Scaling artwork basics: registration errors can be overcome by having precise, targeted, real-time data.
In our company, we deal with so many exotic materials and unusual constructions, the usual approaches to registration --especially scaling artwork--don't apply. Even the materials themselves are completely different from one batch to another.
There are two distinct and unrelated categories of registration error: size and shift. Size is the final size of a specific layer/core/panel after fabrication is complete. Size is affected by several factors, especially shrinkage (or expansion) of the core/prepreg materials during etch and lamination. It is controlled by scaling the artwork to compensate for this anticipated shrinkage (expansion). Such scaling is typically done on a core-specific basis; that is, for a given core, such as Apco polyimide, 0.004", 1 over 1, with 2113 glass, there will be a specific scale factor.
That approach to scaling is not predictable, however, because it ignores other variables in the panel construction that affect shrinkage/growth. In the example above, the core will have completely different shrink/growth characteristics if assembled with different prepregs and cores, different copper geometries, even different press cycles, when used on other jobs; another reason why scale factors will be different from one factory to the next, even when the construction of the panels is the same.
Shift error relates to the position/location of a layer within the panel, relative to the nominal/correct position. These errors are related to tooling, and they occur primarily in three areas of the fabrication process.
1. Front-to-back errors on a single core, for example layers 2 3 or 6-7, etc., occur during dry-film imaging or direct imaging (yes, direct imaging also has its limitations). Causes include inaccurate or defective film targets, inaccurate film punching/taping, loose hinges in glass frames, excessive force used in setup, etc.
2. Core-to-core errors occur in lamination. Causes include inaccurate post-etch punching, inaccurate post-etch punch targets, excessive force used in stackup, loose tooling pins/rivets, uneven press pressures, etc.
3. Panel-to-panel errors occur at the drill. Causes include dull drills, worn spindles, excessively high aspect ratios, incorrect zero-reference, inaccurate post lamination x-ray tooling, etc.
The aforementioned shift/tooling problems can be further separated into a) repetitive/setup errors, which can usually be corrected digitally; and b) random from front-to-back, core-to-core and panel-to-panel, usually caused by worn tooling or operator error.
Solving registration problems can be complex. The key is having precise, targeted, real-time data on which engineers can base good decisions.
FIGURE 1 shows the success a Japanese fabricator achieved in controlling registration on a 42-layer panel, using 0.0016" cores and prepregs. For Figure 1, the thickness was 0.080", hole diameter was 0.012", aspect ratio was 6.4:1. The worst-case shift was 0.0016". Using special software that captures scaling data, the fabricator, OK Print, was able to produce this PCB and hold worst-case true position error anyplace on the panel to under 0.002".
[FIGURE 1 OMITTED]
There is no myth to managing registration on complex constructions, laser vias, additive process, sequential lamination and rigid-flex PCBs when precise, targeted data are available.
TOM R. PAUR is president of PerfecTest (perfectest.com). He can be reached at firstname.lastname@example.org.
|Printer friendly Cite/link Email Feedback|
|Author:||Paur, Tom R.|
|Publication:||Printed Circuit Design & Manufacture|
|Date:||Feb 1, 2004|
|Previous Article:||The roller coaster of cyclicality: want to see the future? Start by using 12/12 rate of growth curves.|
|Next Article:||'One source' fits all: modeled as a "mini Flextronics," this group uses the latest equipment and the engineering practices for its competitive...|