Write-on-the-Fly vs Step-and-Repeat Lithography: How Encoder Feedback Governs Each Mode

Two architectures, two definitions of position accuracy

In a step-and-repeat tool, the stage carries the substrate to a field position, the servo settles, and the exposure proceeds while the mechanics are nominally still. Position feedback matters in two places: bringing the field to its commanded location, and holding it there against vibration and drift during the exposure. Accuracy is a settled property, and the encoder's dynamic behavior matters mainly through the servo loop, not through the exposure itself.

Write-on-the-fly inverts this. The stage never stops; the pattern is emitted in register with the moving substrate. Every element of the pattern must be placed at the moment the substrate passes the exposure zone, so the encoder's reading of instantaneous position is converted directly into pattern geometry. A timing skew between the position sample and the modulator output becomes a placement error; an encoder update that arrives late leaves the pattern generator guessing; velocity variation changes the effective dose per unit length. The encoder is no longer merely reporting position—it is pacing the exposure.

How the synchronization chain actually works

In a hardware-synchronized write-on-the-fly system, the encoder's incremental signal is processed into position samples that trigger or clock the pattern data path. The scale pitch sets the natural spatial sampling grid; interpolation multiplies it into finer position steps. The pattern generator maps each incoming data element to the position interval during which it should be exposed, and the modulator switches accordingly. The whole chain—scale, readhead, interpolation, synchronization electronics, data formatter, modulator—has a total latency, and latency that is constant can be compensated; latency that varies cannot.

This is why the quality of the synchronization matters as much as the encoder's datasheet accuracy. A high-resolution encoder with jittery interpolation produces pattern wobble no error map can remove, because the wobble happens within a single field at writing speed. Conversely, a modest-resolution encoder with a clean, deterministic timing chain can deliver excellent pattern fidelity. Buyers should ask vendors to describe the chain explicitly and to state which encoder configuration was used for the placement data they present.

• Spatial sampling grid set by scale pitch and interpolation ratio
• Constant latency is compensable; variable latency converts directly to placement error
• Signal quality at writing velocity governs interpolation validity
• Ask which encoder configuration produced the vendor's placement evidence

What fails in each mode, and how to detect it

Step-and-repeat errors concentrate at field boundaries: overlay metrology between fields shows placement offsets, and stitch quality at the boundary reflects the relative placement of adjacent fields. Because exposure happens at rest, dose is uniform within a field unless intensity or focus drifted. These errors are localizable with standard overlay tools, and corrective action—error mapping, settle tuning—follows familiar stage-qualification practice.

Write-on-the-fly errors follow the motion. Placement error grows along the scan direction where encoder latency or velocity ripple accumulates; dose varies with the instantaneous writing speed, so line width changes across fields written at different accelerations; line ends at scan reversals inherit the worst dynamic conditions. Detection therefore requires dynamic tests: write test patterns at production speed and inspect stitch, placement and linewidth along and across the scan direction, then repeat at reduced speed—an error that shrinks with speed points at the synchronization chain rather than at static geometry.

Choosing the mode for your workload

The economics follow from the pattern structure. Workloads with many small, dense fields—photonic waveguide arrays, sensor grids, memory-like arrays—are dominated by settle time in step-and-repeat mode, and write-on-the-fly can cut writing time substantially because the stage never stops. Workloads with sparse, large-area geometry, or processes where per-field alignment and focusing are critical, often run as well or better with the simpler timing of step-and-repeat, and the same tool may support both modes with the encoder synchronization chain engaged only when writing in motion.

The deciding evidence is dynamic, not nominal. Ask each candidate vendor for: writing time on your actual layout class, placement and stitch metrology taken with writing in motion at the quoted throughput, and the encoder configuration and synchronization latency behind those numbers. A tool that documents these cleanly can be qualified for either mode; a tool that answers only with static stage specifications should be treated as unproven for write-on-the-fly regardless of its datasheet resolution.

Specification questions to put to the vendor

Reduce the mode decision to verifiable questions. For write-on-the-fly: what is the encoder update rate and total synchronization latency during writing, is the latency constant, and what velocity range does the tool write over with placement held within specification? For step-and-repeat: what is the settle time per field at the field size and acceleration you will run, and what residual position drift occurs during a typical exposure? In both modes: which scale accuracy grade and interpolation configuration is installed, and can the vendor supply placement data measured with the same configuration?

Finally, align the acceptance test with the mode. A factory acceptance test for a write-on-the-fly tool must include patterns written in motion at production settings and inspected for stitch, placement and linewidth—static positioning verification alone will pass a tool whose dynamic synchronization is marginal. Agreeing on this before purchase costs nothing; discovering it after installation costs a tool.