Termination torque QA field guide for electrical crews

Termination torque, and why it is the failure nobody had to have

Termination torque is how tight you make an electrical connection, measured against the value the equipment manufacturer published for it. Get it right and the connection carries current for forty years and never makes news. Get it wrong in either direction and it fails, and the failure is one of the most common avoidable ones in the trade.

There are two ways to be wrong, and both end at the same place. Too loose and the contact area between conductor and lug is smaller than it should be, so resistance is higher, so the joint runs hot under load. Heat oxidizes the contact faces, oxide raises resistance again, and the connection cooks itself looser in a slow loop until it arcs, opens, or starts a fire. Too tight and you crush the conductor, shear the strands, strip the threads, or cold-work the metal so it cannot hold clamping force. A crushed conductor has less metal carrying current and a stripped lug has no clamp at all. Both run hot. Both fail.

So there is a right number, it lives between those two failures, and it is not a feel. The connection that an electrician tightens by hand and calls good is the connection an infrared camera finds glowing six months later. The whole discipline is small: use the value the manufacturer gave, apply it with a tool that actually reads that value, and write down that you did. The rest of this guide is how that gets done and proven.

Does the NEC require a torque wrench?

Where the manufacturer gives a numeric torque value, recent NEC editions require you to use a calibrated torque tool to reach it. This lives at 110.14(D). The 2017 edition introduced the calibrated-torque-tool language, and later editions reworded it toward requiring an approved means to achieve the indicated value, with a calibrated torque tool as the default and manufacturer-described alternatives, such as shear-head bolts or breakaway devices, allowed where the instructions provide them. Confirm the exact wording against the edition the jurisdiction has actually adopted and any local amendments.

Read what that changed. For years torquing terminations was best practice that good electricians did and rushed crews skipped. It is now a code requirement wherever a value exists, which means an inspector can ask how the connection was torqued and what tool reached the number, and a hand-tight lug with no torque behind it is a violation, not just sloppy work.

The requirement is conditional on a value existing. If the equipment or the instructions give a number, you owe the calibrated tool and the number. If no value is provided anywhere, the code points you elsewhere for a value, which is the fallback table question covered below. The thing the rule does not let you do is read the number, own the tool, and then run the screw down by feel anyway.

Where do you get the torque value?

You get it from the equipment, in this order: the value printed on the equipment label or the wiring diagram inside the cover, the value stamped on the lug or connector itself, and the value in the manufacturer's installation instructions. Those are the same value, and it is the value the equipment was listed to. Use it. Do not reach for a generic chart when the gear in front of you tells you its own number.

This matters more than it sounds. The connection was tested and listed under UL to perform at the manufacturer's torque, not at an average from a table. A breaker lug, a meter socket, a panel main, and a terminal strip can each call for different values for the same conductor size, because the geometry and the metal of the connection differ. The label on the panel door is not decoration. It is the spec, and it overrides anything you carry in your head.

When the number is missing, work the order before you give up. Check the inside of the dead-front, the connector packaging, and the manufacturer's published instructions online by catalog number. Manufacturers will give the value over the phone if it is genuinely not printed anywhere. Only when there is truly no manufacturer value do you fall back to the generic tables, and you note that you did, because the next person needs to know the number came from a table and not from the gear.

The generic torque tables, and when you may use them

When the manufacturer provides no value anywhere, the NEC points you to published tightening-torque tables as a fallback, commonly the tables in the informative annex of UL 486A-486B, the standard for wire connectors, and a parallel set has appeared as an informative annex in recent NEC editions. These are organized by conductor size and by the slot or socket size of the screw, because the screw size, not just the wire, drives how much the connection can take. Confirm which annex your adopted code edition references before you cite it on a submittal.

Use them as the fallback they are, not as a default. The values are general figures meant to cover connectors that came without instructions, and they are deliberately conservative across a range of products. The instant the equipment gives its own number, that number wins, because the listing is to that number and the table is not.

The other thing the tables teach is scale, and it keeps you from gross errors. Small device and breaker screws land in pound-inches, often in the low tens. Larger mechanical lugs run from tens into the low hundreds of pound-inches. Big bolted bus and stud connections move into pound-feet. Knowing the unit and the order of magnitude for the connection in front of you is how you catch a tool set to the wrong scale before you ruin a lug.

The calibrated torque tool and keeping it calibrated

The tool is a torque screwdriver or torque wrench that reads in the unit the connection calls for, in a range that brackets the value. The two common kinds are the click type, which slips or clicks when it reaches the set value, and the dial or electronic type, which shows the reading as you pull. Both work. The point of either is that it stops you at a known number instead of a guess.

Match the tool to the value, not to what is on the truck. A torque screwdriver reading 10 to 50 pound-inches is the right instrument for breaker and device screws. A wrench rated in pound-feet for bus bolts is the wrong tool for a small lug, because the bottom of a wrench's range is its least accurate part and you cannot feel a 25 pound-inch target on a tool built for 200 pound-feet. Use the instrument whose range sits the target near its middle.

Calibration is the part crews forget, and it is the part that makes the number real. A torque tool drifts with use and abuse, so it carries a calibration certificate and a recalibration cadence, commonly annual or per the manufacturer's interval, and a dropped tool gets pulled and rechecked regardless of the calendar. An inspector or commissioning agent can ask for the calibration certificate, and a click wrench that nobody has checked in three years is producing numbers nobody should trust. The certificate is what turns your torque record from a claim into evidence.

Technique: the seated conductor, the sequence, the smooth pull

Before the tool touches the screw, the conductor has to be right. The strands go in fully, past the contact area, so the clamp lands on conductor and not on insulation, and nothing got nicked or cut back during stripping. A conductor that is short of full insertion, or one with half its strands trimmed to fit, fails the torque test no matter how perfectly you hit the number, because the number was never the only thing holding the joint.

Use the right bit and socket, square and fully seated, so you drive the fastener and not the corners of its head. A cammed-out Allen socket or a slipping screwdriver bit reads torque into rounding the fastener instead of clamping the conductor, and now you have a stripped head and a connection you cannot prove. On a connection with more than one bolt, like a multi-bolt lug or a bus splice, tighten in a sequence, snug all the fasteners first and then bring them to value in a cross or star pattern, so the joint seats evenly instead of cocking on the first bolt you finish.

Then pull smooth. Torque is a slow, steady pull up to the click or the target reading, not a jerk. A snap loads the joint past the set value before the tool can react, so a click wrench that reads correctly still over-torques a connection if you yank it. One smooth pull to the click, then stop. You do not give it the extra quarter turn for luck. The extra quarter turn is how the conductor gets crushed.

Should you re-torque electrical connections?

No, not unless the manufacturer specifically calls for it. This surprises people, because the instinct is that snugging an old connection back up makes it better. It does the opposite. A connection that was torqued correctly and has since seated and relaxed slightly is at the clamping force it was meant to hold. Run a torque tool back to the original value on that joint and you are now adding force on top of a connection that has already set, which can over-tighten it, damage the conductor or the connector, and void the listing.

There is a real difference between verifying and re-torquing, and it is worth being precise about. Re-torquing means setting a tool to the spec and tightening until it clicks, which adds rotation and force. Verifying tightness, where a maintenance standard allows it, is a more careful operation. NFPA 70B, the recommended practice for electrical equipment maintenance, addresses this directly and cautions that checking an existing connection back to the full specified value with a torque tool can result in an improperly terminated conductor, and where a verification torque is used it is held below the original, commonly not exceeding about 90 percent of the manufacturer's initial value, so the tool detects a loose joint without driving a good one tighter. Confirm the current 70B edition for the exact wording.

So the field answer is clean. At install, torque once, to the value, and mark it. After that, you do not chase tightness with a wrench unless the manufacturer's instructions tell you to re-torque on a schedule, which some specific products do. To find a connection that has loosened in service, you look at the witness mark and you scan it under load. You do not re-torque the whole panel as preventive maintenance, because that practice creates more bad connections than it fixes.

Marking the torqued connection: the witness mark

After a connection is torqued to value, you put a paint stripe across the fastener and onto the body it threads into. That stripe is the witness mark, and it does two jobs at once. It shows the connection was torqued, so an inspector walking the panel can see at a glance which screws have been done and which were missed. And because the brittle line breaks if the fastener turns, it flags later movement: a witness mark with a clean offset between the screw half and the body half is a connection that has backed out since it was marked.

The product is a torque-seal or tamper-indicating lacquer, sold under names like DYKEM Cross-Check and the Markal markers, in a squeeze tube or a paint pen. You lay the stripe so it spans the moving part and the fixed part. When it cures it is brittle, so any rotation cracks the line and leaves the two halves out of register. The color is up to the shop, but pick one that stands out against the gear and stay consistent so the marks read fast.

The mark is not a code requirement, and nobody should pretend it is. What it is, is the cheapest QA habit in the trade and the one inspectors quietly love, because it lets them confirm a panel of connections in seconds instead of breaking down and re-checking each one. On a job where torque has to be proven, the witness mark plus the torque record is what proof looks like.

Field example: torquing and recording a 200 A panel

A 200 A panel goes in on a commercial tenant fit-out, and the QA is set up before the first conductor lands. The torque values come off the panel: the wiring label inside the dead-front lists the main lug value, in this case 250 pound-inches as printed on that label, and the branch breakers carry their own smaller screw value, around 30 pound-inches as marked. Those are the numbers for this panel. A different panel can read different numbers, which is exactly why you read the label and do not assume.

The crew sets a pound-inch torque wrench for the main and a torque screwdriver for the breaker screws, both with current calibration certificates on file. Each conductor goes in fully seated, strands intact, then each connection gets one smooth pull to the marked value. A witness stripe goes across every fastener as it is finished, so a glance down the panel shows which screws are done.

Then it gets written down: the panel ID, each connection, the conductor, the spec value and where it came from, the tool serial and its calibration date, the date torqued, and that the witness mark was applied. When the panel is energized and loaded during commissioning, an infrared scan reads the terminations under load and confirms none of them run hot. That record, the torque plus the witness mark plus the thermal scan, is what a commissioning agent signs off, and it is what answers the question if anything goes wrong later.

Aluminum terminations need more care, not the same care

Aluminum loosens itself, and that is the whole problem. Aluminum and the copper or steel of the lug expand and contract at different rates as the connection heats and cools under load, and aluminum cold-flows, meaning it slowly deforms and creeps out from under a clamping force that copper would hold. So an aluminum connection torqued correctly can relax over time in a way a copper one does not, which is why aluminum terminations have a worse history of loosening, overheating, and failing.

The defenses are specific. Use a lug listed for aluminum, marked AL or AL/CU, never a copper-only lug on an aluminum conductor, because the listing and often the plating and the barrel geometry are designed for aluminum's behavior. Where the connector manufacturer calls for it, apply the antioxidant joint compound, also called an oxide-inhibiting compound, to the contact surfaces. Aluminum grows an insulating oxide film almost the moment it hits air, and the compound and the act of making the connection break through and seal out that film so the metal-to-metal contact stays low resistance. UL 486B addresses joint compound for aluminum connections; follow the connector instructions for whether and where it is required, because some connectors ship pre-filled and some require you to add it.

Torque it to the aluminum value with the same no-re-torque discipline as copper, and lean on the witness mark and the infrared scan to catch the creep, rather than re-torquing the connection on a schedule. An aluminum lug that shows a broken witness mark or reads hot under load is the one to open, clean, and re-make, not the whole panel. The grounding side of an aluminum service has its own connector-listing and corrosion concerns covered in the grounding electrode system and bonding guide.

The lug and the conductor: listed for the size and the metal

Torque only matters if the lug was right for the conductor in the first place. The connector has to be listed for the conductor size and material it is clamping: the right range of AWG or kcmil, and AL or AL/CU for aluminum or CU for copper only. A lug crammed with a conductor larger than its range, or holding more conductors than it is listed for, is a bad connection at any torque, because the geometry the value was set for does not exist.

There are two families. A mechanical lug clamps the conductor with a screw or bolt, and that is the connection you torque to a value. A compression lug is crimped onto the conductor with a tool and a die, and there you are not torquing, you are crimping with the correct die for that lug and conductor. The die has to match the connector, by the index or color the manufacturer marks, and the standard behind both is UL 486A-486B. A crimp with the wrong die is loose or over-crushed the same way a bad torque is, and you inspect it the same way you would inspect any connection: full conductor insertion, the right number of crimps, no cracks, and many compression lugs have an inspection window that lets you confirm the conductor bottomed out before you crimped.

Watch the conductors-per-lug count. A lug is listed for a specific number of conductors, usually one unless it is marked for more, and doubling up two conductors under a single-conductor lug is a violation and a hot connection waiting to happen. When two conductors have to land, you use a connector listed for two or you split to two lugs.

Dissimilar metals at the joint

Put two different metals together in a connection and you have set up a place for corrosion, and corrosion is just resistance growing where you cannot see it. Copper against aluminum is the classic pair: in the presence of moisture the two metals form a galvanic cell, the less noble metal corrodes, and the joint slowly goes high-resistance until it heats and fails. This is why an aluminum conductor on a copper-only lug is not just a listing problem, it is a corrosion problem.

The fix is built into the listing. An AL/CU connector is designed and often plated, commonly with tin, so copper and aluminum can meet through a compatible surface instead of bare metal on bare metal, and the antioxidant compound seals moisture and oxygen out of the joint. The plating and the compound are not optional dressing. They are what keeps the dissimilar-metal connection from corroding itself apart.

The same logic runs through grounding and bonding connectors buried in soil or concrete, where dissimilar metals and moisture are guaranteed and the connector must be listed for both metals and for the environment. When a connection joins two metals, the question is always whether the connector and the compound were chosen for that pair, because torque cannot save a joint that is corroding from the inside.

How do you verify a torqued connection actually holds?

You verify it in two steps that catch two different failures. The first step is at install: torque to the value with a calibrated tool and apply the witness mark, which proves the connection was made correctly the day it was made. The second step is under load: an infrared thermography scan, with current flowing, confirms the connection runs cool, which proves the joint is actually low-resistance and not hiding a defect the wrench could not feel.

The two steps are not redundant, because a connection can pass the torque check and still be bad. A nicked conductor, a contaminated contact face, a hidden crack in a crimp, or a lug that was right at value but still high-resistance for a reason you cannot see at install will all read fine to the torque tool and then run hot once current flows. The infrared scan is the only one of the two that tests the connection doing its actual job. Loose and high-resistance terminations are exactly what an infrared survey is built to find, and the infrared thermography inspection guide covers how that scan is run, the load it needs, and the NETA criteria that sort the find.

So the strong QA on any connection that matters is torque it, mark it, and scan it under load at commissioning. The torque and the mark are the workmanship record. The thermal scan is the proof under load. A connection that was torqued to value, carries a clean witness mark, and reads cool under real current is a connection you can sign for.

Bolted connections, NETA, and the low-resistance test

On the big stuff, bus joints, large bolted lugs, switchgear and busway connections, torque is half the acceptance and a resistance test is the other half. NETA acceptance testing, in its standard for electrical power equipment, calls for bolted connections to be torqued to the manufacturer's values and for the connection resistance to be measured with a low-resistance ohmmeter, the ductor or DLRO, which reads the joint in micro-ohms. Confirm the section and the criteria against the current NETA edition.

The resistance test catches what torque alone cannot on a large bolted joint. A bus splice can be torqued to value and still be high-resistance from corrosion, contamination, or plating damage on the mating faces. The ductor measures the actual resistance through the joint, and the common acceptance approach is comparison: a bolted connection that reads more than about 50 percent above similar connections at similar values gets investigated, because the outlier is the one with the hidden problem. NETA does not set a single universal maximum; the comparison to sibling connections is the test.

This is where torque QA, the resistance test, and the infrared scan converge into commissioning power QA. The bolted joints get torqued to value and witness-marked, the ductor proves them low-resistance cold, and the infrared scan proves them cool under load. On switchgear, busway, and any large bolted connection feeding critical load, that stack is the acceptance, and it is the same discipline scaled up from a breaker screw to a bus bar.

The torque record an inspector accepts

A connection that was torqued perfectly and never written down is, to an inspector or a commissioning agent, a connection that might not have been torqued at all. The record is what converts your work into something a third party can accept, and it does not have to be elaborate. It has to be complete and tied to the connection.

For each connection that matters, the record names the equipment and the specific connection, the conductor size and material, the spec torque value and where it came from, the tool used and its calibration date, the date the connection was torqued, and that the witness mark was applied. Where the connection was later verified under load, the infrared result rides with it. That set lets a reviewer confirm the value was right, the tool was trustworthy, the work was done, and the joint runs cool, without re-opening anything.

The reason this is worth the few minutes is the question that comes later. When a connection runs hot a year out, or a load drops, or an insurer asks, the record is what answers whether the joint was ever made right. A torque record with a calibration date and a witness mark is evidence. A note that says torqued is not.

What does a missed torque actually cost?

It costs a callback at best and a fire at worst, and the cheap part of the curve is gone the moment current starts flowing. A connection skipped or run by feel reads fine at install and energizes fine, so it passes the day it goes in. The cost shows up later, on someone else's clock, which is exactly why the corner gets cut: the person who skips the torque is rarely the person who pays for it.

Trace the bill. The skipped lug runs hot under load, an infrared survey finds it as a hot connection on a feeder, and now there is a return trip, an outage to de-energize and re-make the joint, and a re-scan to prove the fix. That is the lucky version, where the survey catches it. The unlucky version is the connection that nobody scanned, which arcs, opens the circuit, drops the load, or ignites the enclosure. A loose termination is one of the most common origins of an electrical fire, and it traces straight back to a connection that took an extra thirty seconds to do right and did not get them.

Set against that, the QA is almost free. A calibrated tool, the manufacturer's value, a paint stripe, and a line in a record cost minutes per connection at install. The estimate that prices torque QA into the work looks slightly higher than the one that does not, and it is the one that does not generate the callback. On the connections that matter, the torque and the record are the cheapest insurance on the job.

What to document

The documentation exists so that anyone after you can confirm the connection was made right without taking it apart. Capture enough that a reviewer can reproduce the value, trust the tool, and verify it was checked under load.

Record the equipment and the specific connection, the conductor size and material, the spec torque value and its source, the tool and its calibration date, the date torqued, whether the witness mark was applied, and whether the connection was infrared-verified under load. Where you fell back to a generic table because no manufacturer value existed, write that down too, because the source of the number is part of the record.

Common mistakes

• Tightening by feel with no calibrated tool, then calling a hand-tight connection good.
• Using a generic torque table when the equipment label or lug gives its own listed value.
• Re-torquing a connection that was never specified to be re-torqued, which over-tightens a seated joint.
• Skipping the antioxidant compound on an aluminum connection where the manufacturer requires it.
• Putting a copper-only or wrong-range lug on the conductor, so the value was set for a connection that does not exist.
• Leaving no witness mark, so nobody can tell which connections were torqued or whether one has moved.
• Cutting or short-inserting the conductor strands, so the joint fails no matter how perfect the torque.
• Jerking the wrench instead of a smooth pull, which over-torques the joint past the set value.
• Never recording the value, the tool, or the calibration date, so the torque cannot be proven later.

Field checklist

Standards and references

The NEC, NFPA 70, is where the requirement lives. Termination tightening is at 110.14(D), which requires reaching the manufacturer's indicated numeric torque value, with recent editions calling for a calibrated torque tool or another approved means the manufacturer provides. The exact wording shifted between the 2017 and later editions, so confirm it against the edition the jurisdiction has adopted and any local amendments. Article 110 also carries the general aluminum-conductor requirements that drive the AL and AL/CU listing of connectors.

The connector listing standard is UL 486A-486B for wire connectors, which is where the connector's performance, the AL/CU marking, the joint-compound requirement for aluminum, and the fallback tightening-torque tables in the informative annex come from. The connection is listed to the manufacturer's value under that standard, which is why the equipment value governs over any generic chart. Equipment manufacturers' installation instructions govern the actual value and any re-torque schedule, and they win wherever they are stricter than the general code.

On the maintenance and acceptance side, NFPA 70B, the recommended practice for electrical equipment maintenance, addresses verifying existing connections and cautions against re-torquing a seated connection back to the full specified value. NETA acceptance and maintenance testing specifications cover bolted-connection torque and the low-resistance ohmmeter test on bus and large connections. NFPA 70E governs the safety of working the connection energized. Confirm every section, edition, and value against the documents the jurisdiction and the project have actually adopted before citing them on a record.

Units, terms, and conversions

Torque shows up in a few units across a label, an instruction sheet, and a metric drawing, and reading the wrong one is how a connection gets ruined. The same value can read as pound-inches on a breaker, pound-feet on a bus bolt, and newton-meters on an imported product.

Small connections are almost always pound-inches, written lb-in or in-lb. Large bolted connections move to pound-feet, lb-ft. There are 12 pound-inches in a pound-foot, so a value read in the wrong unit is off by a factor of 12, which is the difference between a correct connection and a crushed one. Metric sources use newton-meters: one pound-foot is about 1.356 newton-meters, and one pound-inch is about 0.113 newton-meters. Match the tool's unit to the value's unit before you set anything.

lb-in / lb-ft / N-m

Pound-inches and pound-feet are the common torque units; 12 lb-in equal 1 lb-ft; newton-meters are the metric unit, with 1 lb-ft about 1.356 N-m

NEC 110.14(D)

The code section requiring terminations to be tightened to the manufacturer's value, with a calibrated torque tool or approved means where a value is given

Calibrated torque tool

A torque wrench or screwdriver, click or dial type, that reads a known torque value and carries a current calibration certificate

Witness mark

A brittle paint stripe across a torqued fastener and its body that shows it was torqued and cracks if it later moves

Antioxidant compound

Oxide-inhibiting joint compound applied to aluminum connections where required, to break and seal out the aluminum oxide film

AL / AL-CU

Connector markings for conductors the lug is listed to terminate: aluminum only, or aluminum and copper

Ductor / low-resistance ohmmeter

An instrument that measures a bolted connection in micro-ohms to find a high-resistance joint that torque alone cannot reveal

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What does an infrared inspection find?
• What is the difference between grounding and bonding?
• Should you use back-stab or screw terminals on a receptacle?
• What is a three-way match in accounts payable?
• What is arc energy reduction?