Desktop Pick-and-place: overview

[rx8pilot]

Automated P&P for 1-off prototypes is rarely useful unless there are a large number of relatively few parts.
The time taken to set everything up will usually be way longer than to manually place, once you have a good manual system.
It certainly isn't useful enough for it to be worth anyone selling a machine aimed at 1-offs.

Manually placing with a foot-operated vacuum pen can be pretty quick - couple of seconds per part when placing parts from the same strip on a board you're familiar enough with to know where they go without looking it up.

Think about what you need to to when manually placing a board, up to the point where you actually place a part - find the right part, take parts out of tapes, make sure it's the right way round etc.   
For an automated system you not only need to do that, but them tell the system where it goes, and put it somewhere it can pick it up.

For more complex PCBs, a small P&P can be useful for placing things that there are lots of - decoupling caps, pullup resistors, LED arrays etc. but for anything there's only a few of, doing it manually is going to be much quicker, with a little practice.

I have done the whole range from manual to P&P - this is generally a pretty good description of reality. Before I had a P&P, I used a variety of methods to get parts on the boards but spent a considerable amount of time getting the manual process about as fast as it could possibly be. As it turns out - after you get a system worked out and dialed in - you can populate even some fairly complex boards faster than you can even get the feeders loaded on a P&P. After that, you would still have to program, tweak, run a double-stick tape pass, etc. For me, the biggest improvement was how I organized the parts ahead of time and how I set them up for the manual assembly. The first few boards felt like slow painful death. After some time, it was not so bad. As my little operation grew, I had days where I was placing parts 10-12 hours straight which at an average of 400-500 parts per hour on a REALLY good day (you have to include breaks).

The pick and place system initially made it way worse - especially when I needed to do 2-3pcs of 5 different designs. It probably took 4x the time at least because I was always fiddling with something. After getting the P&P workflow figured out including the myriad of quirks that every system has - things sped up for the small batches, but not by very much. The biggest improvement by far was my own personal sanity. It is easier to deal with a P&P process all day than picking up tiny parts repetitively for 10-12 hours. I still, of course, wanted more speed which required more process development and more feeders. My current setup represents a lot of effort in process development to achieve a useful assembly throughput. One critical element for me was to have a LOT of feeders so that I can have 4-5 jobs loaded and ready to roll at any given moment.

The last new design I did (last week) was the first time I attempted P&P assembly of a prototype - it was probably about the same effort as doing it manually. The good news is that the prototype most likely will only have a handful of minor changes before I run 50 of them. The machine is 98% ready now for the final design. If I only did prototypes - my P&P would be a total waste of time. If I only had a handful of feeders, it would be a waste of time. The only reason it is a good thing is becuase I put a TON of effort into getting it all worked out an have a lot of feeders that afford some flexibility and effeciency. Also, I got a used commercial system - a fixer upper. Although it has some years on it - it was designed to run 24/7 and is quite consistant. After I figured out its personality - it just works even with BGA's and 0201's. Big capacitors, big inductors, trays, cut tape, tubes, etc - no problem.

[jmelson]

I have done the whole range from manual to P&P - this is generally a pretty good description of reality. Before I had a P&P, I used a variety of methods to get parts on the boards but spent a considerable amount of time getting the manual process about as fast as it could possibly be. As it turns out - after you get a system worked out and dialed in - you can populate even some fairly complex boards faster than you can even get the feeders loaded on a P&P. After that, you would still have to program, tweak, run a double-stick tape pass, etc.

I have a big P&P machine here (Philips CSM84).  i never "tweak" positions, and only use the double stick tape when making calibration adjustments to the machine.  I just did a special job for a customer that was only 4 boards, but they had about 300 parts/board.  It did take me about an hour to set up the feeders and load them on the machine.  But, after that, I cranked out each board in about 6 minutes, including the paste stencil application.  No errors, no possibility of parts in the wrong place, etc.  These were boards I'd made before, so I had the solder stencils, P&P program and all ready to go.

I also don't manually program the machine.  I wrote a C program to read in the placement file from the CAD system and write out the file for the P&P machine.  All I have to do is decide what part to put in which feeder location.

Now, for one-offs, I would not use the P&P machine, but for even 4 modestly complex boards, there's no question let the machine do it!

Jon

[rx8pilot]

I did not mean that I tweak the positions, just all the other things. Like I got some tapes where the parts were rotated 90deg from the last few times I ran the same parts. There are other little things, that need checking and small changes. Any new parts that have not yet been defined have to be entered. So far, in 18 months, I have not yet set up a new design that did not have something that needed some attention.

In relative terms, I am new to P&P and still have a lot to learn. So far, I have figured out how to automate the programming so I never have to do any of that manual. I also (recently) synchronized our internal part numbers with the CAD system so that it is a few keystrokes to get from design to a programmed machine. If faced with 4 boards that are 300 parts - I would certainly setup the machine for it. Not sure how many unique parts you are talking about, but I doubt I could get the machine placement ready in an hour. By the time I find/organize parts, load a feeder and verify with the BOM, theta noted, etc....it seems to be about 5mins per part (loading the feeder is maybe 90 seconds). Then I get them in the machine, and verify that I did not swap the 100nf and 10nf caps, which of course is an easy mistake. So 20 parts, programming, checking, verify - maybe 2-3hrs at my current speed for the 1st board.

[jmelson]

In relative terms, I am new to P&P and still have a lot to learn. So far, I have figured out how to automate the programming so I never have to do any of that manual. I also (recently) synchronized our internal part numbers with the CAD system so that it is a few keystrokes to get from design to a programmed machine. If faced with 4 boards that are 300 parts - I would certainly setup the machine for it. Not sure how many unique parts you are talking about, but I doubt I could get the machine placement ready in an hour. By the time I find/organize parts, load a feeder and verify with the BOM, theta noted, etc....it seems to be about 5mins per part (loading the feeder is maybe 90 seconds). Then I get them in the machine, and verify that I did not swap the 100nf and 10nf caps, which of course is an easy mistake. So 20 parts, programming, checking, verify - maybe 2-3hrs at my current speed for the 1st board.

I use a lot of the same parts on different boards (currently, I am making about 15 different boards).  So, most of the part orientation is built into my program that converts the file formats, and the list of parts that is fed to it.

I've been doing this since 2007, and learning how to be more efficient about it.
The only part that I ever had to change orientation on was when switching to a different 4-pin optocoupler that was put in the tape 180 different.  otherwise, there are JEDEC standards for part orientation in tapes, so almost everybody always does it the same way.  (Sometimes I change between parts in tubes and parts on tape, and that entails a 90 degree change in the program.  But, everything seems to be moving toward tapes, now.)

Jon

[mikeselectricstuff]

IME probably the biggest time-saver is to make sure that your PCB parts library is set up so that part names and, in particular, rotations relative to the tape, are consistent with what the P&P machine uses.
That way you only need to verify rotations the first time you use a part - after that you know it will be the right way round.

Then write some simple software that takes the PCB software's P&P file and converts it into the P&P machine's native format. Add fiducials as components with pick/place points, so your import utility can include the board references in the P&P file.

You want to have different part types for the same size of chip resistors, caps, polyfuses, LEDs etc., as the different parts of the same nominal type may have different vision requirements.

Try to minimise the BOM  as much as possible to reduce feeders.

Aside from loading parts into feeders, my job setup time is typically 5-10 minutes max. The thing that takes the most time is setting up a new part that needs the vision parameters setting up - if there's only a few new parts I'd usually hand-place them.


 

[rx8pilot]

On my last design, I did a lot of that legwork Mike just described. Really looking forward to my next design to see how well it works. To make things a bit more challenging, I am transitioning to 0402/0201 from 0805/0603. My remaining stock of old designs use big parts and all the new use the little stuff. Lots of feeder changes from one size to another, but only a limited number left.

Can hardly wait til I have flushed all the old boards.

Sent from my SM-G892A using Tapatalk

[mrpackethead]

I concur with mikes comments.     I actually use our PNP lines for prototypes yes, one offs.. we've got it down to the point that unless it was just a handful of parts then it just not worth it.   This is for a couple of reasons.

(a) our desings are using parts that are allready on the machines..   Weve been really careful about pcking parts that are all ready there.

(b) our librarys partnames etc all match what are already on the PNP line.

(c) programmign the machines is easy, just pass the PNP files from atltium to some python code ( via a web server ) it splits it top and bottom, and also between machines ( we have two machines inline )..   

(d) we have a really strict way about how we setup our pcb panels with tooling, and fids and all that stuff.   

On our machines theres a couple of manual changes on the rails that need to be made to swap between boards.

Our system will also tell me if there is enough stock to run the job, and where additional parts are ( if we have them in stock )..   And for jobs that have one offs where we will only have a few parts for a one off, then we can DNP it really easily )..

its taken a LOT of time and effort to get here, but its paid off.  But the efficiency gain only occurs when to start right at design.

The next task is to interface the serial ports of the the machiens via a terminal server, so we can program them 'online'.. 

[rx8pilot]

its taken a LOT of time and effort to get here, but its paid off.  But the efficiency gain only occurs when to start right at design.

This comment, in my opinion, is the big takeaway.

For P&P to be fast for small batches and prototypes - it is possible but not easy for someone that just got a machine. I feel like I still have quite a bit of effort to get everything all worked out to the point where I am quick. We have quite a bit of work coming up and each board design is an opportunity to push the process to a higher level. The good news is that each design I run through the system is faster than the previous.

Like @mrpackethead said - efficiency starts at the beginning of the design. A part numbering scheme has bee critical for us. The scheme defines the way the parts are stored, ordered, filtered, etc...and each system in your system has to compatible with that part number scheme. When I sort my BOM by part number - it naturally groups the parts by category, size, value, etc that matches the way they are stored. This makes it far easier to pick parts to be loaded in the machine. When I add a new part to the design, I can see what is 'normally' in the P&P and try to make it work. Whatever final component is chosen - the part number pushes all the way through to the BOM, P&P, DigiKey, inventory, and accounting. We just recently got all this organized so still waiting to see how much faster we get from design lock to P&P setup.

[mrpackethead]

Something that is helping me, but i hate doing is using checklists.   theres enough stuff to do, and it has to be done in the right order as well. 

its taken a LOT of time and effort to get here, but its paid off.  But the efficiency gain only occurs when to start right at design.

This comment, in my opinion, is the big takeaway.

For P&P to be fast for small batches and prototypes - it is possible but not easy for someone that just got a machine. I feel like I still have quite a bit of effort to get everything all worked out to the point where I am quick. We have quite a bit of work coming up and each board design is an opportunity to push the process to a higher level. The good news is that each design I run through the system is faster than the previous.

Like @mrpackethead said - efficiency starts at the beginning of the design. A part numbering scheme has bee critical for us. The scheme defines the way the parts are stored, ordered, filtered, etc...and each system in your system has to compatible with that part number scheme. When I sort my BOM by part number - it naturally groups the parts by category, size, value, etc that matches the way they are stored. This makes it far easier to pick parts to be loaded in the machine. When I add a new part to the design, I can see what is 'normally' in the P&P and try to make it work. Whatever final component is chosen - the part number pushes all the way through to the BOM, P&P, DigiKey, inventory, and accounting. We just recently got all this organized so still waiting to see how much faster we get from design lock to P&P setup.

[rx8pilot]

Something that is helping me, but i hate doing is using checklists.   theres enough stuff to do, and it has to be done in the right order as well. 

As a pilot - checklists have become a normal in my world. For the most part, there is no individual task that is difficult, but if you forget about it - things go south quickly. Flipping the gear lever just before landing is easy - getting distracted during approach and landing is even easier.

[mikeselectricstuff]

Something that is helping me, but i hate doing is using checklists.   theres enough stuff to do, and it has to be done in the right order as well. 

As a pilot - checklists have become a normal in my world. For the most part, there is no individual task that is difficult, but if you forget about it - things go south quickly. Flipping the gear lever just before landing is easy - getting distracted during approach and landing is even easier.

..like forgetting to put fids on both sides of a PCB that's populated on both sides.... just did that last week!

[MUDGEL]

As a beginner hobbyist who wants to run before he can even crawl, this has been a really revealing discussion. In essence, I would love to be getting prototyping work from others to process on my own equipment. Ive just realised how far ahead of myself I am to even be considering this. Thanks again for enlightening me.

[jmelson]

..like forgetting to put fids on both sides of a PCB that's populated on both sides.... just did that last week!

My CSM84 can use either traditional solid fiducials or a PTH.  I still have some boards that use the mounting holes as fiducials.  I don't see much difference in placement accuracy.

Jon

[mrpackethead]

yes, all sorts of thigns are possible, but its easier to put them in there.  My processing scripts look for some FIDS with very specifically named component names.  ( FIDBOT1, etc etc ), which it then rips out and automatically creates the fiducial info.

..like forgetting to put fids on both sides of a PCB that's populated on both sides.... just did that last week!

My CSM84 can use either traditional solid fiducials or a PTH.  I still have some boards that use the mounting holes as fiducials.  I don't see much difference in placement accuracy.

Jon