FANUC 0i-F waveform diagnosis: why a fault trace can omit the warning signs

After an intermittent stop, a quiet trace can easily be read as proof that a channel was healthy. First establish whether the recording covers the instant under investigation. A capture that retains only data after an alarm may omit the preceding change entirely. This guide checks whether the evidence is adequate, rather than deciding component condition from one graph.

Scope and safety

The reference is the FANUC Series 0i-MODEL F Maintenance Manual B-64605EN/01, section 1.6, particularly printed pages 58–59 and 65–66. This applies only to 0i-F configurations whose functions and screens match that edition, not automatically to F Plus or other series. It reviews existing records and acquisition plans; it provides no parameter-writing, servo-tuning or fault-provoking test procedure. Qualified personnel must arrange any new acquisition and machine movement under the machine builder's safety procedures. Follow emergency procedures first if danger arises.

Put the trigger back on the timeline

The manual distinguishes JUST, AFTER and BEFORE windows. JUST records the specified period after [TRACE] is pressed. Once tracing has been started, AFTER retains the period after the trigger event, while BEFORE retains the preceding period. Page 59 also allows a delay in BEFORE that moves the end beyond the event. The BEFORE label alone therefore does not establish that every stored sample predates the alarm.

Mark three points during review: tracing started, trigger occurred, recording ended. Add the time of the abnormal behavior described by the operator. If that instant lies outside the retained window, write “not covered by the recording,” rather than “did not occur.” If trigger occurrence is unconfirmed, do not describe a graph as a complete fault-triggered capture.

Longer coverage is not finer observation

This edition specifies waveform and signal sampling cycles from 2 to 4096 ms in multiples of 2 ms, with approximately 32700 data points. Each channel uses one point per sample; signal acquisition collectively counts as one channel, not one per signal. At 4 ms, the manual illustrates about 130 s for one waveform channel, versus about 26 s for four waveform channels plus signals.

Adding measured items can therefore shorten the retained history for the same incident. A longer sampling cycle extends coverage but gives a coarser view of brief changes. Our review approach is to define the timing question first, then assess whether the existing acquisition plan can answer it. A sufficiently long recording does not by itself establish reliable capture of short pulses. These capacities describe acquisition, not acceptable fault duration.

Hypothetical examples: missing history and missed pulses

Suppose an alarm occurs at time 0 and a related signal changes at −20 ms. An AFTER capture may exclude that precursor. A quiet post-alarm trace cannot rule out the earlier change. This is a hypothetical timing example, not a measured repair case or manufacturer case study.

In a separate, purely sampling example, instantaneous readings every 4 ms may miss a 1 ms pulse that begins and ends between adjacent readings. This illustrates the limits of discrete evidence; it does not claim to verify how this CNC internally latches a particular signal. Signal retention and trigger detection require separate examination of the applicable PMC logic and documentation. Absence from the plotted record is not proof of electrical absence.

Missing from the screen may mean missing from the selected view

Page 66 permits simultaneous measurement of up to 32 signals, while a signals-only screen shows up to nine and a waveform overlay shows up to four. A handover screenshot is consequently not a substitute for the complete acquisition list. Check which addresses were recorded and which are displayed before stating that a particular signal did not change.

A useful handover package includes original data, acquisition-setting screenshots, CNC model and software version, axis/path mapping, trigger description, complete alarm text and capture time. Preserve the raw file rather than overwriting it to tidy a plot. Mark signal meanings that remain unverified.

Conclusions that respect the evidence

1. The window excludes the incident: report insufficient time coverage and leave channel condition unresolved.
2. The window fits but sampling is coarse: say that brief changes remain possible; do not prescribe replacement.
3. The acquisition is complete but screenshots are partial: review the remaining existing views or file rather than provoking the fault again.
4. A related change is visible: report sequence and sampling limits, then verify the cause against alarms, PMC logic and machine documentation. Being earlier in time does not establish root cause.

Source

FANUC Series 0i-MODEL F Maintenance Manual B-64605EN/01, section 1.6.2, printed pages 58–59 (windows, sampling and capacity), and section 1.6.3, pages 65–66 (tracing and signal display). This is a third-party mirror of a manufacturer-authored manual, not an official FANUC download site. Timeline checks, hypothetical examples and the handover checklist are editorial analysis, not manufacturer acceptance criteria.