The Drawing That Was Right and the Installation That Was Not
Why this matters
Techs are trained to distrust the print, and mostly that instinct is correct. Prints get superseded, they get drawn for a different option package, they get filed in the wrong door pocket. But the instinct has a blind side: when the print and the panel disagree, "the print is out of date" is only one of the two available answers, and it is the comfortable one. The other answer is that the print is right and somebody wired the unit wrong, and if what they wired wrong is a protective device, the unit has been running without a safety for as long as the installation has existed. This is one of those calls, followed end to end.
The call, and why nobody restarted it
Commercial packaged heating unit, intermittent shutdowns over about six weeks, and on one occasion a maintenance man reported a hot-plastic smell near the unit before it dropped out. Two prior visits by the shop, 1.5 hours each, both closed as no fault found because the unit ran normally while the tech was on the roof.
The smell is the part that changes the sequence. A hot-insulation or hot-plastic smell means something in the unit has been running above its design temperature, and re-energizing to reproduce the fault puts the same energy back into whatever was cooking. So the unit does not get restarted to watch it fail. If the reported smell had been gas rather than hot plastic, this is not the same call at all: everyone leaves the building immediately, nobody touches a switch, nobody turns on a light, nobody uses a phone inside, and the call goes to the gas utility from outside the building.
Containment before anyone diagnoses anything
Open the unit disconnect, apply your own lock and tag before opening the control enclosure, since 29 CFR 1910.333(b)(2) requires the circuit be de-energized and locked or tagged before work on or near exposed energized parts. Then prove it dead the only way that is worth anything: test the meter on a known live source, test the conductors you are about to work near, test the known source again (NFPA 70E-2021, 120.5). A meter that failed silently between the first and second test is why the third step exists.
Then tell the customer, in one sentence, that the unit stays off until the cause of the smell is identified. That sentence is the whole containment decision, and it is easier to say at the start of the visit than after a second event.
The measurement that was true and useless
The prior visits both included the same test, and it is written on both tickets: continuity checked across the high limit, limit reads closed, limit good.
That statement is true. The device was closed and it was good. What it does not establish is whether the device is anywhere in the path it is supposed to protect. A limit switch is not a component you evaluate on its own merits. It is a component whose entire value is positional, and a continuity check across its own two terminals tells you about the contact and nothing about the circuit.
That is the error the whole call turns on, and it is not a careless error. It is the natural error, because a continuity check across a device is fast, it is unambiguous, and it produces a clean result you can write on a ticket.
Reading the print against the strip
The control terminal strip is marked TB1 and carries 10 numbered positions. The unit print, revision letter and date both legible in the title block, and the model on the print matching the model transcribed off the unit nameplate, shows the high limit landed across TB1-4 and TB1-5, with TB1-5 carrying on to the heat contactor coil. Straightforward series arrangement: limit opens, coil drops, heat stops.
In the panel, one leg of the limit is on TB1-4 as drawn. The other leg is on TB1-8, and position 8 on the print feeds an auxiliary output, not the coil.
One conductor, on the wrong one of ten positions, and it is the difference between a protected unit and an unprotected one.
Three observations that separated factory from field
Before concluding anything, the question is whether this arrangement shipped that way or was made later. Three observations, all on the panel itself:
- The conductor landed on position 8 passes through a sheet metal opening with no bushing, while every factory conductor entering that section passes through a factory grommet.
- Its insulation colour matches nothing else in the control section, and its jacket is a different material than the factory control harness.
- Position 5 carries a witness mark under the terminal screw, a bright compressed ring in the plating with nothing currently landed on it. Something was torqued down on position 5 and was later removed.
Any one of those is suggestive. The witness mark is the one that closes it, because a factory that never landed a conductor on position 5 does not leave a compression ring there. Photographed before anything moved, along with the strip as a whole.
Proving position with the path, not the part
The decisive test is a continuity map of the coil path, run cold with the lock still on, and it takes about ten minutes.
Put one lead on the contactor coil terminal and the other on the control transformer's supply landing, then walk the path device by device: lift or isolate one device at a time and record whether the path opens. Every device that opens the path is in the coil circuit. Every device that does not is somewhere else.
Run that way, the coil path passed through the operating control and one other switch, and it did not pass through the high limit. Lifting the limit entirely did not open the coil path. The limit sat on a branch that fed the auxiliary output on position 8, which is why the maintenance man had at some point seen the unit's indicator behave oddly and thought nothing of it.
So the failure was not intermittent at all. The unit had been running with no functioning high limit since whenever that conductor was moved, and the intermittent shutdowns were a downstream protective device catching what the high limit should have caught first. The hot-plastic smell was the interval between those two setpoints, happening repeatedly.
The correction, and how it was confirmed
The correction itself is two minutes: move the conductor from position 8 to position 5, dress it through the factory grommet, torque the terminal to the value marked on the strip or in the equipment instructions, and confirm nothing else was disturbed against the as-found photographs.
Confirmation is the part worth doing carefully, and it is done cold. Re-run the same continuity map on the coil path. The path now opens when the limit is isolated and closes when it is restored, which is the only positive proof that the device is in the circuit it protects.
What does not happen: nobody jumpers the limit to see what the unit does without it, and nobody forces an over-temperature condition on a combustion path to watch the limit trip. You do not induce a fault on a combustion path, a pressurized path, a refrigerant-bearing path, or on anything serving a protective or relief function. If the equipment instructions document a functional trip test for that limit, that documented procedure is the only one that gets used.
Only after the cold confirmation does the unit come back on, and then the check is ordinary: normal startup, normal operation, and the auxiliary output on position 8 now behaving the way the print says it should, which is a second independent confirmation that the conductor is where the print puts it.
What the hours actually say
Two prior visits at 1.5 hours each and this one at 3.5 hours makes 6.5 hours of on-site labor across three visits, all of it on the same fault. The print described the correct arrangement on day one. The first visit had the print available.
The distinction that separates the 3.5-hour visit from the two 1.5-hour ones is not skill and not tooling. It is one choice: testing the path rather than the part. All three visits carry the same information about the limit's contacts. Only the third carries information about the limit's position.
What this generalizes to
Any device whose job is to interrupt something. Limits, safeties, interlocks, flow switches, float switches, pressure switches, door switches, freeze protection. For all of them, the useful question is never "is the device good," it is "what does the load path do when this device opens." A device can be functionally perfect, correctly sized, correctly set, and completely irrelevant because of which two terminals it landed on.
And on the documentation side: when a print and a panel disagree, run the identity check on the print first (model on the print against the model on the nameplate, revision block legible, option package matching what is installed), because that is what tells you which of the two is the suspect. Here the print's identity checked out, which is what turned the disagreement into a finding about the installation instead of a reason to set the print aside. A print you have verified against the equipment is evidence. A print you have not is a hypothesis, and the two get treated very differently.
References
- 29 CFR 1910.333(b)(2), de-energizing and locking or tagging circuits before work on or near exposed energized parts
- NFPA 70E-2021, 120.5, establishing and verifying an electrically safe work condition, including testing the instrument on a known source before and after
- Equipment manufacturer documentation for any functional test of a protective device, which is the only sanctioned way to trip one
- See related: How to Verify Documentation Against the Equipment in Front of You; When the Drawing and the Installation Disagree