Inline Measurement for Connector Pins and Terminals

A connector fails for reasons that are almost never visible in the finished assembly. A terminal sits a tenth of a millimetre too high in the housing and the mating half engages with excess force; a pin row drifts out of position by half a pitch at the far end of a long strip and the board will not seat; a crimp barrel is pressed 0.05 mm too tall and the pull-out force falls below the harness specification after thermal cycling. All three defects are produced by a tool that is wearing, at a rate that is slow enough to pass a first-article inspection and fast enough to ruin an afternoon of production.

This page is about catching those three defects where they are made: at the crimp press, at the connector assembly station, or at the packing station, with a POMEAS HM series inline measuring head looking at the terminal or the pin row as it passes. It covers which connector dimensions a silhouette measurement can hold and which it cannot, why the mounting orientation is the single most consequential decision in the project, the tolerance arithmetic that has to be done before a capability figure is quoted, and where the other three POMEAS product lines are the better instrument. The bench and one-shot cases have their own pages, and the last sections point to them.

Why connectors are a gauge problem before they are an inspection problem

Three properties of connector manufacturing make it different from, say, shaft grinding. The first is that the critical dimensions are small and the tolerances are tighter than the feature: a 0.5 mm pitch board-to-board connector carries a pin position tolerance of a few hundredths of a millimetre, which is a quarter of the pin width. The second is that the part is made at speed and in continuous strip: stamped terminals arrive on a carrier, crimped at hundreds of strokes per minute, and any measurement that samples one terminal per shift is measuring the tool on a good day rather than the process. The third is that the failure is assembled, not dimensional: a terminal that is 0.03 mm out of position may still assemble, and the defect only appears as a return from the customer's end-of-line test.

An inline gauge addresses the third property directly, because it measures every part and therefore measures the trend rather than the sample. That is the commercial argument, and it is worth stating plainly: the value of inline measurement on a connector line is not a tighter number than the bench system produces, it is the part number at which a drifting tool is detected. A crimp press that starts to wear at 09:00 and is caught at 09:15 costs fifteen minutes of scrap. The same press caught at the end of the shift costs the shift.

What it does not do is replace the laboratory. Insertion and withdrawal force, crimp pull-out force, plating thickness, solderability and the high-potential test are not optical quantities and never will be. The inline head sits in a chain that still contains a pull tester and a first-article report; its job is to keep the parts that reach those instruments worth testing.

What a silhouette head can hold on a terminal or a pin row

The HM family measures a shadow. A collimated beam projects the part onto a receiver and the software evaluates dimensions from the outline in a single exposure: 35 µs, no scanning, no moving parts. Everything measurable follows from that sentence, and so does everything that is not measurable. A terminal presented side-on is an excellent subject, because most of its drawing is a set of lengths that live on the silhouette. A connector housing presented from above is a poor subject, because its drawing is mostly depth.

Feature on the connector drawing	Inline silhouette	What governs the result
Crimp height and crimp width of a terminal	Yes, when presented side-on	Both are distances across the optical axis once the terminal is rotated 90 degrees; the crimp barrel must be clear of the carrier strip shadow
Crimp bell-mouth length, cut-off tab length, brush height	Yes, side-on	Axial dimensions on the same silhouette; a burr on the cut-off tab reads as an increase in length
Pin pitch within a row, row spacing	Yes	Full-field exposure has no accumulation error along the row; the datum is the first and last pin, not a stage coordinate
Pin width, pin thickness (one direction), tip taper length	Yes	Pin thickness is measurable only in the direction the head looks; the other thickness needs a second view
Pin length, terminal protrusion beyond the housing face	Yes, side-on with the housing face square to the view	The housing face must be a visible edge; a rounded or chamfered face shifts the reference by a fraction of a pixel
Terminal insertion depth inside the housing	No	Internal to an opaque moulding; a silhouette cannot see it at any magnification
Pin coplanarity across a row (tips in one plane)	Partly, and only in one direction	A side view gives the tip line in that plane; a pin displaced towards or away from the head is invisible in that projection
Crimp compaction, strand placement, wire brush symmetry	No, as a section property	Lies inside the barrel; a silhouette shows the outside of the crimp, not the conductor inside it
Insertion force, pull-out force, contact resistance	No	Mechanical and electrical quantities, not dimensions
Plating thickness, porosity, whisker growth	No	Requires XRF, SEM or a chemical method

Two of those rows decide most connector projects, and both are usually argued about after the purchase order rather than before it. The insertion depth row is the one that hurts: a customer who asks for "terminal seating verification" is often asking for a dimension that no optical instrument can see inside a moulding, and the honest answer is that seating is verified electrically, mechanically by a pull test, or by measuring what protrudes on the mating side. The coplanarity row is the one that is routinely over-promised, and the next section is devoted to it because it is the most common source of a failed acceptance test on a connector line.

The mounting orientation decides the project

A silhouette head has one optical axis, and every dimension it reports is a distance perpendicular to that axis. Rotating the head around the part therefore does not change the instrument, it changes the drawing that the instrument can see. On a terminal press the practical choice is between three views, and they are not interchangeable.

View	What it measures well	What it cannot see	Typical station
Top-down on a strip (head above the carrier)	Pitch, pin width, terminal outline, cut-off tab length	Crimp height, any vertical dimension, anything under the carrier	Stamping die exit, terminal reel inspection
Side-on to the strip (head horizontal, looking across the terminals)	Crimp height and width, bell-mouth, protrusion, pin length, tip line in that plane	Pitch along the row if the terminals overlap in projection; displacement towards the head	Crimp press exit, housing assembly
End-on to a pin row (head looking along the row axis)	Row-to-row spacing, pin thickness in the second direction, tip-line straightness in the other plane	Pitch within the row, which collapses into a single outline	Press-fit connector seat, mating-face inspection

The consequence is arithmetic rather than opinion: a single head gives one projection, so a requirement that spans two planes needs two heads, two stations, or a part that is rotated between two exposures. Projects that fail at acceptance almost always fail here, because the quotation was written for "connector inspection" and the drawing contained one callout in each of two planes.

The side-on view deserves a warning of its own. A crimp terminal on a carrier strip is a chain of parts at a fixed pitch, and when the head looks across the chain the neighbouring terminals overlap the one being measured. The measurement window must therefore be placed between the carrier and the cut-off tab, in the gap the press leaves open, and the window has to be tall enough to contain the tallest feature plus a margin for the strip position tolerance. That is a mechanical design task on the fixture, not a software setting, and it is the reason inline crimp gauges are usually built as a dedicated station rather than clamped onto an existing press frame.

Pin coplanarity: what a silhouette can and cannot prove

Coplanarity is the requirement that all the tips of a pin row lie in one plane, and it is the dimension that press-fit and board-to-board connectors live or die on. It is also the requirement most often mis-scoped, because "coplanarity" sounds like a single number while the measurement is a projection.

With the head side-on to the row, the silhouette contains the tip line as a curve in the image plane. The deviation of that curve from a fitted straight line is a real, useful number: it is the tip-line straightness in that plane, and it catches the common failure of a single pin standing proud or sitting low after a bending operation. What the same image cannot contain is a pin that is displaced along the optical axis, because a pin moved 0.05 mm towards the head projects onto very nearly the same pixels as one that did not move. A row measured only this way can pass a part whose true coplanarity is out of tolerance.

Two honest options exist. The first is a second head at 90 degrees to the first, so that the two projections together constrain both directions; the pair then covers coplanarity except for the component along each optical axis, which is the residual that a laser profile or a dedicated coplanarity gauge exists to measure. The second is to keep coplanarity on a dedicated instrument and use the inline head for the dimensions it holds unambiguously, with the split written into the control plan. The second option is usually the cheaper and the more defensible one, and it is what the lead-frame page in the telecentric line describes for the flat-stamped case, where the coplanarity of a lead frame is a different problem from the coplanarity of a pin row.

Where the inline head earns its place on a press-fit connector is not coplanarity but drift: pitch, pin width, pin length and tip taper all move with punch wear, and all of them are pure silhouette dimensions. Gauging those on every part leaves the coplanarity gauge measuring a connector whose other dimensions are known good, which shortens every argument about a rejected lot.

Sizing the head against a connector family

The published family spans five fields of view, and the trade-off is the whole selection argument: the same receiver resolves a smaller field more finely, so the smallest head holds the tightest numbers while the largest head covers the most part. Selecting a head is a question of the largest span the head must see in one exposure and the tightest tolerance that span must hold, not of buying the best number on the sheet.

Head	Field of view	Repeatability	Position accuracy	Connector-scale fit
HM-1006	φ6 mm	±0.05 µm	±0.2 µm	Fine-pitch board-to-board rows up to about 12 positions at 0.4 mm pitch (derived: 11 x 0.4 + 0.2 mm pin = 4.6 mm), single terminals, crimp barrel sections
HM-1015	φ15 mm	±0.05 µm	±0.3 µm	20 positions at 0.5 mm pitch (derived: 19 x 0.5 + 0.2 = 9.7 mm), FPC and FFC contacts, discrete terminals at the press
HM-1040	φ40 mm	±0.08 µm	±1.0 µm	10 positions at 2.54 mm pitch (derived: 9 x 2.54 + 0.64 = 23.5 mm), circular connector inserts, terminal blocks
HM-1065	φ65 mm	±0.1 µm	±1.2 µm	20 positions at 2.54 mm pitch (derived: 48.9 mm envelope), multi-row headers, harness-side connector bodies
HM-1120	φ120 mm	±0.2 µm	±2.5 µm	40 positions at 2.54 mm pitch (derived: 99.7 mm envelope) or a whole fine-pitch row plus fixture datum in one view

Read the table as a decision, not a ranking. If the requirement is a 0.4 mm pitch row measured to a few microns, the φ6 head is the only sensible choice, and putting a φ120 head on that job buys coverage nobody asked for at twelve times the position uncertainty. If the requirement is the span between the first and the last contact of a 40-way header, no smaller head contains the part at all, and the φ120 is correct even though its position accuracy is the largest in the family - because the alternative is two exposures and a stitching step that reintroduces exactly the accumulation error a full-field measurement avoids.

The arithmetic behind the fit column is deliberately simple, and it is worth doing in the quotation rather than at commissioning: positions minus one, times pitch, plus one full feature width, plus a margin for the fixture locating tolerance. A row that just fits has no room for the part to be off-centre in the fixture, and a head chosen at 100% of the nominal span will be replaced within a month by a larger one.

Do the tolerance arithmetic before you promise a capability

A capability figure is a promise about the ratio of tolerance band to measurement variation, and the measurement variation is the sum of everything between the drawing and the reported number: the head, the fixture, the part, the temperature and the software recipe. On a connector line the head is almost never the dominant term, which is the opposite of what most buyers assume.

Take a pin position requirement of 0.10 mm diameter true position, which is a tolerance band of ±0.05 mm on each axis, on a 0.5 mm pitch row. A conventional gauge-study rule allocates at most 10% of the band to the measurement system, so the entire system - head, fixture, part presentation - must vary by no more than 0.010 mm. An HM-1015 with a published position accuracy of ±0.3 µm consumes 3% of that 10 µm budget; the same head on an HM-1120 with ±2.5 µm consumes 25%. Everything else in the budget is mechanical: if the connector body is located in the fixture with 5 µm of play repeating in the same direction, the fixture has already spent half the budget before the optics are considered.

That is why the first deliverable of a connector inline project is a fixture drawing rather than a head datasheet. The useful questions are how the part is located (a moulded feature, a pin, a datum face), how it is clamped (and whether the clamp distorts the housing), how repeatable the transfer is, and whether the fixture can be re-mastered against a known artefact without removing it from the line. A head that is ten times better than the fixture is money spent on the wrong half of the problem.

One genuine advantage of a full-field exposure deserves the arithmetic, because it is often overlooked in comparison tables. A measuring machine that moves a stage from feature to feature accumulates the stage error over every step; a single exposure measures every feature in one coordinate frame, so the error on a span of 19 pitches is the error on two edges, not 19 times a per-step error. On a 0.5 mm pitch row that is the difference between a span uncertainty that stays near the head's position accuracy and one that grows with the number of contacts - and it is the reason fine-pitch connector metrology moved to full-field systems in the first place.

Cycle time, exposure and the moving part

The published exposure is 35 µs and a single capture can carry up to 100 measurement items, which is what makes inline connector gauging practical: a row of terminals, each with a height, a width and two end positions, is a few dozen items in one shot rather than a few dozen exposures. The consequence is that exposure is never the constraint on a connector line. The constraints are the transfer, the trigger and the computation.

Motion blur is the one optical constraint that a fast line imposes, and it is worth computing rather than assuming. Blur equals part velocity multiplied by exposure time, so a terminal moving at 0.5 m/s during a 35 µs exposure smears by 17.5 µm (derived: 500 mm/s x 35 µs), which is larger than the entire gauge budget in the example above; at 0.1 m/s the same exposure smears by 3.5 µm, which is tolerable for a 10 µm budget. The practical answers are the ones connector builders already use: measure during a dwell, trigger on the press encoder rather than on a timer, or accept a stop-and-go transfer at the gauge station. A strobe does not defeat blur on a continuous strip; it only freezes what the sensor sees for the duration of the exposure.

The second constraint is the handshake. An inline gauge that reports after the part has left the station can still reject it downstream, but it cannot stop the press, and a reject that has to be found by a downstream sensor is a reject that will occasionally be missed. The useful design puts the measurement inside the machine cycle: the press dwells, the head exposes, the result is returned before the press strokes again. That is a control-integration requirement, and it decides the field of view that is affordable, because a larger head has no penalty in exposure time but does have one in the mechanical envelope around the tool.

What actually breaks in a connector plant

The failure modes that damage an inline connector gauge are rarely optical, and they are worth listing because each one has a design answer that costs less than the downtime it prevents.

Oil and stamping lubricant are the first. Terminal strip arrives coated, and a film on the head window or on the backlight diffuser scatters the collimated beam, which softens both edges of the shadow symmetrically and therefore biases every width in the same direction. The signature is a slow drift of all widths together, which is easy to mistake for tool wear. A positive-pressure air purge on both the head and the illuminator, plus a window that can be replaced in under a minute, turns a shift-losing event into a scheduled task.

Bright tin plating is the second. A reflowed bright tin surface is specular and rounded at the edge, so the sub-pixel edge algorithm finds the transition at a slightly different place depending on the angle of the light that reaches it; matte tin is diffuse and more stable, gold is more stable still. The consequence is not that tin cannot be measured, it is that the lighting recipe and the edge threshold belong to the part number, under revision control, and must be re-qualified when the plating finish or the supplier changes. A recipe copied from a gold-plated sample to a tin-plated one will read consistently and be consistently wrong.

Thermal drift is the third, and on connectors it has two sources. The instrument drifts with the ambient temperature of the cabinet, and the part itself changes size between the press and the gauge: a glass-filled polymer housing with a typical coefficient in the 20-30 ppm/K range (industry handbook value, not a POMEAS specification) moves a 10 mm pitch span by about 1.0-1.5 µm over a 5 K change (derived). The fix is procedural rather than optical: master the gauge on a reference artefact at the start of each shift, keep the reference at the same temperature as the parts, and record the master value so that instrument drift and part growth can be separated when a trend appears.

Static and handling damage are the fourth. A head mounted on a press frame sees the press's shock, not the shop floor's vibration, and a bracket that rings at the press frequency will produce a measurement that is stable when the press is idle and noisy when it is not. Mounting the head on a separate column tied to the fixture rather than to the press frame, and triggering after the dwell, removes most of it. Connector lines also run at high static potential near the strip, and the head's cabling should be routed as the machine builder's ESD practice requires rather than as the shortest path.

Where the inline head stops and the other lines start

POMEAS builds four product lines, and a connector can legitimately be inspected by any of them. The choice is decided by where the part is in its process and what the question is, not by which line is being promoted.

Line	Instrument	When it is the right answer for a connector
D - inline measurement	HM series head at the station	Every part, at the press or assembly station, dimensions that drift with tool wear: pitch, crimp height, protrusion, pin length
A - zoom optics	Motorized zoom lens with a camera	One station that must cover several connector families at different magnifications, with an operator or a robot presenting the part; visual plus dimensional judgement
B - telecentric	Telecentric lens on flat stamped parts	Lead frames and flat contacts where the requirement is high-magnification 2D geometry with no perspective error, measured off-line or in a dedicated station
C - flash measurement	One-shot flash measurement system	Batch QA in a tray or on a glass stage, many parts per exposure, the first-article and audit role that keeps the inline gauge honest

The boundary that matters commercially is between D and C. A flash system measures a connector lying still on a stage, dozens at a time, with a field large enough for the whole part and a software library that knows what a connector looks like; it is the correct instrument for incoming inspection, for the first article and for the audit that proves the line is still capable. The inline head measures one part at a time in the machine, with a field just large enough for the feature family, and no operator. Buying the inline head to do the flash system's job produces a station that is slower and less flexible than the bench instrument; buying the flash system to do the inline head's job produces a queue, a sampling plan and the defect that ships.

A commissioning checklist for a connector gauge

The list below is the one that separates a station that survives a year from one that is bypassed in a month. It is short on purpose.

• Write the feature list with the drawing callout numbers, and mark every callout that the silhouette cannot hold, before the purchase order rather than at acceptance.
• Decide the view for each callout, and count the planes: one callout in each of two planes needs two heads or two stations.
• Size the head on the largest span plus one feature width plus fixture margin, not on the nominal part.
• Draw the fixture, including how the part is located, how it is clamped, and whether the clamp deflects the housing.
• Compute the gauge budget: 10% of the band, then subtract fixture repeatability before subtracting the head.
• Fix the trigger to the machine encoder or the dwell, and compute the blur at the real line speed.
• Put the lighting recipe and edge threshold under revision control per part number, and re-qualify them when the plating finish changes.
• Install an air purge and a field-replaceable window on both the head and the illuminator.
• Mount the head on a bracket tied to the fixture, not to the press frame, and check the reading with the press running.
• Define the master artefact, the re-master interval and the record that separates instrument drift from part growth.

Frequently asked questions

Related reading

• Inline dimensional measurement system: HM series overview and published specifications
• HM series online image measuring heads: five models, ranges and selection
• Inline measurement for machined parts: gauging at the machine
• Inline measurement for automotive powertrain parts
• Inline measurement for semiconductor lead frames and packages
• Zoom lens for connector pin inspection: FOV, sampling and focus
• Telecentric lens for lead-frame coplanarity
• Flash measurement for connectors

Product starting points

• HM telecentric measurement system for in-line inspection - head, illumination, controller and software as a station
• HM-1006, φ6 mm field - fine-pitch rows and single terminals
• HM-1015, φ15 mm field - 0.5 mm pitch rows, FPC contacts, crimp barrels
• HM-1040, φ40 mm field - 2.54 mm pitch headers and terminal blocks
• HM-1065, φ65 mm field - multi-row headers and connector bodies
• HM-1120, φ120 mm field - 40-way headers and whole-row spans