An inspection sensor looks through a fence toward a wall. A handheld instrument is aimed near a narrow post with a larger surface behind it. In both scenes, a plausible distance may belong to something other than the intended object.
A valid report establishes only what its documented checks establish. It does not automatically identify the physical surface that produced the accepted value. Before selecting a module for a mixed scene, define the intended object, the acceptable result and the evidence needed to distinguish it from nearby surfaces.
Separate three questions
What does the application want to measure? Write this in physical terms: the front of a package, the visible wall behind a fence, or the intended inspection surface. “Distance to the target” is incomplete if different people would choose different objects in the same scene.
What does the device report? Obtain the actual output definition for the delivered model and mode. Some ranging architectures can distinguish several return candidates and apply a selection rule. That general capability does not establish that a particular ERDILRF module exposes multiple returns, offers a selectable rule or provides diagnostic data.
What will the host accept? A parser can validate a frame while the application still rejects its distance as unsuitable or ambiguous. Keep protocol validity, measurement availability and application acceptance separate in the log.
Return-selection labels need a definition
Terms such as nearest, strongest and last appear in ranging discussions, but they answer different questions. They are useful prompts for an interface review rather than features to assume from a module name.
| Term | Clarification needed |
|---|---|
| Nearest | Nearest among which detected and qualified candidates? What happens to a weak foreground return? |
| Strongest | How is strength evaluated, and does the output identify the surface the application intends to measure? |
| Last | Last among which accepted candidates and within what documented measurement conditions? |
| Single distance output | What, if anything, is documented about selection and invalid-result behavior? Which questions remain unanswered? |
Do not infer a selection algorithm from one successful demonstration. A change in reflectivity, angle, illumination or geometry can alter the returns available to the receiver. A consistent number in one arrangement does not establish a fixed selection policy for all scenes.
Build a comparison where the surfaces are distinguishable
Choose a fixture in which the foreground and background positions are independently known using a reference method suitable for the required accuracy. The separation must be sufficient for that reference and the measurement uncertainty to distinguish the surfaces. This is a requirement for interpreting the test, not a universal minimum spacing.
Record the module and firmware identity, mode, mounting, host optics, reference method and scene geometry. Describe materials and surface finish without assigning an unmeasured reflectivity percentage from color alone. Keep individual reports, unavailable readings and host rejection reasons.
| Scene | Question to investigate |
|---|---|
| Intended surface alone | Can the chosen configuration produce useful results under the recorded conditions? |
| Background alone | What readings occur when the intended foreground object is absent? |
| Foreground and background together | Which surface is consistent with each result, and how often is the outcome ambiguous or unavailable? |
| Controlled movement near an edge | Does the reported result change as the visible portion of each surface changes? |
| Representative illumination or surface changes | Does the result still satisfy the application’s acceptance rule? |
| Final enclosure and window | Does the installed optical path change the observed outcomes? |
Change one factor at a time where practical, and state any factors that changed together. These are proposed comparisons, not ERDILRF test results or a guarantee that every module can identify both surfaces.
Count outcomes without inventing confidence
Classify the recorded outcomes against the independent references and a predeclared uncertainty allowance. Useful categories include consistent with the intended surface, consistent with another known surface, unavailable, and unresolved. Keep the total attempt count so discarded readings do not disappear from the evaluation.
A value that is consistent with a known surface is evidence about that test arrangement; it does not reveal the device’s internal processing. An intermediate value should not automatically be described as an average of two optical returns. Several causes may remain possible until further evidence is available.
If the documented interface supplies only distance and limited status, record those fields faithfully. Do not create a “signal strength” or “confidence percentage” field from numeric stability alone. Request additional diagnostics only after confirming that they exist for the selected configuration.
Use the result to make a design decision
When the installation frequently returns the wrong surface, review the geometry and use case before adding smoothing. Averaging a stable background distance does not turn it into a foreground measurement. Changes to mounting, aperture clearance or the intended observation area require their own checks.
The application may need to withhold an ambiguous reading, ask an operator to confirm the observation, or reconsider the sensor arrangement. Define that behavior explicitly rather than presenting every accepted serial frame as a confirmed target measurement.
For the related geometry question, the LRF50H target-edge note explains why nominal spot size alone does not establish which surface is reported. Compare ERDILRF modules and their published conditions, or send a scene sketch, intended surface, distances and raw observations for an application review.