Measuring reversal error without calling every error backlash

Measure the difference before naming its cause

When opposite approaches reach different measured positions, it is tempting to write “backlash” on the report. That label may be premature. Start by recording the direction-dependent observation and its conditions. A useful investigation keeps the measurement result separate from the mechanism proposed to explain it.

Newport’s terminology guide distinguishes reversal error, backlash and hysteresis. This article proposes a diagnostic workflow rather than a mechanical classification rule. Use the site’s specification page to keep the separate questions of accuracy, repeatability and resolution clear throughout the investigation.

Establish a direction-controlled baseline

Choose a safe target within the working range, with sufficient clearance for approaches from both sides. Define the excursion, final approach speed, acceleration, dwell and measurement interval. Record the actual sequence, including any controller overshoot, take-up move or automatic compensation. A nominal negative command may still end with a positive final approach.

Measure repeated arrivals from each side without changing the reference or sensor zero between groups. Keep timestamps so thermal drift can be considered alongside direction. Plot the two groups separately and retain their individual observations. A single combined spread can hide a consistent offset between groups, while a difference between two isolated observations says little about repeatability.

Document the sensing point, fixture stiffness, cable routing and payload. If a contact probe is used, consider its orientation and contact force within the measurement assessment. Check for an independent reference movement or fixture motion that could influence the reading. Do not assume the measurement stand is stationary merely because it is outside the moving carriage.

Investigate one influence at a time

After preserving the baseline, choose one controlled variation that addresses a specific hypothesis. A longer dwell investigates whether the observation changes with settling time. A different excursion investigates sensitivity to recent motion history. A revised cable support investigates whether cable loading contributes. Each variation needs its own identifier and a repeat of the original baseline where practical.

Avoid immediately altering mechanical preload or controller compensation. Such changes may require supplier guidance, affect safety or invalidate the original configuration record. First confirm that units, scale, sign and timestamps are correct. Then compare external measurement with available feedback channels, while recognizing that the channels may observe different physical locations.

A position-dependent reversal pattern deserves measurements at additional targets. A time-dependent pattern deserves environmental and warm-up records. Neither pattern, by itself, identifies a failed component. Record candidate explanations and the evidence needed to distinguish them rather than replacing an uncertain diagnosis with confident wording.

Hypothetical example: the direction gap changes with dwell

Imagine an inspection axis whose opposite-direction arrivals appear separated in an early reading, but the separation is smaller after a longer observation interval. No numbers or actual product results are assumed. The first conclusion is simply that the measured behavior depends on when it is observed.

The engineer checks the time trace, controller state and fixture response before deciding whether the effect is mechanical clearance, elastic behavior, residual motion or a measurement artifact. The production dwell remains part of the requirement. A favorable reading after a much longer delay would not automatically satisfy a process that acquires data immediately after arrival.

Practical checklist for a reversal investigation

• Label the final approach direction of every observation.
• Preserve one unchanged baseline before modifying any settings.
• Keep sensor zero, reference establishment and timing explicit.
• Check fixture, probe and cable influences before assigning a component fault.
• Change only one experimental condition per diagnostic comparison.
• Repeat at relevant target positions when the effect varies with travel.
• Separate observed differences from proposed causes in the report.

Include uncertainty and the chosen statistic when reporting the directional difference. NIST’s uncertainty guidance supports stating how that uncertainty is expressed. A credible report may conclude that the present setup cannot isolate the cause. That is still useful: it prevents an unsupported mechanical diagnosis and identifies the next experiment needed before changing the machine or accepting a compensation strategy.