Air-Core Coils: Measure How Distance and Rotation Change Pickup

Build a pair of air-core coils on a nonmagnetic rail. Drive the larger coil with a 100 kHz sine wave and measure the smaller coil at 35 mm, 60 mm and a 90° rotation at the original 35 mm center spacing. Measure the transmit current in every position. That lets you compare geometry even if the generator voltage and coil impedance produce slightly different currents.

The fixture operates at low voltage with one shared instrument return. The receive coil has no external load apart from the measurement probe, and one end is deliberately grounded. The two coaxial positions can be compared with a simple axial-field calculation. The rotated position shows how small a signal this particular fixture and measurement can distinguish.

Wind and position the coils

A 100-ohm limiter and the 32-turn transmit coil feed a low-side 10-ohm sense resistor. Both 10x probe grounds connect only to G0. The 24-turn receiver returns to G0 and feeds CH2. The jig marks 35 and 60 mm between coil center planes and a 90-degree pivot at the near center.
Original wiring and side-view geometry. The same receiver moves along the coil axis or turns around a pivot through its near center. Download the editable SVG.

Wire the shared return and check the drive

  1. With both outputs disabled, connect the center of SIGLENT SDG1032X CH1 through a 100 Ω ±1%, 0.25 W resistor, then the transmit winding, then a 10 Ω ±1%, 0.25 W sense resistor to the generator BNC shield. Use the shield and the low end of the sense resistor as the single physical junction G0. Label the resistor's high end TP_S.
  2. Bring the receive leads back as one twisted pair. Land one wire at G0 and the other at TP_RX on an insulated terminal near G0. Keep any external load off this terminal. Use two compensated RIGOL PVP2350 probes with a RIGOL DHO804. Set each probe's switch and its channel's attenuation to 10×. Connect both ground clips only to G0. Put the CH1 tip on TP_S and the CH2 tip on TP_RX. Keep all protective earth connections intact and keep ground clips off every other node.
  3. Set the generator to sine, 100 kHz, 4.00 Vpp displayed into Hi-Z and 0 V offset. Leave generator CH2 disabled. The generator still has a physical 50 Ω output impedance with Hi-Z selected. At this setting its unloaded sine has a nominal 2.00 V peak. The minimum combined external resistance is 108.9 Ω, so even ignoring the generator's 50 Ω, the nominal steady-state peak current is bounded by 2.00 V / 108.9 Ω ≈ 18.4 mA. The 100 Ω limiter then dissipates about 17 mW or less for this sine setting. The 50 Ω source resistance lowers the current further and the coil reactance can appreciably reduce it. Use the measured shunt voltage for all comparisons.
  4. Enable the output and check CH1 first. It should show a stable near-100 kHz sine across the shunt. For scale, 0.240 Vpp across 10.0 Ω corresponds to 24.0 mApp. If the sense voltage exceeds about 0.27 Vpp or has a clipped or erratic shape, disable the output and check the generator load setting, resistor values, wiring and probe ratios before proceeding. Disable the generator output before moving the receiver or connecting a link.

Use DC coupling and the 20 MHz bandwidth limit on both scope channels. Start around 50 mV/div on CH1 and 10 mV/div on CH2, then choose a range that fits the largest receive waveform without clipping. Use a window that visibly spans at least 20 cycles, with a stable CH1 trigger for driven readings. Use Auto trigger for the drive-off control if the trace otherwise stops updating. Keep the vertical scale, offsets, timebase, bandwidth, probe settings, acquisition mode and measurement window fixed for all three positions and their controls. The DHO804's minimum displayed vertical scale is 5 mV/div at 10×. A stable, resolved trace is required before using a small numerical reading.

Take three runs and two controls per run

Nominal geometry. Enter your measured spacing and angle in the worksheet.
CaseCenter spacingAngle between plane normals
near35 mm0° coaxial
far60 mm0° coaxial
edge35 mm, same center as near90° rotation
  1. For each geometry, take three active readings from stable captures. Record the drive frequency from CH1 and the median of the three peak-to-peak values on each channel. Check that CH2 shows a repeatable sinusoid at the drive frequency when it is visibly resolved. If a trace is clipped, unstable, or the wrong frequency, correct the acquisition and take the run again.
  2. With the generator output disabled, leave the coil and both probes connected. Take three CH2 peak-to-peak readings and record the largest as the off control.
  3. With the output still disabled, fit a short insulated link from TP_RX to G0 at the probe end. Enable the generator, confirm CH1 still shows the expected drive, and record the largest of three CH2 peak-to-peak values as the short control. Disable the output and remove the link before changing geometry. The link belongs only across the receiver terminal and G0.
  4. Calculate the guard as max(1 mVpp, 3 × off, 3 × short). The 1 mVpp term is a worksheet floor. Your same-settings trace and controls determine what the bench setup can resolve. Record the peak-to-peak reading for an unresolved capture in your bench notes. Use ratios only when both relevant traces are visibly resolved and both readings exceed their guards. If a value happens to clear the numeric guard but its trace is unresolved, report that run as unresolved and leave the result out of your comparison.

The off control samples the same receiver and instrument with the source disabled. The powered short control shows what remains at the probe with the receiver terminal shorted to the reference. Electric pickup into the normal, unshorted input can still contribute to the active reading. Keep both controls as evidence and use their maximum to set a guard. A simple subtraction would mix uncorrelated noise and drive-correlated pickup.

Estimate the coaxial signal

Let a be transmitter radius, b receiver radius, z center spacing, and Nₜ and Nᵣ the turn counts. Use metres and μ₀ ≈ 4π × 10⁻⁷ H/m. The on-axis field of a circular loop is proportional to a² / (a² + z²)^(3/2). Approximating that field as uniform over the small receive winding, multiply by its area and turn count. Faraday's law gives the sine-wave amplitude from the time derivative of the flux.

For the nominal geometry, M_axial is about 1.615 µH at 35 mm and 0.647 µH at 60 mm. If the shunt reports 24.0 mApp at 100 kHz, those predict about 24.4 mVpp and 9.75 mVpp. The nominal near/far ratio after normalizing by current and frequency is about 2.50. These are calculated values. The measured radii, turn spacing, center offsets and finite winding widths affect the result. For example, a 5% change in the receiver radius changes its area by about 10%. At the near position the receiver radius is about 0.19 of sqrt(a² + z²), so the uniform-field assumption remains approximate.

For a resolved coaxial trace, compute the apparent mutual inductance as V_rx,pp / (2π f I_pp). For the rotated position use this quotient only to form a response ratio relative to near. The axial calculation applies only to coaxial loops. In the ideal symmetric fixture, rotating by 90° reduces the axial flux through the receiver. In a physical build, angle error, lateral offset, lead pickup and capacitance can leave a residual. A small reading says something about this assembly and its guard under these settings.

The PVP2350 at 10× presents nominally 10 MΩ in parallel with 10 pF at its tip. At 100 kHz that is about 159 kΩ in magnitude. With 15 pF and 105 kHz it is about 101 kΩ. If the receive winding's source impedance at that frequency is below 1 kΩ, the loading error is roughly 1% or less under those bounds. The measurement includes the finite probe load. The scope's bare input capacitance is specified at the scope connector and is already accounted for by using the probe's stated tip impedance with a compatible input. Avoid applying this simple comparison if the receive winding and fixture show signs of resonance, distorted waves or strong capacitive feedthrough.

Analyze your worksheet

Download the original diagram, analyzer, host tests and blank worksheet. Copy observations-blank.csv to observations.csv. Enter peak-to-peak readings in volts, frequency in hertz and geometry in millimetres and degrees. Use the measured radii and shunt resistance in the command. Use 10.0 Ω and note the resistor's ±1% tolerance if you use its nominal value.

python3 coil_coupling.py observations.csv --tx-radius-mm 40 --rx-radius-mm 10 --sense-ohm 10
python3 -m unittest -v test_coil_coupling.py

The analyzer accepts one near, one far and one edge row. It enforces 33–37 mm near and edge, 58–62 mm far, up to 3° coaxial error, 87–93° edge rotation, and no more than 1 mm change from the measured near spacing to the rotated spacing. Its 5–40 mApp and 95–105 kHz checks screen worksheet entries. Use the narrower drive conditions and waveform checks during the experiment. The JSON result records the controls, guard and predicted coaxial voltages, and emits null for ratios whose readings do not exceed the guard. Judge trace quality separately from the numeric check. The included tests run on a computer with synthetic values.

Complete Python analyzer
#!/usr/bin/env python3
"""Analyze three manual readings from the Tricky Circuit air-coil fixture."""

import argparse
import csv
import json
import math
import sys

CASES = ("near", "far", "edge")
COLUMNS = (
    "case", "spacing_mm", "angle_deg", "frequency_hz", "sense_vpp",
    "receiver_vpp", "off_vpp", "short_vpp",
)
MU0 = 4 * math.pi * 1e-7
TX_TURNS = 32
RX_TURNS = 24
FLOOR_VPP = 0.001


def bounded(value, minimum, maximum, label):
    try:
        result = float(value)
    except (ValueError, TypeError):
        raise ValueError(f"{label}: expected a number") from None
    if not math.isfinite(result) or not minimum <= result <= maximum:
        raise ValueError(f"{label}: expected {minimum:g} to {maximum:g}")
    return result


def axial_mutual_h(tx_radius_m, rx_radius_m, spacing_m):
    """Small receiver in the on-axis field of a circular transmitter."""
    receiver_area = math.pi * rx_radius_m**2
    field_per_amp_turn = MU0 * tx_radius_m**2 / (
        2 * (tx_radius_m**2 + spacing_m**2)**1.5
    )
    return TX_TURNS * RX_TURNS * receiver_area * field_per_amp_turn


def analyze(stream, tx_radius_mm, rx_radius_mm, sense_ohm):
    tx_radius_m = bounded(tx_radius_mm, 39, 41, "transmitter radius mm") / 1000
    rx_radius_m = bounded(rx_radius_mm, 9.5, 10.5, "receiver radius mm") / 1000
    sense_ohm = bounded(sense_ohm, 9.9, 10.1, "sense resistance ohm")
    reader = csv.DictReader(stream)
    if tuple(reader.fieldnames or ()) != COLUMNS:
        raise ValueError("unexpected CSV header; use observations-blank.csv")

    rows = {}
    for row in reader:
        case = row["case"]
        if case not in CASES or case in rows or None in row:
            raise ValueError(f"unexpected, repeated or malformed case: {case!r}")
        distance_bounds = (58, 62) if case == "far" else (33, 37)
        angle_bounds = (87, 93) if case == "edge" else (0, 3)
        spacing = bounded(row["spacing_mm"], *distance_bounds, f"{case} spacing mm")
        angle = bounded(row["angle_deg"], *angle_bounds, f"{case} angle deg")
        frequency = bounded(row["frequency_hz"], 95000, 105000, f"{case} frequency Hz")
        current = bounded(
            bounded(row["sense_vpp"], 0, 1, f"{case} sense Vpp") / sense_ohm,
            0.005, 0.040, f"{case} sense current App",
        )
        receiver = bounded(row["receiver_vpp"], 0, 1, f"{case} receiver Vpp")
        off = bounded(row["off_vpp"], 0, 1, f"{case} off control Vpp")
        short = bounded(row["short_vpp"], 0, 1, f"{case} short control Vpp")
        guard = max(FLOOR_VPP, 3 * off, 3 * short)
        omega_current = math.tau * frequency * current
        above_guard = receiver - guard > 1e-12
        normalized = receiver / omega_current if above_guard else None
        model = None
        if case != "edge":
            mutual = axial_mutual_h(tx_radius_m, rx_radius_m, spacing / 1000)
            model = {
                "mutual_h": mutual,
                "predicted_vpp": omega_current * mutual,
                "apparent_mutual_h": normalized,
                "apparent_over_model": normalized / mutual if above_guard else None,
            }
        rows[case] = {
            "spacing_mm": spacing, "angle_deg": angle, "frequency_hz": frequency,
            "current_app": current, "receiver_vpp": receiver,
            "off_vpp": off, "short_vpp": short, "guard_vpp": guard,
            "above_guard": above_guard,
            "normalized_response_v_per_rad_per_s_amp": normalized,
            "coaxial_model": model,
        }

    if set(rows) != set(CASES):
        raise ValueError("expected one near, one far, and one edge case")
    if abs(rows["edge"]["spacing_mm"] - rows["near"]["spacing_mm"]) > 1:
        raise ValueError("edge spacing must stay within 1 mm of near spacing")
    frequencies = [row["frequency_hz"] for row in rows.values()]
    if max(frequencies) - min(frequencies) > 0.01 * rows["near"]["frequency_hz"]:
        raise ValueError("frequency spread exceeds 1% of near frequency")

    key = "normalized_response_v_per_rad_per_s_amp"
    near, far, edge = (rows[case][key] for case in CASES)
    return {
        "cases": {case: rows[case] for case in CASES},
        "near_over_far": near / far if near is not None and far is not None else None,
        "edge_over_near": edge / near if edge is not None and near is not None else None,
    }


def main():
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument("observations", help="CSV with manual oscilloscope readings")
    parser.add_argument("--tx-radius-mm", type=float, required=True)
    parser.add_argument("--rx-radius-mm", type=float, required=True)
    parser.add_argument("--sense-ohm", type=float, required=True)
    args = parser.parse_args()
    try:
        with open(args.observations, encoding="utf-8-sig", newline="") as stream:
            result = analyze(stream, args.tx_radius_mm, args.rx_radius_mm, args.sense_ohm)
    except (OSError, ValueError) as error:
        print(f"Input error: {error}", file=sys.stderr)
        return 2
    print(json.dumps(result, indent=2, allow_nan=False))
    return 0


if __name__ == "__main__":
    sys.exit(main())

Interpret the measurements

Compare the measured near and far responses normalized by 2π f I_pp. If the far response falls below its guard, the numeric near/far ratio is unresolved for that run. If the ratio differs substantially from the model, start by checking turn counts, radius, center spacing, transmitter current and lead routing. Keep all controls and measurements together so that another run can use the same comparison.

For a related experiment with a loaded secondary, see what a 600 Ω load does to a 1:1 audio transformer. The LPS4018 pulse fixture measures one winding's change in incremental inductance during a current ramp.

Check your reasoning

If the far coil gives half the near voltage and the far drive current also halves at the same frequency, what did the normalized comparison show?

The two normalized responses are equal. The measured drop in voltage tracks drive current. With this result, recheck the positions and controls before attributing a difference to spacing.

References