结论：复测范围取决于整改影响了什么，而不是整改了几个地方
测试不合格之后，整改再复测是常规流程。问题是复测多少 ——全部重做成本高、周期长；只测不合格项又可能漏掉整改带来的新问题。
判断的依据是：整改措施影响了设备的哪些特性。 影响到的特性对应的项目要复测，未影响的可以论证免测。
注意是「影响了什么特性」而不是「改了几个零件」。 换一个小电容可能影响整个发射频谱，而换一个外观件可能什么都不影响。
判断的方法
第一步，明确整改措施的具体内容。 改了什么、怎么改的。
第二步，分析整改影响的技术特性。 电气参数、结构、屏蔽、接地、软件行为、热特性。
第三步，列出与这些特性相关的测试项目。 这些项目原则上要复测。
第四步，对未列入的项目做免测论证。 说明为什么不受影响。
第五步，与实验室确认复测范围。
第四步的论证要具体。 写「本次整改不影响其他项目」是不够的，要说明整改措施的作用机理，以及为什么它不改变其他特性。
各类整改的影响范围
| 整改类型 | 通常影响的项目 |
|---|---|
| 增加滤波元件 | 发射、抗扰度、漏电流、耐压 |
| 改进屏蔽 | 发射、辐射抗扰度、可能影响温升 |
| 改进接地 | 发射、抗扰度、漏电流、接地阻抗 |
| 更换电源元件 | 发射、温升、单一故障、效率 |
| 改进散热 | 温升、可能影响发射（风扇） |
| 软件修改 | 取决于修改内容，可能影响多项 |
| 结构加强 | 机械项目、可能影响屏蔽 |
| 更换材料 | 阻燃、生物相容性、机械性能 |
| 更改标识 | 通常只影响标识检查 |
注意「增加滤波元件」的影响面很广。 加一个 Y 电容可以降低发射，但同时增加了漏电流；加共模电感可能改变某些频段的抗扰度表现。所以这类整改后，相关的一组项目都要复测。
软件修改的特殊性
软件修改的影响最难判断，因为它不可见：
只改界面文字。 通常不影响电磁特性和安全。
修改控制逻辑。 可能影响输出特性、单一故障行为、基本性能。
修改时钟或功率管理。 直接影响发射。
修改通信协议。 影响 EMC 相关项目。
修改保护逻辑。 影响单一故障和安全相关项目。
建议软件修改时做变更说明，列出修改内容和可能的影响。没有说明的软件版本变更，实验室通常会要求较大范围的复测，因为无法判断影响。
论证的写法
免测论证建议这样写：
描述整改措施。 具体内容和技术细节。
说明作用机理。 整改措施如何解决原问题。
分析影响范围。 影响哪些特性，不影响哪些。
逐项说明免测理由。 对每个申请免测的项目说明。
提供支持数据。 如果有自测数据支持，一并提供。
「提供支持数据」能显著提高论证的可信度。 比如整改后自测漏电流仍在限值内且与整改前相近，这个数据支撑「漏电流不受影响」的论证。
避免反复整改的做法
反复整改是代价最大的情形——每次整改都要复测，周期一次次延长。避免的办法：
一次整改到位。 不要只针对刚好超标的那一点做最小整改，要留出余量。刚好压到限值以内的整改，批次差异可能又超出。
分析根本原因。 只处理表面现象，问题可能换个形式再出现。
评估整改的副作用。 改前先想清楚会影响什么，避免解决一个问题引出另一个。
整改后先自测。 在送去复测之前自己验证一遍，降低再次失败的概率。
一次处理多个问题。 如果有多项不合格，一并整改再复测，比逐项整改逐次复测省时省钱。
「留余量」这一条最实际。 整改的目标不应当是「刚好通过」，而是「稳定通过」。
与变更管理的衔接
整改本质上是设计变更，应当纳入变更管理流程：
变更记录。 整改内容要有正式的变更记录。
影响评估。 不只是对测试的影响，还包括对生产、对其他市场已获批产品的影响。
文件更新。 图纸、物料清单、技术文件要同步更新。
其他市场的申报。 如果产品已在其他市场注册，整改可能构成需要申报的变更。
最后一项容易被忽略。 为了通过某个市场的测试做的整改，可能影响已在其他市场注册的产品状态。建议整改时一并评估各市场的申报义务。
常见的复测范围争议
实务中容易产生分歧的几种情形：
软件版本更新但功能未变。 委托方认为不影响，实验室可能要求复测发射相关项目。解决办法是提供详细的变更说明。
更换等效元件。 同规格不同品牌的元件，是否等效需要论证。
结构件的非功能性改动。 外观件改动通常不影响，但如果涉及屏蔽或接地就另当别论。
附件增减。 新增附件是否需要重新测试整机。
生产场地变更。 不同场地的工艺差异可能影响产品特性。
这些分歧的根源都是「影响不明确」。 提供充分的信息和自测数据，可以让判断有依据，减少不必要的复测。
复测的组织
确定范围之后，复测的组织也有讲究：
样品要用整改后的。 确认送去复测的样品确实包含了整改措施。
样品数量。 如果复测项目包括破坏性试验，要准备足够样品。
保留整改前的样品。 便于对比分析。
记录整改前后的差异。 包括照片和参数。
一次送全。 把所有需要复测的项目一次安排，避免多次往返。
我们的做法
出现不合格项时，我们会说明我们观察到的现象和可能的原因，但整改方案由委托方决定 ——因为整改涉及设计，需要委托方评估其对产品其他方面的影响。
复测范围方面，我们会根据整改内容给出建议的复测项目清单，并说明判断依据。 委托方如果认为某些项目可以免测，可以提供论证一起讨论。
如果你的产品测试出现不合格需要整改，想先判断复测范围，可以把整改方案发过来一起评估，或者直接联系：132 4819 8029。能力范围见服务介绍，流程见检测流程，更多内容见知识库。
English version
Conclusion. After a test failure, modification and retesting is the normal course. The question is how much to retest: repeating everything costs time and money, while testing only the failed item risks missing problems the modification introduced. The basis for deciding is which characteristics of the equipment the modification affects. Items relating to affected characteristics must be retested, and items relating to unaffected ones can be justified as exempt. Note that this concerns which characteristics are affected, not how many parts were changed: replacing one small capacitor can shift the whole emission spectrum, while changing a cosmetic part may affect nothing.
How to decide. State the modification specifically, covering what was changed and how. Analyse which technical characteristics it affects, among electrical parameters, structure, screening, earthing, software behaviour and thermal characteristics. List the test items relating to those characteristics, which in principle must be repeated. Justify exemption for items not listed, explaining why they are unaffected. And agree the retest scope with the laboratory. The justification must be specific: writing that the modification does not affect other items is insufficient, and the mechanism of the modification and why it leaves other characteristics unchanged must both be explained.
Effects by modification type. Adding filter components affects emissions, immunity, leakage current and dielectric strength. Improving screening affects emissions, radiated immunity and possibly temperature rise. Improving earthing affects emissions, immunity, leakage current and earth impedance. Replacing supply components affects emissions, temperature rise, single fault behaviour and efficiency. Improving cooling affects temperature rise and possibly emissions where a fan is added. Software modification affects items depending on what changed, potentially many. Structural reinforcement affects mechanical items and possibly screening. Material substitution affects flammability, biocompatibility and mechanical performance. And changed marking usually affects only marking checks. Adding filter components has notably wide effects: a Y capacitor reduces emissions while increasing leakage current, and a common mode choke may change immunity behaviour in some bands, so a group of related items must be repeated.
The particular case of software. Software effects are the hardest to judge because they are invisible. Changing only interface text usually affects neither electromagnetic characteristics nor safety. Changing control logic may affect output characteristics, single fault behaviour and essential performance. Changing clocking or power management affects emissions directly. Changing communication protocols affects EMC items. And changing protection logic affects single fault and safety items. Produce a change description with each software modification, listing what changed and the possible effects. Without one, laboratories generally require broad retesting, because the effect cannot be judged.
Writing the justification. Describe the modification with its technical detail. Explain the mechanism by which it resolves the original problem. Analyse the scope of effect, stating what is and is not affected. Justify exemption item by item. And provide supporting data where available. Supporting data raise credibility considerably: measuring leakage after modification and finding it within limits and close to the pre-modification value supports the argument that leakage is unaffected.
Avoiding repeated modification cycles. Repeated cycles are the most expensive outcome, each requiring retesting and extending the schedule. Modify thoroughly the first time rather than making the minimum change to clear a marginal failure, since a result that only just passes may fail again with batch variation. Analyse root causes, since treating symptoms lets the problem return in another form. Assess side effects before modifying, so that solving one problem does not create another. Self-test after modification, before submitting for retest, to reduce the chance of a second failure. And address multiple failures together, modifying once and retesting once rather than iterating item by item. Leaving margin is the most practical point: the aim of modification should be passing reliably rather than passing narrowly.
Connection to change management. Modification is design change and belongs in the change management process. Record the change formally. Assess the effect not only on testing but on production and on products already approved in other markets. Update documentation, including drawings, bills of material and technical files. And consider notification obligations in other markets, since a modification made to pass one market's testing may constitute a notifiable change where the product is already registered elsewhere. The last is easily overlooked: assess notification obligations across markets whenever modifying.
How we handle it. When an item fails we describe what we observed and the possible causes, while the modification is the client's decision, since it concerns design and the client must assess its effect on other aspects of the product. On retest scope we propose a list of items based on the modification and explain the basis. Where the client believes certain items can be exempted, we are glad to discuss a justification.
Send us the modification plan and we will assess the retest scope with you. Phone or WeChat: +86 132 4819 8029.