QKC20190823：讀書會_半導體元件基本失效率 (ISO 26262-11:2018, 4.6)

hlperng

品質學會品質知識社群 (QKC) 讀書會
專題：半導體元件基本失效率 (ISO 26262-11:2018, 4.6)
時間：2019 年 08 月 23 日 (星期五) 16:00 - 18:00
地點：台北市羅斯福路二段 75 號 9 樓 (時代大樓品質學會九樓會議室)
引導：彭鴻霖 會友

hlperng

ISO 26262-11:2018 「ISO 26262 對半導體的應用指南」為道路車輛功能安全性 ISO 26262 系列國際標準的第 11 部分，是 ISO 26262 系列國際標準 2018 年第一次改版時新增的兩個部份之一，共 188 頁。ISO 26262-11 的前身為 ISO/PAS 19451-1:2016 及 ISO/PAS 19451-2:2016。

ISO 26262-10:2018 第 4.6 節「半導體基本失效率」(base failure rate for semiconductors)，提供如何計算及使用半導體元件基本失效率相關議題的澄清、指南、及範例等資訊，作為進行可靠度預估作業的參考。

ISO 26262 依照發生原因，將車用電子產品中的安全相關失效包括：單點故障指標 (single-point fault metric, SPFM) 、潛在故障指標 (latent fault metric, LFM)、及隨機硬體失效機率指標 (probabilistic metric for random hardware failure, PMHF)。其他相關故障指標，包括：殘餘故障 (residual fault)、安全故障 (safe fault) 不會造成違反安全性目標，一般保守假設為 50 %。

定量安全性分析進行計算作業時，一般排除系統性失效，聚焦在隨機硬體失效，基本失效率是主要的輸入資料。


ISO 26262 特別強調系統性失效 (systematic failure) 及隨機性失效 (random failure) 之間的差異。

根據產業資訊來源建立的常設基本失效率計算方法，主要是作為估算半導體元件隨機硬體失效機率指標 (PMHF) 之用：

• 國際電工委員會 (IEC) 發行的 IEC/TR 62380 國際標準第 7 章 。國際標準 IEC 62380:2004 在 ISO 26262 發行之後已由 IEC 宣布過期，由 IEC 61709:2017 取代之。
• 德國西門子公司發行的 Siemens SN 29500 公司標準第 2 部 (SN 29500-2)。SN 29500 係根據 IEC 61709 擬定。
• 法國空中巴士集團 (Airbus) 發行的 FIDES 指引產業標準
  

國際標準對於電子產品可靠度預估的現況：

• ISO 26262-11:2018 會主導工業產品可靠度預估的發展趨勢
• IEC 61709:2017 取代與合併 IEC 62380:2004。
• IEC 61709 與 SN 29500 有密切關聯。
• 法國 FIDES Guide (2009) 的現況不清楚，目前因應工業 4.0 的潮流，德國 Siemens SN 29500 暫時佔優勢。
  

ISO 26262-11:2018 的內容，實質上與 IEC 61508 的關係較為密切。

半導體失效率模型，從結構觀點考量，可分為晶粒 (die) 、封裝 (package) 與過應力 (overstress) 三部分，其中晶粒與封裝的失效模式為隨機性，而過應力失效則屬於系統性且與晶粒關係較大。









原則上，ISO 26262-11:2018 認為 EOS 為系統性失效模式，因此不納入 PMHF 之計算式中。












溫度影響因子 (influencing factor) 或溫度加速因子，π_T
T_ref = 55 °C = 328 K
T_str = 125 °C = 398 K
π_T = 78.6

hlperng

ISO 26262-5 第 8.4.3 節規定，安全性目標訂定車輛安全完整性等級 (ASIL) 為 B、C、D 三個等級的電子產品必須決定硬體零件的失效率推定值。這是車用電子產品必須執行可靠度預估作業的主要依據章節。

8.4.3 This requirement applies to ASIL (B), C, and D of the safety goal. The estimated failure rates for hardware parts used in the analyses shall be determined:
a) using hardware part failure rate data from a recognized industry souorce, or
b) using statistics based on field returns. In this case, the estimated failure rate should have a comparable confidence level at least 70 %, or
c) using expert judgement founded on an engineering approach based on quantitative and qualitative arguments. Expert dudgement shall be exercised in accordance with structured criteria as a basis for this judgement. These criteria shall be set before the estimation of failure rates is made.

EXAMPLE 1. Commonly recognized industry sources to determine the hardware part failure rates and the failure mode distributions include SN 29500, IEC 61709, MIL-HDBK-217F Notice 2, RIAC-HDBK-217 Plus, UTE C80-811, NPRD-2016, EN 50129:2003, Annex C, RIAC-FMD:2016, MIL-HDBK-338, and FIDES 2009 Ed. A. The failure mode distributions e.g. those defined by "Alessandro Birolini - Reliability Engineering" can be used.

NOTE 1. The failure rate values given in these databases are generally considered to be conservative.

NOTE 2. In applying a selected industry source the following considerations are appropriate to avoid artificial reduction of the calculated base failure rate:
- the mission profile;
- tha applicability of the failure modes with respect to the operating conditions; and
- the failure rate unit (per operating hour or per canlendar hour).

NOTE 3. If the confidence level for the failure rate of different hardware parts used in the SPFM and LFM evaluation is significantly different, the metrics will be biased.

NOTE 4. It may still be necessary to scale these statistics-based data from field retruns before using them together with values from other data sources with different confidence levels. See also NOTE 2.

NOTE 5. Failure rates based on field returns can be calculated as described in ISO 26262-8:2018, Clause 14 (Proven in use).

NOTE 6. The criteria for expert judgement can include a combination of heuristic information supported by a combination of field data, testing, reliability analysis and physics-of-failure based simulation approaches while considering the novelty of the design.

NOTE 7. Informative references from international reliablity expert bodies can be used: SAE J1211 "Robustness Validation" - Analysis, Modeling and Simulation provides physical-of-failure (PoF) based failure mechanism models, JEDEC-JESD 89, JEDEC JESD 91, JEDEC JESD 94, JEDEC JEP 143, JEDE JESD 148.

NOTE 8. If failure rates from multiple data sources (as listed in 8.4.3) are combined e.g. in the case the failure rate of different parts are not available from the same source, the failure rate can be scaled using a scaling factor such that the quality of prediction of the different failure rates is equivalent. This scaling can be used if a rationale for the scaling factor between two failure rate sources is available.

EXAMPLE 2. An element failure rate is found only in one source whereas a similar element is available in that and another source. The scaling factor is the ratio of the failure rate from these two sources utilising the same mission profile.

EXAMPLE 3. Failure rates from data handbooks are generally considered to be conservative. If a random hardware failure target value consistent with the use of handbook data is chosen, failure rate derived from field returns can be used by applying an appropriate scaling factor (corresponding to a confidence level more conservative than usual, for example).

NOTE 9. If a suitable scaling factor is not available, separate target values compliant with the SPfM and LFM requirements can be assigned to the different elements under consideration (analogues to 8.4.4).

NOTE 10. For semiconductors, see ISO 26262-11:2018, 4.6 for guidance.