What is this made of? Tapping your way to a material family

Antique dealers do it. Furniture restorers do it. Anyone who has ever suspected a "solid oak" table of being oak-printed foil over MDF does it: rap a knuckle on the surface and listen. Dense, solid material rings; veneer over particleboard thuds; a hollow panel booms. The knock is an impulse, the object answers with a brief decaying ring, and how that ring dies away says a great deal about what the thing is made of.

The Tap test tool captures that ring — the ring-down — through the phone's unprocessed microphone path and extracts the features that separate material families: how sharply the resonant modes are tuned, how fast the high frequencies fade relative to the low ones, how the energy decays overall. Metal rings long and clean. Glass and ceramic ring bright, with fast-fading highs. Solid wood sits in the middle. MDF and plastics damp almost everything almost immediately. These are the same two dimensions — resonance and damping — that the fruit tap test reads on a melon; one physical principle, pointed at furniture instead of produce.

And crucially, the features are about damping character, not pitch. A small brass bell and a big bronze one ring at wildly different frequencies but decay the same way — frequency describes the object's size and shape, decay describes its material. Building on the decay side is what lets one analysis speak to tabletops, countertops and flea-market bowls alike.

Why one tap is not an answer

Here is the honest part, and it's the part the guided activity is built around: a single tap is one opinion. Tap force moves the answer. Striker hardness moves the answer — a fingernail, a knuckle and a ring-clad finger excite different modes. Where you hold the object moves the answer, because your grip is a damper. A single capture cannot tell you whether it measured the material or the way you happened to hit it.

A set of taps can, just by agreeing or not. So the "What is this made of?" activity refuses to conclude from less than three usable taps, and demands that at least two of three point at the same family before that family becomes the run's answer. If the taps disagree, the activity does not pick a winner or average the disagreement away — it asks for another pass. Two taps on the same object disagreeing is information about the protocol, not the material: the striker changed, the grip changed, the spot changed. The right response is to steady the procedure, not to trust a coin flip.

The Tap test also exists as a single capture on the Tools tab — but that's the expert's version, for someone who already taps consistently. The activity is the honest way to run it.

A second opinion from the vibration motor

The activity offers an optional second instrument: press the phone against the object and let its vibration motor probe it while the accelerometer and microphone record the response. This is a genuinely independent measurement — a different excitation, through a different contact, read by different sensors — which is exactly what makes it valuable as a cross-check.

Its verdict is treated as corroboration, not as another vote. If the probe agrees with the taps, confidence rises. If it contradicts them, the run ends with no verdict at all. That is deliberate: two independent instruments disagreeing about the same object means at least one measurement went wrong, and averaging a wrong measurement into a right one produces a confident answer to the wrong question. No verdict is a better result than a blended one.

The reference tap: trading a claim for a comparison

The strongest option in the activity is also the simplest: tap something nearby whose material you already know. The solid-wood shelf next to the suspicious table. The known-granite counter next to the maybe-laminate one.

This changes the kind of claim being made. Without a reference, the classifier is asserting something absolute — "damping like this means wood" — from physics priors that can't know your exact object. With a reference, the claim becomes relative: "this unknown rings like that known thing." A comparison between two taps, made seconds apart, on the same phone, with the same striker, is a far more defensible claim than an absolute one — most of what could bias the measurement biases both taps equally and cancels.

The activity is careful about when the reference is allowed to override, though. The two taps have to genuinely resemble each other in the damping features before the reference's family replaces the prior-based answer — a glass reference says nothing useful about a plank, and a reference match that fired anyway would be worse than none. And the comparison deliberately ignores frequency and mode count: those describe geometry, and matching on them would mean matching "another thing the same size," which is precisely the mistake this tool exists to avoid. When the reference does decide the answer, the verdict says so, so you always know which kind of claim you're looking at.

Where it earns its keep

The use cases are the ones you'd expect from the knuckle-rap tradition, made repeatable: solid wood versus veneer or MDF under a convincing finish; real stone versus laminate printed to look like it; hollow versus solid doors and panels; the flea-market find whose seller's story deserves a second opinion. In each case the phone isn't doing anything your ear can't in principle do — it's doing it the same way every time, and writing the answer down.

Honesty about limits, as always: the answer is a material family, not a species — "solid wood," not "walnut" — and the current classifier is built from physics priors rather than a trained model. Which is where the last piece matters: every tap in a run is saved as an ordinary measurement, filed under the object in the catalog, ready to take a ground-truth label when you eventually find out what the table really was. Runs where the taps disagreed are kept too — a future trained classifier needs to learn when to decline, and those are the examples that teach it. The activity you run today is also the dataset that makes tomorrow's version better.

For the whole family of measurements a phone can make — knocks, hums, vibration, light — start with the tools overview.