Uneven or Inconsistent Output: A Cross-Trade Decision Tree

Why this matters

Inconsistent output (fine one minute, weak the next, strong in one zone and poor in another) is the hardest fault to pin down because it will not sit still for testing. Customers describe it as "sometimes" and techs chase it for hours. The trick is to separate output that varies over time from output that varies by location, then catch it in the act.

Start here: does it vary by time, or by place?

This single question splits the whole diagnosis.

• Varies over time (good then bad at the same point): suspect an intermittent supply, a loose connection, a cycling control, or a temperature-dependent component.
• Varies by location (always weak in one zone, always strong in another): suspect a balancing, distribution, or restriction problem, not the source.

Confirm which one before going further. Take readings at several points and at several times.

Branch 1: varies by location (a distribution problem)

If one area is always short and another always strong, the source is usually fine and the path is unbalanced.

• Check balancing and distribution. A branch with less restriction steals supply from the others. Dampers, balancing valves, or splits set wrong send too much one way and too little another.
• Check for a restriction on the weak branch. A partial clog, kink, collapsed run, or fouled section on one leg starves only that leg while the rest run strong.
• Check run length and sizing. The farthest or smallest-sized branch naturally gets the least. If it was added later or undersized, it will always lag.

If the weak zone clears when you throttle the strong zones, you have a balancing problem, not a unit problem.

Branch 2: varies over time, on a regular cycle

If output pulses or cycles predictably, something is switching.

• Check the control for short cycling or hunting. A controller with no deadband, or an oversized system, turns on and off rapidly so output surges and fades.
• Check staged equipment. A system that drops or adds stages produces a step change in output. Watch the stages while the symptom happens.
• Check for a duty-cycle device. A timer, defrost cycle, or regeneration cycle that interrupts normal operation will dip output on a schedule.

Branch 3: varies over time, randomly (intermittent fault)

Random inconsistency is usually a connection or a marginal part.

• Wiggle test the connections (powered off, locked out, then carefully under power if trained to). A loose lug, corroded terminal, or marginal contact passes full output until vibration or heat opens it, then recovers. This is the top cause of random inconsistency.
• Check for a temperature-dependent component. A capacitor, sensor, or relay that drifts as it warms works fine cold and fails hot, or the reverse. Note whether the symptom tracks runtime or ambient temperature.
• Check supply stability. A source that sags under load, a shared circuit, or upstream demand spikes can make a healthy unit deliver unevenly.

Branch 4: load-driven inconsistency

Sometimes the output is steady but the demand on it is not, so it looks inconsistent at the point of use.

• Check for competing draws. When another fixture, zone, or device pulls at the same time, the shared supply drops everywhere. The symptom appears only when both run.
• Map the timing. If the dip always lines up with another load starting, the problem is shared-supply capacity, not the unit.

What to check first, in order

1. Decide: varies by time or by location.
2. For location: check balancing, then the weak branch for restriction.
3. For regular cycling: check control deadband and staging.
4. For random: wiggle-test connections, then temperature-dependent parts.
5. For both: check supply stability and competing loads.

Cheapest and most common first: a loose connection (random) and an unbalanced distribution (by location) cover most inconsistent-output calls.

References

• Trade-standard balancing and commissioning practice for distribution systems
• Manufacturer control and sequence-of-operation documentation
• OSHA lockout/tagout (29 CFR 1910.147) for safe connection testing
• See related: Universal "The Noise Only Under Load: A Cross-Trade Decision Tree"