Outer-Layer Imaging: Where LDI and Film Diverge on Fine Lines

Let us frame it correctly: outer-layer fine-line capability is not set by etching. It is set by whichever link in the chain bottoms out first. The chain runs: how the pattern is written into the resist (exposure), how fine the dry film can resolve, how much pattern plating widens the line, and how much etching eats off the sides. The shop floor verdict is often "etching cannot hold it", while the link that actually bottomed out sits further upstream.

Exposure method — film contact printing versus LDI direct imaging — is the entrance to that chain. It determines how much margin the other three links still have to spend. This article works through the accuracy budget of each method, adds the two constraints outer layers carry beyond inner layers, and sets out when changing exposure method actually buys anything.

Sunking PCB publishes trace capability in tiers: 4/4 mil standard, 3/3 mil advanced, 2/2 mil R&D, with trace width tolerance at ±20%. What follows explains why exposure method shifts from optional to mandatory as you move from the standard tier toward advanced and R&D.

1. Outer layers are not the same problem as inner layers

Inner-layer exposure works on a core: dry film on copper, expose, develop, etch, done in one forming step. The central tension inner-layer is layer-to-layer registration — every layer's pattern must line up after lamination, and the difficulty lives in scaling compensation and the target system.

Outer layers add a step that changes the geometry of the line itself: pattern plating. The pattern is imaged first, the areas that will keep copper are opened up, copper and hole copper are plated to thickness, tin goes on as an etch resist, the resist is stripped, and the original base copper is etched away.

Two constraints follow that inner layers never face:

1. The dry film has to survive plating. It is not only an etch mask; it is also the sidewall of the plating channel. The narrower and taller those resist walls get — and they get taller because they must contain the plated copper thickness — the more readily they deform, lean over, or leave residue at the base during developing and plating.
2. Etching removes base copper, but the line has already been plated up. The etchant works downward and sideways at once, so the sidewall acquires a slope and the top and bottom widths differ. The data has to carry the line slightly wide on purpose. That is etch compensation.

So outer-layer imaging is not "print the artwork accurately". It is "write an artwork that downstream processes will deliberately deform, pre-shaped by the predicted amount of that deformation." Exposure owns only the first step, but its error gets amplified by everything after it.

2. The accuracy budget of film contact exposure

Film (photo tool) contact exposure is mature, inexpensive and fast in volume. Its error sources can be itemised.

Dimensional stability of the film itself. Film is an organic substrate and its dimensions move with temperature and humidity. However well the room is controlled, the film changes size between plotting, fixing, storage and use. This error scales linearly with panel size — the same relative distortion becomes a larger absolute deviation on a larger panel. Sunking's standard maximum panel is 600×570 mm with 610×1100 mm in the advanced tier, and the long axis absorbs most of this cost.

The contact gap. Contact exposure assumes film and dry film are pressed together; in practice there is always some air, dust, panel waviness and copper surface roughness in between. Any gap gives light a diffusion path, line edges blur, and edge acuity on fine features drops directly. Vacuum draw-down compresses this gap but does not remove it.

Stacked registration on two sides. Both outer sides register independently and both must align to the inner layers. The film route needs the operator or a semi-automatic system to bring two films onto the panel targets, and the errors of the two sheets add.

Wear on the tool. Film abrades, scratches, and picks up debris. Once a sheet has run some number of panels, the defects it creates show up at the same location panel after panel — which is also a useful diagnostic signature: defects that repeat at a fixed location across panels point to the film or the targets, not to exposure parameters.

Summed, these errors are entirely adequate at coarser widths. The problem is that most of them do not shrink as the line shrinks. Panel size, room conditions and the contact gap stay the same order of magnitude while the trace gets finer, so the relative margin disappears quickly on fine lines.

3. What LDI removes, and what it takes on instead

LDI (Laser Direct Imaging, also called direct write or maskless exposure) scans a laser beam across the dry film from data, with no physical mask.

What it removes is clean:

• No film, therefore no film scaling with temperature and humidity, no wear, no debris, no tool storage and control.
• No contact gap, so no diffusion path from contact printing and a steeper line edge.
• No film plotting and no manual alignment step. A revision is a data change, not a new photo tool. On low-volume work and on designs that revise repeatedly, this is often a bigger win than the accuracy itself.

What it takes on is equally concrete:

• Scaling compensation moves from "estimate a value when plotting the film" to "measure it before exposure." This is LDI's main accuracy advantage: compensation is adjustable per panel or per batch. Measure the targets, derive the actual distortion of this panel, then write the data stretched by the measured amount. It does require a sound target system, enough measurement points, and a compensation model matched to the material.
• Registration accuracy transfers to the scanning system. The machine has to know where the panel is and how far it has skewed, then correct the scan path. That is a machine-and-calibration capability rather than an operator-skill capability.
• Imaging quality depends on data-to-resist matching. The laser drives a reaction in the dry film by delivered energy, so the resist's sensitivity, spectral response and thickness have to suit the source. Changing dry film type usually means re-establishing exposure energy.
• Throughput behaves differently. Direct imaging is a scan with a per-panel exposure time; contact exposure covers the whole panel at once. The cost models differ, and the larger the batch and the fewer the revisions, the stronger film's cost advantage.

One sentence on the division of labour: film puts the accuracy burden on consumables and manual alignment; direct imaging puts it on measurement and modelling. Fine lines want the latter, because consumable error cannot be corrected panel by panel while measured error can.

4. Dry film: perfect exposure is wasted if the resist cannot resolve

Dry film photoresist carries the result of exposure. On fine lines it has two independent failure modes that pull in opposite directions.

The resist walls will not stand up. In a pattern-plating flow, the resist that remains after developing defines the areas that will later be etched away, and the plating channels sit between those walls. The narrower and taller those walls become, the more easily they lean over during developing rinse or deform under plating solution flow. A leaning wall closes a channel, which reports as a short or as insufficient spacing.

The channel floor does not develop clean. Developer exchange in a narrow channel is poor and unreacted resist can remain at the base. Residue blocks plating, which reports as discontinuous deposit, undersized trace width, or an open.

Both modes point at one conclusion: thicker dry film is not safer, and thinner is not better either. It has to satisfy two conflicting requirements at once — contain the copper thickness that pattern plating will deposit, and still hold an acceptable aspect ratio on fine features. Higher copper requirements (heavy copper, high current, hole copper on high-order HDI) push the resist thicker, and thicker resist has a worse aspect ratio on fine lines. That is one of the root reasons outer-layer fine lines and heavy copper are hard to have together, and why minimum trace width on heavy-copper boards generally does not reach what conventional boards achieve.

Exposure method matters here indirectly but genuinely: direct imaging produces steeper edges and a more controlled energy profile, so the developed resist sidewall sits closer to vertical and the same resist thickness holds a finer feature. Film contact exposure carries a transition zone at the edge from the contact gap, giving a more sloped sidewall after development and less usable margin on fine lines.

5. Pattern plating and side etch: the line changes shape twice more

What gets written is not the final line. It is reshaped twice downstream.

First, pattern plating widens it. Copper grows in the resist channel and does not only grow upward; near the top of the resist it spreads laterally. Uneven current density also makes the deposited thickness vary across the panel — edge versus centre, near large copper areas versus around isolated fine traces. This is exactly why plating uniformity feeds directly into trace width tolerance. We covered the release criteria in PCB electroplating quality inspection and release; here we only take its conclusion: the spread in plated thickness is part of the spread in trace width.

Second, etching narrows it and gives it a slope. While the etchant bites down through the base copper it also bites sideways, so the cross-section is a trapezoid rather than a rectangle. The finer the line and the thicker the copper, the larger that slope is as a fraction of the width. The remedy is to carry the line wide in data by a compensation amount so it lands on target after etching. On heavy copper that compensation is substantial; the review method is in heavy copper etch compensation review, and this article does not repeat it.

Reading these two deformations together with exposure explains a common puzzle: why the same data on the same exposure tool delivers different trace widths at different copper thicknesses. Exposure only guarantees that the back-calculated artwork gets written accurately; the artwork itself is back-calculated from copper thickness, plating scheme and etch compensation. Change the copper and the back-calculation changes.

This is also what ±20% trace width tolerance actually means. It is not the accuracy of exposure. It is the delivered band after the spreads from exposure, pattern plating and etching stack up.

6. When changing exposure method actually buys something

New equipment is not a universal answer. The question is which link bottoms out first.

Cases where moving to direct imaging clearly pays:

• Widths at 3/3 mil and finer, especially R&D-tier 2/2 mil. In this range film's fixed errors — scaling and contact gap — take too large a share of the budget.
• Large panels. Scaling error grows with size, and per-panel measured compensation targets exactly that.
• HDI outer layers with small capture pads. Microvia-to-pad margin is tight, so registration error carries more weight. Sunking runs HDI to 5 stages (5+N+5) with laser-drilled minimum holes of 0.1 mm standard and 0.075 mm advanced; the design inputs for that class of board are in HDI build-up copper and plating design inputs.
• Frequent revisions at low volume. The gain is cycle time rather than accuracy, from skipping photo tool generation.
• Mixed-dielectric stackups and materials with unstable scaling. Measuring per batch beats a frozen film value.

Cases where the gain is limited:

• 4/4 mil on a modest panel. Film still has adequate margin here, and the bottleneck is usually dry film, plating uniformity or etching — changing exposure does not address it.
• Fine lines on heavy copper. What bottoms out first is resist aspect ratio and side-etch compensation, not exposure resolution. The copper-thickness-and-width combination has to be validated as feasible first.
• Yield problems that look random in position and correlate with batch. That pattern suggests material, pre-treatment, developer or etchant variation rather than imaging method.

A workable triage rule is to read the spatial distribution of the defect:

• Defect repeats at a fixed location across panels → check the photo tool, the targets and the data. Imaging side.
• Defect follows the panel edge or clusters in a region → check scaling compensation and plating current distribution.
• Defect correlates with line width and not with location (all fine lines poor, coarse lines fine) → check resist resolution and etch compensation, not exposure registration.
• Defect drifts between batches → check material scaling, developer and etchant condition.

7. Where outer-layer imaging closes the loop

Imaging results have to be measurable, or the compensation model has nothing to iterate on.

Trace width measurement. Measure actual width after the outer-layer pattern is complete and compare it to the data target. The difference is the current bias of the compensation model. Group that number by location on the panel instead of taking a single average — location-dependent spread points to plating and scaling, while a location-independent bias points to the compensation value itself.

Outer-layer pattern inspection. Opens, shorts, insufficient width, nicks and pinholes are caught by AOI against the data. The item-by-item criteria are in the outer-layer pattern inspection checklist and are not repeated here.

Electrical test. Sunking runs 100% electrical test on every order, by flying probe or fixture, alongside 100% FQC inspection and per-batch reliability testing across 10+ quality gates. Electrical test catches continuity outcomes; it cannot catch a parametric deviation such as "trace is undersized but still conducts". Width measurement and AOI are therefore not replaceable by electrical test.

Registration verification. Outer-to-inner alignment is confirmed by X-ray target measurement and by microsection. IPC-6012 Class II and Class III set explicit criteria for annular ring, dielectric thickness and plating, and the acceptance class has to be fixed at quotation — Class III margin requirements tighten which trace width schemes remain available.

8. What to state at RFQ and engineering review

Stating the following at quotation and engineering review removes most of the back-and-forth on fine-line work:

1. Minimum trace and spacing, and where it occurs. A panel that is fine-line throughout and a panel where only the BGA escape area is fine-line are two different difficulties and two different costs. Marking the fine-line region is far more useful than a single global minimum.
2. Finished outer copper thickness. This sets the pattern plating amount, the resist thickness and the side-etch compensation, and it is the first constraint on fine-line feasibility. Copper and width must be assessed together.
3. Whether trace width carries a special tolerance. Default is ±20%. If the design has impedance-controlled traces, impedance tolerance constrains width tolerance in turn and needs separate declaration; see tolerance capability for the published basis.
4. Panel size and the array scheme. Large and elongated arrays distribute scaling error differently, which feeds directly into imaging method selection. Sunking's thickness range is 0.2–6.0 mm with a 5×5 mm minimum board size; extreme dimensions should be flagged up front.
5. Layer count and whether it is HDI. Layer count (40 standard production, 64 maximum) and HDI stage count set registration difficulty and capture pad margin.
6. IPC acceptance class. Fix IPC-6012 Class II or Class III before engineering review. Class III narrows the available schemes.
7. Whether engineering may adjust for manufacturability. Pad trimming, compensation values, escape angle handling. Giving the scope of that authority lets engineering optimise without repeated confirmation cycles.
8. Expected revision count. If several revisions are likely, the cycle-time gain of direct imaging belongs in the total cost, not just the per-exposure cost.

With those in the data package, the manufacturing capability page is useful for checking which combinations sit in the standard tier and which need advanced or R&D-tier assessment. For context on where a single-sided fine-line job lands, see fine line 6 mil trace single-sided manufacturing.

9. Summary

Fine-line capability is the capability of whichever link bottoms out first, and exposure is only the entrance. Film contact exposure concentrates its error in consumables and alignment, and that error does not shrink as the line shrinks. LDI moves the error into something measurable and correctable per panel, which is why it has more margin on fine features. But the two things outer layers carry beyond inner layers — resist that must survive pattern plating, and etching that must be compensated — are frequently the real bottleneck.

Before changing exposure method, triage once on the spatial distribution of the defect: fixed repeating location points at imaging, correlation with line width points at resist and compensation, batch drift points at material and chemistry. Get the triage right and the investment lands on the link that actually bottomed out.

FAQ

Is LDI always more accurate than film exposure?

On fine lines usually yes, but not because "a laser is more precise than light". The error structure differs. Film's dominant errors come from consumable dimensional change and the contact gap; those are fixed in magnitude and do not shrink with the line. Direct imaging moves the error into target measurement and a compensation model, which can be corrected per panel or per batch. If the width is not fine and the panel is not large, film still has adequate margin and switching will not produce a visible yield change.

My fine lines are failing. Will changing exposure method fix it?

Triage first. Check whether the defect correlates with location or with line width. Repeating at a fixed location across panels points to the photo tool, targets or data — the imaging side. All fine lines poor while coarse lines are fine points to resist resolution and etch compensation, where a new exposure tool helps little. Clustering along the panel edge or in a region points to scaling compensation and plating current distribution. Drift between batches points to material and chemistry. Only the first case and the scaling case show a clear gain from moving to direct imaging.

Are the exposure difficulties the same on outer and inner layers?

No. Inner layers form in one step and the central problem is layer-to-layer registration. Outer layers add pattern plating, where the dry film is not only an etch mask but also the plating sidewall, and where etching removes base copper after the line has already been plated up. Outer layers therefore carry resist aspect ratio and side-etch compensation as two additional constraints. The same exposure tool generally supports a tighter width inner-layer than outer-layer.

Why does the same Gerber deliver different trace widths at different copper thicknesses?

Because the artwork written into the resist is itself back-calculated from copper thickness. Copper sets the pattern plating amount and the side-etch amount, so the compensation value changes with it. Exposure only guarantees the back-calculated artwork is written accurately. That is why finished copper thickness and minimum trace width must be quoted together — width alone is not enough to judge feasibility.

Is thicker dry film the safer choice?

No. The resist has to satisfy two conflicting requirements: contain the copper that pattern plating deposits, which wants it thick, and keep resist walls standing on fine features without leaning, which wants it thin. Heavy copper requirements push the resist thicker, and thicker resist has a worse aspect ratio on fine lines, raising both residue and lean-over risk. This is one of the root reasons heavy copper and fine lines are hard to have together.

Is ±20% trace width tolerance the accuracy of the exposure step?

No. It is the delivered band after the spreads from exposure, pattern plating and etching are stacked. Exposure determines how accurately the artwork is written, plating uniformity determines how much the line is widened and whether that is consistent across the panel, and etching determines how much is taken off. Improving any single link will not narrow the tolerance proportionally.

What is the single most important thing to state when quoting a fine-line job?

Where the minimum trace width occurs, together with the finished outer copper thickness. Fine lines across the whole board and fine lines in one escape region are two different costs, and copper thickness decides whether the width is achievable at all. With both supplied, engineering can decide between standard, advanced and R&D-tier assessment, and can select the imaging method.