different QFNs and 3 wsons. these all come in tray.

Quote
But there are so many variants in how parts ID pin 1, it may be too much hassle to program in.

no , that's why you need to use the camera as teach-in. here is how the juki did it. you would mark , in a separate column, the parts that have a polarity. (this can be exported from the cadprogram. i simply added a field to the library that said 'polarized with true or false)
when the program wast first ran the machine would roll to the part location and halt. the monitor would show the camera image with four buttons. At this point you looked at the screen how the part sits in the feeder. then you click a button. the machine picks the part and rolls to the placement position. all you had to do now is hot the rotate button (rotated 90 degrees per click so max 3 clicks and you were bang on,until it sat right. the machine would then memorise this rotational offset.
the pick and place software only used centroid data and the rotation angle given from cad. the machine would do that once for every feeder marked as 'polarized'
done. In essence on the first run the machine sorted the parts , polarized first , placed one of each type with your assistance and then shot the rest on. the teach data was saved in the placement file.

after placement of  a part the camera woudl show theplaced part. if it was wrong it could pick it back up , rotate 90 degrees and place again. so just in case you messed up you coudl still correct.

the sequence was :
look at feeder. click ok
machine picks part and goes to placement coordinates. camera shows part above placement position. click rotate until it matches, then click ok.
machine places and retracts nozzle. camera stays in position. if correct click ok, if not correct click rotate . repeat until correct.

now, since in this machine the camera psotion is offset you could go straight to place and show placed part. by then clicking 'clockwise 90' , 'counterclockwise 90' or '180' you can adjust. so teach in could be done very simple. you wouldnt care how stuff sat in the tapes or sticks. during first run you set it visually.
subsequent runs can be launched with or withour re-teach ( if you had run a different project or had changed tubes or reels you could click the affected slot and it would re-teach that one only.
Professional Electron Wrangler.
Any comments, or points of view expressed, are my own and not endorsed , induced or compensated by my employer(s).
 

Offline free_electron

  • Super Contributor
  • ***
  • Posts: 8515
  • Country: us
    • SiliconValleyGarage
Re: O-M-G cheap pick and place machines coming!
« Reply #64 on: July 31, 2014, 04:06:03 pm »
something so clever that I slap my forehead and go "why didn't I think of that?".
like a gravity detector with requiring a frequency counter. :) a ble module with a accelerometer and position sensor woudl do. hold module , take positional reading , let go , acceleremoter nog gives attracting vector. display on smartphone : your feet are that way ... Great for when you are so drunk (or stoned) and don't remember which way is up.
Professional Electron Wrangler.
Any comments, or points of view expressed, are my own and not endorsed , induced or compensated by my employer(s).
 

Offline Karl Lew

  • Newbie
  • Posts: 9
Re: O-M-G cheap pick and place machines coming!
« Reply #65 on: July 31, 2014, 04:29:49 pm »
For me, the fun challenge of FirePick Delta has been figuring out how to integrate vision with motion in an open, extensible way. Effective PnP motion ideally requires a model informed by vision (or other sensors) in real time to handle real world discrepancies from plan. This sets it apart from 3D printing, which can be done as a batch job after hellish calibration. From a software engineering perspective, a REST architecture offers a lot of benefits in solving this coordination, since REST can drive GUI as well as machine-to-machine interaction.

With a REST foundation, the question then boils down to what the REST protocols should be for hooking all that up together. FireREST-CV is the first such protocol and it is based on FireSight, which is a JSON pipeline wrapper we developed for OpenCV (i.e., vision recognition as configuration vs. program):
https://github.com/firepick1/FireREST/wiki/FireREST-CV

To prove the utility of FireREST-CV as a REST service, we snapped it into OpenPnP, an existing desktop application developed by Jason von Nieda (http://openpnp.org/). It works. Neil took it to OSCON last week. FireREST-CV also allows us to use mobile devices to watch what's going on while OpenPnP does orchestrates the PnP job. But that's still not enough.

Moving beyond FireREST-CV is FireREST-CNC, which is a REST wrapper for GCODE. The REST web server is the RPi, which we choose for its ubiquity, but there is really nothing that ties FireREST to that particular SBC. This REST architecture allows the motion-vision coordination to happen in one of two places: 1) on the desktop, or 2) on the embedded SBC. The former reflects current state and we are using OpenPnP to validate that use case. The latter represents the future of the micro-factory as we see it. Indeed, Jason has kindly offered to make "headless OpenPnP" available under LGPL for further exploration. For now, given the  SBC challenges (e.g., 10hrs to compile OpenCV), we'll gradually tackle that complexity use case by use case, starting with calibration, evolving the software with the SBC's.
 

Offline Mechatrommer

  • Super Contributor
  • ***
  • Posts: 11518
  • Country: my
  • reassessing directives...
Re: O-M-G cheap pick and place machines coming!
« Reply #66 on: July 31, 2014, 05:27:35 pm »
that is one problem. which leads me to think about another potential problem... also component drop during movement. i would strongly suggest there is a vacuum integrity check from pick to place.
that is one problem. which leads me to think about another potential problem...

if the component drops on the pcb, we need the vision comes looking for it, trash or replace it in the feeder... to avoid damaged board due to short circuit of misplaced/dropped component that sticked on it. also to avoid stacked components on dropped one resulting in failure placement and the machine keep placing the failed until they are all piling up, just like failed 3d printer printing in the air stuff plastic piles on the nozzle until it hits something.

there should be also mechanism to detect if the dropped component bounces out of the board to skip searching the every inches/corner of the board which takes/waste time, or either fire the alarm to call the operator to indicate there is problem... ERR108: dropped component cannot be found. but then, calling for manual intervention in an unsolved problem is unprofessional isnt it? hmm ::)
« Last Edit: July 31, 2014, 05:36:20 pm by Mechatrommer »
Nature: Evolution and the Illusion of Randomness (Stephen L. Talbott): Its now indisputable that... organisms “expertise” contextualizes its genome, and its nonsense to say that these powers are under the control of the genome being contextualized - Barbara McClintock
 

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13677
  • Country: gb
    • Mike's Electric Stuff
Re: O-M-G cheap pick and place machines coming!
« Reply #67 on: July 31, 2014, 05:33:41 pm »
Depends on how well taped they are, I've seen SOT-23s with quite a lot of variation, and even some 0603s with a fair amount of wobble space in the tape.
that is a problem. true. another thing i ran into is tapes not consistently feeding the right amount. so the part sits offset on the nozzle leading to placement errors. for 1206 not a big problem. the moment you reflow it self adjusts. 0402 ? disaster ..
vision needs to compensate for that.

My pick/place also auto adjusts its pick position based on offsets seen when visioning previous parts so it homes in on the nominal position

Not really necessary as long as your speed curves are safe - may be useful if you're pushing for max speed.  It would be useful to be able to detect a head crash though
[/quote] that is one problem. which leads me to think about another potential problem. tall components like SMD electrolytics. when that head moves it better be up high as i don't want it to run into already placed components

You'd normally travel at max height - if clever you'd auto-set this based on the height of the tallest part in the job to reduce wasted up/down time
Quote
also component drop during movement. i would strongly suggest there is a vacuum integrity check from pick to place. venturi + pressure sensor. a venturi doesn't cost anything. a venturi + pressure sensor will add max 10$ to the bom. well worth the effort. it can be put in as an optio , but provide an analog channel that is being monitored during travel and can have min and max set.
mis-picks & picks from empty slots are much more common than drops - you may want different behaviour though - mis-pick=retry, drop during travel = stop & alert user

Vacuum sensing could be useful but not foolproof. it may be doable more cheaply by sensing vacuum pump current draw. I can "hear" when my P&P mis-picks from the change in pump tone.



Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Offline Karl Lew

  • Newbie
  • Posts: 9
Re: O-M-G cheap pick and place machines coming!
« Reply #68 on: July 31, 2014, 05:43:39 pm »
Mike, this surprised me and it wasn't a use case I had thought about:

Quote
mis-picks & picks from empty slots are much more common than drops - you may want different behaviour though - mis-pick=retry, drop during travel = stop & alert user

What have you found to be the root causes for mis-picks from empty slots?
 

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13677
  • Country: gb
    • Mike's Electric Stuff
Re: O-M-G cheap pick and place machines coming!
« Reply #69 on: July 31, 2014, 06:06:23 pm »
Mike, this surprised me and it wasn't a use case I had thought about:

Quote
mis-picks & picks from empty slots are much more common than drops - you may want different behaviour though - mis-pick=retry, drop during travel = stop & alert user

What have you found to be the root causes for mis-picks from empty slots?
Usually, running out of parts, obviously this gets caught later with vision
Also tape cover not peeling, or splitting
Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Offline Karl Lew

  • Newbie
  • Posts: 9
Re: O-M-G cheap pick and place machines coming!
« Reply #70 on: July 31, 2014, 06:24:05 pm »
Quote
Also tape cover not peeling, or splitting

Ah. That's interesting. For FirePick I's semi-automatic feeder, the operator would peel back the tape for say, 15 components, and immediately notice a tear. For FirePick Delta, ripped cover tapes would probably happen most with initial setup conducted by operator, but could also happen during prolonged feed. I've noticed that excessive handling of tape leads to dimples in the tape which might (?) provide a weakness for rips to start. Also, Neil's feeder is 3D printed, which creates a slightly rounded edge that hopefully (?) might not trigger the rip as much as a sharp metal edge. We'll watch out for rips 'n tears, in prolonged use. Thanks!
 

Offline free_electron

  • Super Contributor
  • ***
  • Posts: 8515
  • Country: us
    • SiliconValleyGarage
Re: O-M-G cheap pick and place machines coming!
« Reply #71 on: July 31, 2014, 09:59:32 pm »
mispicks can happen because a part simply vibrated out of its pocket. it can be clinging to the peeled tape, vibration could have cause it to stand up or pop out altogether.

i just read : operator would peel back tape. WHOA !! you better have a tape winder on those feeders or the machine is useless !
the idea of a pick and place is that you press start and go do something else. having to babysit the feeders is a waste of time.

that is why you do vacuum integrity check. nozzle down, vacuum on , wait for detectable drop to a preset value ( pumpdown of the line. if the pick is half then you will never meet the vacuum threshold : eject part ( head travels to the 'bucket' where it drops the part ) and then does a retry. 3 retries : alert operator.

even if you don't install vacuum sensor by default make sure we have access to a spare analogue input and can set a threshold on that value after pick.
pick part , wait for threshold ( with timeout ) then go do the place. retry 3 times.
Professional Electron Wrangler.
Any comments, or points of view expressed, are my own and not endorsed , induced or compensated by my employer(s).
 

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13677
  • Country: gb
    • Mike's Electric Stuff
Re: O-M-G cheap pick and place machines coming!
« Reply #72 on: July 31, 2014, 10:19:11 pm »
We will also offer a second up-looking camera for those that have the cash and want the performance.  We're using cellphone camera modules, they're high-resolution, easy to multiplex, and allow us to access all of the hardware functions like shutter speed, ISO, and hardware shutter, unlike USB pen cams and some webcams.
Surely these cameras are so cheap now that it would be worth having both as standard instead of  messing with mirrors?
Quote
And of course, if someone still wants to use stencils, that's totally cool too.  If you've got 'em, use 'em.  But having paste dispense for small quick stuff is still something many will find value in.
Absolutely.
Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 

Offline Neil Jansen

  • Newbie
  • Posts: 9
Re: O-M-G cheap pick and place machines coming!
« Reply #73 on: July 31, 2014, 10:22:27 pm »
i just read : operator would peel back tape. WHOA !! you better have a tape winder on those feeders or the machine is useless !
the idea of a pick and place is that you press start and go do something else. having to babysit the feeders is a waste of time.
That's true only for FirePick 1, which was Karl's original design.  My design, FirePick Delta, has full-auto cover tape winders.  So far, I'm extremely happy with them.

Quote
that is why you do vacuum integrity check. nozzle down, vacuum on , wait for detectable drop to a preset value ( pumpdown of the line. if the pick is half then you will never meet the vacuum threshold : eject part ( head travels to the 'bucket' where it drops the part ) and then does a retry. 3 retries : alert operator.

even if you don't install vacuum sensor by default make sure we have access to a spare analogue input and can set a threshold on that value after pick.
pick part , wait for threshold ( with timeout ) then go do the place. retry 3 times.
FirePick Delta SMT nozzles are interchangeable, and do include a spare analog input pin for pressure sensing (or vacuum pump current monitoring via ina139 current amplifier).  I have a few pressure/vacuum sensors laying around that I bought for such a purpose, but it'll be a month or two before I have a chance to play around with this.
 

Offline mikeselectricstuff

  • Super Contributor
  • ***
  • Posts: 13677
  • Country: gb
    • Mike's Electric Stuff
Re: O-M-G cheap pick and place machines coming!
« Reply #74 on: July 31, 2014, 10:23:40 pm »
the idea of a pick and place is that you press start and go do something else. having to babysit the feeders is a waste of time.

3 retries : alert operator.
3 retries - give up on that part and get on with the next one, until it has nothing more it can do, then  alert operator!
Youtube channel:Taking wierd stuff apart. Very apart.
Mike's Electric Stuff: High voltage, vintage electronics etc.
Day Job: Mostly LEDs
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf