The case for automated PnP

Chat about pick and place machines, reflow ovens, assembly techniques and other SMT tips & trix

Moderators: adafruit_support_bill, adafruit

Please be positive and constructive with your questions and comments.
ai_dude
 
Posts: 21
Joined: Fri Dec 11, 2009 1:49 am

The case for automated PnP

Post by ai_dude »

Hi all,

Newbie to this forum, about to invest in some PnP equipment and evaluating options. I recently got a decent benchtop conveyor oven, so it's really the PnP machine I need now. Smallest parts I use currently are 0603, SC-70 (slightly smaller than SOT-23), SSOP and SMA's, but I'd like to go to 0402 and some small resistor arrays in the future.

Here's my evaluation so far...
Tweezers and soldering iron method: I use this currently but there are some issues -- slow, not the prettiest looking joints, backaches from hunching over to see the top of the PCB, and since I've not been able to find anyone locally who can do these, I'm doing all the SMD stuff currently. But the boards function well.

Syringe dispensing + Vacuum-pickup pen: I got one of these recently (as part of a hot-air rework tool), but only tried a few resistors. I need to spend some time with this method and time myself. I expect the syringe dispensing will be a tedious process though. We'll see.

Stencil print + Vacuum+pickup pen: I need to get a stencil to test this, but I really expect this to be a good speed improvement over syringe + vacuum pen.

Manual PnP: I'm really wondering how much speed advantage I'd get with these. One major concern is that for parts on a carousel (some part will be ), many parts will fall upside down, and it's difficult to pick up each with tweezers and flip them over. In my current tweezers method, I dump the components on an old piece of paper (so that it's not perfectly flat, but no creases) and use the right-side-up parts first, then tap the paper to flip some of the upside-down parts over, etc, etc. With the manual PnP units I've seen, it seems there's not much space to do that.

Automated PnP: I'm actually pretty committed to this already, so this comparison is really to help me justify it. :wink: The advantages are of course that I only have so much time, and this will let me get more done, with relatively minimal supervision. And it would be much easier to get someone else to supervise it, while I get back to running my business. However, I'm also considering building my own PnP machine, as I really want a compact benchtop type system, and there's not many of those on the used market. New units are cost prohibitive. I mentor a high-school robotics team and I know some of the kids are ready to help with this project, as we're not in competition season currently.

FWIW, APS lists some average placement rates for different methods in their manual Pnp system documentation. They show Tweezer = 75 to 125cph, Vacuum pen = 100 to 150cph, and Manual Pnp = 400 to 600 cph. I expect a bit of bias, but apparently there's quite a bit, as my tests show that I can do 240cph with tweezers (0603's, some 0805's, and SOT-23's). I will compare the same board with the same component set using the syringe dispensing + vacuum pen method soon and report back.

Cheers,
-Neil.

alex_dubinsky
 
Posts: 188
Joined: Wed Jul 30, 2008 5:17 pm

Re: The case for automated PnP

Post by alex_dubinsky »

You ask about justifying a pick-and-place machine, but you do not say how much work you are doing?

The centerpiece of my setup is a stereo microscope, like the kind jewelers use. Your hands gain superpowers. 0201 parts? Sure, I can do that.

Besides that, I use a motorized vacuum pen with syringe solder paste. Solder paste isn't tedious at all. Something like five seconds per pad. Your oven sounds great, but I found a $40 Milwakee heat gun to do reflow very well.


I actually do not know much about automated p&p, but isn't it 1) ridiculously expensive? and 2) require a considerable amount of set-up time to program the CAM file? If you're doing one-off boards, I'm not sure you'll actually save any time at all.

ai_dude
 
Posts: 21
Joined: Fri Dec 11, 2009 1:49 am

Re: The case for automated PnP

Post by ai_dude »

About 50 to 200 pcs of each board at a time, each with about 20-30 SMDs.
In the past year, I've placed an estimated 60,000 to 70,000 SMDs :shock: , and thankfully this number is growing.

It's not just a matter of speed, but that my time is better spent elsewhere, and outsourcing to CEMs has its own issues (not cost-effective for these relatively short runs, long lead times during the quote cycle, and having to float a lot of cash to run 6 months to 1 year of products at once, which is only doable with some products anyway). I've been looking to hire an SMD-capable solderer, but no luck for over a year now.

Nowadays I get backaches due to being hunched over a desk for these long periods. Planning on doing a test tonight of soldering on a tilted table (my standalone laptop desk) to see if this eases the back problems.

I've also used a reflow tool to solder boards, but that's a hobbyist process, and not the proper profile for commercial-grade products. I would be wary of doing this for pressure sensors and other profile-sensitive components.

Yes, a new machine would be ridiculously expensive, ($30K+), and not viable for what I do, but I don't need those extra-speedy modern machines, and given the state of our economy now, I've found used well-known commercial machines (Juki etc) with a 30-day warranty for $6000 to $9000. I'm also investigating the DIY PnP option also at this point. My calcs tell me $3000 is realistic, but I need to make sure I'll have the time for this. Until I completely figure out those details, I am tinkering with the option of getting a manual PnP to help alleviate my workload for now, hence my desire to compare speeds.

The PnP CAM file (component positions and orientations) is called a Centroid file, and my PCB program (Eagle) created this for me with a few clicks, so no prob there.

alex_dubinsky
 
Posts: 188
Joined: Wed Jul 30, 2008 5:17 pm

Re: The case for automated PnP

Post by alex_dubinsky »

How long would you say a 5,000 SMD run takes you to do by hand?

$6k to $9k is actually reasonable, given the volume you do. I am surprised you are suggesting building your own. I'd think the work for that will exceed the $5k savings. It's not just making an X-Y table. It's getting the picking and the placing to work flawlessly.

Btw, could you explain to me a bit about what a manual p&p does? Which part of the process does it automate?

ai_dude
 
Posts: 21
Joined: Fri Dec 11, 2009 1:49 am

Re: The case for automated PnP

Post by ai_dude »

Part of the reason for considering DIY is that I will get exactly what I want -- a compact benchtop machine. The compact ones aren't easily available on the used market. And I have a couple unique ideas that I'd love to prove would work. Building it will be fun, but time is an issue nowadays. But I also mentor a high-school robotics team, and we're not going to this season's competition, so some of the students want to help build the machine. The only thing that intimidates me here is vision alignment -- I'm an ex software developer, so I know I can do it, but really don't have time to start learning image processing, and my students won't be able to do this part. That would be necessary for really really fine pitch (0201's), but not sure if it's required for 0603's. I'm planning on running some tests with a vacuum pickup (syringe with vacuum pump) mounted on my CNC mill, to see how reliably it will pickup each time.

I did a time test recently, and got 240cph with tweezers w/soldering iron. But that's fully isolated from the rest of the world and I know I can't this for 8 hours straight, not even 4 hours. I'd say realistically over a longer period, I'd get 200cph average, and perhaps 3 hrs straight would be enough. That means 5000 SMDs would take me 25 hrs, and about 8 days.

How long would you take to do 5000 SMDs with vacuum pickup and reflow tool?

Manual PnP does not automate anything, but makes hand placement easier by keeping the pickup nozzle vertical on an X-Y arm system. And because it's supported, it's supposedly more stable, so easier, and hence quicker. There are holders for reels, tubes, etc, and an arm rest. Some brands have optional cameras, so the operator does not have to hunch over to see/align the components. But with the price of these new, I'd quicker go for a used fully automatic machine. See some examples here...
http://www.apsgold.com/pick-and-place/m ... ace-system
http://www.lpkf.com/_images/2601-lpkf-protoplace.jpg
http://madelltech.com/M2-1.html

freaklabs
 
Posts: 37
Joined: Wed Nov 04, 2009 7:42 pm

Re: The case for automated PnP

Post by freaklabs »

Just a warning. DIY PnP is not as easy as it seems. It requires the simultaneous control of motors and pneumatics and very fine resolution due to the small size of the parts. It really depends on how much time you're willing to spend on it. If you're doing 200 boards a month with 30 SMDs on it, you can justify it by the time and effort saved, the fact that your eyes won't go bad in the near future, and that you can increase your production capacity by a factor of about 10. If you do a DIY, you will definitely learn a lot about PnP, but probably gonna be rough to get exactly right.

scsi
 
Posts: 30
Joined: Sun Jan 10, 2010 9:09 pm

Re: The case for automated PnP

Post by scsi »

Do it yourself! Especially if you can build a decent gantry sysem or already have one. This way I'm not going to be alone in this business... :)

Here's I'm testing the machine with 0402 resistors at ~1000cph rate: http://www.youtube.com/watch?v=OAwdO_bTOTA

Software is ready-built by Madell - the only PPM manufacturer who was willing to sell their current software package to a stranger. It uses top and bottom vision for alignment and does so quite nicely. Yes, Madell's SX feeders do use weight peeling, so what? I don't mind moving those weights once every several hours and it reminds me that these weights is the most expensive part in SX feeders (read one dollar).

In short, their software is very DIY friendly and everything can be adjusted. You can build your XYZ table as large or as small as you want and still be able to use all their options.

The primary reason I went for the DIY option is because I was already building a fast and accurate CNC milling machine for plastic parts but then decided to add a PP function and found the software with vision that works. There is simply no room for two large machines in the garage I'm doing it in and the combination of functions was a must. The secondary reason was cost and finally - fun! :)

And in this video you can see how fast it can turn plastic into small chips after a short brake with the nozzle changer test: http://www.youtube.com/watch?v=bYfD0UUKyek

Regards,
Kyryl

freaklabs
 
Posts: 37
Joined: Wed Nov 04, 2009 7:42 pm

Re: The case for automated PnP

Post by freaklabs »

PnP looks good. Congratulations on getting everything to work together to place 0402s. Must not be easy.

scsi
 
Posts: 30
Joined: Sun Jan 10, 2010 9:09 pm

Re: The case for automated PnP

Post by scsi »

Thanks freaklabs! In fact the 0402's was easy once I ruled out the statics that caused the parts to pop prematurely from their sockets. According to Madell folks, the SX feeders work with 0201's no problem. The thing advances the tape and then the camera looks for the part that it is about to pick up. Works flawlesly.

ai_dude
 
Posts: 21
Joined: Fri Dec 11, 2009 1:49 am

Re: The case for automated PnP

Post by ai_dude »

Wow, very nice. I'm also in the middle of a DIY PnP build, and so far we (I have 3 robotics kids doing this with me as an internship) have the frame built, and this week we'll get the linear rails/shafts/bearings in place for the Y- and X-axes. I'm getting exited already.

Cheers,
-Neil.

scsi
 
Posts: 30
Joined: Sun Jan 10, 2010 9:09 pm

Re: The case for automated PnP

Post by scsi »

Very good Neil. Try to minimize the backlash by as much as possible in X and Y axes. Are you going to use ballscrews or leadscrews? Steppers or servo?

ai_dude
 
Posts: 21
Joined: Fri Dec 11, 2009 1:49 am

Re: The case for automated PnP

Post by ai_dude »

Neither... belts :) Individual belts on each side of the Y-axis. For the X-axis, I'm considering ballscrews though.

BTW, how much was the software? I was thinking software would be very easy to write, except for the vision part, so I may consider third-party software.

I have a CNC mill already, but thinking my PnP gantry system will be the base for a 3D printer eventually.

Cheers,
-Neil.

scsi
 
Posts: 30
Joined: Sun Jan 10, 2010 9:09 pm

Re: The case for automated PnP

Post by scsi »

Will the belts be driven by a common motor or independently? Belts are fast and all, tighten them and preload well. If two sides of the Y axis are driven independently then you have to deal with the skew that is inevitable. Not sure if Madell's software supports skew correction, but they can use two home marks to at least check for skew in Y axis. One of their larger machines has double motor option on the Y.

I too used to do some programming in the past and also though that it's got to be easy to write something like this and use some 3rd party vision library. Turns out a good library from Cognex or Dalsa with an SDK and stuff will run you well over $1K. Add some good frame grabber and you're at $2K minimum. Madell has sold me their software plus a grabber board for roughly the same amount. As you can guess, it has to be a particular frame grabber from a very reputable vendor with vision libraries and also a very particular motion controller. The latter might be familiar to you - DeskCNC. Thanks to this I can turn the machine from PnP into a CNC mill simply by starting a different program. I bought the DeskCNC software from IMService directly along with their motion controller. Madell then deducted the cost of the controller from their package price.

ai_dude
 
Posts: 21
Joined: Fri Dec 11, 2009 1:49 am

Re: The case for automated PnP

Post by ai_dude »

Hmmm...interesting.

One stepper, with a single long shaft driving both sides... http://tinyurl.com/yexcmg5

I was thinking that using linear encoders on the bed would get me very good accuracy, but that would not compensate for picking up parts non-centered. As an alternative, I may go with laser centering, but need to research that still.

FWIW, I know there are some open-source image-processing libraries, but have not researched that yet.

Cheers,
-Neil.

scsi
 
Posts: 30
Joined: Sun Jan 10, 2010 9:09 pm

Re: The case for automated PnP

Post by scsi »

Neil, I don't see how you are going to accomplish belt tensioning from looking at the sketch. You may want to add that later.

Also, with linear encoders you have to use a motion controller that has position control loop and uses the encoder feedback. Can do it all in software but the speed and reliability will be questionable. Most of servo drives will take the feedback and some have position control loops. Some can even take step/dir commands as an input but none will tolerate the flexibility of the belt in between the motor and the encoder. Keep that in mind if you are seriously looking into linear encoders.

In my case the gantry came with high-end linear motors and encoders with 0.1um resolution and is very accurate. Weights one ton (literally) with the stand it sits on. Both - weight and accuracy as well as speed are a total overkill for the PnP machine but that's how it's built. Originaly the gantry was set up for the high precision die bonding process and the sub-micron resolution was paramount.

In my opinion, vision based or laser based centering is a must in PnP machine unless you are looking into using some mechanical centering mechanisms.

What is your plan for the feeders?

Regards,
Kyryl

Locked
Please be positive and constructive with your questions and comments.

Return to “SMT (Surface Mount Tech)”