The Connection Checks Worth Doing Before You Close It Up

Why this matters

Every shop has a version of the pre-close list and most of them are too long, which is the same thing as not having one. A list of eighteen items gets skimmed at the end of a long day, the tech ticks the ones that are easy to tick, and the items that actually mattered are the ones that took real effort. The result is a list that is complete on paper and caught nothing.

The fix is not a longer list or a stricter one. It is a sort with a criterion sharp enough that most candidate checks come off. What belongs here is not "the important checks", because plenty of important checks belong later. What belongs is the checks that stop being possible, or stop being cheap by a wide margin, the moment the panel goes on, the insulation goes back, or the ceiling closes.

The criterion, and the two bins

For each candidate check, ask one question: can I do this later at similar effort?

• If no, it goes on the pre-close list. Access is about to be lost, evidence is about to become unreadable, or the system is about to leave the state the check requires.
• If yes, it comes off, and it goes on the commissioning or first-service list instead.

The sort is the deliverable, not the list. Two shops with the same trade and the same equipment will sort a handful of items differently based on how their work is sequenced, and both can be right. What they cannot do is skip the sort and inherit somebody else's list.

Neither bin is ranked: the item that catches the most on a rooftop is not the item that catches the most in a crawlspace, so any ordering you put on either bin is your own site's, because the item that catches the most on a rooftop is not the item that catches the most in a crawlspace.

What comes off the list

This is the useful half, because these are the items that crowd out the ones that matter.

A loaded voltage drop across a termination. This is one of the best checks in the trade and it cannot be a pre-close check, because it requires the circuit energized and carrying real load, which is precisely the state you do not have while the panel is open and you are working in it. Taking it later is not a compromise, it is the only way to take it. Put it on the commissioning list with a blank for the current alongside the blank for the millivolts, since a drop reading without its current is not a measurement.

A second torque pass "for luck". Many listed connectors are designed and tested to be set once. Going around again is not a check, it is a modification to a listed assembly, and on a conductor that has already begun to cold-flow it sets a new preload on a smaller cross-section. Where the manufacturer's instructions or a maintenance specification do call for a re-check at a stated interval, that instruction governs and it belongs on that interval's list, not this one.

Anything that requires the system to be at operating temperature. Expansion travel at a guide, clearance at a penetration under load, thermal images of a terminal. All of these need the system hot, and the system is not hot while you are still building it. What does belong now is the cold reference: the scribe mark and the recorded temperature that make the hot reading interpretable later.

A general visual sweep with no stated criterion. "Looks good" has no failing condition, so it cannot detect anything, and it occupies a line a real check could use.

What has to happen now

Each of these fails the "can I do it later" question for a specific and stateable reason, and each catches something specific.

• Witness marks across every fastener that will be behind a panel or above a ceiling. Once closed, there is no way to distinguish a screw that has backed off from one that was never set. The mark makes the next visit a five-second read instead of a judgment call. It catches loosening, and it also catches the fastener that was never torqued at all.
• The termination record: conductor material, terminal marking, and where the torque value came from. Whether a terminal is identified for the conductor material is readable now and not readable through a closed enclosure. NFPA 70 (National Electrical Code) Article 110, in the edition your jurisdiction adopted, requires connectors identified for the material and installation to the listed torque, and OSHA gives listed-equipment instructions force at 29 CFR 1910.303(b) for general industry and 29 CFR 1926.403 for construction. This item catches the copper-only lug on an aluminium conductor, which is unfixable later without redoing the work.
• Achieved anchor embedment, anchor count and base material at every support point. The planned embedment is on the drawing. The achieved embedment exists only in the memory of whoever held the drill, and it is the number that decides what the point can carry.
• Support clearances and wear surfaces, at every point the run will move. Wear pad present, shield under the strap where the clamp bears on insulation, nothing within a fraction of an inch of a strut. These become invisible behind insulation and lagging.
• The cold reference for anything that will move. A scribe mark against a fixed bracket with the pipe temperature written next to it. Costs under a minute per point, and without it a hot reading a month from now is a number with nothing to compare against.
• Strain relief, meaning nothing hanging on a connection. Every conductor, tube and flexible connector between an anchor point and its termination should be slack, and no termination should be carrying weight. Once the enclosure is closed, a conductor bearing the weight of its own cable looks exactly like one that is not.
• A staged pressure or leak test with the joint visible. Visible is the operative word: the value is not the pressure, it is that you are looking at the joint while it is under it.
• Photographs with something in frame that gives scale and identity. A photo of an anonymous lug in an anonymous enclosure is not evidence. A photo with the panel label and a scale reference is.

Two of these carry real hazard, so do them properly

Insulation resistance testing on conductors that will be inaccessible. This test applies a high DC voltage to the conductor. De-energize, lock, tag and prove dead under 29 CFR 1910.333(b)(2) for general industry or 29 CFR 1926.417 for construction, proving live-dead-live per NFPA 70E-2021, 120.5, which reaches you through your employer's electrical safety program in the edition it adopted. Disconnect connected electronics before applying the test voltage, because the test will damage them. Confirm nobody is in contact with the conductor at either end before you press the button, including a second tech at a remote junction, and discharge the conductor after the test and confirm it discharged, because a long run holds a capacitive charge afterward.

Pressure testing. Use the test medium and pressure the specification or the adopted code names. A pneumatic test stores enormously more energy than a hydrostatic test at the same pressure, so where a pneumatic test is permitted it is done with a regulator and a relief device sized for the test, with the area cleared and nobody standing at a cap, a plug, a blind or a closure while pressure is applied. Never use oxygen as a test medium under any circumstance. Bring pressure up in stages with a hold at each, because the point of the staging is to find the defect at the lowest energy that will show it.

The list, filled in for one job

A rooftop unit changeout: power termination in the unit's disconnect, a refrigerant line set with four brazed joints, four support points anchored into a concrete curb pad, and a condensate connection.

The crew ran the sort and got eight items into the pre-close bin and five into the commissioning bin. Time on the pre-close bin, measured: 14 minutes for the whole unit, most of it the photographs and the anchor log.

• Witness marks on the two power terminations and on all four support clamps. Two minutes.
• Termination record: 6 AWG copper into a lug marked for copper, 75 C, torque per the marking on the lug, set with a calibrated driver. Two minutes.
• Anchor log: four points, base material confirmed as the curb pad, one anchor each, achieved embedment recorded per point. Three of the four matched the plan. The fourth did not, and that is the finding below. Four minutes.
• Support clearances: wear pads at both guide points on the line set, shields under both clamps that bear on insulation, nothing touching the curb edge. Two minutes.
• Cold reference: scribe marks at the two guides with the ambient recorded. One minute.
• Strain relief: whip slack at the disconnect, nothing hanging on a lug, condensate line supported independently of the trap. One minute.
• Staged pressure test on the line set with the joints visible and unwrapped. Held at each stage per the specification. Counted separately from the 14 minutes because it is its own operation.
• Photographs: each brazed joint before wrapping, the open disconnect with the panel label in frame, all four anchor points. Two minutes.

Deferred to commissioning: loaded voltage drop with current recorded, hot travel read against the scribe marks, thermal image of the disconnect at load, superheat and subcooling, and the condensate flow check with the unit running.

The one it caught

The fourth anchor point came up short. The tech drilling it had hit steel at roughly two thirds of the planned depth and set the anchor at the depth the hole gave rather than relocating, which is a four-second decision that would have been invisible ten minutes later. Because the log has a column for the achieved depth rather than a tick box, it was written down, and because it was written down it was questioned before the crew came off the roof. That point was re-drilled clear and reset while the lift was still there.

Put the trade in the right units, because it is easy to flatter. The pre-close bin cost 14 minutes of labor that was already on the ticket. The alternative was a return trip: two techs, a lift, and roughly 3.5 hours of shop time that nobody pays for, plus a rescheduled slot. Those are not the same kind of hour, so 14 minutes against 210 is not a return on anything. It is a warranty-avoidance trade, and the honest way to say it is that a quarter of an hour of sold time stood in for most of a half-day of unsold time, once.

That "once" matters too. The list does not catch something on every job, and a shop that judges the list by its hit rate will drop it in a quarter. Judge it by whether the things it does catch are the things you cannot fix later.

What moves an item between bins

• A joint that will be buried, encased, wrapped or above a hard ceiling moves several commissioning items forward, because "later" for that joint is not a service call, it is demolition.
• A connector whose instructions specify a re-check interval moves the re-torque from "off the list entirely" onto that interval's list, with the manufacturer's number in the blank.
• A system someone else will commission raises the weight of the cold references and the photographs, because what you wrote down is your only channel to that person, and it makes identity in every frame non-optional.
• Work under a specification that requires documented tests removes your discretion for the tests it names. Your sort applies to everything it does not name.

References

• NFPA 70 (National Electrical Code), Article 110, in the edition your authority having jurisdiction adopted, for connectors identified for the conductor material and installation to the listed torque.
• 29 CFR 1910.303(b), general industry, and 29 CFR 1926.403, construction, requiring listed equipment to be installed and used in accordance with its listing and labeling.
• 29 CFR 1910.333(b)(2), general industry, and 29 CFR 1926.417, construction, for de-energizing and verifying before insulation resistance testing; NFPA 70E-2021, 120.5, for live-dead-live, binding through the employer's electrical safety program in the edition adopted.
• The adopted mechanical, plumbing or fuel gas code and the project specification, which own the test medium, test pressure, staging and hold times.
• See related: How to Make a Terminated Connection That Lasts; How Spacing and Anchorage Decide What a Support Carries; The Leak Check After Every Connection.