Why a Held Bend Loses Its Piezo Voltage

Press the tip of a piezoelectric film and hold it against a fixed stop. The voltage changes during the press and then fades while the film is still bent. Move the electrical load and you can change how quickly the recorded pulse disappears, even though the stop and the film remain in place.

Use a TE Connectivity LDT0-028K film and a 10× oscilloscope probe to measure this effect. A fixed capacitor makes the source easier to characterize, and a removable 2.2 MΩ resistor provides the comparison. The article gives calculated expectations and a runnable analyzer. No physical measurements have been recorded.

What discharges while the film stays bent?

The film contains polyvinylidene fluoride (PVDF), a polymer that produces charge when its strain changes. In a simple electrical model, a changing strain supplies charge to the sensor capacitance and to external capacitance. Once the motion ends, a resistor connected across the sensor lets that charge leave. A constant displacement can coexist with a voltage that approaches the pre-press baseline.

Let CT = Cfilm + C1 + Cprobe. When movement has stopped and leakage is dominated by the connected resistance, V(t) − Vbaseline = A exp(−t/τ) and τ = ReqCT. The 10× probe contributes nominally 10 MΩ and 10 pF at the tip. With R1 fitted, Req = 1/(1/10 MΩ + 1/2.2 MΩ) ≈ 1.803 MΩ.

TE gives 0.37–0.62 nF at 1 kHz for exact orderable 11028362-00, with about 0.48 nF typical in its example. With the 4.7 nF capacitor and the probe, nominal total capacitance is 5.19 nF, giving about 51.9 ms with the probe alone and 9.36 ms with R1. Taking the manufacturer capacitance range, 4.7 nF ±5%, probe 10 pF ±5 pF and both resistance tolerances of ±1% gives model ranges of about 47.9–56.3 ms and 8.64–10.14 ms. That calculation excludes film creep, contact motion, extra parasitics and contamination. The software uses deliberately broader experimental screens.

Under the same simple model, the ratio is τprobe/τshunt = 1 + Rprobe/R1 ≈ 5.55. The common capacitance cancels. The ratio tests an electrical loading prediction, and the fit uses a separate amplitude for each run. The initial amplitude also depends on how fast the tip was pressed and whether it slid. It is not a force or displacement measurement.

Build a low-travel fixture

A padded clamp leaves 16 millimeters of piezo film free and a rounded insulating pusher limits travel to 0.5 millimeters. Electrically a fixed 4.7 nanofarad capacitor and optional 2.2 megohm shunt share the two film tabs with a 10-times oscilloscope probe.
The clamp and stop stay fixed through all four captures. The electrical comparison changes only the 2.2 MΩ branch. Full-size PNG · Editable schematic and fixture drawing
ItemExact part and connection
FilmTE Connectivity LDT0-028K, orderable 11028362-00, solder-tab clear, no added mass. Assign one tab A to S and the other tab B to G0, and record the orientation. A and B are this lab’s labels.
C1KEMET C315C472J5G5TA, radial C0G ceramic, 4.7 nF ±5%, 50 V DC, between S and G0 for every capture. Mount close to the two film terminals.
R1Vishay Beyschlag MBB02070C2204FCT00, 2.2 MΩ ±1% axial film. One end goes to G0 and the other to S through a short removable link.
ScopeRIGOL DHO804 and PVP2350 probe, both set to 10×. Probe tip to S; ground clip to G0 only. One channel, direct current (DC) coupling, no external power.

Secure the tabs and their short wires so they cannot move at the joints. Support the film in padded insulating jaws, leaving 16 mm from the last clamp edge to the tip. Keep the film flat with no finger pressure on the free area. Use a rounded insulating pusher on a guide and adjust a rigid stop to admit at most 0.5 mm of tip travel from first contact. Check travel against a gauge before attaching the film. Hold the pusher against the stop without moving the film sideways.

The 0.5 mm stop is a small chosen fixture stroke. TE does not specify it as an allowable strain rating. The datasheet’s example reports about 7 V open-circuit at 2 mm under its own setup and more than 70 V for large flexure. Keep C1 attached, use a compensated 10× scope probe, and begin with a display range that can show several volts on either side of baseline. A 50 V C1 is a component rating, not a permissible sensor-output target. Do not crease, twist, strike or hand-flex the strip, run without the fixed C1, or connect the signal directly to a logic pin. If the clamp slips or the film shows damage, stop and correct the fixture before retesting.

Record the same press under each load

  1. With the scope input at DC coupling, capture without pressing and confirm a steady baseline. Set a time record long enough to include at least 40 ms before and 80 ms after the end of the press. Use sample mode, a 10× channel setting and the same vertical scale for all four captures. Choose a suitable edge trigger polarity after seeing the first small press.
  2. With the shunt link open, move the pusher promptly to its stop and hold for at least 150 ms before slowly withdrawing. Do not hit the stop or release while collecting the tail. Save the raw memory waveform to comma-separated values (CSV) using the scope’s waveform export. Keep voltage at the probe tip, as displayed with the correct 10× setting.
  3. Inspect the full waveform and identify the end of the press or contact transient from the capture and the observed motion. Align that time to 0 ms. The pre-press −35 to −20 ms region must be quiet and entirely before the movement. Use a guided stroke that ends within 20 ms of first movement. If the press was slower or this region contains motion, mark that capture unresolved and repeat without striking the stop. Do not move 0 ms independently to improve an exponential fit. There should be no slide, rebound, release or continuing motion inside either fitting window. If the end of motion cannot be aligned consistently or the shunt tail still rings at +4 ms, mark the capture unresolved and repeat without moving the fixed fit window. TE’s free-ring example at 16 mm gives about 20 ms for five cycles. The held pusher changes the mechanical boundary, so that free-ring example does not predict settling in this fixture.
  4. Release, return to the same baseline, and close the shunt link only while the film is at rest. Repeat the press. Then open it again for a second probe-only capture and close it for a second shunt capture. Use the same geometry, polarity and probe throughout. Allow at least 0.5 s and verify baseline after each release before recording again. Keep both repetitions even when they disagree.
  5. Convert time to milliseconds and tip voltage to millivolts. Put all four traces in one copy of the blank worksheet. A uniform 0.5 ms sample step after reducing the raw memory data is suitable; no interval can exceed 1 ms. Preserve time order, report the quiet pretrigger and capture through +80 ms. If reducing a fast capture, select real acquired points in time order after visually inspecting the raw data for ringing and clipping.

Read the fitted tails

Run python3 piezo_tail.py observations.csv. For each capture, the analyzer takes the median voltage from −35 to −20 ms as the baseline, then subtracts it and fits a straight line to the natural logarithm of the tail’s magnitude. The time constant is the negative reciprocal of the logarithmic slope. The fitting interval is 8–75 ms for the probe-only state and 4–22 ms under the shunt. The fit windows hold the same expected fraction of each type of decay to reasonable approximation.

A fit must stay at one sign and above the larger of 8 mV or four times the pretrigger span. The pretrigger span must be at most 4 mV and every captured point within ±2 V. The tail must fall by at least 0.45 in natural-log units, and its maximum absolute log-domain residual must be at most 0.12. These software checks do not detect all clipping or motion. Examine those on the raw scope trace before analysis. At the calculated 9.36 ms time constant, staying above 8 mV through +22 ms would need about 55 mV at +4 ms under the model. A noisier baseline raises that requirement. A weak pulse can leave the result unresolved even when the circuit is wired correctly.

The fitted time constants must lie within 35–70 ms with the probe alone and 6–14 ms with the shunt. The two fits for each load must agree within 20% of their average, all four press tails must share a polarity, and the entire slow-to-fast ratio across the repetitions must be within 4.2–7.0. The slow-to-fast range uses the extrema of just two captures on each load. It is an observed comparison, not a confidence interval. A good-looking trace that disagrees on its repeat stays unresolved.

Exit 0 means that all four traces support the fixture’s loading model under these screens. Exit 3 gives a valid but unresolved report, and exit 2 reports an invalid data file. A short fitted time can come from a dirty board, extra sensor leakage, a 1× probe setting or continued motion. If the fit fails, keep the raw files and check the wiring, mechanical hold and probe before changing any thresholds. A fitted tail says how the pulse behaves under this small mechanical stroke; it does not calibrate the film for acceleration, static force or displacement.

Complete analyzer

Download the analyzer or get the fixture, worksheet and synthetic host tests. The analyzer uses the Python standard library and accepts either press polarity. The tests run on generated waveforms and do not claim a physical result.

#!/usr/bin/env python3
"""Fit piezo-voltage tails under two probe loads and report repeat agreement."""
import argparse
import csv
import json
import math
from pathlib import Path
from statistics import median

HEADER = ("run", "time_ms", "voltage_mv")
RUNS = ("probe_only_1", "shunt_1", "probe_only_2", "shunt_2")
WINDOWS_MS = {"probe_only": (8, 75), "shunt": (4, 22)}
TAU_LIMITS_MS = {"probe_only": (35, 70), "shunt": (6, 14)}
EPS = 1e-9


def read_samples(stream):
    reader = csv.reader(stream, strict=True)
    if tuple(next(reader, ())) != HEADER:
        raise ValueError("Expected header: " + ",".join(HEADER))
    runs = {name: [] for name in RUNS}
    for row in reader:
        if len(row) != len(HEADER) or row[0] not in runs:
            raise ValueError(f"Unknown run or wrong field count at line {reader.line_num}")
        run, time, voltage = row
        t, v = float(time), float(voltage)
        if not (math.isfinite(t) and math.isfinite(v) and abs(t) <= 200 and abs(v) <= 1e5):
            raise ValueError(f"Nonfinite or out-of-range sample at line {reader.line_num}")
        points = runs[run]
        if points and not 0 < t - points[-1][0] <= 1.0 + EPS:
            raise ValueError(f"Time must increase in steps up to 1 ms at line {reader.line_num}")
        points.append((t, v))
    for name, points in runs.items():
        if len(points) < 121 or points[0][0] > -40 or points[-1][0] < 80:
            raise ValueError(f"{name}: capture must cover -40 to +80 ms at 1 ms or finer")
    return runs


def fit_line(xs, ys):
    x_bar = math.fsum(xs) / len(xs)
    y_bar = math.fsum(ys) / len(ys)
    slope = math.fsum((x - x_bar) * (y - y_bar) for x, y in zip(xs, ys))
    slope /= math.fsum((x - x_bar) ** 2 for x in xs)
    return slope, y_bar - slope * x_bar


def fit_run(name, points):
    mode = name.rsplit("_", 1)[0]
    start_ms, end_ms = WINDOWS_MS[mode]
    pretrigger = [v for t, v in points if -35 <= t <= -20]
    baseline = median(pretrigger)
    baseline_span = max(pretrigger) - min(pretrigger)
    result = {"fit_window_ms": [start_ms, end_ms], "baseline_mv": baseline,
              "baseline_span_mv": baseline_span, "tau_ms": None, "polarity": None,
              "max_log_residual": None, "reasons": []}
    reasons = result["reasons"]
    if baseline_span > 4.0 + EPS:
        reasons.append("baseline span")
    if any(abs(v) > 2000 + EPS for _, v in points):
        reasons.append("capture magnitude")
    fit = [(t, v - baseline) for t, v in points if start_ms <= t <= end_ms]
    polarity = 1 if median(v for _, v in fit) >= 0 else -1
    result["polarity"] = polarity
    floor = max(8.0, 4 * baseline_span)
    if any(polarity * v < floor - EPS for _, v in fit):
        reasons.append("tail below signal floor or changes sign")
        return result
    times = [t for t, _ in fit]
    logs = [math.log(polarity * v) for _, v in fit]
    slope, intercept = fit_line(times, logs)
    if slope >= 0 or logs[0] - logs[-1] < 0.45 - EPS:
        reasons.append("tail does not fall enough")
        return result
    tau = -1 / slope
    result["tau_ms"] = tau
    residual = max(abs(y - (slope * t + intercept)) for t, y in zip(times, logs))
    result["max_log_residual"] = residual
    if residual > 0.12 + EPS:
        reasons.append("tail differs from one exponential")
    lo, hi = TAU_LIMITS_MS[mode]
    if not lo - EPS <= tau <= hi + EPS:
        reasons.append("time constant outside fixture screen")
    return result


def analyze(runs):
    results = {name: fit_run(name, runs[name]) for name in RUNS}
    reasons = []
    if len({record["polarity"] for record in results.values()}) != 1:
        reasons.append("press polarity inconsistent")
    spans = {}
    for mode in WINDOWS_MS:
        taus = [results[f"{mode}_{repeat}"]["tau_ms"] for repeat in (1, 2)]
        if any(tau is None for tau in taus):
            continue
        low, high = sorted(taus)
        spans[mode] = [low, high]
        if (high - low) / ((high + low) / 2) > 0.20 + EPS:
            reasons.append(f"{mode} repeats differ by more than 20%")
    ratio = None
    if len(spans) == len(WINDOWS_MS):
        slow_lo, slow_hi = spans["probe_only"]
        fast_lo, fast_hi = spans["shunt"]
        ratio = [slow_lo / fast_hi, slow_hi / fast_lo]
        if ratio[0] < 4.2 - EPS or ratio[1] > 7.0 + EPS:
            reasons.append("entire repeat ratio does not match fixture screen")
    return {"supported": not reasons and all(not row["reasons"] for row in results.values()),
            "runs": results, "repeat_spans_ms": spans, "probe_only_to_shunt_ratio": ratio,
            "reasons": reasons}


def main():
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument("observations", type=Path)
    args = parser.parse_args()
    try:
        if args.observations.stat().st_size > 1024 * 1024:
            raise ValueError("CSV exceeds 1 MiB")
        with args.observations.open(encoding="utf-8-sig", newline="") as stream:
            report = analyze(read_samples(stream))
    except (OSError, ValueError, csv.Error) as exc:
        parser.exit(2, f"Input error: {exc}\n")
    print(json.dumps(report, indent=2, allow_nan=False))
    return 0 if report["supported"] else 3


if __name__ == "__main__":
    raise SystemExit(main())

Challenge

If you double C1 without changing either resistance, does the ratio of the two fitted time constants double?

In the simple model each time constant increases by the ratio of the new total capacitance to the old one. That factor cancels when dividing them. For the same rapidly delivered charge, the larger capacitance makes the starting voltage smaller, so the existing signal floor may become harder to meet.

References

The RC lesson derives the exponential in a resistor-capacitor circuit. The piezo sounder sweep electrically drives a sounder across frequencies. Here the film itself produces a pulse during a bounded bend, and the voltage is fitted while the bend is held.