October 21, 2026 · Technical Guide
The qualification report you asked for proves something specific: that a design, built in a particular configuration on a particular day, met a specification under a defined set of severities. It says nothing about the carton on your dock this morning. Receiving inspection is the only step in the chain that connects the two — and in most organisations it is the step that quietly collapses into opening the box, counting the pieces and filing the Certificate of Conformance.
Both directions of failure cost money. Under-inspect and a lot of connectors with 40 mΩ contacts, a substituted finish or a missing interfacial seal goes into a harness that is expensive to rework and impossible to re-inspect once potted. Over-inspect and you hipot every unit, stress insulation that was already tested at the factory, and add days of dock time to a part that was built correctly — while still missing the one defect that sampling would have caught, because you are testing the wrong characteristic.
The way out is not more inspection. It is a documented plan whose depth is set by three inputs: how severe the defect consequence is, how large the lot is, and what the supplier's history justifies. This guide sets out that plan in the order it should be executed — documents first, then physical and dimensional, then electrical — because each layer decides whether the next one is worth doing.
Inspection layers are not interchangeable. Each answers a different question, fails in a different way, and costs a different amount of time per lot.
| Layer | Question it answers | Effort | Typical escapes it catches |
|---|---|---|---|
| 1. Documentation | Is this lot the thing I ordered, built by the process I qualified? | Minutes per lot | Wrong specification revision, wrong finish class, substituted contact or backshell, expired or ambiguously aged elastomer, broker stock with no chain of custody, quantity shortfall |
| 2. Physical & dimensional | Do the parts in the box match the drawing? | Minutes per piece on a sample | Damaged threads or keyways, missing or wrong seals, wrong contact count or gender, plating substitution, machining burrs, foreign object debris, shipping damage, illegible or wrong marking |
| 3. Electrical | Will these parts carry current and hold voltage? | Seconds per piece, needs a bench | High contact resistance from contamination or thin plating, insulation-resistance and dielectric failures from moisture or cracked inserts, open or high-resistance shell bonds, intermittent contacts |
Run them in order, and treat a layer-1 failure as a stop. If the CoC names a different specification revision than the purchase order, you are not inspecting the part you designed for — you are inspecting an unknown part very carefully. Electrical testing an unidentified lot is the most expensive way to learn that the finish class is wrong.
Documentation is where a receiving inspection is won, because it is the only layer that can catch a substitution the hardware itself conceals. The minimum set for a MIL-spec circular connector lot:
Four document conflicts should stop a lot before anyone opens a single bag: a specification revision that does not match the order; a CoC part number that differs from the ordered part number in any character, including the finish letter and the keying letter; a date code so old that the elastomer's storage life is in question; and a finish certificate for a class you did not order. Each of those is a conversation with the supplier, not an inspection decision.
Acceptance sampling is the part of receiving inspection most often applied from memory and most often applied wrongly. Three points clear most of the confusion.
First, AQL is not a permitted defect rate. The Acceptable Quality Level is the quality level at which, under the plan, a lot is accepted with roughly 95 percent probability during normal inspection. A plan set at 1.0 percent AQL will sometimes accept a lot worse than 1.0 percent defective and will sometimes reject a lot better than it. That is not a flaw in the plan; it is what sampling does. Writing “AQL 1.0” on a drawing and then treating a single defect as a supplier failure is a misunderstanding of the tool, and writing “AQL 0” is asking for a plan that the tables do not contain — if you want zero acceptance, say so and use a zero-based plan.
Second, the standard has a lineage, so quote the edition. MIL-STD-105E was withdrawn and lives on as ANSI/ASQ Z1.4; ISO 2859-1 is the international twin, and it has a 2026 revision, so a plan should name the standard and the edition it was read from. The three inputs are the same in all of them: lot size, inspection level, and the AQL applied to each defect class. Inspection level II is the default; levels I and III trade sample size for discrimination; the special levels S-1 to S-4 exist for expensive or destructive checks and give very little discrimination by design.
Third, the real engineering decision is the defect classification, not the AQL number. Sorting characteristics into critical, major and minor is what makes the plan meaningful, and it is the part that must come from the design authority rather than from the quality department's template.
| Class | Example for a circular connector | Usual treatment |
|---|---|---|
| Critical | Contact retention failure, damaged master key or keyway, cracked insert, wrong contact gender, shell thread damage that prevents full mating | 100 percent screen, or a zero-acceptance plan — a sampling plan that can accept one critical defect is not appropriate |
| Major | Contact resistance above limit, insulation resistance below limit, dielectric failure, missing interfacial or rear seal, wrong finish class, wrong contact count, illegible part marking | Sampled, commonly 0.65–1.5 percent AQL, on the characteristics that are testable per piece |
| Minor | Cosmetic plating discolouration, marking position, minor moulding flash, packaging condition that did not damage the part | Sampled, commonly 2.5–4.0 percent AQL; worth tracking as a trend, not worth rejecting a lot over on its own |
The mechanic is a two-step lookup: lot size plus inspection level gives a sample-size code letter, the letter gives the sample size, and the AQL column gives the acceptance and rejection numbers for that sample. For general inspection level II, normal inspection, the sample sizes run like this — the lot-size bands are the part people misread, so check them against the table rather than against a remembered number.
| Lot size | Sample size (level II) | What it means in practice |
|---|---|---|
| 2–8 | 2 | Sampling is meaningless at this volume — inspect 100 percent |
| 9–15 | 3 | Inspect 100 percent; the sample would be most of the lot anyway |
| 16–25 | 5 | 100 percent below roughly 50 pieces is usually cheaper than the paperwork for a sample |
| 26–50 | 8 | First volume where a plan starts to pay for itself |
| 51–90 | 13 | Typical pilot-build lot |
| 91–150 | 20 | |
| 151–280 | 32 | |
| 281–500 | 50 | Common production lot for a single part number |
| 501–1200 | 80 | |
| 1201–3200 | 125 | Typical distributor stock lot |
| 3201–10 000 | 200 | Above this, move to level I unless the part is critical |
| 10 001–35 000 | 315 | |
| 35 001–150 000 | 500 | |
| 150 001–500 000 | 800 | Sample size grows far more slowly than the lot — that is the point of the plan |
| over 500 000 | 1250 |
Read the acceptance and rejection numbers from the AQL column for that code letter, or generate the plan with a free 105E/Z1.4 calculator rather than from a table on a blog — including this one. Then make the plan defensible in four habits:
Four measurements cover most of what goes wrong electrically, and all four fit on a bench that costs less than a rejected harness. The acceptance limits below are the ones Airoadcon publishes for its own production testing; they are typical of the MIL circular family rather than unique to one supplier.
| Test | Method / instrument | Typical acceptance | A failure usually means |
|---|---|---|---|
| Low-level contact resistance (LLCR) | Four-wire Kelvin milliohm meter, test current 100 mA maximum; per contact, not averaged — MIL-STD-1344 Method 3002 / EIA-364-23 | ≤ 5 mΩ per contact | Contamination or oxide film, insufficient contact force, thin or missing plating, wrong contact seated, damaged spring member |
| Insulation resistance (IR) | 500 VDC insulation tester, reading at 2 minutes — MIL-STD-202 Method 302 / EIA-364-21 | > 5,000 MΩ at 25 °C | Moisture in the insert, cracked or contaminated insulator, conductive debris, wrong elastomer, damaged seal path |
| Dielectric withstand (DWV) | AC or DC hipot, at sea level and where required at simulated altitude — MIL-STD-1344 Method 3001 / EIA-364-20 | Typically 1,500 VAC at sea level for the family, derated with altitude | Insert porosity, contamination, insufficient creepage, moisture, a defect that IR alone may not reveal |
| Shell-to-shell bonding | Four-wire milliohm measurement across mated shells, and across mated shell plus backshell — MIL-STD-1344 Method 3007 | < 2.5 mΩ for EMI-rated connectors | Plating damage on the coupling or accessory thread, seal or gasket between shells, missing grounding fingers, backshell not torqued |
Three sequencing rules make the results mean something. Measure resistance before you hipot: a dielectric test polarises insulation and can leave residual charge that skews the next insulation-resistance reading, and on some inserts it can alter the very surface films the contact-resistance test is looking for. Never judge contact resistance with a handheld multimeter, which drives enough open-circuit voltage to break through the films that cause the failure you are trying to find. And test in the configuration you are buying: a bare connector and a mated pair with a fitted backshell have different shell-bond and sealing behaviour, and a limit quoted for one is not evidence for the other.
Where altitude matters, the DWV value belongs to the altitude, not to the part in isolation — the same connector that holds 1,500 VAC at sea level has a much lower limit at 70,000 ft, and the derating curve is in the product specification. That relationship, and the way contact current rating falls with bundle size and ambient, is covered in the derating guide, and the measurement mechanics in the connector testing guide.
Visual and dimensional inspection is cheap and catches a surprising share of escapes, provided the inspector knows what is being asked of them and what is invisible.
What it catches: thread damage and cross-threading from shipping; burrs and machining marks on coupling threads; a damaged master key or keyway (a safety function, not a cosmetic feature); contact recess and protrusion that reveals a mis-seated or wrong contact; a missing, torn or deformed interfacial seal or rear grommet; foreign object debris in a cavity; a bent shell or deformed flange; marking that is wrong, illegible or not permanent; lockwire holes that are blocked or oversized; accessory thread condition where the backshell will land.
What it cannot catch: plating composition — cadmium, zinc-nickel and electroless nickel can look similar enough to fool an experienced eye, and the difference is a compliance question, so verify with XRF on a sample; subsurface voids and internal contact damage, which need X-ray or cross-section; contact retention force, which needs a pull test; seal integrity under pressure; and hardness or heat treatment of anything.
Dimensional checks split the same way. Thread gauges (GO/NO-GO) and keyway position can be checked per piece and should be. Contact spacing, cavity position and plating thickness are sample-and-instrument checks — optical or CMM, and a plating-thickness measurement on a sectioned or XRF sample — not something to claim per piece from a visual pass.
For connectors bought outside the manufacturer's authorized channel, the inspection question changes from “is this lot conforming?” to “is this even the part?”. The governing documents are AS5553 for designers and producers, AS6081 for distributors, and AS6171 for the test methods used to detect and quantify a suspect part.
| Indicator | What it looks like | Check that settles it |
|---|---|---|
| Broken chain of custody | No manufacturer lot code, or a lot code that cannot be traced through the distributor's paperwork | Ask for the full custody trail back to the manufacturer; if it stops at a broker, treat the lot as suspect |
| Resurfacing / “blacktopping” | Sanding or abrasive marks under the plating, edges rounded, marking re-applied over a polished area | 10–20× magnification at the marking and at shell edges; XRF for plating composition |
| Incoherent date codes | Mixed date codes in one bag, a date code that post-dates the shipment, or an implausibly old part at a new price | Compare every piece's marking to the lot documentation, not just the bag label |
| Finish mismatch | Plating looks right, certificate says otherwise, or the ordered finish is the restricted one | XRF on a sample per finish; check the finish callout against the specification table |
| Internal anomalies | Nothing visible externally | X-ray for contact position, solder voids or debris; DPA cross-section on a sample where the consequence justifies it |
| Performance drift | Parts that pass visual and fail electrically, or pass at low current and heat at rated current | LLCR and DWV on the sample; retention pull test; a rated-current temperature-rise check on a few pieces |
Disposition matters as much as detection. Suspect counterfeit parts should be quarantined, documented and dispositioned — not returned to the distributor as though they were a normal non-conformance, because that puts them back into circulation and, in a defence programme, into a reporting obligation. The procurement side of this problem — how a part becomes obsolete in the first place, and how last-time-buy and qualified-substitution decisions are made — is in the obsolescence and last-time-buy guide.
Separate the three things that a failing lot can mean, because they have different owners and different paperwork.
Two rules are worth writing into the procedure. Reworked lots are re-tested on the characteristic that was reworked, not on the characteristic that was easiest to measure. And a part that has already been through a repair cycle is not a receiving-inspection question at all — it belongs to the rules in the field repair and re-termination guide.
None of the above is enforceable unless the purchase order asks for it. These clauses cost nothing at quotation and prevent most arguments at receipt.
Documents first, because they are the only layer that catches a substitution the hardware hides. Then physical and dimensional checks on a sample sized from the lot, with the critical characteristics screened at 100 percent. Then four electrical measurements on a bench, in the order that keeps the readings honest. Classify defects before choosing an AQL, quote the edition of the sampling standard, use the switching rules, and put the plan on the record. Where the parts did not come from the manufacturer or an authorized distributor, add counterfeit screening and treat an unbroken document chain as a finding in itself. That is a receiving inspection that costs minutes per lot and is defensible in an audit, a customer's supplier review, or a failure investigation two years later.
Airoadcon supplies J599 (MIL-DTL-38999), ZH83723, MIL-DTL-26500, specialty military and circular connector families with 100 percent electrical testing on every unit — continuity, insulation resistance, dielectric withstand and shell conductivity, with no AQL sampling shortcut on the electrical screen — a lot code traceable to raw material heats, plating bath logs and test records, and a Certificate of Conformance with every shipment.
Send us the receiving inspection plan you have to satisfy — the characteristics, the sampling standard and edition, the acceptance limits and the documents your quality organisation needs — and we will confirm which of them are already covered by our production test data for the series you are considering, and quote lot acceptance data and test reports with the quotation. Our quality and test capability page lists the in-house electrical, mechanical and environmental methods, including the LLCR, IR, DWV, shell-bond, salt-spray and thermal-cycling limits referenced above. Free evaluation samples available.
Email: info@airoadcon.com | WeChat: 18991927716