How to Trace a Circuit on a Schematic

Why this matters

Tracing a circuit is the difference between diagnosing and swapping. A tech who can follow a control path on paper, predict what the meter should read at each point, and then confirm it, will find the open device on a call. A tech who cannot will replace the most likely-looking component and hope.

The trace itself is not the hard part. The hard part is keeping track of what you have proved, because a trace done in your head loses its findings the moment the customer asks a question. This article works the trace as a written artifact you fill in as you go, so that at any moment you can point at a line and say what you know and how you know it.

The trace log

Before you probe anything, draw six columns on a page or a tablet note. You will fill one row per test point.

The columns are load-bearing in that order. Predict before you measure. A reading recorded without a prediction beside it teaches you almost nothing, because you cannot tell a normal value from an abnormal one until you have committed to what normal was supposed to be. The Agree column is where diagnosis actually happens: a row that disagrees is either your fault, your print's fault, or the machine's fault, and the note column is where you say which.

Step 1: Make the circuit safe, then decide how you will work it

Open the disconnect, apply your own lock and tag, and prove dead with live-dead-live: verify the meter on a known live source, test the conductors, verify the meter again on that same known source. For panels, branch circuits and energized conductors the governing lockout and verification duty is 29 CFR 1910.333(b)(2), not 29 CFR 1910.147, which excludes electrical-hazard exposure on utilization equipment; the live-dead-live proving sequence is NFPA 70E-2021, 120.5.

Most control tracing is voltage tracing, which means the circuit has to be live for the readings to exist. Decide that before you start, not halfway through: wear what your employer's assessment requires, keep your free hand out of the enclosure, and clip your common lead to a fixed reference so you are moving one probe rather than two. If any part of the same job opens a compressor, an accumulator or a pressurized vessel, that stored-energy isolation is 29 CFR 1910.147 and takes its own lock.

Resistance and continuity tracing is the safer alternative and it demands the circuit be dead and isolated at both ends, otherwise a parallel path elsewhere in the ladder gives you a false continuity reading.

Step 2: Define both ends before you touch the middle

Write the two endpoints in your log first. Every control trace runs from a source (the transformer secondary, the supply side of the control fuse) to a load (a coil, a valve, a board input). Everything in between is a series of devices that either pass or block.

If you cannot name both ends on the print, you do not have a trace yet, you have a wander. Naming them also sets your reference: on a control circuit, the source's common leg is the reference for every voltage reading you take.

Step 3: Establish the reference and confirm the circuit is alive at all

First two rows of the log are always the same. Meter from the source's hot leg to the source's common. Predicted: full control voltage. Then meter from the load's return terminal to that same common. Predicted: near zero.

Those two rows prove that the control circuit is energized and that the return leg is intact. If the first row fails, you have no control power and the trace you were about to run is the wrong trace. If the second row fails, the fault is in the return path, and every measurement you take on the supply side will read normally while nothing works. Techs lose whole afternoons to that one.

Step 4: Half-split rather than walk

The instinct is to start at the source and step down the string one device at a time. Resist it. On a series string of devices, walking it sequentially takes up to as many measurements as there are devices. Splitting it in half each time takes about the base-2 logarithm of that count, rounded up.

For a string of 8 devices in series, sequential walking is up to 8 measurements; half-splitting is at most 3, because 2 to the third power is 8. That is roughly 2.7 times fewer measurements for the same certainty, and the gap widens as strings get longer.

Half-splitting works when the path is a simple series chain. It breaks in three cases, and you need to spot them on the print before you commit: a parallel branch (either leg can carry, so an absent reading downstream does not localize), a board-driven output rather than a physical contact chain (the board decides, so the fault may be an input the board is reading elsewhere), and a shared common that is itself open (which makes every device look open).

Step 5: Read the contact convention before you interpret any row

Contacts on a schematic are drawn de-energized and at rest. A contact drawn open may be a normally-open device that is supposed to close when its condition is met, and finding it open in a machine that is not running is not a fault.

That is why the Note column exists. "Open, but the machine is off and this device closes on proof of airflow" is a complete row. "Open" alone will send you to the parts truck.

Across a closed contact you expect near zero volts, because there is no meaningful drop across a good closed contact. Across an open one you expect the full circuit voltage, because the whole potential appears across the break. That inversion catches people: the device reading the largest number is the one that is not working.

Step 6: Fill the rows and let the Agree column carry the diagnosis

Work your split points, one row each. Do not summarize as you go and do not erase a row that turned out to be irrelevant, because a row you proved good is the reason you will not re-test it in twenty minutes.

When a row disagrees, do not immediately condemn the device. Ask the three questions in order: did I measure the point I thought I measured, does the print describe this build, and only then, is the device faulty. The first two are free to check and the third costs a part.

A filled-in trace log

A no-cool call. On the print, the compressor contactor coil sits at the end of a series safety string of 8 devices between the control fuse and the coil. Control circuit is in the 24 V class, common on the transformer secondary. Devices are numbered 1 through 8 from the source.

Six rows total. Two of them were the fixed reference rows from step 3, so the localization itself took the 3 split measurements the math predicted for a string of 8, plus one confirmation across the suspect device. Walking the string device by device from the source would have taken up to 8 measurements to reach the same place.

The reasoning behind the fourth row. Reading near zero after device 6 while reading full after device 4 is what brackets the fault. Everything upstream of the reading point is proved to pass, everything downstream is unproved, and the break sits between the last passing point and the first non-passing point. That bracket is the entire method.

What the log then told the tech to check. Device 6 on this print is a pressure-actuated safety, drawn normally open, closing when its condition is satisfied. So an open reading on a machine that is not fully running is expected, and the row's note says so. The real question became whether the condition that should close it was present. It was not: the device was doing its job and reporting a genuine problem upstream in the system it monitors. Condemning device 6 as a failed switch would have produced a machine that ran with its protection defeated.

What would change the answer. If device 6 had been drawn normally closed, an open reading with the machine at rest is a fault in the device or its wiring, and it gets tested directly. If the string had contained a parallel pair anywhere between devices 4 and 8, the half-split bracket would be invalid across that pair and you would test both legs individually before splitting further. If the second row had read full control voltage instead of near zero, you would have abandoned the string trace entirely, because an open return makes every device in the chain measure as though it were open.

How to verify the trace before you close the panel

Re-run one passing row. Pick a row you marked Agree early and re-measure it now that you have a conclusion. If it still reads what you recorded, your reference clip has not moved and your earlier rows are still valid. A common lead that fell off partway through invalidates every row after it, and this catches that in one measurement.

Force the state and watch the row flip. If your conclusion says a device should close when a condition is met, create that condition through the machine's own controls and confirm the row you predicted actually changes. Never induce a fault on a combustion path, a pressurized path, a refrigerant-bearing path, or anything serving a relief or protective function; if the only way to prove your conclusion is to defeat a safety, the conclusion has to be proved another way.

Hand the log to the print. Mark each proved-good device on the schematic with a tick and the open one with a circle, photograph it with your log, and attach both to the job record. If the fault returns, the next visit starts from six known rows instead of from nothing.

References

• 29 CFR 1910.333(b)(2), work practices for electrical work including lockout and verification of de-energized parts
• 29 CFR 1910.147, the control of hazardous energy, where the same job also isolates mechanical or stored energy
• NFPA 70E-2021, 120.5, establishing and verifying an electrically safe work condition
• See related: How to Read a Ladder Diagram; The Symbol Classes Worth Knowing Across Trades; The Difference Between a Schematic and a Wiring Diagram