Sandustry Guide and Wiki

machines

Sandustry Filter Rejection Routing Field Guide

This field guide treats filter rejection routing as a measurable factory boundary rather than a screenshot recipe. It defines the intended result, the safest commissioning test, the failure evidence worth preserving, and the decision needed before scaling.

Early Access · observation-led field manual · Checked 2026-08-22

Define the Filter Rejection Routing outcome

Start with one observable outcome: keep rejected material visible and prevent it from re-entering the selected output. Write it beside the module before changing geometry. The acceptance statement must name what enters, what leaves, where byproducts go, and which state counts as safe. For this task, the strongest live signal is that clean selected and rejected samples always reach different labeled destinations. Do not replace that evidence with machine animation, a screenshot, or a ratio copied from another build. Early Access can change balance and component behavior, while the outcome remains testable. Record the game build, save context, starting inventory, and the exact location of every observation so a later run can be compared rather than remembered.

Build a controlled test boundary

Keep the first test away from the main factory. Use two independent collectors with overflow and quarantine on the rejection side. Limit the amount of material or liquid that can enter, clear every destination, and preserve access above, below, and beside the likely obstruction. Label inputs and outputs even when their direction seems obvious. The boundary should protect valuable storage and unrelated production if the test fails completely. Check the current official Wiki for documented names and roles, but treat exact capacity, timing, and geometry as local observations unless an official source states them. Photograph or sketch the empty state so settled pixels and level changes remain visible during diagnosis.

Run a repeatable commissioning test

The core commissioning action is to run selected, rejected, mixed, full-selected, and full-rejection samples separately. Use the same measured input for comparisons and wait long enough to distinguish a sustained result from a startup burst. Observe the entire path rather than only the final collector. Record start time, first useful output, stable interval, warning level, and anything left outside the intended destination. Stop immediately if containment or maintenance access is threatened. Repeat the clean run once before introducing a blocked or starved state. A single successful batch establishes a candidate configuration; repeated success plus controlled failure and recovery establishes a module that is useful in production.

Recognize the specific failure pattern

The important failure for this page is one full destination causes backflow or redirects material across the interface. Preserve the scene before digging, rotating blocks, or opening a wall. A changed scene removes evidence about where material accumulated first. Compare upstream and downstream checkpoints: a rising upstream level with low downstream acceptance points toward a constrained interface, while an empty upstream pocket suggests supply. If both swing, observe a longer interval and reduce the batch. Community posts are useful for naming symptoms and possible checks, but they do not prove that the same cause exists in this build, save, geometry, or operating level.

Measure useful output

Measure accepted selected output and contained rejection across the same interval. Count only material or liquid that reaches the intended usable destination. Do not count temporary movement, animation cycles, or pixels that later fall into quarantine. Run equal intervals after clearing buffers and record the slower repeatable value, the current build, and the layout state. Enter conservative observations in the production planner and keep them separate from official facts. Also measure the time from normal operation to the warning line during a downstream stop. Throughput explains normal capacity; warning time explains whether the design gives a player or control system enough time to act safely.

Practice shutdown and recovery

Use a written recovery sequence: isolate input, clear each destination independently, and retest clean samples. Protect controls and escape routes before saving material. Never open the lowest wall under a full store or reservoir without empty containment beneath it. Move uncertain recovery material to a labeled quarantine bay instead of returning it automatically to feed. After the obstruction is removed, inspect every interface, restore normal destination capacity, and restart with a tiny clean sample. A recovery is complete only when the module returns to its documented normal range without new mixing, overflow, repeated manual intervention, or damage to a neighboring process.

Decide whether to redesign or scale

Use the evidence to connect mixed production only after direction and both full states are recorded. Change one boundary at a time and rerun the same acceptance test. More machines do not repair a buried outlet, full collector, unstable liquid level, or shared transport limit. If the module succeeds, duplicate the smallest proven version and keep each copy independently isolatable. Then test the shared supply, collection, byproduct, and control paths because they can become the new bottleneck. Preserve the original measurement sheet so expansion can be compared against a baseline instead of judged by appearance or a short period of apparently faster movement.

Maintain the guide after updates

Keep a compact maintenance card for Filter Rejection Routing: purpose, build, source links, inputs, outputs, normal levels, warning states, observed throughput, shutdown order, recovery result, and last checked date. After an Early Access patch, read current official notes first and recommission the smallest isolated module before opening the full factory. Retire values that no longer reproduce and explain what changed. When sharing a layout, distinguish official documentation, your measured result, and community-reported possibilities. That boundary keeps the page useful without converting one player's save-specific observation into an unsupported universal rule.

Sources and evidence notes

Frequently Asked Questions