Get the app

ToneWood · Experiment journal

Measurement Method Validation

Split the same piece of wood into A and B, then use repeated measurements and different models to examine how well the results agree.

View the live data table

01Why split the wood?

While developing ToneWood, I wanted to understand how much a phone-based tap measurement reflects the wood itself, and how much depends on specimen geometry and the calculation model. I split each source board into A and B, recorded their dimensions, mass and vibration frequencies, and accumulated repeated tests and model comparisons.

A pair from the same board gives the comparison a shared starting point. Different dimensions can produce different frequencies. The question is whether the inferred sound speed and elastic modulus remain close after accounting for each specimen’s geometry and density. Natural variations in density and grain can also contribute to an A/B difference.

One source board
ASpecimen A
BSpecimen B
Measure eachCompare sound speed c and modulus E
Experimental design. Each group retains the link between the source board and its A/B specimens.

02What did I test?

For each specimen, I measure length, width, thick and thin edges, and mass, then record flexural tap tests. Each measurement record retains five tap frequencies so repeatability can be examined alongside the final result. The table also preserves measurement times and App algorithm versions.

I compare the current App algorithm with Euler–Bernoulli and Timoshenko beams, a Mindlin variable-thickness plate, a candidate surrogate, and a 3D solid model. Each makes different assumptions. Imported simulations retain their complete measurement snapshots and parameters; 3D results remain blank until an experiment has been imported.

I also add longitudinal vibration measurements as a different point of comparison. The relation c = 2Lf₁ applies when the free-rod model and fundamental longitudinal mode are appropriate. Peaks captured by the App remain candidates whose mode needs experimental confirmation. The ultrasound column is reserved for independent laboratory measurements; unmeasured values are left blank.

03How are the results interpreted?

Are repeated taps stable?
First examine the spread of five frequencies on one specimen. If repeatability is poor, check support, impact position, recording and mode selection before interpreting model differences.
Do the A/B specimens agree?
Use the symmetric difference Δ = 200 × |A − B| / (A + B), without treating either piece as an absolute reference. Compare inferred speed and modulus across geometries, rather than frequency alone.
Do different methods agree?
Differences relative to the longitudinal method use longitudinal speed as the denominator: positive values mean the model is higher. Agreement between models describes their behavior under these inputs and assumptions; independent measurements are needed to assess accuracy.

This is an ongoing validation record. Stable repeat measurements, agreement between paired specimens, and agreement between models provide different kinds of evidence. They do not by themselves establish accuracy for every species, dimension or geometry. Further measurements, mode identification and laboratory comparisons will continue to appear in the table below.

Continuously updated records

Data validation table

The same table and calculations as the Data Validation page, showing the experiments published with this report.

Updates every 15 secondsWaiting to sync

Loading the data table…