Cable assemblies are the part of an electronic order that buyers treat casually. The BOM for a control panel might be forty line items; the harness is one of them, usually the cheapest, and often the one nobody engineers properly. Then the units arrive, the acceptance test fails on intermittent faults, and it turns out the crimp height on 2,400 contacts drifted because nobody set a pull-force standard and nobody checked it after the first fifty pieces. We have cleaned up that exact mess more than once. This is the process we now insist on before a harness purchase order leaves our desk.

The drawing is the product — everything else is negotiation

A harness without a proper drawing cannot be quoted, built or inspected consistently. "Same as last time" works until the person who built it last time leaves. A usable harness drawing contains, at minimum: a to-scale layout with overall length tolerances, a wire schedule listing every circuit with gauge, colour, insulation type and length, a connector table with manufacturer part numbers and pin assignments, and a bill of materials for sleeving, tape, heat-shrink, labels and loom. Branch lengths need their own tolerances — overall length alone tells the assembler nothing about where a breakout should sit.

Length tolerance is where most disputes start. On a 300 mm jumper, ±5 mm is normal and costs nothing. On a 3-metre cabinet harness, ±5 mm is unreasonable and will be quoted with a price premium or ignored. State tolerances per circuit in the drawing, not once in a note. We also require the drawing to define the measurement method — end-to-end pin to pin, or pin to a marked datum — because two assemblers measuring the same harness "correctly" will otherwise produce parts that don't fit the same panel.

Close-up inspection of assembled electronic components
Inspect what you can measure. Crimp height, pull force and continuity are quantifiable; "looks fine" is not.

Crimp quality is a process, not an inspection

The single biggest cause of harness field failures is bad crimping, and it is almost always a process-control failure rather than a lazy operator. A correct crimp has a specific combination of conductor crimp height, insulation crimp height and bell-mouth geometry. Those are set on the applicator and checked with a micrometer on samples. Once set, the setting drifts with tool wear and material batch variation. Good factories check crimp height at the start of every shift and after every reel change, and they log the values.

Pull-force testing is the other half. Each terminal type has a published minimum tensile strength — a 22 AWG contact might be specified at 60 N, an 18 AWG at 110 N. Ask for the test record, and make sure the sampling matches your volume: we require five pieces at setup, then periodic pulls through the run. A supplier who cannot produce a pull-test log is not controlling the process; they are hoping.

ItemWhat to specifyWhy it matters
Crimp heightValue + tolerance from terminal datasheetControls contact resistance and mechanical strength
Pull forceMinimum N per terminal, sampling planCatches tool wear and bad material early
Wire gaugeAWG + strand count, not just AWGFlexible vs stranded changes life in moving applications
InsulationMaterial + temperature ratingPVC vs XLPE changes ampacity and heat resistance
Connector brandExact MPN, or approved alternate listCounterfeits look identical and fail at mating cycles
Test dataContinuity + hi-pot report per batchProves the shipped batch, not the sample

Connectors: counterfeit risk is real and expensive

Branded connectors get copied. We have seen housings that mate fine on the bench and fail insertion-force specs in the field, and terminals with plating that measures thinner than catalogue. For any safety-relevant or high-cycle connector, specify the exact manufacturer part number, require the supplier to buy from an authorised distributor, and ask for the distributor invoice on the first order. If a quoted price for a known connector is 40% under everyone else, you are not getting that connector.

Where a genuine alternate is acceptable, list it explicitly. "Equivalent" as a blanket permission is how you end up with a different terminal plating on half the batch. Approved-alternate lists shift the decision to you, where it belongs, and let the factory buy competitively within a range you have already approved.

Worth the money: Ask for a first-article inspection report on the initial batch — dimensional checks on length, crimp height and pull force, plus a continuity map, with the physical sample retained. It costs USD 100–300 depending on harness complexity and catches drawing misreads before 5,000 pieces are built to the wrong revision.

Testing: continuity is the floor, not the ceiling

For low-voltage signal harnesses, a continuity and mis-wire test on a dedicated tester is the baseline and should be 100% on production runs. The tester compares against the drawing netlist and flags opens, shorts and wrong-pin conditions that a visual check will never see consistently. For power or mains-carrying assemblies, add a hi-pot (dielectric withstand) test — typically 1,000 V AC or the applicable standard for one minute, or a reduced time at elevated voltage — and require the report. Insulation resistance testing matters for anything running in damp environments.

When a harness carries shielded pairs for signal or data, specify the shield termination method and whether the shield is grounded at one end or both. Getting that wrong produces noise problems that show up as intermittent faults and are maddening to debug in the field. State it on the drawing, and require the tester to verify shield continuity separately from the signal circuits.

Marking, packaging and the last metre

A harness that cannot be identified cannot be installed correctly. Every assembly should carry a label with the part number and revision, and every circuit should be identified at both ends by the numbering scheme defined in the drawing. This is not decoration — it is what lets a technician at the destination troubleshoot without a wiring diagram in hand, and it is the difference between a twenty-minute repair and a two-day one.

Packaging protects the connectors, not the wire. Damaged contacts from a loose bundle rattling in a carton are a real defect source. Specify that connectors are capped or bagged, that bundles are coiled to a minimum bend radius appropriate to the cable type, and that rigid conduit sections do not press against contact faces in transit. For sea freight, add desiccant to cartons carrying any bare or plated contacts to avoid oxidation during the four to six weeks in a container.

What we watch before the order goes out

Three signals tell us whether a harness supplier will be reliable. They ask questions about the drawing — clamping methods, routing, environment — which means they read it. They quote pull-force and crimp-height control in their process description without being prompted. And they volunteer that some aspect of your spec is over-engineered or hard to build as drawn, rather than accepting everything. A supplier who agrees to every detail without comment is usually one who has not thought about how to make it.

None of this is exotic. It is the standard discipline of a competent cable house, and there are many of them in China and across Asia. What separates them from the cheap ones is not the machine or the operator; it is whether anyone in the building owns the process and keeps the records. Specify like it matters, ask for the data, and the harness stops being the part of the order that ruins the shipment.