BGA Rework Profile Library: Eight Recipes for Common Package Sizes

Rework profiling on a one-off basis is how mid-volume shops burn boards. Every time the operator dials in a new profile from scratch — eyeballing the board thickness, glancing at the BGA size, guessing at the soak temperature — the chance of damaging an adjacent component or under-reflowing a corner ball goes up. We maintain a library of eight base profiles indexed by three variables, and 95% of incoming rework jobs map cleanly to one of them.

The three indexing variables, in order of impact on the profile:

• Package size (mm × mm) — controls the thermal mass under the head. A 7mm × 7mm BGA needs a fundamentally different ramp than a 27mm × 27mm one.
• Ball pitch (mm) — controls the solder volume and the void clearance requirements. 0.4mm and 0.8mm pitch react differently to ramp rate.
• Substrate thickness and copper weight — controls how fast heat propagates from the head to the bottom of the joint. A 0.8mm board responds faster than a 2.4mm board with 2oz inner planes.

"Eight profiles is enough to cover every BGA we've seen in three years. Tuning the ninth profile is usually a sign the operator picked the wrong starting point — not a new package class." — Pioneer Horizon process engineer

The rest of this article walks the eight base profiles, the thermocouple placements we use to validate each one, and the tweaks that turn a base profile into a job-specific one without restarting the development cycle.

The library lives on the rework station PC as a JSON file with eight named profiles. Each profile has soak duration, soak temperature, ramp rate to peak, peak temperature, time-above-liquidus (TAL), and cool-down rate. Operators select by package class; the station auto-populates the parameters.

Compact convection profiles (hot air, single-sided heat)

• Profile A — small QFN/CSP, 5–10mm, 0.4–0.5mm pitch. Soak 150°C for 60s, ramp 1.5°C/s to 235°C peak, TAL 60–75s, cool 2°C/s. Use for small power QFNs and CSPs on 1.0–1.6mm boards.
• Profile B — small-medium BGA, 10–17mm, 0.5–0.8mm pitch. Soak 155°C for 75s, ramp 1.2°C/s to 240°C peak, TAL 70–90s. Workhorse profile for most MCU BGAs.

Medium IR-assist convection profiles

• Profile C — medium BGA, 17–23mm, 0.8–1.0mm pitch, thin board. Bottom IR preheat to 130°C, soak 160°C, ramp 1.0°C/s to 240°C peak, TAL 80–100s.
• Profile D — medium BGA, same range, thick board (≥1.8mm) or heavy copper. Bottom IR preheat to 145°C, soak 165°C for 90s, ramp 0.8°C/s to 240°C peak, TAL 80–100s. The slower ramp accommodates the larger thermal mass.

Large BGA dual-side profiles

• Profile E — large BGA, 23–30mm, 0.8–1.0mm pitch. Bottom IR preheat to 150°C plus top convection, soak 165°C for 90s, ramp 0.8°C/s to 235°C, TAL 90–110s.
• Profile F — large BGA, same range, on 8+ layer board with thermal planes. Bottom IR to 155°C, soak 170°C for 100s, ramp 0.7°C/s to 240°C, TAL 90–120s, slow cool 1.5°C/s.

Specialty profiles

• Profile G — fine-pitch FPGA/SoC, 23–35mm, 0.6mm pitch or finer. Same envelope as Profile F but with extra-tight thermocouple monitoring on three corners. Used on $400+ FPGA replacements.
• Profile H — POP (package on package) restacking. Two-stage profile — first stage detaches POP, second stage reattaches with fresh paste. Special handling for the upper memory stack.

Profiles A through D account for roughly 75% of the rework jobs we run in a typical month. E and F account for another 18%; G and H for the remainder.

A profile is only as good as the validation behind it. We validate every profile with thermocouples — not on the part being reworked (you can't reuse the validation board), but on a paired sacrificial coupon laid out to match the customer's actual board stackup. The placement is what separates a usable validation from a misleading one.

Standard four-point placement

• TC1 — joint under the BGA centre. Drilled from the back of the coupon into the centre ball pad. This is the slowest-heating joint and the one most at risk of under-reflow.
• TC2 — joint under a corner ball. Same drill technique. Corner balls heat fastest and are most exposed to peak-temperature overshoot.
• TC3 — top of the package, glued near the centre. Captures what the head is seeing.
• TC4 — adjacent SMD passive, 5mm from the BGA edge. Validates the thermal stress on neighbouring components. Anything reading above 200°C on this TC during a rework is a red flag for collateral damage.

What we record

1. Peak temperature on each TC.
2. Time-above-liquidus on TC1 and TC2 (the joint TCs).
3. Delta-T between TC1 and TC2 — the corner-to-centre temperature gradient. Anything above 12°C suggests profile asymmetry; we tune the head height or air flow to close it.
4. Cooling rate from peak to 150°C on TC1.

The validation gets a one-page printout that travels with the profile in the library. When the same package class shows up six months later on a different customer's board, we don't re-validate from zero — we cross-check the existing profile data against the new board's stackup and re-run only if the stackup is meaningfully different.

About one rework job in five doesn't drop cleanly into a base profile. The base profile is still the right starting point — the job just needs one or two specific adjustments. We catalogue the common tweaks so an operator can apply them in a known, reversible way without inventing a new profile.

Heavy thermal-plane underlay

Symptom: centre-ball TC reads 5–8°C below corner-ball TC during validation. Cause: solid copper plane under the BGA draining heat. Fix: increase bottom IR by 10°C, extend soak by 15s. Re-validate. Usually closes the gradient inside one iteration.

Adjacent tall component blocking airflow

Symptom: corner-ball TC reads 5°C above centre during validation. Cause: nearby connector or shield can deflecting hot air. Fix: top nozzle reposition, increase head height by 1–2mm, slightly raise top airflow. Re-validate.

Mixed BGA + adjacent leaded component to protect

Symptom: TC4 on adjacent component above 215°C. Cause: package being reworked is too close to a heat-sensitive neighbour (electrolytic cap, plastic connector). Fix: kapton shielding on the neighbour, top airflow reduction by 10%, slightly tighter head focus. Document the fix in the job-specific profile note.

Underfilled BGA being removed

Symptom: BGA refuses to lift cleanly even at full TAL. Cause: underfill bonded to the substrate. Fix: pre-heat soak extended to 90s, peak hold extended to 110s, and a mechanical assist (vacuum pickup + light prying force). On very stubborn underfills we cut the underfill with a precision blade before the heat cycle.

• For tweak documentation we use a delta-format: base profile name + a JSON patch describing the changes.
• Tweaks that recur three times become candidates for a new base profile.
• The library's been stable at eight profiles for two years now — most "new" jobs are tweaks, not new profiles.

For the underlying voiding diagnostic when rework is being driven by void rates, see our BGA voiding article.

Not every BGA failure is rework-economical. The honest conversation we have with customers is when rework is the right answer versus when board replacement is cheaper. For low-value boards the answer is usually replace; for high-value boards rework wins even at lower yields.

Our rework yield envelope by profile

• Profiles A–D — 96–99% first-pass yield on validated profiles, less than 8% void rate post-rework.
• Profiles E–F — 92–96% first-pass yield, less than 10% void rate.
• Profile G (fine-pitch FPGA) — 88–93% first-pass yield, less than 12% void rate. The lower yield is structural — the package is large, the ball count is high (1700+ on some FPGAs), and the cost of an additional 4% yield is more than the board itself.

When we recommend replacing the board instead

1. Board unit cost < $200 and BGA unit cost < $40 → replace the board, scrap the failure.
2. BGA is single-sourced and on allocation, while the board is cheap → strip the BGA, scrap the board.
3. Two prior rework cycles on the same site → the pad has been thermally stressed enough that a third reflow is reliability-suspect. Replace.

Cost breakdown for a typical Profile F rework

• Setup and profile selection — 25 minutes.
• Component removal, pad clean-up, paste application — 35 minutes.
• New BGA placement and reflow — 12 minutes.
• X-ray inspection and electrical verification — 20 minutes.
• Total: roughly 90 minutes per board. Indian-market cost: ₹4,500–6,000 per rework including consumables and X-ray time.

If you'd like us to review a recurring rework problem on one of your active lines — what we'd recommend in terms of profile, board changes, or replace-vs-rework — send the failure analysis and we'll come back inside three working days. For broader rework economics on fine-pitch parts, our fine-pitch BGA rework article covers the reball-vs-replace decision in more depth.