Printed Circuit Board Assembly & PCB Design SMT Electronics Assembly Manufacturing Forum

Printed Circuit Board Assembly & PCB Design Forum

SMT electronics assembly manufacturing forum.


Placement Program Tweaking

#17897

Placement Program Tweaking | 15 October, 2001

First, I�d like to kiss-up to the people that are likely to be responding to this question. Thank you for contributing to my recent query about placement program control.

OK. Let�s talk about �good practices� in placement programming. Throw in any "best practices" that you have, if you get ambitous.

Here�s the background, say we build our placement program and fiducial locations from CAD data, like we�re supposed to do. If things go bad, one of the following happens: * Everything is off position. * One [or very few] component[s] is [are] off position, but all others are on.

EVERYTHING IS OFF POSITION: Fix the machine (i.e., calibrate) and not the CAD locations, write a nonconformity report, etc.

ONE COMPONENT IS OFF POSITION: Our operators tend want to "tweak" their pick and place machines or their programs. We preach that this is wrong, but also recognize that when you release the dogs of war, thing don�t always get done according to Hoyle. On the other hand, �tweaking� is infinitely easier and operators live for the moment. So if they or others repeatedly see this same problem, it�s not a concern of this moment, but of a future moment. The problem becomes one they will deal with, if and when they have to deal with it.

So as a middle ground, we allow �tweaking� providing the operator writes a nonconformity report.

What do yall do?

reply »

Ken Bliss

#17901

Placement Program Tweaking | 15 October, 2001

Hi Dave, Although I am just a blathering salesman trying to sell carts, I am not offering any here. I do have a few thoughts on your CAD positioning issue and tweaking. We do have full CNC sheetmetal puching equipment in our facility and it operates basicly the same way a pick and place machine operates. We make it absolutly positively mandatory that any tweaking that needs to be done is done live. Meaning that when a programming error is found engineering is notified immediately and they must immediatly get with the operator and work together to modify or correct the program. What happens if you don't in our experience is someone forgets to change the master program (although the written order was issuesd).and the order is repeated the next day, week, or month and guess what, we have an error that looks just like it did last time. Using this method people get more focused on solving the problem in our experience and moving faster as they are creating down time on a very expensive machine. In our facility this correction is usually done in about 5-10 minutes max.

reply »

Ivan Rojas

#17902

Placement Program Tweaking | 15 October, 2001

Well Dave ,,, ,, first thing I should do ,, is check my CAD data , and make sure match with the program in the pick and place machine ,, right now I work with Siemens and GC-Place ,,, In my case it may be other things to look for ,,, if is everything is off ,, like I said I will look in GC-Place and see if it match with VMPS from My Siplace line ,, then look my Fiducial ,, if is close to %100 perfect ,,, if it is only one part ,, I will look for the GF and test it and make sure the pick up is clean and not touching the feeder magazine . And finally I will do the calibration. ( most of the time this happen because and operator make a mistake , like living a feeder open ) Also I can said that I work with MYDATA and Quad and all they work the same way . I hope this help you a little bet of what you are looking for . Ivan R.

reply »

Michael Trybula

#17904

Placement Program Tweaking | 16 October, 2001

At my previous job I inherited several lines where tweaking of a few components was allowed. The first thing I did was stop that practice. The problem with tweaking a program is that you are fixing the symptom and not the root cause. When the root cause is fixed the symptom is that the program appears to have shifted again. This means that operators will hate having maintenance done on their machines.

In order to control your process and monitor equipment performance accurately you must program from CAD data and keep the operators out of the program. You are right that everything off is usually a calibration issue. One or a few components off may be calibration (one head only) or a damaged nozzle (or similar items). Unless the problem is investigated and corrected immediately, more work is created because every changeover needs tweaking of the program.

When I first implemented this program the operators resisted. It took a little while to straighten out all the programs and remove the "tweaks". Once this was done and the machines had a proper maintenance schedule, the operators preffered this mode of operation becuase they knew the program would work right when loaded and not be dependent on who was the last operator to run the program.

If you allow tweaking you may even find that operators are making a "secret" copy of the program so that no other operator changes their program. This means that there is no control over programs and ECOs do not get onto the floor correctly.

reply »

#17905

Placement Program Tweaking | 16 October, 2001

Hi Dave,

tweaking, oh yes, we did like it to have that job done and not have people sitting around waiting for work or beeing able to deliver on schedule. Nonconformity report, ...ummmmh, weren�t that big that we thought we needed it, actually the tweaking was done by the same person that later looked for the real cause of that trouble. I tended to store those "golden boards" which we used to check the programming done with CAD-data for the first time and used them to look after the trouble when the job was done or the schedule did allow it. There may be many causes and you don�t have, in most cases, time to go through all procedures, just keep this thing running. It was a bit more difficult when you tweaked the machine data and not only the program data (mostly when all was off, the check had to be done before the next job, although setting up the new job and looking if tweaking (the same way?) would be necessary could tell a lot about were to focus on.

Just a tip, looking for dirt (especially all lenses and sensors) didn�t take much time and solved a lot of problems.

Wolfgang (the great tweaker)

reply »

Boca

#17912

Placement Program Tweaking | 16 October, 2001

When only a few components are off placement?

You'd like to say no tweaking of any kind, anywhere, by anybody (but me). But you're not there to run to their rescue all hours of the day and nite (almost but not quite), and getting product out the door is what pay the bills, sooooo .... If a technician is available, hav'em check pickup and so on, as someone else mentioned. But let them tweak the machine copy if they need to, to get product moving, and then submit the change request for the program.

Do NOT let them save the machine copy over the original copy. Bin there done that. So have others commenting here have also.

Put some tweaking guidelines in place. Such as the sequence of calling floor support (giv'em a chance at it) who can do them, and how many they can do.

Boca

reply »

#17923

Placement Program Tweaking | 17 October, 2001

Dave, Like everyone else,I don't generally allow it. If only one component is off and that head is doing another component of the same type with no probs, ( pick-up is OK ) then I just say "don't ask me - move the bloody thing". I usually find it's just a case of - sh_t in: sh_t out. My protel handler only deals with pads so if an engineer makes up a component that consists of pads/tracks/ fills etc it will throw out a wrong co-ordinate. One trick I always use is to make the first two placements in every program the same as the fiducial data picked up from a dummy feeder. These two placements are generally skipped. If I do have a problem then I unskip them and do a fiducially corrected scan. If it reads the fids ok at fid correction but then doesn't go exactly to the centre of the marks that are your two first mount data then your fid data is incorrect and so will everything else will be out. You can also use this method to do a very quick head offset check. Do the fid correction and tell the head you want head N to do the dry placement run (on the dummy placement fids ) if your camera scan has told you the fid data is OK. Eventually I will convince the world that the first two mount data should be your fid data but skipped. You can also check if there is any difference between the two. If there is then someone has changed the fid data, probably because they were too lazy to change the data origin to get the fids right in the middle of the camera/sensor. Bad set up also plays a part. BUT if I have to I'll tweak and not save. I'm lucky that I generate all my own data and don't inherit any from customers, - "but it ran perfectly at shopX". To those of you do, my commiserations. Darby

reply »

#17963

Placement Program Tweaking | 19 October, 2001

I used to love tweaking. I tweaked and tweaked until I got a master degree in tweaking. But than I had to teach others how to program a board and avoid tweaking. After all it does not go quite conform with ISO and other quality approaches. I investigated more time to avoid the root causes of tweaking. Meanwhile, and this is not a sales pitch, my students get a board right in the first run at least nine out of ten times. Therefor, I may spend an hour or so to �massage� the CAD data in an Excel spreadsheet until it fits my preference. Most of the machine programs can rotate or mirror image the placement positions until two hands are not enough to explain in which direction you have to move a position if you want to change it. The board designer looks through the board from the top side to define the positions on the bottom side, but I ( my brain ) can not do it. Then he puts the datum in a registration hole. It looks very nice on his monitor with a nice white cross in the dark center of the hole, but when I try to measure from there to a placement position I can not align my caliper on the hole. I always put the datum on the bottom left side of the board edge, even when I have to create some extra X and Y columns in my spread sheet and even when the edge is notched out. Note, all coordinates are positive. Of course the fiducial positions have to be shifted by the same amount. Now check again with the caliper and measure from the board edges to the fiducial. Check one or two placement positions, if it matches with the corrected coordinates. Check the placement angles for IC�s and tantalums. Not right? Get another column in your spreadsheet. When you are done with this data preparation check the time. How much was it? How long do you need for tweaking in comparison? How close do you get to the initial set up cost of $ 300 you quoted the customer? NO MORE TWEAKING!!!!

reply »

#18043

Placement Program Tweaking | 31 October, 2001

In theory, no tweaking is necessary. Unless you happen to work with equipment old enough to vote! We have several lines with old MK chip shooters. of course, we use the same program for all the lines. There is no way to calibrate each machine so that 1 program will give perfect placement no matter which machine it is loaded in. So the operator is expected to tweak the program or the offset because the program he is using was on machine A last week and now he uses the same program on machine B, which places component slightly differently.

When will there be machine smart enough to detect and correct automatically any placement offset?

reply »


CAL

#18059

Placement Program Tweaking | 1 November, 2001

DAVE- Sorry for the late response.... as I was outta da country fer a bit. Wolfgang has a great point...If your machine has linear scales or encoders and they are dirty (even a spec of dust) this could have a huge impact on the positional data (esp. on higher resolution machines). Dirty scales coupled with a servo motor controller that is not "tuned" correctly could cause problems. Does the "off part(s)"happen consistently at the same spot? If so check hardware. Does the "off Part(s)" happen consistently with the same head? If so check hardware. Is the "off part(s)" consistent in any one direction? I assume you are using two fiducial marks in opposite corners, can you add 1 or two more fiducials to see if the placements gets any better? Check the components.. If the component lead termination is not symmetrical, thus causing a false or in correct size value, the vision system can off-set this "false" value adding it to the placement data having it placed incorrectly (Esp. if your acceptance tolerance window is large). Hope I relayed that correctly.

If you have not gathered I am not fond of tweaking. If a machine is spec'd at -/+ 90um @4 sigma it should perform at this level.

Hope this helps, Cal

reply »

Dave

#18062

Placement Program Tweaking | 1 November, 2001

Ditto for what most people have said. If you have good equipment/ good data / 3 or more fiducals..everything should be OK. Maybe dirt or something got bang'd. tweaking is evil,but if nothing else works??? dave

reply »

#18066

Placement Program Tweaking | 1 November, 2001

I disagree on the dirty glass scale offset theory. If you have a dirt spot on the scale, every time the axis comes over the spot it misses counts. After a few movements it will not find the components in the feeder anymore, because the failure adds up. I do agree that contaminated components or fiducials can lead to an offset, nevertheless most of the problems can be foreseen without tweaking.

reply »

#18079

Placement Program Tweaking | 2 November, 2001

Here's my 2 cents.......

I worked for a small company that manufactures pick and place machines and had the task of testing/troubleshooting and the dreaded calibration. These machines used vision for part orentation and have what I considered awesome encoders so I would expect that part placement would have been perfect every time. Not so....

We initially were placing real QFP's on a board covered with double sided tape and would place the part, make a visual inspection, make a correction, put the part back in the tray and do it again. This went on for every head/spindle at various rotations all day long. Sometimes into the next day. Oh yes, it was fun to tweek, as long as the machine responded and placed that part beautifully right through final inspection. But what to do when things didn't work out?

The first thing I looked at was the QFP. We would run several different sized ones once we got the first one to place correctly and I believe if the parts were perfect they would have placed the same all the time. Bent leads due to their manufacture or operator handling were a big killer. Glass slugs were next tried, but after the first couple broke cost became an issue. We eventually started using "perfect" parts made out of plastic and milar. They were cheap and no bent leads.

The next thing was the test board. Double sided tape isn't the greatest thing to hold parts onto a board and I found that parts would actually move and rotate after a while. An airisol spray can also be used but after a while more parts will stick to your hands then the board. I also found that all boards are not created equal. They call it board streach and skew, and if I had it my way they would have to put local fiducials for every part from 1206's to fine pitch. I stuck with one "perfect" board until it was so beat up that fidutials were not detectable. I also retaped the board a couple times a day.

The next thing that was an issue for me was the conveyor system. You can have awesome placements and then the board is transported to the oven and it is violently dumped from one machine to another. I think of it as someone shipping glassware and writing "Fragile" on the box. This word is motivation to a handler to throw it harder. What I'm getting at is solder paste and glue won't hold a part over a poorly aligned conveyor system.

The last thing I question is the oven. Parts move big time. Normally it's a good move as the solder "sucks" the leads over the pads.

So in defense of the machine if all outside varibles could be kept consistant tweeking would be a thing of the past. Tweeking may not be a good thing for the people who maintain the machine but for those responsible for pushing out product it keeps them employed. So I say tweek away but be responsible.

reply »


CAL

#18082

Placement Program Tweaking | 2 November, 2001

Ahhhh America.... It is great to agree to disagree.

reply »

ii-feed SMD Intelligent Feeder

ii-feed SMT Intelligent Feeder