TPC — Tonewood Parameters Characterisation

Quick start and user guide · Rev 1.9

Measure a tonewood sample, inspect its resonance and damping, and use its density to estimate a starting soundboard thickness. TPC supports the measurement and target-mass methods presented in Mastering the Sound of the Acoustic Guitar and in the workshops.

Measurement records individual taps and calculates density, longitudinal dynamic Young’s modulus, sound radiation coefficient and Q. Soundboard thickness uses the sample’s density, a soundboard area and your target unbraced mass to calculate a starting uniform thickness.

Use the results to compare samples and test conditions. A high Q or radiation coefficient is not, by itself, a quality score for a piece of wood or a prediction of the completed guitar’s sound.

Before you begin

Prepare the sample and supports

Use a rectangular, unbraced sample with reasonably uniform thickness. Enter its Length along the grain, Width across the grain, Mean thickness and Mass, using millimetres and grams. Measure thickness at several positions and enter a representative mean. Replace the initial example values with your own measurements.

TPC uses the first free–free longitudinal bending mode: the sample bends along its length. “Longitudinal” here refers to the direction used to estimate Young’s modulus; this is not a longitudinal compression-wave test.

Follow the support diagram below Sample. Place the supports approximately 22.4% of the length from each end. From each end and Support spacing give the positions calculated for your entered length. Leave the ends free rather than clamping the sample.

Tap the broad face near one free end, as shown by the arrow, and keep the microphone near the vibrating sample without touching it. Let the plate settle on its supports before measuring the background. Avoid bouncing, rubbing against the fixture or a second contact from the tapping tool.

In Settings → Sample details, record Support method as Jig / elastic supports, Hand-held or Other. Add species, source and relevant notes. The support method is descriptive metadata: selecting it does not apply a correction to Q.

Prepare the microphone

Connect the microphone and open Settings. Choose Microphone and Input channel. Use Channel 1 for a single-channel input.

Pressing Start requests microphone permission. If the microphone’s name is not initially available, start and stop once, then return to Settings and select it. The app requests echo cancellation, noise suppression and automatic gain control off; the reported settings appear after the microphone opens. These are browser reports, not independent verification of the complete recording chain.

Work in a quiet room and keep microphone position, input gain, supports and tapping method consistent. The input meter appears only during acquisition. Its level is dBFS, referenced to digital full scale, not calibrated sound pressure level. TPC Rev 1.9 does not provide the microphone-calibration-file and SPL-calibration controls found in Sustain Explorer.

Microphone capture requires HTTPS or localhost. Recording, analysis and export generation run locally in the browser; the app does not upload the recordings. The font may load from an external resource.

Name the measurement and record the conditions

Enter an identifiable Sample ID, such as “Spruce-001”. Enter temperature in °C and relative humidity in % RH beside the fullscreen button when known. These are manually entered details, not sensor readings; leave unknown values blank.

For a new sample or a linked A/B procedure, use New, immediately after Stop. Choose Single sample or Bookmatched pair · A / B, enter Sample / pair ID, and confirm Create new. The bookmatched procedure is explained below.

A name is a label, not a new dataset. Editing Sample ID renames the current recordings. It does not clear them or begin another sample. Use New when changing specimens, and save the existing ZIP first when its audio must be retained.

New reuses the latest explicitly entered Length, Width and Mean thickness, together with the environmental entries, microphone settings and starting design values. Mass is cleared so the new piece must be weighed. Check all retained values before starting; they are starting entries for the new specimen, not measurements of it. Cancelling the New dialog leaves the current recordings and inputs untouched.

The context line beneath the graphs identifies the current sample and, for linked measurements, its member. Before recording it reads no acquisitions. The first acquired or imported tap establishes the measurement date, which remains stable when exporting or correcting inputs.

Use full screen and automatic settings

Select ⛶ at the top right to use the whole application fullscreen. The enlargement control inside a graph enlarges only that graph. A browser file dialog may interrupt fullscreen; use the header button to return.

Leave Settings → Advanced resonance search → Minimum frequency / Maximum frequency empty for the geometry-based automatic search. No FFT length, smoothing or fitting-window selection is required for the normal procedure.

For a practice run, Settings → Test & diagnostics → Run demo generates three synthetic taps. These are demonstration signals, not measurements of the named sample. Save your work first: the demo replaces the current working tap recordings.

1. Measurement

Record the sample

1. Check the current sample. Select Measurement. Confirm Sample ID, dimensions, mass and the calculated support positions. For a bookmatched pair, check that the intended A or B member is selected.

2. Measure the background. Press Start and remain quiet while the centre display reads Measuring background — keep quiet. Do not tap or move the sample during this initial measurement, which takes approximately one second. Wait for Ready — tap once.

3. Tap once and let it ring. Use one light, clean impact near the free end. The app verifies an impact, captures the vibration and displays Recording — let it ring. Do not tap again while it is capturing or analysing.

4. Follow the next Ready prompt. After analysis, the app retains the recording and either includes the usable tap or explains why it was not used. Wait for Ready — tap again before the next impact. If the plate is still moving or ringing, let it settle.

5. Collect repeatable taps. The first three accepted taps are included automatically. Acquisition stops when three are included, and the Results panel shows their combined result and quality status. Three included taps do not automatically guarantee a confirmed result: repeatability or recording warnings may still need attention.

Stop remains available. It releases the microphone and retains the capture in progress, although an interrupted ring-down may not contain enough information for a valid result. After stopping, the acquisition button becomes Add tap when recordings already exist. This adds to the current sample; it does not start a new specimen.

How long should the sample ring?

Let the app finish whenever practical. Capture normally ends when the recorded signal has remained sufficiently quiet, with a maximum of approximately 4.5 seconds after the trigger. Do not damp the plate just because the main oscillation looks small.

A complete capture and an accepted decay fit are different things. A recording can be retained but rejected because its usable decay range is too short, its envelope is irregular or the mode cannot be isolated. Keep the original recording for inspection rather than interpreting every unavailable result as a failed microphone capture.

Inspect, include or replace taps

Click a tap beneath the graphs to inspect it. Both Recording and the analysis graph follow that selected tap. Blue identifies the inspected tap; a ✓ identifies one included in the combined result. Selection for viewing and inclusion in the calculation are separate actions.

Use Exclude from result to remove a valid tap from the combination without deleting its audio. Use Use in result to include another valid tap. A maximum of three taps is included at once; exclude one before replacing it. Rejected taps remain inspectable but cannot be forced into the result with this button.

To record a replacement, exclude the unwanted tap, press Add tap, and follow the background and Ready prompts again. Delete tap removes the selected recording after confirmation and recalculates the result from those still included.

One or two valid included taps already produce a provisional result and can be stored. Three remain the recommended number for checking repeatability.

Add a note to a tap

Stop acquisition and select the recording, then press Note, beside Use in result / Exclude from result. Enter the observation in Recording note and choose Save note. For example: “opposite face upward”, “hand-held” or “lighter impact”. Note • indicates that this recording has a note. Reopen it to edit; clear the text and save to remove it. Cancel leaves the previous text unchanged.

Each tap can have its own note of up to 600 characters, including rejected or excluded recordings. Notes stay with their recordings when switching taps or A/B members and when reanalysing. They do not change the analysis or create a separate measurement group.

Save note saves the text in the working measurement, not to disk. Export a measurement ZIP to retain every tap note. Included-tap notes also appear in result JSON and stored summaries; changing one can require Update stored sample. The PDF includes a short excerpt of the selected tap’s note. The Excel/CSV export keeps its existing 18 columns and does not add a note column.

Correct dimensions without repeating the taps

Stop acquisition, then correct Length, Width, Mean thickness or Mass. TPC retains the recordings and their stored frequency and damping analyses, and recalculates density, E, radiation coefficient and target thickness.

This is for correcting the description of the same measured sample. If the plate has actually been thinned, shortened or otherwise changed since recording, retain the previous ZIP and make a new measurement of its new condition. The old recording does not describe the altered plate.

Load existing WAV recordings

Select Load… and choose one or more uncompressed PCM or IEEE-float WAV files. Check the current sample inputs first: imported taps are added to that sample. Use New beforehand when the recordings belong to another specimen.

Include quiet before the first impact, one clean impact per ring-down, and its complete decay. The app can separate sufficiently spaced taps within a WAV; distinct files for individual taps make the intended grouping unambiguous.

Multichannel WAVs are averaged to mono during import. The microphone Input channel setting does not select a WAV channel. Export a mono WAV containing the intended signal when channels contain different signals.

Imported recordings are checked for impact timing before the usual resonance and damping analysis. The app cannot reverse compression, noise suppression or other processing already present in a file.

Read the recording and resonance graphs

Recording

Recording shows the original captured signal for the selected take, including its available pre-impact portion. Time zero is the beginning of that recording. The impact marker identifies the detected tap within it.

During acquisition, this panel shows the incoming signal. Retained data consist of individual tap recordings and their backgrounds, not an unlimited continuous recording of the entire listening session.

A red CLIPPING warning and red waveform identify excessive input level. The live warning remains visible for at least about one second; a retained clipped take also shows a warning when inspected. Reduce the input gain or tapping strength and repeat the affected tap.

A tap can occasionally remain accepted when clipping is confined to the immediate impact and ends before the analysed interval. It is then marked Check · impact clipped. Check · near clipping identifies a level close to the limit. These are warnings about the recording, not the wood; reduce the level for the next capture rather than aiming for clipped impacts.

Resonance fit

Resonance fit shows the isolated bending-mode oscillation in blue and a fitted response in orange. Its horizontal axis is Time after tap, unlike the Recording graph’s time from the start of the file.

Fit start and Fit end mark the interval used to estimate damping. They do not mark a cut in the recording or the start of the FFT observation. The blue signal is paler outside that interval, but it remains available. The orange curve is solid within it and dashed outside it: the dashed portions are extrapolation, not additional fitted observations.

The interval is selected from the decay envelope, not from the geometric centre of the recording. The analysis excludes an initial period of at least five cycles or 15 ms, whichever is longer, and first establishes a conservative fit beginning near −10 dB relative to the early envelope reference.

In Rev 1.9, it can extend that interval back to approximately −5 dB when the additional early decay is regular and its slope agrees with the later fit. The extension must pass consistency checks on the early section and on the combined interval. early interval checked appears beneath the graph when this earlier start is used. If those checks fail, the conservative start is retained rather than forcing the early transient into the estimate.

The lower endpoint is unchanged: normally near −30 dB, or earlier to maintain about 10 dB above the estimated noise floor. For example, with an estimated 35 dB signal-to-noise ratio, the lower limit becomes approximately −25 dB. The earlier start does not replace an unrecorded lower endpoint or guarantee a result from a truncated ring-down.

A slowly decaying mode may take a substantial time to reach the starting level. A late fit marker therefore does not mean that the preceding waveform was lost. The raw Recording and complete filtered ring-down remain available; the display spectrum uses its own post-impact observation and short edge taper.

Use Auto or Full to inspect the complete retained filtered ring-down, including the later tail. The initial view already includes the fitting interval. You can inspect the boundaries but cannot move them manually in Rev 1.9. Zooming does not alter the fitting result. An older ZIP can still show its original fit interval until you run Reanalyse saved taps; loading alone does not apply the new interval-selection rule.

Spectrum

Choose Spectrum for the unsmoothed frequency-domain view of the recorded post-impact ring-down. The identified-mode marker indicates the frequency selected by the analysis. This spectrum is calculated from the recorded signal, not from the isolated band-pass-filtered oscillation shown in Resonance fit.

The initial view is centred near the mode with a 50 dB vertical range. Zoom out or choose Full to reveal the complete analysed frequency band and inspect other peaks and the background. Full refers to the available band after analysis downsampling—typically extending to about 6 kHz—not necessarily the original microphone’s entire bandwidth.

The spectrum uses short cosine edges, capped at 5 ms at each end, with no frequency smoothing. Its reference is the selected modal region, expressed in relative dB rather than SPL. The observation interval, taper and level reference remain fixed when you zoom. A stronger peak outside the initial region can therefore appear above 0 dB.

Frequency zoom does not change the analysed time interval. A finer plotted FFT grid does not create resolving power absent from the recorded observation. Use the spectrum to inspect mode identity and nearby content, not to force every specimen to have a particular peak shape.

Understanding the results

Main parameters

For several included taps, TPC takes the median frequency and median amplitude time constant, then calculates the combined E, Q and radiation coefficient from those values and the current sample inputs. It does not average the recorded waveforms together. The Results panel can therefore differ slightly from the selected tap’s graph readout.

Secondary parameters and repeatability

Resonance · f₁ is the identified bending frequency. Time constant · τ describes the amplitude envelope: in an exponential model, amplitude falls to 1/e of its initial value after τ seconds. Log decrement · δ describes the logarithmic amplitude decrease per cycle. The app uses Q = πf₁τ and δ = 1/(f₁τ).

Envelope fit is R² × 100 for the fitted envelope interval. A value near 100% means that the observed interval closely follows the fitted straight line in decibels. It is not the percentage accuracy of Q, E or density, and it does not prove that the correct mode or support condition was selected.

Repeatability reports the half-range of included frequencies in hertz and the half-range of Q relative to the median Q. With one tap it reads One tap. These are descriptive spreads between repetitions, not calibrated measurement uncertainties.

Quality states

Read Check in Results and the explanation beside the selected tap. Hover over these entries for further detail. The selected tap’s check and the combined result’s check are separate: one Good tap does not remove a warning on another included tap. An excluded but otherwise valid tap still has an accepted analysis; a rejected tap does not. These states describe the measurement, not the sample’s suitability for a particular guitar.

Support losses, geometry, environmental conditions and input processing can influence the measurement. A repeatable hand-held result and a repeatable jig result are not automatically interchangeable. Record the support method and compare measurements made under consistent conditions.

Measure bookmatched halves

Bookmatched pair · A / B

1. Prepare the pair. Select New → Procedure → Bookmatched pair · A / B, enter a pair name and choose Create new. This creates a new working set; it does not convert an existing single-sample recording into half A. Save the previous set first.

2. Measure A. Check the copied dimensions, enter half A’s measured mass, and complete its tap sequence. Its recordings and result belong to A.

3. Switch to B. After acquisition stops, select B beside Sample ID. A remains retained. An unmeasured B inherits the latest explicitly entered Length, Width and Mean thickness, rather than the values present when the pair was first created. Check those dimensions and enter B’s own mass before recording. The two halves have independent masses, recordings, results and thickness targets.

The same rule applies to either half while it is still unmeasured. Simply viewing an already recorded half does not change the latest-entry defaults. Once a half has recordings, switching back restores its own geometry and results; it does not substitute the other half’s dimensions. Make any correction explicitly while that half is selected.

4. Compare and save. Use Compare to see both sets of results and B’s percentage differences relative to A. Copy both results for Excel places the available results in separate rows. The app does not combine the two frequencies or moduli into one result.

Store each member separately when both numerical summaries are wanted in Saved samples. Save measurement + audio · ZIP saves the complete linked set, regardless of which member is displayed.

2. Soundboard thickness

Apply the target-mass method

1. Check the density source. Stop acquisition and select Soundboard thickness. Current sample density and the sample name identify the input being used. For a linked set, the active member supplies the density. This calculation needs valid sample dimensions and mass, but does not require E or Q.

2. Choose the soundboard area. Select a reference outline: Dreadnought ≈ 1620 cm², OM ≈ 1430 cm², 00 ≈ 1330 cm² or Classical ≈ 1400 cm². Use Custom area and enter your actual area when available. Editing the area switches the selection to Custom area.

These are approximate gross-outline areas before the soundhole is removed, not areas calculated from the displayed icons. There is no automatic soundhole subtraction. The entered area and target mass must refer to the same construction stage.

3. Enter Target unbraced top mass. The app calculates Starting uniform thickness immediately:

Thickness (mm) = 10,000 × target mass (g) / [density (kg/m³) × area (cm²)]

The selected body shape supplies an area reference, not an automatically recommended target mass. Use the target appropriate to your design method.

4. Keep the design result. With a valid tap result available, select Store sample, or Update stored sample if the measurement was already stored. This preserves body area, target mass and starting thickness with that numerical record. Save the ZIP as well to retain the complete working measurement. The thickness calculation itself works before tapping, but storing a result or exporting its PDF/CSV requires a valid tap result.

This is a starting uniform thickness, not a final stiffness or structural-safety target. Density sets the starting thickness; brace tuning remains a later operation. Start and Stop are hidden in this view, and the app prevents switching here while acquisition is running.

OM reference example

The notes in this view give an illustrative OM reference: a 1,430 cm² gross-outline area, 2.7 mm uniform thickness and 400 kg/m³ density correspond to 154.44 g, or approximately 154 g before cutting the soundhole.

Mass (g) = 1,430 cm² × 0.27 cm × 0.400 g/cm³ = 154.44 g

This is an OM reference example, not a verified Martin factory specification or a stiffness recommendation. It does not fill the target-mass field automatically. Choose your own target and keep its construction stage consistent with the area; the resulting thickness uses the density of your currently selected sample.

Graph controls

Move to the sliders icon in the lower-right corner of either graph, or click it, to reveal the controls.

With a graph focused, A runs Auto, 0 resets, Enter opens Limits, and Ctrl/Cmd + Z undoes a graph-view change.

These controls change the view, not the measurement. In TPC, dragging on Recording does not select a new FFT or fitting interval. Choosing another tap resets the detail view to that tap’s defaults.

Store, reopen and export

Store a numerical summary

After at least one valid tap is included, select Store sample below the workspace. Saved samples opens the browser-local collection, including starting thickness when calculated. Details shows the date, tap count, design inputs, support method and measurement check.

For an already stored measurement, the button becomes Update stored sample. Changes to inputs, taps, an included tap’s note or thickness targets show Stored result needs update. Click Update to replace that record; it is not silently overwritten. Each member of a linked pair has its own stored summary.

Store sample does not save the audio. The collection keeps up to the latest 100 summaries in the same browser when local storage is available. Export the collection regularly. Deleting a saved summary does not delete the current tap recordings; deleting a current tap does not automatically update an earlier stored summary.

Save the complete working measurement

Stop acquisition, wait for analysis to finish, then select Export → Save measurement + audio · ZIP.

The ZIP retains all included, excluded and rejected tap recordings; their available backgrounds, analyses and individual tap notes; sample inputs, environmental details, sample notes and support method; design targets; measurement identities and dates; graph views; latest-entry dimension defaults; and the Saved samples collection. Linked sets include both members, even when only one is currently displayed.

The filename uses the Sample ID, or the pair name for linked sets. Keep the ZIP intact. Open it through Load… or Export → Load saved measurement…. Loading replaces the current working set, including both linked members when present. Check the selected microphone before resuming acquisition.

Current saved analyses are restored rather than automatically recalculated. Very old archives without the required stored analysis may be recalculated during import and labelled legacy analysis updated. Keep the original file when migrating older work. Save a new copy after updating a session, and reopen it in Rev 1.9 rather than assuming that earlier versions preserve all current fields and behaviour.

Previously exported Normal/Flipped sessions can still be loaded and inspected without losing either recording series. Their legacy member labels remain, but that procedure is no longer offered in New. For new recordings, use the per-tap Note to describe an orientation or support condition.

The app retains up to 60 tap recordings per member and limits the complete ZIP to 155 MiB. Save regularly and remove unneeded recordings before reaching these limits. Closing or reloading the page is not a substitute for exporting the working set.

Choose the export destination

In compatible browsers, Export → Choose export folder… authorises a destination for subsequent exports during the current page session. Existing filenames receive a suffix rather than being silently overwritten.

Ask where to save each file opens a native save dialog for each export. Use browser downloads returns control of the destination to the browser. Unavailable folder/save-dialog controls remain disabled; ordinary downloads remain the fallback.

With direct file writing, the app reports Saved only after writing completes. Cancelling or failing an export does not mark the measurement saved. With ordinary browser downloads, ZIP download requested · check your browser means that the download was requested, not that TPC has verified the file on disk. Check the file before choosing New or closing the page.

Copy results to Excel

Choose Export → Copy current result for Excel, then paste into Excel. Use Copy saved samples for Excel for the stored collection, or Compare → Copy both results for Excel for linked measurements. If clipboard access is unavailable, the app opens a selected-text box for manual copying.

The 18 columns are, in order: Date; Sample ID; Length (mm); Width (mm); Mean thickness (mm); Mass (g); RH (%); Temperature (°C); Density (kg/m³); f1 (Hz); Q; Fit (%); Dynamic E (GPa); SRC (m4/kg/s); Body shape; Body area (cm²); Target mass (g); Starting thickness (mm).

Units remain in the headers, numeric cells contain numbers and missing values are blank. Settings → Export preferences → Excel decimal separator selects Automatic, Point or Comma. Copies use tabs; CSV uses semicolons with decimal commas, or commas with decimal points.

Tap count, provisional status, detailed quality checks and tap notes are not extra spreadsheet columns. Included-tap notes appear in result JSON and stored-summary JSON; the full ZIP retains notes for every retained tap. Keep the appropriate file when that evidence is needed, and update stored summaries before exporting their latest results, notes and thickness calculation.

Export a report, image or individual recording

Measurement report · PDF produces one A4 portrait page for the active sample/member. It includes sample details and conditions, the combined material results and repeatability, the selected recording and its isolated-resonance fit, target thickness when available, and a short notes section containing sample notes and the selected tap’s note. Long notes may be excerpted to keep the report to one page; the ZIP retains their full text.

Select the tap intended for illustration before exporting. The graphs describe that selected tap, while the main numbers combine the included taps. The report does not include every recording or the complete A/B comparison; export the other member separately when required. Use Selected graph · PNG to preserve a Spectrum view or another particular graph view.

The menu also provides Result · CSV, Measurement data · JSON, Saved samples · CSV/JSON and Selected tap · WAV. The WAV contains the selected unnormalised mono recording with its available pre-impact audio, not the entire listening session.

PDF, PNG, spreadsheet copies and result JSON do not replace the measurement ZIP. Save the ZIP to reopen audio, change tap inclusion and continue working.

When a measurement needs attention

No result after tapping

First check Sample ID, dimensions, mass, microphone and channel. Select the retained tap and read its reason. No clear impact — movement or noise only asks for a settled sample and one distinct tap; Multiple impacts in one recording — tap once and let it ring asks for separate, uninterrupted ring-downs.

Rev 1.9 checks for sufficient vibration after each candidate impact, helping prevent small preliminary disturbances from being counted as additional taps. It does not simply keep the loudest event: more than one independently confirmed impact is still rejected. Very weak real impacts can still be missed, and the check is not a guarantee that every disturbance will be classified correctly.

If an older recording was rejected because its impact time was wrong or a preliminary disturbance was counted as another tap, try Reanalyse saved taps before repeating it. Recovery still requires enough recorded decay after the real impact. A relocated onset cannot restore a missing tail.

For insufficient signal or background-noise warnings, reduce noise and adjust microphone position. For insufficient decay range, let the plate ring until capture finishes. For nearby or ambiguous modes, check support position and inspect the full Spectrum. An irregular or curved decay can remain unsuitable for a single Q even when a peak is clearly visible.

Do not increase tapping strength until clipping occurs merely to obtain a number. Rejected recordings remain useful evidence of what happened.

Use a manual search range only when necessary

If the automatic mode choice is consistently wrong, stop acquisition and enter both Minimum frequency and Maximum frequency under Advanced resonance search, based on the spectrum and the intended bending mode. The maximum must exceed the minimum.

This restricts where the app searches; it does not force an acceptable decay or make a support resonance into a material mode. Changes apply to new taps. Clear both fields to return to automatic search.

Reanalyse retained recordings

Save a copy of the ZIP first. Then open Settings → Test & diagnostics → Reanalyse saved taps and confirm Reanalyse.

The app rechecks impact timing and runs the analysis on the retained audio using the current sample and search settings. In Rev 1.9 this applies both the revised impact-confirmation check and the guarded earlier fit start. It can change the reported frequency, Q and derived results; it is not merely a display refresh. Loading a saved ZIP alone leaves its existing analyses unchanged.

Original recordings, backgrounds and tap notes are kept. Deliberately excluded previously valid taps remain excluded; newly recovered taps can fill vacant inclusion slots, up to three. If three taps are already included, a recovered recording becomes available for inspection without silently replacing one of them.

For a linked set, this command applies to the active member. Switch members and repeat only when needed. Inspect the new results, update any stored summary separately and export a new ZIP. Reanalysis cannot reconstruct a tail that was never recorded.

Keeping useful comparisons

Keep sample orientation, supports, microphone position, gain and excitation consistent, and record temperature and humidity. Compare the full evidence: geometry, identified mode, fitted interval, repeated taps and any warnings—not one isolated number.

TPC estimates properties under the stated bending-test assumptions. A clear fit and a repeatable result support the measurement, but do not remove the effects of the test arrangement or turn the result into an automatic instruction for the finished instrument.