Connector-to-PCB Interface: PC Tail Fit, Coplanarity and Solder-Joint Fatigue

The Joint That Belongs to Nobody's Drawing

A board-mount connector is bought as a connector. The board is bought as a board. The joint between them is specified by the connector datasheet's recommended footprint, drawn by the PCB designer as a land pattern, and built by an assembly house to its own reflow profile — three documents, three owners, and no single drawing that says what the finished joint has to survive.

That gap is where board-level connector failures come from, and they do not look like connector failures. The part is fine. The board is fine. The joint cracks at the fillet heel after four hundred thermal cycles, or the tails never sat flat so half the rows got a partial fillet, or the hole was 0.2 mm oversize so the solder formed a donut around the tail instead of a fillet up it. In every one of those cases the incoming inspection passed, the first-article paperwork is complete, and the failure appears in the field as an intermittent on a channel nobody can reproduce on the bench.

This article treats the connector-to-board interface as its own design problem: what the tail has to do, how the three governing dimensions relate, why coplanarity is arithmetic rather than opinion, and where solder joints actually crack. It assumes the termination type is already chosen — if you are still deciding between a crimped wire, a solder cup and a board tail, read the termination selection guide first.

Three Jobs, and the Failure Each One Produces

A through-hole connector tail does three unrelated jobs at once, and each one fails in a different way. Separating them is the whole trick, because a fix aimed at the wrong job makes the other two worse.

Most specifications address only the first job. Continuity and a hipot pass prove current can flow on the day of test. Nothing in a standard build sequence proves the joint can absorb 0.04 mm of differential movement a thousand times.

What the tail actually has to do

The tail is a pin that has been deliberately made the weakest mechanical element in the chain. That is not an accident of cost. If the joint is the compliant element, the strain lands in a ductile solder fillet instead of in the plated barrel, the contact spring or the insert. Make the joint too strong and too stiff — a fat fillet, a short thick tail, a tail that fills its hole with no clearance — and the same displacement is transferred into the barrel wall, where a crack is a guaranteed open rather than a measurable resistance drift.

Two rules follow, and they are worth stating plainly because both get violated on real drawings. First, the tail is not a fastener. Solder has useful tensile and shear strength at room temperature and almost no creep resistance at 125 °C; anything that loads it continuously will eventually move it. Second, the length of unsupported tail between the board surface and the insert is a lever. Every extra millimetre of standoff multiplies the bending moment that a sideways nudge on the connector applies to the fillet.

Tail Diameter, Finished Hole and Annular Ring

Three dimensions govern the joint, and they are set in three different places: the tail diameter by the contact design, the finished hole diameter by the fabricator's drill and plating, and the pad diameter by the land pattern.

The hole is where most of the trouble starts. A hole that is too small will not accept the tail without forcing it — and forcing it means the part rides up on its tails, the opposite row never seats, and the connector is now out of coplanarity by exactly the interference you created. A hole that is too large leaves a radial gap the solder has to bridge on its own; the fillet then forms on the pad around the tail rather than climbing the tail, the barrel fill is incomplete, and the joint has a stress concentration at the point where the solder meets the tail instead of a smooth transition. The classic symptom is a joint that looks like a donut and passes visual inspection under a low-power loupe.

The annular ring is the third number: pad diameter minus hole diameter, divided by two. The ring is not decoration. It is the area that transfers load from the solder into the board laminate, and it is the toe the fillet lands on. Designs built purely to minimum fabrication rules often carry a ring of a few hundredths of a millimetre, which is enough for a via and not enough for a part that a technician will pull on.

Those bands are typical practice, orientation only — they are not a specification, and they are not a substitute for the connector's own recommended layout. The authority is the recommended PCB footprint on the connector drawing. Start there, then negotiate with the fabricator: if their minimum drill is larger than the footprint allows, the answer is a different board process or a different tail, not a land pattern invented in the CAD seat.

Two other hole details matter more than they are usually given credit for. Plated wall thickness — the usual 20 to 25 µm — is the current path and the mechanical anchor in one; a thin wall cracks longitudinally during thermal cycling and the joint becomes a resistance that drifts. And hole wall quality matters: a burr or a smear at the entry prevents the tail seating fully, which is invisible after reflow and shows up as one tail that is 0.3 mm high on an X-ray.

Tail Length, Board Thickness and Barrel Fill

Tail length is chosen against board thickness, and the usual mistake is to treat it as a single number. What matters is the protrusion beyond the far-side pad: enough tail to form a fillet, not so much that the solder wicks up the tail toward the insert.

A protrusion on the order of half a millimetre to one millimetre gives the fillet something to climb. Less than that and the joint is a flat pad-to-pad bridge with no fillet at all — a joint that survives the factory and fails in service. More than a couple of millimetres and two things happen: solder wicks up the tail, which can travel into the contact bore or bridge to a neighbouring tail on a fine pitch, and the protruding stub becomes a lever that catches on handling fixtures.

Board thickness drives the part number, not the process. A connector specified for a 1.6 mm board will not give acceptable protrusion through a 2.4 mm backplane, and shimming it up with washers to gain tail length destroys the coplanarity you were trying to protect. Order the long-tail variant or a different shell; the arithmetic does not bend.

Barrel fill is the fraction of the plated hole occupied by solder, and every acceptability standard sets a minimum for it. The reason to care is not cosmetic: an unfilled barrel means the tail is unsupported through the thickness of the board, so the first bending moment bends it at mid-barrel, exactly where the plating is thinnest at the knee of the hole. A half-filled barrel with a beautiful top fillet is a joint that fails at the wrong end.

Straight, Right-Angle and Straddle: Geometry Sets the Tolerance

The same footprint rules apply to all three, but the tolerance stack does not, and the difference is what makes right-angle parts the ones that bite.

Right-angle connectors are also where the load path matters most, because the body hangs off the board and every cable that enters it applies a moment at the row nearest the edge. A right-angle part retained only by its solder joints is a design that has decided to use solder as a structural fastener. It will work in the lab and fail at the harness shop.

Coplanarity: The Arithmetic Behind “It Sits Flat”

Coplanarity is the band within which all tail tips lie when the part is measured unconstrained — no fixture pressure, no screws pulling it down. It is usually specified as a single number, and that number is meaningless without the span it applies over, because what actually matters is the angle the part can tolerate.

The relation is simple trigonometry. A tilt of θ across a row span of S produces a tip deviation of S × tan θ. Invert it and the budget collapses faster than most people expect:

A one-degree tilt — which nobody would call a defect when looking at the part — already consumes roughly 0.09 to 0.18 mm of the allowed band depending on how many rows the part has. Add the tail-length tolerance from the contact drawing and the board's own bow and twist, and the arithmetic is unforgiving: fabrication limits for board warpage are well under one percent of the diagonal, but on a 200 mm board one percent is already two millimetres of bow across the panel.

What happens when the tails do not sit flat is the part of the story that never makes it into a report. The connector rocks on the high row. The low row's tails enter at an angle and scrape the barrel plating on the way in, which is a slow corrosion site. Reflow then pulls the body down, so the joint that was open is now closed under preload and the joint that looked perfect is carrying the preload. Both survive the factory. One of them has a fatigue crack seeded before the assembly ever left the building.

The measurement that would catch it is not the one usually made. Pressing the part down and looking at it proves nothing, because the pressing is what makes it flat. The part belongs on a granite plate with the gap under each corner measured, or in a fixture with indicators on the four corners of the shell, unclamped. Chamfered or pointed tail tips exist precisely to absorb a fraction of a millimetre of this error, and they only work if the tails are allowed to find their own position — which means the hardware goes in snug and untorqued before soldering and is only torqued after, in the sequence the installation drawing gives. The torque and anti-decoupling guide covers why the sequence matters on the coupling side; the same logic applies on the board side.

Why Solder Joints Crack: It Is Nearly Always CTE

Board-level joints do not usually fail from vibration in a well-supported design. They fail from thermal expansion mismatch, because the two things being soldered together grow at different rates and the joint is the only element that can absorb the difference.

Run the numbers on a realistic case. An aluminium shell 50 mm long clamped to a board that is rigid in-plane: a 100 °C excursion gives a differential of about 8 ppm/°C × 50 mm × 100 °C, or roughly 0.04 mm. That is the displacement the joint set has to absorb — not once, but every cycle, in both directions, for the life of the equipment. A solder fillet can absorb it. A solder fillet can absorb it a few thousand times. What it cannot do is absorb it while also being asked to hold the connector against its mating force and a cable load, because the same element is then carrying steady stress and cyclic strain at once.

The crack almost always starts where the geometry changes: at the fillet heel on the board side, or where the fillet meets the tail. That is a stress concentration by construction, and it is also the point most affected by how much solder was deposited. A fat, convex fillet looks reassuring and is the wrong answer mechanically — it raises the stiffness, drops the strain tolerance, and moves the concentration to a sharper transition. A concave fillet with a smooth transition is both the acceptability criterion and the better fatigue design, which is a rare case of the cosmetic and the structural agreeing.

Lead-free assembly makes the same joint less forgiving. SAC alloys are stronger and stiffer than tin-lead at temperature, so they transmit more of the mismatch into the intermetallic layer and the pad instead of creeping. The joint is not weaker on day one; it is less able to relax.

The Load Path: Solder Is Not a Fastener

Quantify the load before deciding the joint can take it. A multi-contact circular connector's mating force runs on the order of a few newtons per signal contact, so a hundred-way part needs a few hundred newtons of push to mate and a similar pull to break — and the technician applies it by hand, at an angle, every time. If the only thing reacting that force is a row of solder fillets, the design has handed a hand-applied shear load to the weakest element in the assembly.

The test is blunt and takes a minute: with the hardware removed, could a technician mate the connector without stressing a joint? If the answer is no, the hardware is doing the work the joint should never have been asked to do. Note that this is a design-level rule, not an assembly instruction — tightening the screws harder does not convert solder into a fastener, it just preloads the joint before it starts cycling.

Press-Fit: The Solder-Free Option, and When It Fits

Compliant-pin — press-fit — termination replaces the solder joint with an interference fit that deforms the pin, not the hole. The pin has a slotted eye that closes as it enters a hole slightly smaller than the pin's widest section, and the resulting cold weld at the hole wall is gas-tight. There is no reflow, no flux, no cleaning, and no thermal excursion to distort the assembly.

It also moves the tolerances onto the board. Compliant pins typically demand a finished hole tolerance of about ±0.05 mm, a controlled drill and no entry burr, and a defined board thickness range, because the pin has a limited compliance band. Press force per pin runs from several newtons to a few tens of newtons depending on pin size, so a large backplane connector needs a press with a real force budget and a support fixture that backs up every pin — pressing into an unsupported board is how pads get driven off.

Solder still wins in four situations, and they are common: low volume, where the tooling and hole-tolerance control cost more than the joints; mixed SMT and through-hole boards going through one reflow; heavy hermetic bodies whose glass seals should not be reflowed twice; and any joint that must be reworked in the field by someone with a soldering iron. Press-fit is repairable at the factory — the pin is pressed out and replaced, once, in the same hole — but it is not something a field technician improvises.

Inspection: What You Can See and What You Cannot

Visual inspection at the bench catches fillet defects and nothing else. It is still the only method that can be applied to 100 percent of joints, which is why the criteria have to be tight enough to be worth applying.

Everything below the surface needs a different method, and each has a blind spot. X-ray shows barrel fill and voids as density, and is poor at judging fillet shape on a tail that also has a contact body above it. Cross-sectioning shows fill, intermetallics and cracks — and destroys the joint, so it belongs at first article and at any process change, not on deliverable hardware. Mechanical shear or pull testing on a coupon tests the process, not the part. The electrical checks — continuity and low-level contact resistance — are the only ones that detect the joint after it has degraded, and they need a baseline taken at build if the number is to mean anything; the DWV, insulation resistance and contact resistance guide covers how to establish that baseline.

Two practical points. First, do the destructive section once, at setup, on a board built with the production stencil and profile — a section cut from a hand-soldered sample proves nothing about the line. Second, if the field failure is intermittent, the joint you should suspect is the one on the row that was hardest to seat, not the one that looks worst in the photograph.

The Design Review Checklist

Twelve questions that catch most of what goes wrong, in the order the answers become expensive:

• Does the land pattern match the connector's recommended PCB footprint, dimension for dimension — hole, pad, pitch, and the position of any keying or polarising feature?
• Is the finished hole diameter inside the fit band for the actual tail diameter, and has the fabricator confirmed it rather than the drill nominal?
• Is the annular ring large enough for a loaded feature, not a via?
• Does the tail protrude through the thickest board in the tolerance range by at least half a millimetre?
• Is the longest row span's coplanarity measured unclamped, and does the remaining band survive board bow and twist?
• What reacts the mating force — hardware, or the solder?
• What reacts the cable load on a right-angle part — backshell, or the fillet?
• Is the shell material's expansion mismatch to FR-4 acceptable at the part's longest dimension over the specified temperature range?
• Is there a board stiffener or local rib under a large connector on a thin board?
• Is the hardware torque sequence defined, and does it come after reflow?
• Is barrel fill called out, and is there a first-article section to prove it?
• For press-fit, has the hole tolerance and press force been agreed with the fabricator and the assembly house, not just with the connector supplier?

What to Put in the RFQ

The joint is a three-party interface, so the RFQ has to be three-party as well: connector supplier, board fabricator, assembly house. Items that belong in the documents rather than in a conversation:

• Connector part number and the tail style — straight, right-angle, long tail, or compliant pin — because the same shell comes in all of them.
• Board thickness range the part must accommodate, worst case both ends.
• Specified finished hole diameter and tolerance, plus pad diameter, as a controlled requirement rather than a fabrication default.
• Acceptability standard and class for the solder joints, named in full.
• Barrel fill requirement and the inspection method that will prove it, including whether sectioning is expected at first article.
• Retention hardware and its torque value and sequence.
• Thermal cycling or thermal shock requirement for the assembled board, if the equipment has one, quoted as a range and a cycle count.
• Baseline low-level contact resistance data with the test method and the ambient conditions, so a later drift is measurable.

Airoadcon supplies the connector half of that interface with board-mount options across the range: J599 (MIL-DTL-38999) shells with M39029 removable contacts including PC-tail variants, MIL-DTL-24308 D-subminiature connectors with PCB terminations, Micro-D and nano-D board-mount families, and the J30J micro-rectangular series with straight and right-angle board tails alongside its crimp variants.

Frequently Asked Questions

Can I solder a connector tail by hand if the board is already populated?

You can, and the joint will be worse than a reflowed one in ways that are hard to see. Hand soldering a multi-row connector heats the shell unevenly, so the rows cool at different rates and the fillets differ across the part; it also makes it very hard to keep every tail seated while wetting. If hand repair is unavoidable, treat the repaired joint as a new process: verify seating before soldering, use the same alloy, inspect every fillet on the reworked rows, and record which joints were touched so a later failure is traceable.

My connector sits flat when I press it, but rocks when I let go. Is that a defect?

Yes, and it is the defect that matters most. Coplanarity is defined unconstrained. If the part only sits flat under finger pressure, the tails are outside the band, and whatever closes the gap during reflow is preloading a joint. Measure the gap at each corner on a flat plate, then compare the total against the tail-length tolerance and the board's actual flatness.

Is a bigger solder fillet stronger?

Stronger in a single pull test at room temperature, weaker in fatigue, which is the load that actually kills board joints. A larger fillet increases stiffness, so the same differential expansion produces higher stress at the fillet-to-tail transition. The fillet you want is smooth, concave and consistent — the shape the acceptability standards describe — not the shape that looks most generous.

Why do joints fail on the return path first?

Usually because the return tails carry the highest current and see the biggest self-heating, so the joint runs hotter than its neighbours and creeps faster. The second reason is diagnostic neglect: return and ground tails are frequently not instrumented, so the first measurement anyone takes of that joint is the one that reads open.

Does conformal coating help joint fatigue life?

It helps against corrosion and contamination and does very little for mechanical fatigue. A coating that is stiff enough to restrain the joint will also transfer the mismatch somewhere else; a coating that is compliant enough not to restrain it provides almost no mechanical benefit. Specify coating for the environment, not for the fatigue problem.

How do I tell a fatigue crack from a cold joint?

By history and by location. A cold joint is present at build, usually dull or grainy, often on one consistently poor position — typically the hardest to heat. A fatigue crack appears after cycles, starts at the fillet heel or the fillet-to-tail transition, and is often accompanied by fine cracking in the plating on the tail. If the joint looked good at first article and failed after thermal cycling, it is fatigue, and the fix is in the geometry and the retention, not in the iron.