The Repair LibraryRead · Learn · Master

Thermal Profile Design for BGA Reflow

The chapter pauses its operations to build the thing both of them fly: the profile itself, designed at Professional depth from the inputs upward instead of copied from the last job that worked. A reflow profile is five stages with reasons: the ramp, bounded in degrees per second because thermal shock cracks ceramics and a too-fast climb outruns the board's ability to heat evenly; the soak, where the whole board arrives at one temperature and the flux does its activation work — the stage that large boards stretch and small ones abbreviate; the climb to peak, deliberate and narrated; the peak itself, set as liquidus plus a bounded margin — enough above the alloy's melting point that every joint in the field genuinely wets, never so much that the package's ratings and the neighbors' tolerance pay for insurance nobody priced; the time above liquidus, kept inside its window because every extra second grows intermetallics and spends the budget; and the cool-down, bounded like the ramp because stress written into freezing joints is stress the failure catalog collects later. The design method works from inputs: the alloy the case file named sets liquidus; the board's mass and layer count stretch or shrink the soak; the package's heat lifetime caps the peak and counts the exposure; the neighbors and the moisture state write the abort lines. The output is the profile card the volume has been flying since the reballing chapter's attach pass — now built rather than inherited: stages, targets, ramps, and abort lines on one card, written before any heat exists. And the section's Professional core is the verification law generalized: a designed profile is a hypothesis — it becomes real on an instrumented run, thermocouples narrating what the recipe actually did at the package, the board, and the neighbor — and only the measured, iterated, verified profile earns the library and the board class it serves. Recipes are folklore until a junction has watched them work.

ProfessionalMedium Risk23 min read

What You Will Learn

  • You will learn the five stages and their reasons — ramp, soak, climb, peak with its margin, time above liquidus, and the bounded cool.
  • You will learn the design inputs — alloy, board mass, package budget, neighbors, moisture — and how each writes its line of the card.
  • You will learn the peak margin — liquidus plus enough, never plus insurance nobody priced.
  • You will learn the verification law — multi-point instrumented runs that turn a recipe into a measured reality.
  • You will learn the failure signatures — what too-fast, too-short, too-low, too-long, and too-quick-cool each write into a field.

What You Will Be Able To Do

  • You will be able to design a profile card from a case file's inputs — stages, targets, ramps, abort lines.
  • You will be able to set a peak as liquidus plus a bounded, justified margin.
  • You will be able to bound time above liquidus and both ramps with reasons, not habits.
  • You will be able to verify a designed profile on an instrumented run and iterate it honestly.
  • You will be able to read a field's defects back to the profile stage that wrote them.

Required Tools

  • The profile-card template and the case file — design starts from the inputs the file already holds
  • Thermocouples, several, with tape — verification runs are multi-point or they are partial
  • The heat platform the profile will fly on — hot air rig or IR station, because profiles are designed for their delivery
  • Donor boards matched to the patient's class — hypotheses verify on instrumented stand-ins first where they can
  • The profile library and its revision notes — verified cards accumulate; folklore does not

When NOT to Attempt This

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

  • You are not experienced with the specific repair type described here.
  • You do not have professional-grade equipment for this procedure.
  • The device has sentimental or high monetary value and you cannot afford a mistake.
  • You have not successfully completed this repair on a sacrificial device first.

Section Overview

The chapter pauses to build what both its operations fly: the profile, designed from inputs instead of inherited (reflow-ovens-when-and-why). Five stages, each with a reason. Bounded ramp, equalizing soak, deliberate climb, the peak, the bounded cool — reasons a bench can say out loud (temperature-and-airflow-finding-the-right-settings). The peak margin is set, not guessed. Liquidus plus enough for every joint to wet — never plus insurance the package's ratings pay for (bga-ball-composition-and-alloys). Time above liquidus stays inside its window. Every extra second grows intermetallics and spends the budget. The profile card is the output. Stages, targets, ramps, abort lines — built from the case file's inputs, written before any heat (bga-removal-with-ir-rework-station). And the verification law governs. A designed profile is a hypothesis until an instrumented run narrates it — multi-point, iterated, then the library. Stages, inputs, margin, card, verification — the design entire.

Why This Matters

Every heat operation the volume has flown ran on a profile someone designed, and this section is where the bench stops inheriting them (reflow-ovens-when-and-why). This matters because the stages have reasons, and reasons transfer: a bench that knows why the soak exists can stretch it for a six-layer board without a forum's permission — the recipe-copier is helpless the first time the board differs from the recipe's (temperature-and-airflow-finding-the-right-settings). This matters because the peak margin is where fields live or die: too little and the far corner pillows exactly as the failure catalog described; too much and the package's budget, the neighbors, and the laminate pay for insurance nobody priced — the margin is a designed number with a cited reason (bga-ball-composition-and-alloys). It matters because the failure signatures close the loop: a popcorned package reads back to the ramp, a pillowed corner to the soak or peak, excessive intermetallic dullness to the time above — the profile's stages are where field defects come from, and reading them backward is how profiles improve. It matters because verification is what separates design from confidence: the card that has never been flown instrumented is a guess in good handwriting — the §5.2 first-article law generalizes to every profile this section builds (bga-removal-with-ir-rework-station). And it matters because the library compounds: every verified card per board class is a repeatable operation the bench never designs twice — the Professional product, accumulating. Design from inputs, verify with junctions, library the result — the section in one sentence.

Required Prerequisites

  • Thermocouple tape and several junctions — to instrument verification runs at package, field, and neighbor (bga-removal-with-ir-rework-station)
  • Blank profile cards or the template page — to design on paper before any element warms (reflow-ovens-when-and-why)
  • A donor board of the patient's class for the instrumented run — to verify hypotheses on stand-ins where the class allows (bga-ball-composition-and-alloys)
  • The library's revision notes — to record what each iteration changed and why, because unexplained revisions are folklore returning (temperature-and-airflow-finding-the-right-settings)
  • The bench's heat platform — hot-air rig or IR station — to design for the delivery the profile will actually ride (bga-removal-with-ir-rework-station)
  • Donor boards spanning mass classes — a thin single-purpose board and a thick multilayer — to feel what mass does to soak and ramp (reflow-ovens-when-and-why)
  • Multiple thermocouples with a logger or the station's inputs — to narrate verification runs at more than one point (temperature-and-airflow-finding-the-right-settings)
  • The case files from the chapter's operations — to design from real inputs rather than invented ones (bga-ball-composition-and-alloys)
  • The bench camera — to file cards, traces, and iterations beside their board classes

Real-World Applications

Profile design pays wherever inherited recipes run out, and they always run out. A bench facing a six-layer telecom board no library card fits designs from inputs: SAC liquidus from the file, a stretched soak for the copper mass, the peak margin justified against the package's remaining budget — and the first instrumented run shows the field arriving together where a copied phone-board recipe would have starved the planes (reflow-ovens-when-and-why). A technician whose reballed field came back with one pillowed corner reads the defect to its stage: the corner over the ground plane never reached the peak the card promised — the next iteration stretches the soak and re-verifies, and the failure taught the profile instead of the customer (bga-ball-composition-and-alloys). A shop inheriting a forum recipe that 'always works' flies it instrumented once before trusting it: the trace shows a ramp at twice the bounded rate and a peak flirting with the package's ceiling — the recipe worked by luck, and the redesigned card works by measurement (temperature-and-airflow-finding-the-right-settings). And a bench asked why its profile library is worth anything answers with the revision notes: every card verified, every change cited, every board class served by measured reality — the library as the shop's accumulated judgment, written down (bga-removal-with-ir-rework-station). The confusions this prevents: a phone recipe starving a telecom board, a pillowed corner blamed on flux, a lucky recipe trusted past its luck, and a library that is folklore with formatting.

Common Challenges

  • The numbers tempt memorization. Liquidus values, ramp bounds, soak windows — the section teaches reasons because reasons transfer: the bench that knows why survives the board the numbers never met (reflow-ovens-when-and-why).
  • The margin wants to grow. A little more peak feels like insurance — it is spending: budget, neighbors, and laminate all pay, and the margin is a designed number with a cited reason or it is anxiety with a dial (bga-ball-composition-and-alloys).
  • Verification feels redundant after careful design. A well-built card invites trust — but design predicts and junctions measure; the gap between them is exactly what the instrumented run exists to find (bga-removal-with-ir-rework-station).

Safety Notes

Risk Level: Medium. Design is desk work; verification runs are real heat under the platform's full law.

Professional Tips Before Starting

  • Design on the card, not in the controller. Paper first — stages, targets, reasons — the controller gets the card's numbers after the card can defend them (reflow-ovens-when-and-why).
  • Instrument more than the target. Package edge, open board field, and the most vulnerable neighbor — three junctions tell the run's whole story; one tells a third of it (bga-removal-with-ir-rework-station).
  • Write the reason beside every number. A card that cites its inputs survives its author — and the next bench inherits judgment instead of numerology (temperature-and-airflow-finding-the-right-settings).

The Design — Stages, Inputs, Margin, Verification

Recap and Frame

Two operations now fly profiles; this section builds them, and the building is the Professional skill (bga-removal-with-ir-rework-station). The stage vocabulary arrives assembled. Soak and ramp from the hot-air fundamentals, profile shape from the oven chapter, the narrator's law from the thermal section — the pieces exist; this section makes them a method (reflow-ovens-when-and-why). The alloy targets arrive known. Liquidus by chemistry — the eutectic's sharp point, SAC's band — from the metallurgy chapter that has anchored every heat decision since (bga-ball-composition-and-alloys). The settings craft arrives trained. Temperature and airflow as deliberate choices was Volume 3's lesson at nozzle scale — this section is the same deliberateness, structured across five stages and a whole board (temperature-and-airflow-finding-the-right-settings). What is new is design as the deliverable. Not flying a card — building one: from inputs, with reasons, toward verification. And the order is the method's order: stages understood, inputs gathered, card written, hypothesis flown, reality measured, library earned. Hold the frame — pieces into method — and the design begins.

The Stages — Five, Each With a Reason

A profile is five stages, and the section teaches them as reasons rather than numbers (reflow-ovens-when-and-why). The ramp is bounded. Degrees per second held modest — because thermal shock microcracks ceramics and sudden gradients stress everything laminated, and because a climb that outruns the board's conduction heats the surface while the mass lags, arriving uneven everywhere it matters (temperature-and-airflow-finding-the-right-settings). The soak equalizes and activates. A held plateau below liquidus where the board's thermal mass catches up — planes, thick regions, and shadowed corners arriving at one temperature — while the flux does its activation work: large boards stretch the soak, small boards abbreviate it, and the pillowed far corner is usually a soak that ended early. The climb is deliberate. From soak to peak on the card's schedule, narrated — the stage where patience is cheapest and improvisation most tempting. The peak wets every joint. Liquidus plus the bounded margin — high enough that the field's coldest corner genuinely wets, and no higher: the margin is designed, cited, and small (bga-ball-composition-and-alloys). Time above liquidus is a window, not a target. Long enough for full wetting and collapse, short enough that intermetallic growth and the package's budget stay respected — the clock the §4.5 pass flew and this section now sets. The cool is bounded like the ramp. Stress written into freezing joints is collected later — the cool-down descends at its own cited rate, and the stillness law owns the whole descent (bga-removal-with-ir-rework-station). The starting envelope is industry's, then the inputs adjust it. Typical opening bounds — ramp near one to three degrees per second, soak sixty to one hundred twenty seconds below liquidus, SAC peaks commonly in the 235–250 °C band at the joint and always capped by the package's rated classification temperature, time above liquidus forty-five to ninety seconds, cool at four to six degrees per second or gentler — the envelope the datasheets refine, never gospel. Bounded, equalized, deliberate, wetted, windowed, descended — the stages entire. Every stage answers a question a field can fail — learn the reasons, and the numbers become derivable from the datasheets.

The Inputs — The Case File as Temperature

The card is designed from inputs, and every input already lives in the case file (bga-ball-composition-and-alloys). The alloy sets the frame. Liquidus from the chemistry the file named — continuity or conversion — anchoring the peak, the window, and the whole card's vertical scale. The board writes the soak. Mass, layer count, plane area — the six-layer telecom board and the thin sensor strip want different plateaus, and the design reads the board's build before borrowing anyone's timing (reflow-ovens-when-and-why). The package caps the peak. The heat-lifetime budget bounds what this exposure may spend — a twice-reworked package designs to a gentler ceiling than a fresh one, and the exposure is counted on the card before it is spent on the bench (bga-removal-with-ir-rework-station). The neighbors and moisture write the abort lines. The vulnerable connector's rating, the flanking BGA's conduction risk, the bake status the §4.1 line verified — each becomes a named stop on the card, because mid-heat is no place to draft policy (temperature-and-airflow-finding-the-right-settings). The flux writes the soak's chemistry line. Activation temperature band and duration come from the flux or paste datasheet — the soak serves the chemistry as much as the copper, and the datasheet is an input like any other. The delivery platform shapes the ramps. Hot air and IR deliver differently — the same card flies differently on each, and the design names its platform. Alloy, board, budget, neighbors, platform — the inputs entire. A card whose every line cites an input is a design — one that cannot is a recipe wearing the format.

The Verification — Hypothesis, Junctions, Library

The designed card is a hypothesis, and the section's Professional core is refusing to trust it untested (bga-removal-with-ir-rework-station). The instrumented run is multi-point. Junctions at the package edge, the open board field, and the most vulnerable neighbor — taped per the standing law, logged through the run — because the card made promises about all three, and one junction can only audit a third of them. On donors, the gold standard goes further. A junction routed under the package at a ball row — through a drilled board, secured with high-temperature attachment — reads what the edge can only estimate, because edge-to-ball offsets run ten to twenty degrees on large packages, and the offset the donor measures travels with the card for the boards that cannot be drilled. The stand-in flies first where it can. A donor of the patient's class carries the first run — hypotheses verify on boards whose failure costs nothing, and the patient inherits only profiles that already measured well (bga-ball-composition-and-alloys). The trace is read against the card. Measured ramps against designed bounds, the soak's actual plateau, the peak the corner junction really saw, the time above liquidus the field really spent — its minimum audited at the coldest junction, its maximum at the hottest — every gap between intent and measurement is the iteration's work list (reflow-ovens-when-and-why). Iteration is cited, not vibed. Each revision names what changed and why — the soak stretched twenty seconds because the field junction lagged the package's — and the revision notes are the library's memory (temperature-and-airflow-finding-the-right-settings). The library holds verified cards only. Per board class, with platform named, revisions cited, and the first-article law satisfied — the §5.2 discipline generalized to every profile the bench will ever build. And the failure signatures feed back. A popcorned package indicts the ramp or the bake, a pillowed corner the soak or margin, dull overgrown joints the window, cracked ceramics the shock — field defects read backward to stages, and the reading improves the card that caused them. Instrumented, staged-in, read, cited, libraried — the verification entire. Recipes are folklore until a junction has watched them work — the library is where watched recipes live.

Common Mistakes

  • Copying the last board's card. A recipe that worked borrowed for a board it never met — inputs differ; the design reads this board's build, budget, and chemistry before any number transfers (reflow-ovens-when-and-why).
  • Growing the margin for comfort. Peak nudged up because anxiety asked — the margin is designed and cited; comfort spends budget, neighbors, and laminate (bga-ball-composition-and-alloys).
  • Verifying with one junction. The package edge alone auditing promises made to three places — multi-point or partial: the field and the neighbor were promised things too (bga-removal-with-ir-rework-station).
  • Revising without citation. The soak stretched 'a bit' after a bad run — uncited revisions are folklore returning through the library's front door (temperature-and-airflow-finding-the-right-settings).
  • Trusting the card on the patient first. A fresh hypothesis flown on the customer's board — stand-ins exist; the patient inherits measured profiles, not confident ones.

Troubleshooting Guidance

Profile faults read backward from the field: cracked, uneven, pillowed, dull, stressed. If ceramics crack or a package popcorns: the ramp or the bake — audit the measured climb against the bound, and the §4.1 moisture line before it (temperature-and-airflow-finding-the-right-settings). If the field arrives uneven — one corner late, one region proud: the soak — stretch the plateau until the field junction and the package junction arrive together, and re-verify (reflow-ovens-when-and-why). If joints pillow or corners never wet: the peak margin or the soak's evenness — the coldest junction's trace says which: a corner that never reached liquidus indicts the margin; one that reached it late indicts the soak (bga-ball-composition-and-alloys). If joints come out dull, grainy, and overgrown: the window — measured time above liquidus against the card's, because dull fields are usually long fields (bga-removal-with-ir-rework-station). If the trace and the card disagree everywhere: the platform — the card designed for one delivery flying on another, or a junction placement that audits nothing: verify the instrumentation before revising the design. The throughline: every field defect is a stage's signature — read backward, revise cited, re-verify instrumented.

Verification & Testing Methods

Confirm the design craft before the placement section builds on it:

  • [ ] I can name the five stages with reasons a bench can say out loud — and bound both ramps with causes, not habits.
  • [ ] I can design a profile card from a case file's inputs — alloy, board, budget, neighbors, platform — with every line cited.
  • [ ] I can set a peak margin as liquidus plus enough, justified against the coldest corner and the package's ceiling.
  • [ ] I can bound time above liquidus and read dull, overgrown fields back to a window that ran long.
  • [ ] I can verify a card multi-point on a stand-in, iterate with cited revisions, and library only what junctions watched work.

Then try the practice exercises below — design and instrumented verification on donors; scenarios differ from the quiz.

Practice Exercises

  1. Design one card from a real file (6 minutes, a case file, the template). Build the profile card from inputs — alloy's liquidus, the board's mass read honestly, the package's budget ceiling, the neighbors' abort lines, the platform named — with every number's reason written beside it (bga-ball-composition-and-alloys).
  2. Instrument and fly the hypothesis (7 minutes, a class-matched donor, three junctions, the platform). Tape junctions at package edge, open field, and the most vulnerable neighbor; fly the card under the platform's full law; and log all three narrations through every stage (bga-removal-with-ir-rework-station).
  3. Read the trace and iterate (7 minutes, the run's log, the card). Compare measured reality to designed intent stage by stage — ramps, plateau, peak at the coldest junction, window — write the cited revision for every gap, and fly the revised card once more if the donor allows (reflow-ovens-when-and-why).
  4. Library the verified card (5 minutes, the library). Enter the card per board class with platform named, revisions cited, and the verification trace attached — then read one field defect from the chapter's history backward to its stage as the library's first diagnostic note (temperature-and-airflow-finding-the-right-settings).

These core steps — the cited design, the instrumented flight, the read-and-iterate, and the libraried card — are tested in the Chapter Quiz at the end of this chapter, where a score of 80% is required to continue.

Key Takeaways

  • A profile is five stages with reasons — bounded ramp, equalizing soak, deliberate climb, designed peak, windowed time above, bounded cool — and the reasons are what transfer to the board the numbers never met (reflow-ovens-when-and-why).
  • The profile card is the case file arriving as temperature: alloy sets liquidus, board writes the soak, budget caps the peak, neighbors and moisture write the abort lines — every line cited or it is a recipe in the card's format (bga-ball-composition-and-alloys).
  • The peak margin is liquidus plus enough for the coldest corner — designed, cited, and small, because comfort spends budget, neighbors, and laminate (temperature-and-airflow-finding-the-right-settings).
  • Time above liquidus is a window, not a target — long enough to wet and collapse, short enough to spare intermetallics and budget — and dull, overgrown fields are usually long fields (bga-removal-with-ir-rework-station).
  • The verification law generalizes: hypotheses fly instrumented and multi-point on stand-ins, iterate with cited revisions, and only junction-watched cards earn the library — recipes are folklore until measured.

Skills Learned

  • You can now design a profile from inputs with every line cited.
  • You can now set margins and windows as reasons, not habits.
  • You can now verify a card multi-point and read its trace honestly.
  • You can now iterate with citations and library only what measured.
  • You can now read field defects backward to the stages that wrote them.

Glossary Additions

  • profile card — the written constitution of a heat operation: the five stages — bounded ramp, soak, climb, peak, cool — with their targets, the time-above-liquidus window, the abort lines, and the platform it was designed for, all on one card written before any element warms. A designed card cites an input for every line — the alloy's liquidus from the case file, the soak stretched for this board's mass, the peak capped by this package's remaining budget, the abort lines written from the neighbors and the moisture state — which is what separates design from recipe: a card whose numbers cannot name their reasons is folklore in the card's format. The card is flown as written, revised only with citations, and earns the profile library only after an instrumented run has watched it work.
  • peak margin — the designed distance between an alloy's liquidus and a profile's peak: enough above the melting point that the field's coldest corner — over the ground plane, under the shadow — genuinely wets and collapses, and no more, because every degree past enough is spending: the package's heat-lifetime budget, the neighbors' tolerance, and the laminate's stress all pay for insurance nobody priced. The margin is a cited number, justified against the coldest junction's measured arrival rather than against anxiety — a pillowed corner indicts a margin — or a soak — that ran too small, while a spent budget and a slumped neighbor indict one that grew for comfort — and the discipline is the same both directions: designed, cited, small.
  • time above liquidus — the window a reflow spends with the field molten: long enough for every joint to wet fully and the collapse to complete, short enough to bound the intermetallic growth that every molten second feeds and the package-budget spending that every exposure counts. The window is set on the profile card and audited two-sided on the verification trace — the minimum at the coldest junction's crossing, because wetting must complete everywhere, and the maximum at the hottest junction's, because intermetallic growth and the package's budget are spent where the field ran warmest — and its failure signatures read cleanly: fields that pillow or never collapse ran the window — or the peak — short, while joints that come out dull, grainy, and overgrown ran it long. The §4.5 attach pass and both removal crafts fly this window; this section is where it is designed rather than inherited.

Suggested Next Sections

Must read next:

  • BGA Placement and Alignment — Section 5.4 takes the rebuilt package back to its board at Professional precision: seating on the lands, the optical and mechanical aids, and the witness discipline that catches rotation while it is still free to fix.

Recommended: