Works In Test Mode But Not Normal Operation Decision Tree
Why this matters
The equipment runs perfectly in test mode and fails in normal operation. This is one of the most common diagnostic patterns across trades, and one of the most misread. The temptation is to declare the equipment "fine" because you saw it run on the bench. The reality is that test mode is missing inputs that normal operation includes. Find the missing inputs and you find the fault.
The framework is straight isolation: identify what test mode bypasses, then re-introduce each bypassed element until you find the one that breaks the system. This applies whether you are working on a furnace board, an elevator controller, a refrigeration rack, a pool pump, an irrigation system, or a CNC.
Symptom presentation
- HVAC board runs perfectly in service-mode/W-call jumper test but locks out in thermostat-driven operation
- Garage opener runs from the wall button but not from the remote
- Boiler fires when you jumper the safety chain but trips when you reinstall it
- Refrigeration compressor runs on manual contactor pull-in but not on temperature-controller call
- Pool pump runs on direct switch but not on timer call
- CNC tool runs in jog mode but not in cycle
- Elevator runs on inspection but not on car call
The common thread: test mode bypasses inputs that normal operation requires.
Quick checks
- What does test mode actually bypass? Read the service literature, not the lore. Common bypasses: safety chain, ambient sensors, demand signal, network communication, interlock chain, time-of-day clock, fault-history lockout.
- Reproduce normal operation conditions deliberately. If normal operation needs a thermostat call, install one and watch. If it needs a network token, restore the network.
- Note exactly what happens at the moment normal operation fails. Hard lockout? Soft fault? No response at all? Brief start then drop?
- Read the panel for fault history that was masked in test mode.
- Compare test-mode and normal-operation timing. Test mode often skips delays, soft-starts, and pre-purge sequences. Failures during those sequences are invisible in test.
Isolation tree
Apply in order, reintroducing one bypassed element at a time.
Stage 1: Inputs
What inputs does normal operation depend on that test mode generates internally or bypasses?
- Demand signal (thermostat, controller call, network command, dry contact)
- Permissive inputs (pressure switch, flow switch, door switch, end-of-travel limit)
- Sensor feedback (temperature probe, level sensor, current transformer, position encoder)
- Network or bus traffic (Modbus, BACnet, CAN, proprietary serial)
- Safety chain (interlocks in series whose continuity is required)
Reintroduce each input one at a time. Test the system after each. The one that breaks the system is the failed input.
Stage 2: Sequencing
Normal operation often includes timed sequences that test mode skips:
- Pre-purge or pre-flush
- Soft-start ramps
- Lockout after N failed retries
- Anti-short-cycle timers
- Recovery delays after fault
If the equipment runs briefly and then drops, the fault is in one of these timed elements. Watch the panel during the sequence and identify which step it fails at.
Stage 3: Load
Test mode often runs unloaded or at a fixed reduced load. Normal operation imposes the full load:
- Pump against full discharge head
- Compressor against full suction lift
- Motor against the driven equipment
- Heater against the actual demand
A unit that runs unloaded and fails loaded points to a load-side fault: stuck valve, blocked filter, restricted line, bearing seizing under load, undersized supply.
Stage 4: Environment
Normal operation includes the ambient conditions:
- Hot enclosure (thermal limit trips that do not fire on the bench)
- Cold start (capacitor or board behavior changes)
- Voltage sag at peak hours
- Vibration from adjacent equipment
If test passes cold and fails after the unit has been running, you have a thermal fault. If test passes during off-hours and fails at peak, you have a power-quality fault.
Stage 5: Control logic
If inputs, sequencing, load, and environment all check, the fault is in the control logic itself: a corrupt parameter, a stuck flag in non-volatile memory, a board that runs the manual code path but fails the auto code path. At this point a factory reset and reconfigure from a known good parameter set, or board replacement, is usually the resolution.
Confirming the diagnosis
The signature of a correctly-found fault: in the bypassed-input state the equipment runs, you reintroduce one specific input, the equipment fails again at the same point. Single-variable isolation is the proof. If reintroducing the input reproduces the failure and removing it restores function, repeatably, you have the fault.
If the failure does not reproduce on demand, you have not isolated. You have observed. Keep going.
Never run equipment in normal mode with a safety element jumpered out as a "test." Test-mode jumpers are diagnostic, not operational. A jumpered safety left in place is a path to fire, refrigerant release, scald injury, or worse. Remove every diagnostic jumper before leaving the site, and verify by physical inspection, not just by walking past the panel.
Next steps
Document the bypassed-input pattern in the customer record. "Runs in test mode, fails when thermostat call introduced; isolated to low-voltage transformer dropping under load" is a record the next tech can use. "Intermittent" is not.
If the fault required parts to repair, leave a clean disposition: what is bypassed temporarily (if anything), what is locked out, when you will return. Do not leave a customer site with safety inputs disabled to "keep it running."
References
- ISO 13379-1 Condition monitoring and diagnostics of machines, data interpretation and diagnostics techniques
- NFPA 70B Recommended Practice for Electrical Equipment Maintenance, isolation testing
- ACCA Standard 4 Maintenance of Residential HVAC Systems
- OSHA 29 CFR 1910.147 Control of hazardous energy (lockout/tagout for jumpered states)
- IEC 60812 Failure modes and effects analysis