Impact Hammer Testing

The Impact Hammer Testing module is a guided, end-to-end workflow for acquiring Frequency Response Functions (FRFs) with an instrumented hammer and one or more accelerometers, then sending the measured FRF matrix straight to Experimental Modal Analysis (EMA). It covers geometry and point planning, sensor setup, live impact acquisition with real-time quality feedback, validation checks, and export. The workflow and default parameters follow established impact-testing practice.

When to use this module: use Impact Hammer Testing when you want to measure FRFs from a structure with a hammer. If you already have computed FRFs (UFF, CSV, HDF), import them directly via Step 2: Import Test Data instead.

The module presents seven numbered steps in a step bar at the top of the window. Use Next / Previous (or click a step) to navigate. Completed steps are marked, and the current step is highlighted.

Step 1: Method

Choose how you will rove around the structure. This determines what stays fixed and what moves between measurements.

Method	What stays fixed	What moves	Best for
Roving Hammer (default)	The accelerometer(s) at a fixed reference point.	The hammer — you strike each excitation point in turn.	Most impact tests; easy to keep sensors mounted and undisturbed.
Roving Response	The hammer at one fixed driving point.	The accelerometer(s) — you move the sensor across response points.	Hard-to-reach or symmetric structures; fewer hammer locations.

Symmetric point locking: in Roving Response, the fixed excitation acts as the reference for every roving sensor position, so the resulting FRF row/column labels stay consistent and reciprocal pairs line up for the reciprocity check.

Step 2: Geometry — Points

Define the test points on a read-only 3D model using two side-by-side tables. A point may appear in both tables — that makes it a driving point (struck and measured at the same location).

Table	Meaning	Columns
Excitation Points (green)	Where the hammer strikes.	Coordinate system, Name, and three direction toggles (D1/D2/D3).
Response Points (blue)	Where the accelerometers measure.	Same columns as the excitation table.

Each point row offers a coordinate system and up to three active directions:

Coordinate system	D1	D2	D3
Cartesian	±X	±Y	±Z
Cylindrical	±R (radial)	±T (tangential)	±Z (axial)
Spherical	±R	±T	±P

Each active direction becomes one Degree of Freedom (DOF). For example, an excitation point named P10 with +X and +Z active produces two excitation DOFs, P10+X and P10+Z. In the 3D view, excitation-only points are green, response-only points are blue, and driving points (in both tables) are orange. Right-click a row to copy it across with Add as Response Point(s) or Add as Excitation Point(s).

DOF and FRF naming

Deepsona labels DOFs and FRFs by their point label, sign, and axis — matching the geometry so names stay readable end-to-end:

Item	Format	Example
DOF	`{Point}{Sign}{Axis}` (axis upper-case)	`P21+Z`, `P10-X`, `P5+R`
FRF channel	`{Response DOF} / {Reference DOF}`	`P21+Z / P10+X`

Tip: if a point has no explicit name, Deepsona auto-generates one as P + point number. Renaming a point updates its DOF and FRF labels everywhere without breaking the downstream curve fit, because the solver tracks DOFs by structured fields, not by the display string.

Step 3: Sensors

Connect your data acquisition (DAQ) hardware and map channels. Click Connect Frontend to scan the device and populate the channel table automatically, then set each channel's role:

Column	Purpose
On	Enable/disable the channel. Only enabled channels take part in the measurement.
Name	Channel name from the device.
Type	Force, Acc, Trial Acc, or Mic. Selecting Force sets the unit to `N`; Acc sets `m/s²`; Mic sets `Pa`.
Sensitivity	Transducer sensitivity (mV per unit) used for scaling.

Force-channel rule (important): the force (hammer) channel is the one that is enabled and whose unit is force (N). Exactly one force channel is allowed — designating a second one reverts the first to Acc. Getting this wrong plots the force signal as a response, so confirm the force channel before measuring.

Step 4: Auto-Determine

Automatic setup observes a few trial strikes and proposes sensible defaults so you don't have to dial them in by hand:

Force channel and response channels are identified.
Trigger threshold is set from the observed force peak (default target ~30% of the peak, clamped to a safe range).
Input ranges are scaled so the impact fills roughly 80% of range without overload.
Exponential response window decay is auto-tuned to the measured ring-down (between 1 dB and 40 dB end-of-block attenuation).

Step 5: Check

Validate measurement quality before committing to the full roving campaign. Three checks are available:

Check	What it verifies	Pass criterion
Driving Point	Counts resonance peaks in a candidate FRF to find the location that sees the most modes (fewest nodal points), so you place the fixed sensor well.	More clearly resolved peaks (default prominence ≥ 6 dB) is better.
Linearity	Re-measures the same DOF at different force levels and compares the FRFs — a linear structure gives the same FRF regardless of strike strength.	Deviation within tolerance (default ≤ 3 dB over a 40 dB dynamic range).
Reciprocity	Compares `H_ij` with `H_ji` (swap hammer and sensor); they should match for a linear, time-invariant structure.	Deviation within tolerance (default ≤ 3 dB).

Step 6: Measure

Live acquisition. Strike the current point; each impact is triggered, windowed, transformed, and averaged into the running FRF. The screen gives immediate feedback so you know whether to keep striking or retake.

Triggering and blocks

A rising-edge trigger (with hysteresis to prevent re-triggering on ring-down) captures a block around each strike, including a short pre-trigger segment so the leading edge of the pulse is preserved. At the default 24 kHz sample rate and 4096-sample block, the frequency resolution is about 5.86 Hz; adjust block size and sample rate for finer resolution or wider bandwidth.

Windows

Window	Applies to	Default	Purpose
Force window (Cut or Tukey)	Force (hammer) channel	Cut, gate offset 0.1 of block	Zeroes the post-impact tail so hammer-cable rattle and rebound noise don't pollute the spectrum.
Response window (Exponential or Uniform)	Response channel(s)	Exponential, 10 dB end-of-block decay	Damps the truncated ring-down to reduce leakage. Uniform applies no decay.

FRF estimators and averaging

Strikes are combined by linear (equal-weight) averaging — the default target is 5 accepted impacts per point. Choose the estimator that matches your noise:

Estimator	Formula basis	Use when
H1 (default)	G_xy / G_xx	Response noise dominates — the usual choice for hammer tests.
H2	G_yy / G_yx	Force (input) noise dominates; more robust at resonance peaks.
Hv	Geometric mean of H1 and H2	Noise is present on both channels.

Coherence is computed alongside the FRF and shown live as a per-frequency quality metric (1.0 = perfectly coherent).

Real-time quality feedback

Each stroke is checked instantly. Problems are shown as flags and played as distinct audible cues so you can keep your eyes on the structure:

Flag	Meaning	Default trigger
Double-Hit	The hammer bounced and struck twice in one block.	A second force pulse above ~10% of the peak (sensitivity 0.9).
Overload	Force or response clipped the input range.	Any sample reaches ~99% of channel range.
Underload	The strike was too weak relative to range.	Force peak below ~5% of range.
Bad Coherence	Averaged coherence is poor in the band of interest.	Mean coherence below the configured floor over repeated strokes.

The Measure screen shows the captured force and response time blocks (with the active windows overlaid), the running FRF magnitude in dB, the live coherence curve, and an accepted-stroke counter (e.g. "Stroke accepted (3/5) — keep striking."). With Auto-Reject on (default), good strokes are accepted automatically and bad ones discarded; turn it off to confirm each stroke manually with Accept / Reject. When the target stroke count is reached, a completion chime sounds and the workflow advances to the next roving point.

FRF display quantities and mode indicators

The measured FRF can be displayed and exported in any of these engineering quantities (all derived from the measured accelerance):

Quantity	Ratio	Unit
Accelerance (default)	a / F	(m/s²) / N
Mobility	v / F	(m/s) / N
Compliance (Receptance)	x / F	m / N
Dynamic Stiffness	F / x	N / m
Mechanical Impedance	F / v	(N·s) / m
Apparent Mass	F / a	(N·s²) / m

A Summed FRF mode indicator (in dB or linear) overlays the power sum of all measured FRFs so you can spot resonance peaks and any frequency band that is missing response energy — a quick completeness check during the campaign.

Step 7: Export

When the roving plan is complete, export the assembled FRF matrix or hand it directly to modal analysis:

Action	Format / target	Notes
Export UFF	UFF Type 58 (`.uff`)	Function-at-a-node FRF; read by all major modal packages.
Export Excel	`.xlsx`	Frequency column plus one column per FRF channel for spreadsheet post-processing.
Export HDF5	`.h5`	Standard HDF5 datasets (not the proprietary HEAD `.hdf`).
Save measured FRF	Deepsona FRF file	Round-trips the full result (roving steps, time blocks, metadata) so you can reload and recompute windows later without re-striking.
Send to Modal Analysis	EMA workspace	Publishes the FRF matrix to Step 3: Algorithms for pole-picking and validation.

Recompute without re-measuring: after loading a saved measurement you can change the force/response windows, FRF estimator, or display quantity and the FRFs are recomputed from the stored time blocks — no need to strike the structure again.