IP3 Fiducials.. How Many should I read?

Okay here is the deal. I am fairly new to the company here, but one thing I have noticed is that we read way to many fiducials, in my opinon anyway. One for example is this. One assembly was reading 17 fids. Smallest pitch was .25mil I say you only need two globals. I changed it to that, they still placed fine, yet everyone else in my department dissagress with me. I heard someone say something about reading locals for ALL fine pitch (.25 and smaller) within 2.5' radius around the part. I have never heard of that before. If I am right, (which I think I am) When should you use locals? If ever.. Please help give me any feedback you can.. thanks..

reply »

I have only used two global fiducials except for paneling. The only time I have ever used local fids is when I have a placement problem then I have activated the local fids for that specific part. Your best bet would be to ask Fuji America.

reply »

Provided your software supports board stretch and the board is stretched in a linear fashion, you should not need fiducials for each image. If you do elect to use local fiducials, Quad recommends a single fiducial. This method is faster and you can get theta from the global fiducials and global fiducials provide a better aspect ratio for theta calculations. Stretch in the board will probably have no effect on theta.

reply »

| Okay here is the deal. I am fairly new to the company here, but one thing I have noticed is that we read way to many fiducials, in my opinon anyway. One for example is this. One assembly was reading 17 fids. Smallest pitch was .25mil I say you only need two globals. I changed it to that, they still placed fine, yet everyone else in my department dissagress with me. I heard someone say something about reading locals for ALL fine pitch (.25 and smaller) within 2.5' radius around the part. I have never heard of that before. If I am right, (which I think I am) When should you use locals? If ever.. Please help give me any feedback you can.. thanks.. | Larry,

Ask five people how many fiducials to read and you'll get five different answers. Not to try to overly validate my opinion or anything, but this worked fine on a board populated with 16 10-mil pitch devices on it:

Read three globals. As far out on the corners of the board as possible. The three will give your machine X,Y, and theta offset. If you think your boards have non-linear stretch, read a few globals in between, to. I believe the IP-3 has mapping capabilities. I know the universal GSM does. But start with three globals.

If I were placing 25-mil on a maintained, calibrated IP3, I wouldn't even bother with locals. I'd go there for 20-mil and below. If the cal and maintenance isn't stellar, then I'd add a local, but not until after I compared the placements with and without that local.

People are hesitant to stop reading fiducials. They seem to provide some type of abstract safety margin that doesn't exist. People should actually be hestitant to add fiducial reads.

Every time you read a fiducial, you're taking away production time and adding yet another adjustment to your placement. I think that someone should justify making an adjustment before making it, rather than having to justify not making an adjustment that he really doesn't know whether or not he needs.

reply »

| Okay here is the deal. I am fairly new to the company here, but one thing I have noticed is that we read way to many fiducials, in my opinon anyway. One for example is this. One assembly was reading 17 fids. Smallest pitch was .25mil I say you only need two globals. I changed it to that, they still placed fine, yet everyone else in my department dissagress with me. I heard someone say something about reading locals for ALL fine pitch (.25 and smaller) within 2.5' radius around the part. I have never heard of that before. If I am right, (which I think I am) When should you use locals? If ever.. Please help give me any feedback you can.. thanks.. | A perfect case for GDT!

Earl Moon

reply »

| Okay here is the deal. I am fairly new to the company here, but one thing I have noticed is that we read way to many fiducials, in my opinon anyway. One for example is this. One assembly was reading 17 fids. Smallest pitch was .25mil I say you only need two globals. I changed it to that, they still placed fine, yet everyone else in my department dissagress with me. I heard someone say something about reading locals for ALL fine pitch (.25 and smaller) within 2.5' radius around the part. I have never heard of that before. If I am right, (which I think I am) When should you use locals? If ever.. Please help give me any feedback you can.. thanks.. | If your machine is calibrated your good to 0.025 mm with the 2048 line camera. otherwise 0.1 mm. Me, however, If there are no other constraining factors on the line I read two fids per image with no locals. I do have some boards which place multiple QFP 308's then I use two locals per device - especially on the longer fabs which tend to suffer stretch at a greater rate than smaller fabs. I guess the quality output should be your driving deteminant. Maby you could design a DOE on a board using two identical devices one with local fids the other without. Good luck.

reply »

| | Okay here is the deal. I am fairly new to the company here, but one thing I have noticed is that we read way to many fiducials, in my opinon anyway. One for example is this. One assembly was reading 17 fids. Smallest pitch was .25mil I say you only need two globals. I changed it to that, they still placed fine, yet everyone else in my department dissagress with me. I heard someone say something about reading locals for ALL fine pitch (.25 and smaller) within 2.5' radius around the part. I have never heard of that before. If I am right, (which I think I am) When should you use locals? If ever.. Please help give me any feedback you can.. thanks.. | | | If your machine is calibrated your good to 0.025 mm with the 2048 line camera. otherwise 0.1 mm. Me, however, If there are no other constraining factors on the line I read two fids per image with no locals. I do have some boards which place multiple QFP 308's then I use two locals per device - especially on the longer fabs which tend to suffer stretch at a greater rate than smaller fabs. | I guess the quality output should be your driving deteminant. | Maby you could design a DOE on a board using two identical devices one with local fids the other without. Good luck. | Boogey! Right on!

Don't we all often run a little test/expermiment to determine fiducial to component pad location accuracy? I mean, don't we, after downloading always "precise" off line programing (as an example), reference primary/global fiducial locations to any of a number of pad locations whether they are chip or fine pitch device types. I do - both for placement and printing. It's too easy not to do so.

After "calibrating" our cameras and other elements/tools as stencil to pad locations (see all the silver and assure cross hairs centered on selected pads) and placement accuracy on calibrated glass or ceramics, don't we see where the need arises for other fiducials. After all, except for mostly minimal board "stretch/shrinkage" there exists little need for more localized fiducials. Boards, stencils, and placement requirements are located as a function of "dumb" Gerber data. Of course, GDT would improve things from the design level.

Just a little (very little) heart felt X-Mas cheer.

Earl Moon

reply »