The Repair LibraryRead · Learn · Master

X-Ray Inspection of BGA

The instrument the blindness made necessary gets its BGA treatment: what a beam passed through a package can honestly show, how to choose the view that answers the question, and where even radiography stays blind. The physics is density: solder absorbs the beam harder than laminate, plastic, or silicon, so every ball images as a dark disc through the package — and the whole read is an exercise in interpreting stacked densities, because a two-dimensional radiograph flattens board, substrate, die, and balls into one overlay. Read straight, the top-down view convicts the geometric defects the failure catalog promised: bridges as dark webs between discs, missing balls as absent discs, misalignment as a field shifted off its land grid, and gross voids as bright holes inside the dark. Read carelessly, it acquits what it never saw — because the beam reports geometry, not metallurgy: wetting quality is invisible, hairline fatigue cracks live below practical resolution, a cold joint images exactly like a good one, and head-in-pillow — the failure X-ray is most often asked to find — can vanish in the top-down view when ball and pillow touch. The section teaches the view ladder that claws some of that back: oblique views that tilt the board until ball profiles, standoff, and pillow seams present themselves; and computed tomography, which slices the stacked ambiguity apart — at a price that the repairability arithmetic must approve. Around the reads it builds the disciplines: void fraction assessed by location and stakes instead of panic, sessions aimed by the die shadow and the peek's flags, the access arithmetic of paying an imaging service versus owning a cabinet, and radiograph claims bounded exactly like optical ones — the beam reports what density showed, and the meters still own the electrical truth.

AdvancedLow Risk23 min read

What You Will Learn

  • You will learn the density read — why solder images dark, how stacked layers overlay in two dimensions, and which defects the top-down view convicts directly.
  • You will learn the view ladder — top-down, oblique, and computed tomography — and which question each answers at what cost.
  • You will learn X-ray's blind spots — wetting, hairline cracks, cold joints, and the head-in-pillow that touches its pillow — and where the meters take over.
  • You will learn the void discipline — void fraction read by location and stakes instead of panic.
  • You will learn the session craft — aiming by die shadow and peek flags, the access arithmetic, and radiograph claims bounded like optical ones.

What You Will Be Able To Do

  • You will be able to read a top-down BGA radiograph and convict bridges, missing balls, misalignment, and gross voids as bounded findings.
  • You will be able to choose the view — top-down, oblique, or CT — that answers a stated diagnostic question.
  • You will be able to state what radiography cannot say and route those questions to the meters.
  • You will be able to assess void fraction by location and stakes and write the call honestly.
  • You will be able to plan and price an imaging session — aim, views, service versus skip — under the repairability arithmetic.

Required Tools

  • Reference BGA radiographs — imaging-service galleries, teardown archives, or the bench's own past session reports
  • The notebook — this section builds the capability table, the void call card, and the imaging plan
  • The donor bin's flagged packages and their stress maps — imaging plans practice on real hypotheses
  • Parts and service pricing sources — the access arithmetic is looked up, not guessed
  • No radiation source at the bench — commercial cabinets are shielded and interlocked, and this section reads images, not beams

When NOT to Attempt This

Do not attempt this section if any of the following apply to you:

  • You are not comfortable working with small surface-mount components.
  • You have not completed the prerequisite sections for this skill.
  • You do not have the required tools in working condition.

Section Overview

The blindness's answer arrives, with its own honest boundary (visual-inspection-limitations-of-bga). The physics is density. Radiographic density images solder dark through package and board — and the top-down read convicts bridges, missing balls, misalignment, and gross voids directly (x-ray-inspection-when-its-needed). The view ladder claws deeper. The oblique view tilts ball profiles, standoff, and pillow seams into presentation; computed tomography slices the stacked overlay apart — at prices the arithmetic must approve (what-is-bga). The blind spots are real. Wetting is invisible, hairline cracks sit below resolution, cold joints image clean, and a pillow-touching head-in-pillow can vanish top-down — the meters keep the electrical truth (bga-failure-modes). The disciplines govern the reads. Void fraction is called by location and stakes, sessions aim by die shadow and peek flags, and radiograph claims stay bounded like optical ones. Density, ladder, boundary, discipline — the seeing's second instrument entire.

Why This Matters

X-ray is the only instrument that will ever show the field's hidden geometry, and the bench that reads it well spends its imaging money like a professional (x-ray-inspection-when-its-needed). This matters because the direct convictions end diagnoses: a bridge webbed between two discs under the die shadow explains the short conclusively — the session that finds it just saved a speculative removal, and the failure catalog's geometric classes are exactly what the beam convicts (bga-failure-modes). This matters because the blind spots ship boards when unknown: 'X-ray came back clean' reads like an acquittal and is only a bounded one — the wetting, the cold joints, and the touching pillow were never in evidence, and the technician who knows that routes the intermittent to the meters instead of back to the customer (visual-inspection-limitations-of-bga). It matters because view choice is question choice: a top-down session cannot answer a standoff question, an oblique can, and CT answers what neither could — ordered wrong, the session bills twice; ordered right, it answers in one pass (what-is-bga). It matters because void panic wastes good boards: voids appear in nearly every real joint, the literature tolerates more than beginners expect, and the location-and-stakes call is what separates a normal thermal-pad void from a corner-ball integrity problem. And it matters because the access arithmetic is part of the assessment: most benches rent the beam by the session — knowing when the question justifies the fee, and what the replacement anchor says about paying it, is repair economics applied to photons. Learn what the beam can testify to — then spend it only on questions it can answer.

Required Prerequisites

  • Nothing consumed — the section spends interpretation and planning (x-ray-inspection-when-its-needed)
  • A notebook page for the capability table — to split what the beam convicts, what the ladder's upper rungs add, and what no view will say (visual-inspection-limitations-of-bga)
  • A page for the void call card — to write the location-and-stakes questions once, calmly (bga-failure-modes)
  • A page for the imaging plan template — to draft aim, views, and claims wording before any session is booked (what-is-bga)
  • Reference radiographs from service galleries, teardown archives, or past session reports — to practice the density read on real images before paying for new ones (x-ray-inspection-when-its-needed)
  • The donor bin's flagged packages with their stress maps — to draft imaging plans against real hypotheses (bga-failure-modes)
  • The §3.1 peek logs — to practice aiming sessions from flags the cheap instrument already planted (visual-inspection-limitations-of-bga)
  • Service pricing from a real imaging bureau — to run the access arithmetic on live numbers (what-is-bga)
  • The bench camera and record — to file every read, call, and plan beside the boards they describe

Real-World Applications

The beam pays for itself when it is aimed, and first at the conclusive find. A shop imaging a console APU after a clean peek and a dead memory bus aims the session at the die-shadow ring the stress map flagged and reads a bridge webbed between two discs — a conclusive geometric conviction that turns a speculative removal into a scoped rework (bga-failure-modes). An imaging session on a laptop controller that top-down called clean earns its oblique add-on: tilted, the suspect corner presents ball profiles edge-on, and one ball shows the seam of a pillow it merely touches — the defect the flat view is famous for missing (x-ray-inspection-when-its-needed). A technician holding a router board whose replacement costs less than a session runs the access arithmetic and skips the beam entirely — the assessment's anchor decides, and the imaging plan's best line is sometimes 'not worth the fee' (what-is-bga). And a technician second-guessing a vendor's radiograph report that waved off 'minor voiding' runs the void discipline themselves: fraction estimated, locations named against the stress map, stakes applied — and finds the vendor's wave-off was right for the mid-field balls and wrong for one corner cluster the note now records (visual-inspection-limitations-of-bga). The confusions this prevents: a session booked with no aim and billed twice, a clean top-down sold as a field acquittal, a normal void population read as a defect, and a fee paid on a board the anchor had already decided.

Common Challenges

  • The overlay confuses new eyes. Board copper, substrate traces, die edge, and balls all stack into one image — the difficulty is learning to assign each density to its layer, and reference radiographs of known-good boards are the fastest teacher (x-ray-inspection-when-its-needed).
  • The clean radiograph over-comforts. A field of tidy dark discs looks like an acquittal — the difficulty is holding the boundary: the view convicts geometry and acquits only what it could have shown (visual-inspection-limitations-of-bga).
  • The upper rungs tempt by default. CT answers everything, so why not always CT — the difficulty is cost discipline: the ladder climbs only when the question demands it, and most questions die on the first rung (what-is-bga).

Safety Notes

Risk Level: Low. This section reads images and plans sessions — the beam lives inside shielded, interlocked commercial cabinets the bench never opens, and the section's own work is desk work.

Professional Tips Before Starting

  • Collect known-good radiographs first. The fastest density education is a healthy board's image beside a suspect's — differences teach faster than absolutes (x-ray-inspection-when-its-needed).
  • Write the question before booking the session. 'Image the SoC' is not a question; 'bridge or no bridge under the die-shadow ring' is — sessions aimed by questions come back with answers (bga-failure-modes).
  • File every session report beside its board's record. The radiograph outlives the repair — and the next fault on the same board starts from evidence instead of memory (visual-inspection-limitations-of-bga).

The Second Instrument — Density, The Ladder, The Boundary, The Disciplines

Recap and Frame

The peek ended at the outer row; the beam goes through the package, and the chapter's second section teaches what that purchase actually buys (visual-inspection-limitations-of-bga). The foundations are laid. Volume 3 introduced the absorbing beam, the imaging modes, and the when-to-reach judgment — this section assumes that grounding and spends it entirely on the package class that made radiography necessary (x-ray-inspection-when-its-needed). The failure catalog is the shot list. Bridges, missing and misaligned balls, voids — the geometric classes are what the beam convicts, and the stress maps say where to point it first (bga-failure-modes). The package anatomy is the overlay's key. Substrate, die, escape routing, standoff — every layer the foundation section named appears in the image, stacked — and reading the stack is reading the anatomy (what-is-bga). What is new is fluency. Not whether X-ray exists but how to read its images like a technician: density by density, view by view, claim by bounded claim. And the section's order is the working order: the read, the ladder of views, the boundary, then the disciplines that keep sessions aimed and honest. Hold the frame — old physics, new fluency — and the second instrument earns its place.

The Density Read — What the Top-Down View Convicts

Everything on the screen is absorption, and the read starts by assigning densities to layers (x-ray-inspection-when-its-needed). Solder owns the dark. The balls image as a grid of dark discs through package and board — solder absorbs the beam hardest of anything in the stack — while laminate, overmold, and silicon pass the beam and image light, and copper traces sit between. The overlay is the price of two dimensions. A radiograph flattens board copper, substrate routing, die edge, and ball field into one image — the escape routing's dogbones and via-in-pad discs appear woven through the ball grid — and the reader's first skill is assigning each feature to its layer before judging any of them (what-is-bga). The direct convictions are geometric. A bridge images as a dark web joining two discs; a missing ball as an absent disc in the grid; misalignment as the disc field shifted off the land pattern's centers; a gross void as a bright hole inside a disc's dark (bga-failure-modes). Ball-shape anomalies read with care. Discs larger and flatter than their neighbors suggest over-collapse, smaller rounder ones suggest under-collapse — real signals, read against the field's own population rather than an absolute, and logged as indications rather than convictions. The claims stay bounded. 'Bridge convicted between discs eleven and twelve, die-shadow ring, top-down view' is a radiograph note that stays true — the same discipline the peek taught, applied to a deeper instrument (visual-inspection-limitations-of-bga). Dark discs, stacked layers, geometric convictions, bounded notes — the density read entire. Read the stack before judging the joint — half of radiograph errors are layer-assignment errors.

The View Ladder — Oblique and the Sliced Volume

The top-down view is the ladder's cheap first rung, and two rungs rise above it (x-ray-inspection-when-its-needed). The oblique view buys the profile. Tilt the board under the beam and the balls present edge-on: profiles show their collapse shape, relative standoff differences appear, and the seam of a head-in-pillow — ball resting on pillow, two masses touching — can separate visibly where the flat view fused them into one honest-looking disc (bga-failure-modes). The tilt has geometry limits. At angle, inner-row balls do not vanish — the beam still passes through them — but their profiles superimpose on the shadows of the rows in front of and behind them along the beam path, and the tangled overlap is unreadable: oblique views work best on the outer rows and at package corners, and a full-field profile survey is not what tilting buys (visual-inspection-limitations-of-bga). Computed tomography slices the ambiguity. Many views reconstructed into a rotatable volume — the overlay problem dissolves: board layers separate from substrate, package tiers separate from each other, and a stacked package's two ball fields finally read as two fields (what-is-bga). CT is the expensive answer. Session time and fees climb steeply, and the arithmetic gates the rung: a consumer board rarely justifies it, a data-bearing or unobtainable board sometimes does, and the question always decides. Stacked packages force the ladder. Package-on-package overlays two grids in every flat view — tier-separating questions are oblique-or-CT questions by construction, one reason stacked rework sits high on every ladder the volume owns. Flat, tilted, sliced — each rung priced by its question — the ladder entire. Choose the view from the question, not the menu — most questions die on the first rung, and the arithmetic loves them for it.

The Radiographic Boundary — What No View Will Say

The beam's testimony has a hard edge, and it falls exactly where beginners assume it does not (visual-inspection-limitations-of-bga). Geometry, never metallurgy. The image reports where matter sits, not how it bonded: a joint's wetting quality is invisible, an intermetallic layer's health is invisible, and a cold joint — solder present, bond absent — images exactly like its healthy neighbor (x-ray-inspection-when-its-needed). Hairline cracks live below the read. The fatigue cracks the failure catalog cares most about are fractions of a ball's diameter, their planes lying broadside to a top-down beam so the beam crosses only a micron-wide gap — practical bench-service radiography does not resolve them, and a cycling-cracked field routinely images clean (bga-failure-modes). The famous fugitive is the touching pillow. Head-in-pillow with ball resting on pillow reads top-down as one dark mass — the defect X-ray is most often asked to find is one it often cannot show flat, which is why the oblique exists and why 'X-ray clean' never closes a pillow hypothesis by itself. The acquittal is bounded by the view. A clean radiograph acquits the defects that view could have shown — bridges, missing balls, gross voids, misalignment — and nothing else: the phrase 'X-ray came back clean' means the geometric classes are excluded, not that the joints are good (what-is-bga). The meters keep the electrical truth. Opens, resistive joints, and intermittents route to the next section's instruments — the beam narrows hypotheses; the electrical truth convicts them. No metallurgy, no hairlines, no touching pillows, bounded acquittals — the boundary entire. X-ray shows where the solder is — the meters say whether it is a joint.

The Disciplines — Voids, Aim, Access, Claims

Four disciplines turn radiography from a picture into an instrument (x-ray-inspection-when-its-needed). Voids are called, not feared. Small voids are normal solder metallurgy — flux outgassing frozen in place — and nearly every real joint carries some: the call is the void fraction — how much of the disc's area reads bright — judged by location and stakes: modest fractions in mid-field balls are routinely tolerable, large fractions clustered at one joint or stacked at a pad interface deserve the note, and corner or die-shadow balls earn stricter reads because the stress map says they carry more (bga-failure-modes). The stakes set the threshold. Published guidance tolerates more voiding than beginners expect, and the bench applies it the way the inspection law always has — stakes-based: a hobby board's tolerable is not an avionics tolerable, and the note records fraction and location so the reader applies their own stakes (what-is-bga). Sessions are aimed, not sprayed. The die shadow, the peek's flags, and the symptom story choose where the beam looks first — an aimed session answers in one pass what an unaimed one bills twice for (visual-inspection-limitations-of-bga). Access is arithmetic. Most benches rent the beam — per-session fees against the board's anchor and payload decide when imaging is bought, and 'not worth the fee' is a professional imaging plan. Claims stay bounded. Radiograph notes name the view, the aim, the convictions, and the bounded acquittals — 'top-down, die-shadow ring: no bridges, no missing balls; wetting and hairlines not assessable in this view' — the same sentence discipline the whole chapter runs on. Called voids, aimed beams, priced access, bounded claims — the disciplines entire. The radiograph is testimony, not verdict — the disciplines are what make it admissible.

Common Mistakes

  • Reading the overlay as one layer. A via-in-pad disc convicted as a void, a substrate trace read as a bridge — assign every density to its layer first; half of radiograph errors are layer errors (x-ray-inspection-when-its-needed).
  • Selling 'X-ray clean' as an acquittal of the field. The bounded exclusion of geometric classes inflated into 'joints verified' — wetting, cold joints, hairlines, and touching pillows were never in evidence (visual-inspection-limitations-of-bga).
  • Booking the beam without a question. 'Image the board' sessions that come back unaimed and inconclusive — the die shadow, the flags, and the symptom story write the question before the fee is paid (bga-failure-modes).
  • Void panic. A normal population of small voids read as a condemned field — the call is fraction, location, and stakes; the note records all three (what-is-bga).
  • Climbing to CT by default. The expensive rung bought for a first-rung question — most questions die on top-down; the ladder climbs only when the question demands it.

Troubleshooting Guidance

The radiograph troubleshoots by re-reading: misassigned, misaimed, overclaimed. If a feature will not resolve into a defect or a normal: find it on a known-good reference image first — the healthy board beside the suspect settles most ambiguous reads in seconds (x-ray-inspection-when-its-needed). If the session came back inconclusive: audit the question, not the service — an unaimed session is a planning failure, and the re-book names the ring, the view, and the hypothesis (bga-failure-modes). If top-down and symptoms disagree: believe the symptoms and climb one rung or route to the meters — the flat view's acquittals are bounded, and an intermittent with a clean radiograph is the boundary working, not a paradox (visual-inspection-limitations-of-bga). If a void call will not settle: write fraction and location and let stakes decide — the note that records both serves every future reader; the panic that records neither serves nobody (what-is-bga). If the arithmetic keeps refusing sessions: that is the arithmetic working — boards whose anchors forbid fees get meter-first diagnosis, and the beam waits for the board that justifies it. The throughline: radiograph disputes resolve by naming the view and its bounded reach — the same move the whole chapter makes.

Verification & Testing Methods

Confirm the second instrument before the meters arrive:

  • [ ] I can explain radiographic density and assign a radiograph's features to their layers before judging any joint.
  • [ ] I can convict bridges, missing balls, misalignment, and gross voids from a top-down view as bounded findings.
  • [ ] I can choose between top-down, oblique view, and CT from a stated question — and defend the cheap rung when it suffices.
  • [ ] I can state the radiographic boundary — no wetting, no hairlines, no touching pillows — and route those questions to the meters.
  • [ ] I can call void fraction by location and stakes, aim a session by the die shadow and flags, and write the bounded radiograph note.

Then try the practice exercises below — image-reading and desk work only; scenarios differ from the quiz.

Practice Exercises

  1. Run the density drill (5 minutes, three reference radiographs, notebook). On each image assign the visible features to layers — board copper, substrate routing, die, balls — then log every geometric conviction the top-down view supports as a bounded finding with location, and every ambiguity as exactly that (x-ray-inspection-when-its-needed).
  2. Build the capability table (5 minutes, desk). Three columns in your own words — what top-down convicts, what oblique and CT add, what no view will say — with the meter-routing line under the third column, as the card every imaging decision consults (visual-inspection-limitations-of-bga).
  3. Make three void calls (5 minutes, reference images or sketched disc patterns, notebook). For each: estimate the fraction, name the location against the stress map, and write the stakes-based call — tolerable, noteworthy, or condemning — with the one-line reason the record keeps (bga-failure-modes).
  4. Write one imaging plan (5 minutes, desk, a flagged donor package). For a real donor hypothesis: the question in one sentence, the aim from die shadow and peek flags, the view rung with its price justified against the board's anchor and payload — or the honest 'not worth the fee' — and the bounded claims wording the session report should come back in (what-is-bga).

These core steps — the density drill, the capability table, the void calls, and the imaging plan — are tested in the Chapter Quiz at the end of this chapter, where a score of 80% is required to continue.

Key Takeaways

  • Radiographic density is the whole physics — solder images dark through the stack, the two-dimensional overlay must be read layer by layer, and the top-down view convicts the geometric classes: bridges, missing balls, misalignment, gross voids (x-ray-inspection-when-its-needed).
  • The view ladder climbs by question: the oblique view tilts profiles, standoff, and pillow seams into presentation; CT slices the overlay into a rotatable volume — at prices the arithmetic gates (what-is-bga).
  • The boundary is hard: no wetting, no intermetallic health, no hairline cracks, and the touching pillow can vanish flat — a clean radiograph acquits only what its view could show (visual-inspection-limitations-of-bga).
  • Void fraction is called by location and stakes, not feared — small voids are normal metallurgy, and the note records fraction and place so stakes can judge (bga-failure-modes).
  • Sessions are aimed by the die shadow and the peek's flags, priced by the access arithmetic, and reported in bounded claims — the beam narrows hypotheses; the meters convict them.

Skills Learned

  • You can now read a BGA radiograph layer by layer and convict the geometric classes as bounded findings.
  • You can now choose the view rung from the question and defend the cheap one.
  • You can now hold the radiographic boundary and route metallurgy questions to the meters.
  • You can now call voids by fraction, location, and stakes.
  • You can now plan, aim, and price an imaging session — including the honest refusal.

Glossary Additions

  • radiographic density — the darkness a material contributes to an X-ray image, set by how strongly it absorbs the beam: solder absorbs hardest and images darkest, copper sits between, and laminate, overmold, and silicon pass most of the beam and image light. Because a two-dimensional radiograph stacks board copper, substrate routing, die, and ball field into one overlay, every read begins by assigning densities to layers — the escape routing's dogbones and via-in-pad discs weave through the ball grid, and half of radiograph errors are layer-assignment errors rather than joint-judgment errors. Read correctly, density convicts the geometric defect classes directly: bridges as dark webs between discs, missing balls as absent discs, misalignment as a shifted grid, and voids as bright holes inside a disc's dark.
  • oblique view — the radiographic view taken with the board tilted under the beam so that BGA balls present edge-on instead of stacked flat: profiles show their collapse shape, relative standoff differences appear, and the seam of a head-in-pillow — ball resting on its pillow as two touching masses — can separate visibly where the top-down view fused them into one healthy-looking disc. The tilt is limited by superposition rather than optical occlusion — inner-row profiles overlap the shadows of neighboring rows along the tilted beam path, so oblique reads work best on outer rows and corners rather than as full-field surveys — and the view's place on the ladder is exactly that: the middle rung that answers profile, standoff, and pillow-seam questions the flat view cannot, before the expensive climb to computed tomography.
  • void fraction — the proportion of a solder joint's radiographic area that reads bright, indicating gas frozen into the joint, and the number that turns void observations into calls: small voids are normal solder metallurgy — flux outgassing trapped at solidification — and nearly every real joint carries some, so the discipline is fraction, location, and stakes rather than presence-panic. Modest fractions in mid-field balls are routinely tolerable; large fractions, voids clustered at one joint, or voids stacked against a pad interface earn the note; and corner or die-shadow balls read stricter because the stress map says they carry more. Published guidance tolerates more voiding than beginners expect and varies by application, so the bounded note records fraction and location and lets the board's stakes set the verdict.

Suggested Next Sections

Must read next:

  • Electrical Testing of BGA Connections — Section 3.3 brings the instrument that convicts what the beam and the peek can only suspect: continuity and resistance discipline across a hidden field, signature comparison against known-good boards, and the meter craft that carries most daily BGA diagnosis.

Recommended:

  • BGA Failure Modes — the catalog whose geometric classes the beam convicts and whose stress maps aim it.
  • What Is BGA? — the anatomy every radiograph overlays, escape routing and standoff included.