Fadal 104/D Control: Troubleshooting Program Skips

Field reports on the Fadal 104/D control describe non-deterministic block execution, cutter-compensation errors, premature feed-rate changes, and lost setup data after a restart. Software version 1.05 did not resolve the reported G0 execution problem. Treat an affected machine as unverified until it repeats a controlled test without skipped blocks, path deviations, or changed offsets.

Define the Affected Configuration

The clearest failure report concerns a Fadal 4020 equipped with the 104/D control and software version 1.05. A separate 6030 installation also required replacement of its 104/D control. The evidence does not establish hardware revisions, option combinations, or a complete affected-version range, so do not infer that every 104/D installation has identical behavior.

Recognize the Reported Failure Signatures

The primary incident occurred at random times: the control reportedly stopped continuing through a G0 command and then fed toward an unknown destination. A complete restart temporarily cleared the behavior, but the restart reportedly lost all tool positions and zeroed some height offsets.

Additional reported execution anomalies included skipped program blocks, cutter compensation disappearing before its programmed cancellation move and returning later, the next block's feed rate taking effect early, inconsistent M00 or M01 behavior until program renumbering, delayed or inconsistent single-block activation, and canned cycles that did not operate correctly or at all. These are field observations rather than a manufacturer-confirmed common root cause.

Reproduce the Cutter-Compensation Fault Safely

The following submitted program was intended to machine a 2-inch hole with a 1-inch cutter and make a spring pass. The reported control skipped N8, lost cutter compensation before the G40 cancellation move, gouged the part, and later showed an axis shift. Whether the program is valid under every Fadal programming mode is not established, so first verify its syntax against the documentation for the installed control.

N1 O1
N2 M6 T1
N3 E1 G0 G90 X0 Y0 S2000 M3
N4 H1 Z.1
N5 G1 Z-.1 F50.
N6 G1 G41 G91 X1.
N7 G3 I-1.
N8 I-1.
N9 G1 G40 X-1.
N10 G0 G90 Z.1
N11 M5
N12 H0 Z0
N13 E0 X0 Y0
N14 M2

1. Confirm that the test program, cutter-compensation convention, full-circle arc syntax, and modal behavior are valid for the installed control.
2. Remove production tooling and valuable material from the test. Use the machine's available simulation or non-cutting verification method before permitting motion.
3. Execute the program repeatedly while recording whether N7 and N8 both execute, where G41 ceases to affect the path, and whether G40 cancels compensation at N9.
4. Compare the commanded and observed paths at the end of the program. Any repeatable axis shift, omitted pass, or gouging motion is a failed result.
5. Retain the exact program and execution observations for escalation; do not alter several variables between runs.

Protect Setup Data Before Restarting

A restart was the only reported temporary recovery for the random G0/feed behavior, but it also created a setup-integrity hazard. Before restarting, externally record every active tool position and height offset needed for the job. The evidence does not identify a reliable backup command or file format, so use only a documented method available on the installed control.

After restart, do not resume at the interrupted block. Compare every recorded value with the control, restore only verified values, and re-establish the setup. The machine must remain out of production if tool positions are lost, any height offset becomes zero unexpectedly, or the commanded position cannot be reconciled with the physical setup.

Separate Execution Defects from Other Machine Problems

Not every reported symptom proves a control-software defect. A stall while tapping a 1-12 thread in steel was observed, but the report could not distinguish insufficient torque from belt slip because no load meter was available. Diagnose that event separately from skipped blocks rather than assigning it to the 104/D software without measurement.

Other operational complaints included a tool-release button needing a delay, no loop counter for subprogram repeats, inability to stop the spindle while single block was active, and inability to use Gibbs SFP while the machine side was running. Door-opening behavior also changed: spindle and coolant had previously restarted automatically but later did not. These reports define functions to test, but they do not establish the intended safety behavior or configuration of every machine.

Distinguish Program Storage from Look-Ahead

A 4 MB memory option was described as additional program-storage capacity, not additional look-ahead. Therefore, do not specify that memory upgrade as a remedy for contouring stutter or data starvation without separate evidence about the control's processing and communications path.

The reported workflow used Multi-DNC through RS232. Filtered Mastercam files were reported at 2-3 MB for roughing and 10-30 MB for finishing. Because finishing files exceeded 4 MB, those programs would still require a transfer or DNC strategy rather than relying entirely on that stated storage capacity. The evidence does not provide baud rate, buffer size, look-ahead depth, or filtering tolerance, so those values must be measured or obtained from control documentation.

Evaluate Replacement and Acceptance Paths

Reported remedies ultimately involved removing or replacing the 104/D configuration rather than proving that version 1.05 corrected it. One site operated a remanufactured 3016 with an 88 control and a new 4020 with a 32MP control for three months of continuous use without the previously described faults. Another site reported a 104/D exchange on a 6030. These outcomes support replacement as an escalation path, not a guarantee that any alternative control will fit every application.

Replacement-machine geometry requires independent acceptance. One supplied used 4020 reportedly had enough Z-axis way wear that the table could not be swept within 0.050 inch; a later account described the indicator range as 0.060 inch. Because those two values differ, preserve both as reported observations and perform a fresh geometry inspection rather than treating either as a certified measurement.

1. Provide the machine model, 104/D identification, installed software version, preserved test program, and observed block-level failure to the manufacturer through an official support channel.
2. Request a written corrective-action or replacement proposal that addresses program execution, setup-data retention, communications, and required options.
3. Inspect replacement-machine geometry before accepting installation.
4. Repeat the cutter-compensation program, G0 execution checks, M00/M01 behavior, single-block behavior, canned cycles used by production, and the normal DNC workflow.
5. Release the machine only after repeated runs preserve tool and height data and produce the commanded path without skipped blocks, premature feed changes, or axis shifts.

FAQ