The Substitution Test When You Can't Measure

Why this matters

Sometimes the meter cannot give you a straight answer. The part reads in spec on the bench but fails under load. The board's logic is sealed and not measurable. The intermittent will not show up while your probes are on it. In those cases the substitution test - swap in a known-good part and watch what the symptom does - is the cleanest proof you have. Done well it is fast and certain. Done sloppy it strands you with two unknowns and a new part you cannot return.

What "known-good" actually means

A substitution is only as trustworthy as the part you sub in. "Known-good" means the replacement is proven to work, correct for the application, and not pulled from an unknown shelf.

• Use a new part of the correct rating, or a part you have personally verified works.
• Match the spec: voltage, current, microfarad, pressure, flow, port size, logic family. A close-enough sub that is actually wrong gives you a false result.
• Never sub a questionable part for a questionable part. Two unknowns prove nothing.

Step 1: Define the symptom before you touch anything

Write down, even mentally, the exact failure: "compressor will not start," "valve will not open on command," "no output from the control." You are about to change one thing. If you do not have a sharp before-picture, you cannot read the after.

Step 2: De-energize and isolate

For anything electrical, kill power and prove the circuit dead before you disconnect. (See related: Proving a Circuit Dead the Right Way.) For pressurized or fluid systems, relieve pressure and close the isolation valves. Substitution means handling the part, so treat it like any component swap: lockout where required, verify dead, then work.

Step 3: Change exactly one thing

This is the whole discipline. Swap the suspect part for the known-good part and change nothing else. Do not also reseat a connector, clean a contact, and adjust a setting in the same move. If you change three things and the symptom clears, you have learned nothing about which one mattered.

Step 4: Restore and re-test under the real condition

Energize or re-pressurize and run the system the way it was running when it failed. A part that only fails under load, heat, or full pressure must be tested under load, heat, or full pressure. A bench-good part proves nothing if the fault only appears in service.

Step 5: Read the result honestly

• Symptom clears with the new part: the original part was the fault. Reasonable confidence. If it is a high-cost part, do a quick confirm test on the old one before you call it (see related: Confirming a Part Is Actually Bad).
• Symptom stays: the original part was not the fault. Put the old part back if it is reusable, and move upstream. You very likely have an input fault: missing power, missing signal, a tripped safety, a wiring break.
• Symptom changes but does not clear: you have more than one fault, or the sub is not truly matched. Re-check the match and look for a second problem.

Watch the traps

• The reverse-fault trap. A bad upstream condition can damage the new part the moment you install it (a shorted load takes out the replacement, a stuck mechanism stalls the new motor). If a brand-new part fails instantly, stop. Find what killed it before you sacrifice another.
• The shelf-stock trap. Parts pulled from "the good pile" are not known-good until verified. Old capacitors drift, relays stick, boards arrive dead. Verify or buy new.
• The two-change trap. Already covered, but it is the one that gets everyone. One variable per test.

When substitution beats measuring

Reach for substitution when the part's function is not directly measurable (sealed logic, firmware behavior), when the fault is intermittent and will not hold still for probes, or when a meter reading lands in a gray zone you cannot call. It is also the right tie-breaker after an ambiguous measurement: the meter says "maybe," the swap says "yes" or "no."

Used with discipline - known-good part, one change, real-condition re-test - substitution is not a guess. It is a controlled experiment with a single variable, and that is exactly what diagnosis is supposed to be.

References

• OSHA 1910.147 - de-energize and verify before component handling
• NFPA 70E - safe work practices for electrical component testing
• Trade-standard diagnostic practice: single-variable testing and known-good comparison
• See related: Proving a Circuit Dead the Right Way; Confirming a Part Is Actually Bad