Planar 1959 Step 18 Alignment Tolerance Budget Stackup

# Stack Every Alignment Tolerance Across the Whole Flow Into One Budget: Step 13's Equipment, Carried All the Way Through

## 1. Why One Mask's Alignment Error Was Never the Only Error This Series Had to Survive

Step 13 described the equipment and the per-mask tolerance a single alignment step has to stay under; this step carries that same tolerance all the way across the full process flow, because the contact window from Step 3, the metal pattern from Step 6, and the field-oxide geometry from Step 1 and Step 2 all have to stay in registration with each other simultaneously, not just pairwise, and errors from separate alignment steps do not cancel — they add. A layout margin budgeted against a single mask's alignment tolerance, as Step 13 described it, is only safe if that is the only alignment step in the flow; this series actually runs at least two photolithographic masks after the initial diffusion, and each one's registration error against the wafer's own alignment marks contributes independently to the final uncertainty in where a contact edge or a metal edge actually lands relative to the covered junction underneath it.

$$\delta_{\text{total}} = \sqrt{\delta_{\text{mask 1}}^2 + \delta_{\text{mask 2}}^2 + \delta_{\text{wafer warp}}^2}, \qquad \delta_{\text{total}} < m_{\text{coverage}}$$

where each $\delta$ term is an independent alignment contribution — Step 3's window mask, Step 6's metal mask, and a wafer-flatness term this series has not previously had reason to name — combined in quadrature because the errors are statistically independent rather than systematically correlated, and the whole combined uncertainty has to stay under the coverage margin $m_{\text{coverage}}$ that Step 7's inspection checks directly on the finished wafer.

Three Independent Errors, Combined in Quadrature each alignment step's own tolerance, stacked against the one margin step 7 inspects δ mask 1, contact window δ mask 2, metal pattern δ wafer warp δ total, root sum of squares must stay under mcoverage δtotal = √(δ1²+δ2²+δwarp²) — errors from separate masks never cancel, they accumulate step 13 budgeted one mask; a real flow has to budget for all of them at once

## 2. Real Diagram: What a Tight Budget on a Single Mask Still Misses Across a Full Flow

A layout tuned to just barely satisfy Step 13's single-mask tolerance, mask by mask, can still fail Step 7's coverage inspection on the finished wafer if the independent errors from each mask happen to add in the same direction, because no single mask's own alignment step has visibility into what the other masks already did.

Passing Each Mask Alone Does Not Guarantee Passing the Stack per-mask tolerance satisfied individually, cumulative tolerance still failed EACH MASK, CHECKED ALONE mask 1, within tolerance mask 2, within tolerance STACKED ON THE SAME WAFER combined offset exceeds the coverage margin step 7's inspection catches this only because it checks the finished wafer, not any one mask's own alignment log

## 3. Why This Budget Had Been Implicit Since the Series Named Its First Margin

Step 7's own coverage-margin inequality already included a term for the metal etch's bias, acknowledging that more than one process step contributes error to the final result, but it treated alignment itself as a single lumped number rather than deriving where that number actually comes from. This step fills that gap by tracing the alignment contribution back to its source across every mask this series uses, confirming that the margin Step 1 through Step 7 built up through careful geometric choices was never being spent against one risk at a time — it was always being spent against the combined, compounding uncertainty of a multi-mask flow, whether or not any earlier step named that compounding explicitly.

Step 18 does not add a new mask or a new inspection; it finally traces Step 7's single margin number back to the several independent alignment steps actually spending it.

Take planar 1959 step 18 alignment tolerance budget stackup further

Ask the copilot about this term, or have our engineers assess it against your process.

Ask CFSGPT about this → Request a technical assessment → Open the CFS app