GMSL 通信中断应对指南：如何确保链路锁定

GMSL 通信中断应对指南：如何确保链路锁定

作者：tombrady

设想这样一种场景：你刚完成全新电路板设计，或是满怀期待试用多款评估套件，却发现收发两端器件无法建立通信。调试千兆多媒体串行链路（GMSL）时遇到这类问题，不仅会拉长调试周期，还容易让人束手无策。GMSL是现代设备中成熟可靠的传输技术，可在汽车电子、工业设备、数字医疗、航空航天等各类严苛工况下实现高速、低时延数据传输。为此，采用系统化的方法来调试GMSL链路锁定问题十分关键。本指南梳理了实现链路锁定、排查相关故障所需的思路与工具，保障数据传输稳定、不间断。

GMSL是ADI推出的SerDes链路技术。SerDes是SERializer（串行器）和DESerializer（解串器）的合成词，顾名思义，整套链路由两类器件配对互联，搭建串行传输通道。

调试串行链路会存在特有难点：不能只单独排查单颗芯片，必须从整机系统层面分析。此外，连接两端器件的传输通道，也是缩小故障范围时需要考量的一环。

GMSL链路锁定

无论是摄像头端还是显示屏端的GMSL器件，链路锁定是所有GMSL系统最基础的功能检查项。锁定状态可通过GMSL评估套件上的LOCK指示灯，或是寄存器内的LOCKED状态位读取判断。

Field Monitoring and Maintenance of Functional Safety The ISO 26262 Functional Safety standard not only governs the development phase but also mandates ongoing maintenance after a product is released to production. While ASIL compliance is typically verified at the end of development, manufacturers are also responsible for monitoring products in the field. This includes evaluating any safety implications that may arise from customer returns or failures observed during real-world operation. The added value of the field monitoring is at least in two areas: Safety Aanomalies First, we need to verify that returned parts show no potential safety anomalies. This is an essential step in validating our safety concepts on real-world applications. If anomalies are found, corrective actions must be taken. Based on feedback provided to the development team, updates to the original safety case—specifically, the product's FMEDA and risk analysis—must be made. Lessons Llearned Second, returned parts provide valuable insights and lessons learned for a specific product or product category. As part of the 8D problem resolution process, these lessons must drive preventive measures to avoid future failures. Corrective actions taken for a specific product are documented in the DFMEA and serve as references for future designs. To complete the feedback loop, any new failure modes or corrective actions are transferred to our internal databases. This ensures that updated DFMEA templates automatically incorporate these insights into future product development. Closing the loop with efficient field monitoring processes Figure 1 (below) shows the relevant items for closing the loop and supporting a process for continuous improvement. The main steps of the field monitoring process are: Analyze all automotive 8D reports in a specific period of time with respect to their root causes and document their status with respect to observed failure modes, effects at the application level, and root causes associated with components within the IC. Additionally, classify the associated report by its main root cause categories like “IC Production (Fab)”, “Test Coverage”, “Design” or “Assembly”. Conclusions may be drawn by a dedicated peer review. If corrective actions are required to address failures in areas such as design or test coverage, the corresponding preventive measures must be documented in the risk analysis, specifically in the DFMEA (Design FMEA). For fault types related to IC production or assembly, the PFMEA (Process FMEA) must be reviewed to ensure it includes preventive or detection measures for future occurrences. This step is not unique to Functional Safety—it is a standard part of quality management and aligns with international standards like IATF 16949. In the third step, we evaluate the impact of the observed symptom with respect to the product-specific safety goals. This involves verifying whether the failure mode is already addressed by the existing safety concept and whether a dedicated safety mechanism is in place to detect or mitigate it. Evidence for this is typically gathered through bench evaluation during the failure analysis. If the failure mode is effectively covered by a safety mechanism or if it naturally leads the system to a safe state, it is not classified as a safety anomaly. A team of experts— typically functional safety managers—summarizes the evaluation results and creates a conclusion that is documented in the relevant database or 8D systems. Diagram 3, SmartArt diagram Figure 1: Closing the loop using Field Monitoring Field monitoring processes supporting lessons learned The data extracted from the field monitoring process contains a lot of valuable additional information that can be used for new projects. Any new set of potential failure modes or additional preventive/detective measures can be collected and transferred back to the corresponding risk analysis tools. Items relevant to a DFMEA (Design FMEA) would lead to an update of the associated DFMEA templates supported by our DFMEA tools and lead to a new revision of the template. Users can be automatically informed about template updates, and new projects will always be started from the newest version, considering all relevant information extracted from field data. The same applies to any items pertinent to the FMEDA that could result in an update of the related components within the component reuse libraries. The efficient implementation of the field monitoring process is not only a formal requirement of ISO26262 but also adds significant value by identifying risks related to field returns and closing potential gaps. Additionally, it serves as a valid source of lessons learned.

图1：评估套件上LOCK指示灯的位置

链路锁定，通常称为LOCK，表示GMSL PHY能够有效通信。此状态意味着两条数据通路，即前向通道（解串器接收通路）与反向通道（串行器接收通路），已正确同步。

Figure 2: Asymmetric, Full Duplex Data Directions

图2：非对称全双工数据传输方向

一旦建立链路锁定，视频传输和控制通道功能将立即可用。为了监测链路锁定状态，每个GMSL串行器和解串器都配备有专用的LOCK引脚，提供关于锁定状态的硬件级指示。

此外，也可读取寄存器来验证此状态，如图3所示。

Figure 3: MAX96716A Register for Link A LOCKED Bit

图3：MAX96716A寄存器用于指示链路A LOCKED状态位

多路输入器件，以MAX96724为例，LOCK引脚仅在所有启用的链路都被锁定时才会置于有效电平。若无需使用全部链路，需先禁用闲置链路，才能正确获取LOCK状态。在这类场景中，LOCK引脚仅在所有链路被锁定时才置于有效电平，以指示系统按预期运行。

虽然锁定状态在链路的两端都有反映，但一般只需监测连接到主处理器的端口即可。对于摄像头系统，应监测解串器上的LOCK引脚。对于显示系统，则应监测串行器上的LOCK引脚。

调试步骤

系统在启动时若未能实现链路锁定，说明基本配置或连接存在问题。以下是排查链路锁定问题的检查清单，以帮助识别和解决问题：

  • 核查外部连接：检查GMSL器件的全部物理接线，确保接线牢固、对位无误。
  • 检查GMSL链路的连接，确认GMSL链路的正负端口是否正确连接。
    • 常见问题包括原理图符号不正确，或屏蔽双绞线配置错误。
    • 必要时，可使用万用表进行快速检查，以验证引脚间的连通性和线缆连接的正确性。
  • 确认电源供电正常：各路电源输出需符合规格，满足GMSL芯片的电压要求。
  • 校验配置引脚参数：核对CFG引脚设置，确认器件上电后载入正确的配置。
    • 常见问题包括：施加到CFG引脚分压器的VDDIO电压不正确，或外部连接在启动过程中导致CFG引脚上出现噪声或错误电压。
    • 如果一端以3 Gbps模式启动，而另一端以6 Gbps模式启动，则两端将无法锁定。
    • 在支持隧道模式的部件上，务必确保链路的两端以相同模式（像素模式或隧道模式）启动。模式不匹配将导致数据损坏。
  • 使用外部晶体时，检查晶体振荡器，确认其在25 MHz下正常工作并产生干净的时钟信号。
  • 区分活动和非活动链路：在配置中禁用非活动链路。
  • 查阅器件勘误文档：读取0x0E寄存器内器件版本号，对照勘误文档，确认是否需要进行其他更改。
    • 勘误文档可前往analog.com官网产品页面下载，例如MAX96724勘误表。

若完成以上全部校验仍无法锁定链路，则大概率存在信号完整性问题。排查物理设置中的潜在问题，例如线缆质量或电磁干扰。条件允许时，可单独与链路两端器件通信，分别确认单端器件功能是否正常。

按上述流程逐项排查，即可定位并解决GMSL系统链路锁定故障，保障通信稳定、系统功能正常。

结语

在GMSL系统中建立并维持链路锁定，是保障串行器、解串器正常通信的首要步骤。原理图符号错误、线缆配置错乱、启动模式不匹配、信号完整性缺陷等诸多常见问题，都会导致链路工作异常。但无需担心！按照上文梳理的步骤操作，即可搭建稳定可靠的GMSL传输链路。最后，掌握链路锁定的实现与监测方法，是深度排查、稳定运维整套GMSL系统的基础。

链路成功建立后，我们将讲解如何检测像素时钟进入串行器，以确认数据传输正常。敬请关注后续内容！

查看 GMSL调试系列全部博客文章