FA & Reliability Daily Abbreviations

FA 確認 / 定位 / 排除 / 確認。FA 報告的標準動詞。

• FA confirms gate oxide breakdown as the failure mode.
• FA rules out process contamination — failure is design-related.

OBIRCH 定位 / 找出 / 偵測 / 顯示。OBIRCH 是找漏電 / 短路位置的招牌工具。

• OBIRCH locates the short to the M3-M4 via column.
• OBIRCH detects a hot spot at the wafer center — likely diffusion issue.

TEM 截面揭示 / 顯示 / 確認。Gate oxide / metal void / 介面問題的標配檢驗。

• TEM cross-section reveals 3 atomic layers of gate oxide instead of 5.
• TEM cross-section confirms void in the M3 via — process gap.

HTOL 試驗在 N 小時失敗 / 通過。Reliability qual 必出結論。

• HTOL fails at 800 hours — escalate to design for review.
• HTOL passes 1000 hours with 0 failures across 77 units.

EM 發生 / 加速 / 縮短壽命。先進節點金屬線越窄,EM 越敏感。

• EM accelerates at elevated current density — keep below 1 MA/cm².
• EM lifetime shortens by 10x with each 10°C temperature rise.

ELFR 推估 / 達到 / 超過 / 限於。客戶最在意的 reliability KPI。

• ELFR projects to 50 ppm based on the JANUS dataset.
• ELFR exceeds customer spec of 100 ppm — escalate.

PPM 維持在 / 飆 / 降 / 穩定。Field return ppm 是客戶看 supplier 的硬指標。

• Field return ppm tracks at 35 — better than competitor's 80.
• PPM spiked to 200 on Product Y last quarter — escalation to senior management.

FIB 切片 / 暴露 / 切割 / 銑出。FA 拆 die 的關鍵動作。

• FIB the via to expose the cross-section for TEM analysis.
• FIB sections the M3 metal at three locations to capture the defect.

FA conclusion · standard

FA conclusion on failure: failure mode is mode; root cause is cause; confirmed by evidence.

{失效} 的 FA 結論:失效模式是 {模式};根因是 {cause};以 {證據} 確認。

• FA conclusion on RMA-2026-018: failure mode is gate oxide breakdown; root cause is TDDB from extended high-voltage operation; confirmed by TEM cross-section and EFA voltage mapping.▸ 聽念法
• FA conclusion on RMA-2026-022: failure mode is M3 via open; root cause is incomplete via fill from a CVD chamber excursion; confirmed by TEM cross-section and chamber correlation.▸ 聽念法

reliability projection

Reliability projection on scope: metric = value; based on N units tested over duration.

{範圍} 的 reliability 推估:{指標} = {value};基於 {N} 個樣品在 {duration} 內測試。

• Reliability projection on Product X: ELFR = 65 ppm; based on 77 units tested over 1000 hours HTOL.▸ 聽念法
• Reliability projection on Product Y BEOL EM: lifetime = 12 years at use conditions; based on 30 wafers tested over 6 stress conditions.▸ 聽念法

FA progress update

FA report for failure_id due date: scope is scope; preliminary indication is finding.

{失效 ID} 的 FA 報告於 {日期} 出爐:範圍 {scope};初步指向 {finding}。

• FA report for RMA-2026-018 due 03/18: scope is 5 RMA units from customer X; preliminary indication is gate oxide TDDB.▸ 聽念法
• FA report for #2026-318 due 03/20: scope is 20 units from yield drop event; preliminary indication is M3 via incomplete fill from CVD chamber excursion.▸ 聽念法

reliability failure escalation

Reliability stress test failed at time: failure mode mode; recommend action before milestone.

{試驗} 在 {時間} 失敗:失效模式 {模式};建議在 {milestone} 前 {action}。

• Reliability stress HTOL failed at 800 hours: failure mode gate oxide breakdown; recommend design review before tape-out sign-off.▸ 聽念法
• Reliability stress EM failed at projected 5 years: failure mode metal void at via interface; recommend recipe re-cal before mass production release.▸ 聽念法

customer ppm report

Customer customer field return ppm: current (target target); top failure mode mode at N%.

{客戶} 現場退貨 ppm:{當前}(目標 {目標});最主要失效模式 {模式} 占 {N}%。

• Customer X field return ppm: 30 (target <50); top failure mode ESD damage at 60%.▸ 聽念法
• Customer Y field return ppm: 120 (target <100); top failure mode mold delamination at 45%.▸ 聽念法

Email — HTOL Fail Escalation

Reliability engineer 發給 design team + PQA,HTOL 在 800 hours 出失敗。

Subject: [ESCALATION] HTOL fail at 800h on Product X — design review needed

Hi all,

Heads up on a reliability failure that needs cross-functional attention before tape-out sign-off.

HTOL test on Product X (Rev D), running at 125°C with operating voltage:
- 3 of 77 units failed at 800-hour readpoint.
- Failure mode: gate oxide breakdown on the power management block.
- FA on the failing units: TEM cross-section confirms thinning of gate oxide at one corner of the high-voltage MOSFET stack.

Root cause path:
- TEM shows gate oxide thickness 4.8 nm vs design 5.5 nm at the affected location.
- Process recipe trace shows the M0 oxide grow recipe ran nominal during these wafers.
- Likely cause: localized layout density issue allowing thinner oxide growth in this geometry.

Customer impact assessment:
- HTOL spec is 1000 hours pass. Failing at 800h means our reliability projection (ELFR ~65 ppm) overstates lifetime if customer use case approaches the stress condition.
- Worst case (customer running at max voltage 24/7): field failures could climb to >150 ppm in year 5.

Request:
- Design team: review the layout in the affected MOSFET stack; consider increasing oxide design margin or relaxing layout density.
- Process: investigate whether a Rev E tweak to the M0 oxide recipe could compensate.
- PQA: flag this on the tape-out checklist; we shouldn't release without resolving.

I'll send the full FA report by Friday EOD, and we should sync next week before the staff meeting.

Thanks,
Reliability ENG

相關縮寫:HTOL · FA · TEM · ELFR · PPM · PQA

Conf Call — Field Return FA Briefing

Customer-facing 內部會議,FA 部門 lead 解釋客戶 RMA 樣品的 FA 結論。

Thanks. Briefing on the customer X field return — 5 units that came back as DOA after 6 months in the field.

Initial scope:
- 5 units, all the same product (Product Y, 28-nm SoC, packaged in BGA-484).
- Customer report: died randomly after 4–6 months of field use.

FA results:
- Electrical: all 5 units failed in the I/O ring, specifically the high-speed serializer block.
- OBIRCH on a representative unit: hot spot localized to one I/O column.
- FIB cross-section + TEM: M3 via on the affected column shows incomplete fill — a 30% void at the via top.
- Chamber traceability: all 5 units traced back to wafers run on CVD chamber B during a 6-hour window in week 14 of last year.

Root cause: incomplete via fill from a transient gas flow excursion on CVD chamber B. The fill defect was sub-critical at ship time (passed CP and FT), but EM-accelerated under field use, eventually opening the via.

ELFR / customer impact:
- Wafers from the affected window were shipped to 2 customers (X and Z).
- Total exposure: ~24,000 units.
- Field return rate to date: 5 units / 12,000 shipped = 0.04% (400 ppm). Above customer spec of 100 ppm.

Action:
- 8D draft underway; due to customer X by next Friday.
- Containment: all suspect wafers from the affected window already shipped — no further shipments to contain.
- Long-term: CVD chamber B gas flow SPC tightened; added a per-wafer thickness verification step for via metal CVD.

Open to questions.

相關縮寫:RMA · FA · OBIRCH · FIB · TEM · CP · FT · EM · ELFR · PPM · 8D

Deviation Report — TDDB Margin Tightening

新節點的 reliability qual 中 TDDB 邊際變窄,deviation report Section 4 + 5。

4.0 Root Cause — TDDB Margin Tightening at 7-nm Node

4.1 Symptom
TDDB stress test on Product X (7-nm Rev D) gate oxide shows lifetime projection of 8 years at use conditions, vs the design target of 12 years. Margin tightened by ~35% compared to the prior 10-nm node.

4.2 Investigation Path
- TEM cross-section on stressed units shows oxide thickness 4.8 nm vs design 5.0 nm — within process spec.
- Stress data fitting reveals slope (β) is lower than expected; suggests the oxide quality has degraded, not the thickness.
- Process trace: oxide grow recipe used a new precursor batch starting 11 weeks ago.
- Inline ellipsometry on monitor wafers shows refractive index dropped from 1.49 to 1.46 — points to lower film density.

4.3 Root Cause Statement
The TDDB margin tightening is attributed to lower film density in the gate oxide, caused by the new precursor batch introducing higher impurity levels (suspect: chloride contamination from the precursor synthesis change). Confirmed by RI shift, stress slope analysis, and incoming precursor SIMS.

4.4 Containment & Action
- Switched gate oxide grow back to previous precursor vendor immediately.
- All wafers using the new precursor batch (8 weeks of production, ~2400 wafers) held for enhanced screening; running TDDB on monitor lots from the affected window.
- Customer notification drafted — no shipments yet, but flagged as a potential issue.

5.0 Long-Term Action

5.1 Process
- New precursor batch qualification requires additional inline RI + SIMS check before release.
- Oxide quality metric (RI + density) added to SPC for the gate oxide step.

5.2 Reliability
- TDDB monitor wafers added to every weekly stress test batch — early detection of oxide quality drift.

5.3 Supplier
- SCAR issued to the precursor supplier requesting root cause on the impurity excursion.
- Supplier-side change control to be reviewed in next quarterly supplier audit.

5.4 FMEA Update
- "Precursor impurity drift → oxide quality degradation → TDDB margin loss" added as a critical reliability failure path.

相關縮寫:TDDB · TEM · SIMS · SPC · SCAR · FMEA

讀完了?來測驗看看記住了沒。