IB物理 实验设计 选题 不确定度 数据处理
IB物理IA(Internal Assessment,内部评估)占最终成绩20%,是六门学科里最"可控"的分数来源:它不依赖考场状态,而取决于你的实验设计和写作思路。很多同学把大量时间花在背公式上,却不知道IA真正的评分逻辑已经变了:2025年首考的新大纲取消了独立的Personal Engagement和Communication维度,改为Research Design、Data Analysis、Conclusion、Evaluation四条标准各占6分,共24分。这篇文章按真实操作顺序拆解IA全流程,从选题到数据处理,每一步都告诉你考官在看什么。
The IB Physics Internal Assessment is worth 20% of your final grade at both SL and HL, and it is the most controllable score in the whole course because it tests how you think, not how fast you recall. Since the new syllabus (first assessment 2025), the IA is marked out of 24 across four equal criteria: Research Design, Data Analysis, Conclusion and Evaluation, each worth 6 marks. The old separate criteria for Personal Engagement and Communication have been removed, so engagement is now judged through the choices you make in your design, and clarity is judged inside every section. This guide walks through the whole process in the order you actually work: choosing a question, designing the method, handling uncertainty, processing data, and writing a conclusion that earns the top band.
1. 评分结构:四条标准决定写作重心
新大纲的IA满分24分,每条标准6分,权重完全一样:Research Design看研究问题是否聚焦、变量是否清晰、方法是否可重复;Data Analysis看数据记录、误差处理与作图是否规范;Conclusion看你是否用数据本身回答研究问题,并和公认理论值对比;Evaluation看你能不能指出方法真正的局限,并给出有依据的改进。注意一点:Conclusion和Evaluation加起来占一半分数,这意味着"解释和判断"比"堆数据"值钱得多。报告有3000词的硬上限(表格、图表、计算与参考文献不计入),10小时课堂时间完成,所以每句话都要有信息量。
Under the new guide every criterion is worth the same, so your writing effort should follow the marks. Research Design rewards a focused research question, clearly defined independent and dependent variables, and a method that another student could repeat from your description. Data Analysis rewards a raw data table that shows units and uncertainties, honest processing, and a graph that carries error bars. Conclusion rewards an answer to your own question that is justified by the processed data, interpreted with its uncertainty, and compared with accepted physics such as a textbook value of g. Evaluation rewards limitations ranked by how much they actually affected your result, each paired with a realistic improvement. Because Conclusion and Evaluation together carry 12 of the 24 marks, a neat method with weak reasoning will not reach the top bands. The report has a hard limit of 3000 words, with tables, graphs, calculations and references excluded, and you have about 10 hours of supervised time, so concise writing is part of the skill being tested.
2. 选题:研究问题决定你的天花板
IA最常见的失败不是实验做砸了,而是研究问题本身不合格。一个合格的研究问题要满足三条:自变量和因变量都能量化、范围明确;只研究一个变量关系,不要贪多;问题里有你自己的语境或生活关联。对比两个问题:差的写法是"How does length affect the period of a pendulum?",太宽泛,看不出你想验证什么;好的写法是"当摆长从20.0 cm增加到100.0 cm时,单摆周期T与摆长L的关系是什么,能否由此估算当地重力加速度g?"。后者自变量、因变量、范围和目的全部明确,考官一眼就能判断你的设计思路。若想让Research Design的"个人参与"分更有说服力,选题可以带生活场景:用手机计时研究水瓶底部开孔后水流尽时间与水位的关系,用弹簧和手机壳研究简谐运动周期与质量的关系,用铅笔芯研究电阻与长度的关系,或者给经典单摆实验加入你自己的变量,比如摆角、摆球材料或空气阻力的影响。
The research question decides your ceiling, and most weak IAs fail here before any data is taken. A good question has three properties: the independent variable and the dependent variable are both measurable over a clear range; it isolates one relationship rather than several; and it carries some personal context that you can genuinely explain. Compare a vague question such as "How does length affect the period of a pendulum?" with a focused one: "As the pendulum length increases from 20.0 cm to 100.0 cm, what is the relationship between period T and length L, and can the data be used to estimate the local gravitational acceleration g?" The second version names every variable, fixes the range, and states the purpose, so an examiner can immediately see your reasoning. To make engagement visible in Research Design, anchor the topic in everyday life: time how long a water bottle takes to empty through a hole at different water heights using your phone, investigate how the period of a mass on a spring depends on mass, measure how the resistance of a pencil lead changes with length, or take the classic pendulum and add your own twist such as a controlled study of amplitude, bob shape or air drag. The topic does not need to be exotic; it needs to be yours.
3. 变量识别与方法设计
选定课题后,第一件事是写下三张清单:自变量(你主动改变的物理量)、因变量(你测量的结果)、控制变量(必须保持不变的条件)。以单摆测g为例:自变量是摆长L,从0.200 m到1.000 m取6个值;因变量是周期T;控制变量包括摆角始终小于10度、使用同一个摆球、支架固定、环境无风。方法设计有三个关键数字:自变量至少取5到6个不同值且跨度尽量大;每个值重复测量5次左右,取平均以减少随机误差;正式实验前先做一轮预实验,确认计时方法可行、摆角衰减不严重。测量周期时不要只测一次摆动,而是测10个完整周期再除以10,这样可以把秒表反应时间带来的误差缩小10倍,这是考官最喜欢看到的规范操作。
Once the question is fixed, write three lists before touching any apparatus: the independent variable you change deliberately, the dependent variable you measure, and every controlled variable that must stay constant. For a pendulum experiment to find g, the independent variable is length L over six values from 0.200 m to 1.000 m, the dependent variable is the period T, and the controls include keeping the release angle below 10 degrees, using the same bob, clamping the support rigidly, and working away from drafts. Three numbers make a method credible: at least five or six different values of the independent variable spread across the widest sensible range; about five repeat readings at each value so a mean and a spread are meaningful; and a pilot run before the real data collection to check that timing works and the amplitude does not decay too fast. When measuring a period, time ten complete oscillations and divide by ten rather than timing a single swing, because this divides the reaction-time error of the stopwatch by ten. Examiners recognise this immediately as a deliberate design decision.
4. 不确定度:IA的分析骨架
不确定度不是附加分,而是Data Analysis的骨架:新大纲要求"每一个仪器的不确定度都要被估算并贯穿到数据处理中"。先分清两类误差:随机误差每次测量都略有不同,可以通过重复测量取平均来减小;系统误差总是偏向同一方向,重复再多也没用,只能通过改进方法消除。仪器不确定度怎么定:标尺类通常取最小刻度的一半,电子仪表按说明书标称值,手工秒表计时把反应时间约0.2秒作为不确定度。单次测量记作"读数±不确定度",重复测量的不确定度可用(最大值减最小值)/2或标准差表示,最后保留一位有效数字,测量值与不确定度的小数位必须对齐。例如用秒表测10次全摆动,5轮读数分别为12.62、12.55、12.70、12.58、12.66秒,平均12.62秒,半极差0.075秒,则单个周期T等于1.262±0.008秒,相对不确定度约0.6%。
Uncertainty is not a bonus section; it is the skeleton of Data Analysis, because the new syllabus expects every instrument uncertainty to be estimated and carried through the processing. Distinguish two kinds of error first. Random errors differ slightly every time and are reduced by repeating readings and taking a mean. Systematic errors always push the result in the same direction, so more repeats do not help and only a change of method removes them. For instrument uncertainty, take half the smallest division for a ruler scale, the manufacturer spec for digital meters, and about 0.2 s of reaction time for hand stopwatch timing. A single reading is written as value plus or minus its uncertainty; for repeated readings, use half the range (maximum minus minimum divided by 2) or the standard deviation, then keep one significant figure in the uncertainty and align the measurement to the same decimal place. Example: timing ten oscillations with a stopwatch over five runs gives 12.62, 12.55, 12.70, 12.58 and 12.66 s, mean 12.62 s, half range 0.075 s, so one period T is 1.262 plus or minus 0.008 s, a relative uncertainty of about 0.6%.
5. 原始数据表与作图
原始数据表的格式本身就在送分:每一列要有表头、单位和不确定度,比如"L/cm: 40.0±0.1""10T/s: 12.62±0.20",不要只写平均值而丢掉原始读数。作图记住四件事:第一,自变量在横轴、因变量在纵轴;第二,每个数据点画误差棒(error bars),纵轴误差棒来自T的不确定度;第三,先画原始关系(T对L是一条曲线),再想办法线性化;第四,拟合直线要穿过误差棒,并画出最大斜率线和最小斜率线来估计斜率不确定度。单摆实验的线性化非常经典:T等于2π乘以根号下(L/g),两边平方得到T²等于(4π²/g)乘以L,所以画T²对L的图应该得到一条过原点的直线,斜率就是4π²/g。看到"应该成直线"是数据分析里最让人安心的一步,它说明你的变量选择和理论模型是对的。
The raw data table itself earns marks if it is formatted properly: every column needs a heading, a unit and an uncertainty, such as "L/cm: 40.0 plus or minus 0.1" and "10T/s: 12.62 plus or minus 0.20", and you must show the raw readings rather than only the means. When graphing, follow four rules. Put the independent variable on the horizontal axis and the dependent variable on the vertical axis. Draw an error bar on every point, with the vertical bar coming from the uncertainty in T. Plot the raw relationship first, then look for a way to linearize it. Fit the best straight line through the error bars and add a maximum slope line and a minimum slope line so you can estimate the uncertainty in the gradient. The pendulum is the classic linearization: since T equals 2π times the square root of L over g, squaring both sides gives T² equals (4π²/g) times L, so a graph of T² against L should be a straight line through the origin whose gradient is 4π²/g. When your processed graph comes out as a straight line, it confirms that the variable choice and the theoretical model are correct, which is exactly the reasoning Data Analysis rewards.
6. 数据处理:斜率、误差传递与结论
斜率本身不是答案,斜率背后的物理量才是。以上述单摆数据为例:若T²对L的最佳拟合直线斜率为4.05 s²/m,那么g等于4π²除以斜率,约9.75 m/s²。斜率不确定度怎么传递:先算相对不确定度,最大最小斜率之差除以2再除以最佳斜率;假设斜率不确定度为0.15 s²/m,相对不确定度约3.7%,则g的相对不确定度相同,g约等于9.75±0.36 m/s²。记住两条传递规则:加减运算时绝对不确定度直接相加,乘除或乘方运算时相对不确定度相加。最后把结果和公认值9.81 m/s²比较:两者之差0.06 m/s²远小于不确定度0.36 m/s²,说明实验与理论一致,结论成立;如果偏差大于不确定度,就要回到评估部分解释系统误差来源,比如摆角略大导致sinθ近似失效、空气阻力使周期偏大从而使g偏小。结论段的标准句式是:研究问题的答案是什么、数值加不确定度、与公认值对比、差异是否在误差范围内。
The gradient is not the answer; the physics hidden inside the gradient is. For the pendulum data above, if the best fit line of T² against L has a gradient of 4.05 s²/m, then g equals 4π² divided by the gradient, about 9.75 m/s². To propagate the uncertainty, work with relative values first: the uncertainty in the slope is half the difference between the maximum and minimum slopes, and if that is 0.15 s²/m the relative uncertainty is about 3.7%, which carries directly to g, giving g equal to 9.75 plus or minus 0.36 m/s². Two propagation rules cover most IAs: for addition and subtraction, add absolute uncertainties; for multiplication, division and powers, add relative uncertainties. Finally compare your value with the accepted 9.81 m/s². Here the gap of 0.06 m/s² is far smaller than the uncertainty of 0.36 m/s², so the experiment agrees with theory and the conclusion is supported. If the gap exceeds the uncertainty, the Evaluation section must explain the systematic sources, such as a release angle that makes the small angle approximation sinθ ≈ θ less valid, or air drag that lengthens the period and pulls g down. A top-band conclusion follows one sentence pattern: state the answer to the research question, quote the value with its uncertainty, compare it with the accepted value, and state whether the difference lies within the uncertainty.
7. 评估与提分清单:常见丢分点全排查
评估部分最忌讳两个词:human error和more accurate。考官想看到的是按影响大小排序的局限,以及每条局限对应的具体改进。以本实验为例:影响最大的是计时反应误差,改进是改用光电门或Phyphox声学计时;其次是摆角控制,可以在释放点放一个量角器挡板保证每次角度一致;再次是空气阻力,改用密度更大的球并减小摆幅;最后是尺子读数视差,读数时视线垂直刻度。每条局限都要说清它让结果偏大还是偏小,以及为什么。最后是丢分点排查:研究问题太宽导致变量失控;原始数据没有不确定度;误差棒缺失;有效数字位数不一致;结论没有和理论值对比;评估只写空话;报告超过3000词;引用他人方法或数据不注明。特别提醒,数据造假属于学术不端,IB会抽查比对,宁可误差大也不可编数据。IA是20%里最稳的20%,按Research Design、Data Analysis、Conclusion、Evaluation的顺序逐条自检,你就能把该拿的分全部拿到。
The two words that sink an Evaluation are "human error" and "more accurate", because they describe nothing. Examiners want limitations ranked by how much they actually affected the result, each paired with a concrete improvement. In the pendulum experiment the largest limitation is reaction-time error in timing, improved by a light gate or the acoustic timer in the Phyphox app; next is amplitude control, improved by fixing a protractor stop at the release point so every swing starts at the same angle; then air drag, reduced by choosing a denser bob and smaller swings; and finally parallax when reading the ruler, reduced by reading with the eye perpendicular to the scale. For every limitation say whether it pushes the result high or low and why. End with a checklist of the classic mark-losers: a question too broad for its variables; raw data without uncertainties; missing error bars; inconsistent significant figures; a conclusion that never compares with the accepted value; an evaluation full of empty phrases; a report over 3000 words; and borrowed methods or data without citation. One warning matters more than all of these: fabricating data is academic misconduct, and IB moderators compare submissions, so an honest result with large uncertainty is always better than invented numbers. The IA is the most stable 20% of your physics grade, and if you self-check against Research Design, Data Analysis, Conclusion and Evaluation in order, you collect every mark you have earned.
Cambridge International AS ALevel Physics Coursebook 正版剑桥国际AS ALevel考试物理教材带线上账号 英文原版¥557.0📱 去京东购买京东精选需要一对一辅导?Need One-on-One Tutoring?
A-Level / GCSE / IB / KS2-KS3 全科辅导,IB Physics IA 设计与实验指导
或截图后用「扫一扫 → 相册」选择本图