The Final Project

The final project asks you to do two things that the problem sets cannot: to measure something nobody handed you the answer to, and to explain what you found to somebody who does not know any physics.

The second is the harder one, and it is the one most likely to matter to you after the course ends.

What the Project Is¶

Pick a question about musical sound that you can actually answer with a measurement, make the measurement, and present the result in ten minutes to an audience of your classmates, treated, for the purposes of the talk, as intelligent people who have not taken this course.

That is the whole specification. The constraints hiding inside it are worth making explicit.

It must be a question, not a topic. “The physics of the violin” is a topic, and a talk on a topic is a summary of things other people found out. “Does bowing closer to the bridge really add high harmonics, and how much?” is a question. You can be wrong about a question, which is what makes answering one worth ten minutes of an audience’s attention.

You must measure something yourself. Not necessarily something new, almost certainly not something new, but something you recorded, timed, counted, or weighed. A project whose evidence is entirely other people’s published figures is a literature review, and it does not meet the outcome.

The measurement must be possible with what you have. A phone, a laptop, and free software are enough for most of the topics below. If your question needs an anechoic chamber, choose another question, or find the version of it that a carpeted bedroom can answer.

Choosing a Question¶

The reliable move is to take something you already know from experience and ask how much. Three examples of the transformation:

Thirty starting points¶

Instruments. Plucking position and harmonic content · Why a wound bass string works · The inharmonicity of a real piano string, measured · How much a guitar’s body resonance actually adds · What a mute does to a trumpet’s spectrum · The clarinet’s missing even harmonics, verified · Where a recorder’s register hole has to go · The end correction of a real pipe · How a drum’s pitch changes with head tension · Why a struck bar sounds different from a plucked string

Tuning and perception. Is a real piano stretched, and by how much · How close is a barbershop quartet to just intonation · How accurately can you tune a fifth by beats · Your own frequency discrimination against frequency · The missing fundamental, demonstrated on yourself · How many cents out does an interval have to be before you notice · Whether the dissonance of an interval changes with timbre

Voice. Formants of your own vowels, measured · How the vocal tract changes between a spoken and a sung vowel · Whether you can find your own singer’s formant · How high a soprano can go before the vowel becomes unidentifiable

Rooms and recording. Reverberation time of three rooms · Room modes in a small bedroom · What a polar pattern really looks like · How much a lossy codec actually changes a waveform · Where aliasing becomes audible · Whether a loudspeaker’s frequency response matches its specification · How stereo localization depends on delay versus level · What a hall’s early reflections do to clarity

The Schedule¶

The Week 7 proposal exists because the most common way this project goes wrong is a question that cannot be measured, discovered in Week 13. Say in the proposal what would count as failure; if you cannot name a result that would disappoint you, the question probably has no content.

Giving the Talk¶

Ten minutes is about 1,200 words. That is very little, and it is enough, if you spend it on the right things.

Spend the first minute on the question, not on background. The audience does not need the history of acoustics. They need to know what you wanted to find out and why anyone would care. Play the sound if you have one, thirty seconds of a guitar plucked in two places does more than a slide about harmonics.

Show your own data, large. One graph you made, with axes labeled in units a non-specialist can read, beats six graphs from papers. If a number is the point of the talk, put that number on the slide by itself.

Say what you found, including if it was not what you expected. A measurement that disagreed with the prediction is a better talk than one that agreed, as long as you say what you think went wrong. Agreement is only interesting if the audience believed disagreement was possible.

Translate every technical term or drop it. You may use one or two, if you define them in plain words the first time. “The spectrum: that is, a list of which pure tones are present and how strong each one is” costs you four seconds and keeps the room with you. An undefined “spectral centroid” loses it permanently.

Do not read the slides. If a slide is readable in three seconds, it is doing its job.

What a Good Talk Sounds Like¶

Three tests, applied by the audience rather than by you:

• Can a listener state your question afterwards, in their own words?

• Can they say what you did to answer it?

• Do they believe the answer, and could they say why?

A talk that is technically flawless and fails all three has not met the communication outcome. This is worth saying plainly because it is a genuinely different skill from the one the problem sets develop, and it is the one that transfers.

Rubric¶

Note the weighting: the explanation is worth as much as the physics. That is deliberate, and it is the point of the exercise.

A Note on Honesty¶

Report what you measured, including the run that did not work and the data point you cannot explain. If your result disagrees with the textbook, say so and offer a reason: a systematic error in your method is a perfectly respectable finding, and identifying one is real physics. Quietly dropping the inconvenient half of your data is the one thing in this project that cannot be recovered from.