The harness test report that arrives with a shipment is usually one page: net numbers, three columns of "PASS", a date and a stamp. It is worth almost nothing in that form, and not because the supplier is hiding something. Each of the three standard electrical tests answers a narrow question, and a page of PASS marks hides which question was actually asked.
Continuity finds opens and miswires. Withstand-voltage — hipot — finds gross insulation faults, and stresses the assembly a little every time it runs. Insulation resistance is the sensitive one, the test that notices a problem before it becomes a failure, and the one most often reduced to a tick-box. None of the three, run perfectly, will find a bad crimp — which is the single most useful thing to know here, because the bad crimp is the defect that comes back eighteen months into the field.
Continuity: right net, wrong threshold
A continuity test drives a small voltage through every conductor path via a fixture that maps each connector position to a tester channel. It checks two things: every net in your wire list is connected end to end, and no two nets are connected to each other. Run on 100% of units it eliminates a whole family of build errors — a wire in the wrong cavity, a terminal not seated, a strand bridging two positions, a branch left off.
Three details decide whether it does that.
The threshold. Every continuity tester has a resistance value below which it declares "connected". Set at 10 Ω — a common default — a terminal barely touching, a crimp holding by three strands and a perfect joint all read PASS. Ask what the threshold is in ohms, and put it on the report. For power conductors ask separately for a four-wire milliohm reading against a stated limit; that is the only version of this test that says anything about joint quality.
The fixture. The test adapter is itself a wiring harness, built by the same shop, and if it is miswired the tester passes miswired product all day with total consistency. The defence is a known-bad sample — a deliberately faulted harness, one open and one short — run at the start of each shift to prove the tester detects both. Ask whether that is done, and ask for the record.
The map. The test program has to be built from your wire list, and somebody has to check that the translation was done correctly. If you are supplying a data wire list with the drawing, as described in the piece on control-cabinet wire numbering and labelling, ask for the tester's net list back and diff it against yours before the first article is approved. Ten minutes, once, catches an error that would otherwise be in every unit.
What continuity cannot see: intermittents (unless the harness is flexed while under test), insulation damage that has not yet become a short, wrong wire gauge, wrong terminal plating, and — as above — a mechanically bad but currently conductive crimp.
Withstand-voltage: gross faults, at a small cost to the part
A withstand-voltage or dielectric test applies high voltage between conductor groups, and between conductors and any shield or chassis connection, watching for breakdown or leakage above a set trip. It looks for insulation that is not there: a jacket pinched under a cable tie, a stray strand across a barrier, a nick from a mis-set stripping blade, thin clearance inside an overmoulded boot.
For machinery, IEC 60204-1 gives the reference practice most European customers will expect: a test voltage of twice the rated supply voltage or 1000 V AC, whichever is greater, at power frequency, applied for approximately one second between the power circuit conductors and the protective earth circuit. Your own product standard may say something different, and that is the point — the number has to come from a standard you name, not from whatever the tester was last set to.
Four things to establish before you accept a hipot column on a report:
- AC or DC, and at what voltage and dwell. These are different tests with different sensitivities, and the numbers belong on the report next to the result.
- Between what and what. A test run only from all conductors to shield will not find a conductor-to-conductor fault. Ask whether the test is run net-to-net as well as net-to-ground, and if not, why not.
- The leakage trip, in milliamps, and how it was chosen. This is the detail that separates a real test from a ritual. Every cable assembly has capacitance, and under AC the tester sees a continuous capacitive charging current that grows with cable length and with shield area. On a long shielded assembly, a tight leakage trip produces false failures — so the trip gets raised. Raise it far enough and the test can no longer see a genuine partial fault. Ask what the trip is set to and what the assembly's own charging current measures at that voltage; the gap between the two is the actual sensitivity of the test.
- How often it runs. Each application stresses the insulation, and repeated tests at full type-test voltage can extend a small void rather than reveal it. That is why routine production testing is commonly specified at a reduced voltage compared with a design-qualification test, and why re-testing a returned assembly at full voltage is a deliberate decision rather than a default.
What withstand-voltage cannot see: insulation that is thin, contaminated or damaged but still survives the applied voltage. It is a pass/fail question asked at one stress level. That is where the third test differs.
Insulation resistance: the sensitive one, and the numbers you should demand
Insulation resistance testing applies a DC voltage — 500 V DC is the usual figure for low-voltage circuits, and IEC 60204-1 uses that value with a minimum of 1 MΩ between the power circuit conductors and the protective earth circuit — and measures the resulting resistance in megohms.
Its value is not the pass/fail line. It is that the reading is a continuous quantity responding to conditions no other test notices: moisture in a jacket, flux or salt residue on a terminal, a hairline split in insulation, contamination from a dirty bench. A harness family that normally reads in the thousands of megohms and starts reading 30 MΩ is telling you something has changed in the process — the wrong cleaning agent, a humid week, a new reel of wire — even though every unit still passes a 1 MΩ limit by a factor of thirty.
This only works if you get the numbers. A report that says "IR: PASS" has thrown away the entire diagnostic content of the test. Specify that measured values are logged per unit or per lot, with the test voltage and the ambient conditions, and ask for the raw log rather than a summary. If your supplier's tester cannot export values, that is itself useful information about what the bench can and cannot do for you.
Ask for the log from the sample build and again from the first production lot, then compare the distributions. Readings an order of magnitude below the samples are a process change worth discussing before shipment.
What none of them find: the crimp
Every test above passes current through a joint or across an insulator at the moment of testing. A crimp that is mechanically wrong but currently conductive is invisible to all three.
The three ways crimps go wrong, and why the electrical bench misses them:
- Under-crimped. Compression is too low to make a gas-tight metal-to-metal interface. Contact resistance today is fine, so continuity passes. Over months air and moisture enter the interface, oxide forms, resistance climbs, and under vibration the joint frets open — the classic field failure that arrives after the warranty conversation.
- Cut or missing strands. A mis-set stripping blade shaves strands, or strands fold outside the barrel during insertion. Conductor cross-section drops, current rating drops with it, and a milliamp continuity test sees nothing at all.
- Over-crimped. The barrel is compressed past the point where the copper work-hardens or cracks. It measures beautifully and fails on vibration.
What finds these is a different bench: crimp height, pull-force and cross-section analysis — crimp height measured against a target and tolerance for that terminal and wire size, pull-force to a table by conductor size, and a micrograph showing compression ratio, absence of voids and a proper insulation-barrel grip. Crimp-force monitoring on an automatic press catches missing strands in-line, the only one of these that scales to 100%.
The comparison, in one table:
| Test | What it proves | What it cannot see | What the report must show |
|---|---|---|---|
| Continuity | Every net connected end to end; no unintended net-to-net connection | Intermittents, high-resistance joints, insulation damage short of a short, bad crimps | Resistance threshold in ohms, 100% or sample size, tester net list, known-bad verification record |
| Low-resistance (4-wire mΩ) | Joint and conductor resistance on power circuits against a numeric limit | Mechanical crimp quality when new | Measured mΩ per net, stated limit, test current |
| Withstand-voltage (hipot) | Absence of gross insulation faults and inadequate clearance at the applied stress | Marginal insulation that survives; anything below the applied voltage; the test itself stresses the part | AC/DC, voltage, dwell time, leakage trip in mA, which nets tested against which |
| Insulation resistance | Insulation condition as a measured value — contamination, moisture, hairline damage | Mechanical faults; open circuits; crimp quality | Measured MΩ per unit or lot at a stated DC voltage, not "PASS" |
| Crimp height / pull-force / cross-section | Mechanical integrity of the termination | Wiring errors, insulation faults | Measured crimp height with target and tolerance, pull-force values by conductor size, dated micrographs |
The logic is the same one that divides in-circuit from functional testing on a board, set out in the CMH guide to PCBA testing: ICT, functional and burn-in: each test has a defined blind spot, and the job is to stop two tests sharing the same one.
One more note on sequence. If your product is a flexing cable rather than a fixed loom, the electrical tests only mean something after the mechanical abuse — insulation resistance measured on a cable that has just finished a bend-cycle run tells you far more than the same reading on a fresh sample. That protocol is covered in the piece on braided charging cable strain relief and overmoulding.
Common questions
Should I require 100% testing or is a sample enough?
Continuity should be 100% on any harness with more than a handful of nets — it is fast, cheap, and the defects it catches occur randomly rather than as a lot-wide trend. Withstand-voltage and insulation resistance are more often sampled, precisely because dielectric testing stresses the part. State the split in the purchase order rather than assuming it, and if a supplier says everything is 100% tested, ask for cycle time per unit and check the arithmetic against their stated output.
The supplier sent a certificate instead of a report. Is that acceptable?
A certificate is a statement; a report is data. Ask for the log: net list, thresholds, measured values where the test produces one, equipment identity, calibration date, operator, date, and the lot or serial range covered. Fix it as a shipment deliverable in writing at quotation stage — the wider paperwork list is in the CMH guide to documents to ask a China supplier for before your first order.
Where does salt-spray testing fit in?
Nowhere near this list, which is worth saying because salt-spray chambers often sit on the same equipment roster. Salt spray is a corrosion test on materials and finishes — terminal plating, shell coatings — and speaks to environmental durability, not to whether this week's harness is wired correctly. If salt-spray results arrive in place of electrical test data, those are answers to two different questions. Ask which standard and how many hours, and file it as evidence about plating.
What to ask the supplier next
Shenzhen harness shops vary widely in what their outgoing bench can actually do, and the honest way to find out is to ask about settings rather than about equipment. Anyone can own a hipot tester; fewer can tell you what the leakage trip is set to and why.
Huisheng Electronics is a reasonable example to run these questions against. Its supplier-published equipment roster lists more than twenty production and test machines, and the outgoing-QC bench described on it covers terminal cross-section analysis, tensile and insertion-force testing, withstand-voltage and insulation checks, salt-spray and cable-continuity testers — alongside automatic cut-strip-crimp machines, servo crimping presses and benchtop crimpers on the production side. On paper that spans both halves of this article: the electrical tests, and the crimp checks that cover their blind spot.
What the roster does not settle is settings, coverage and paperwork. Test voltage, dwell and leakage limits are not stated; whether continuity runs on 100% of units or a sample is not stated; whether a per-lot report with logged values ships with each consignment is not stated. Those are not reasons to walk away — they are the three questions for your first email, and the sort of thing to confirm on a live video walkthrough of the bench rather than by message. Stated terms of MOQ from 1,000 pieces, samples in 7–10 days and volume in 15–25 days leave room to run a sample lot and read its data first.
The list to send:
- What resistance threshold is your continuity tester set to, in ohms?
- Is continuity run on 100% of units? If sampled, at what sample size?
- Do you verify the tester each shift with a known-bad sample, and can I see that record?
- Will you send the tester's net list so I can diff it against my wire list?
- Is the withstand-voltage test AC or DC, at what voltage, for what dwell time?
- What is the leakage trip set to, and what does an assembly of this length measure at that voltage?
- Is the dielectric test run net-to-net, or only net-to-ground?
- At what DC voltage is insulation resistance measured, and will you log measured megohm values rather than pass/fail?
- What crimp height target and tolerance do you use for each terminal and wire size, and how often is it measured?
- Can you supply pull-force values and a dated cross-section micrograph with the first article?
- Will a per-lot report with logged values ship with each consignment, in what format?
- What is the calibration date of each tester used on my parts?
Put the answers next to the pre-shipment inspection scope you plan to use — the CMH guide to product inspection before shipment covers how to fold electrical test verification into a third-party visit — and check them against the documented picture on the Huisheng Electronics factory profile, which sets out the registry check, the published equipment roster and the stated trade terms, and is explicit about what still has to be asked for.