At 120 Hz, a 100 µF capacitor can have roughly 13 Ω of impedance magnitude even when its equivalent series resistance is a small fraction of an ohm. Add 0.22 Ω in series and the magnitude barely moves. Measure resistance and reactance separately and the same change is much easier to see.
This experiment uses one Panasonic FR-A capacitor, a four-terminal-pair LCR meter and a bypassable resistor. An independent short and a resistor-only control help distinguish the intended real-resistance increment from a poor contact. The procedure and the analyzer are ready to run. No physical measurements have been recorded.
The meter reports the series-equivalent real and imaginary components at each frequency. The real part of the capacitor fixture is its equivalent series resistance at that test condition, including any residual fixture error. Negative reactance corresponds to capacitive behavior. As frequency increases, lead and internal inductance can contribute positive reactance, so this experiment keeps the signed reading.
Z(f) = R(f) + jX(f)|Z(f)| = √(R(f)² + X(f)²)Xideal(f) = −1/(2πfC)
For nominal 100 µF, ideal reactance is −13.26 Ω at 120 Hz, −1.592 Ω at 1 kHz, −0.1592 Ω at 10 kHz and −0.01592 Ω at 100 kHz. These are calculations for an ideal capacitance. The real capacitor adds loss and inductance. The Panasonic September 2025 catalog lists 0.130 Ω maximum impedance magnitude for this orderable at 100 kHz and +20 °C. The fixture uses its own bias, contacts and screening rules. Its results do not establish production compliance with the Panasonic limit.
As a numerical example at 120 Hz, take R = 0.080 Ω and X = −13.26 Ω. Adding 0.220 Ω to the real part makes the magnitude change from about 13.2602 Ω to 13.2634 Ω, only 3.2 mΩ. The real part increases by the full 0.220 Ω. At a frequency where reactance is small, the magnitude becomes much more sensitive to that same change.

| Item | Part and setup |
|---|---|
| C1 | Panasonic EEUFR1E101, FR-A, polarized radial aluminum electrolytic, 100 µF ±20%, 25 V, 6.3 × 11.2 mm body, 2.5 mm lead pitch. Positive lead goes to P; striped negative lead goes to LOW. |
| R1 | Vishay Dale WSL2512R2200FEA, Power Metal Strip 2512, 0.220 Ω ±1%, 1 W at its stated rating condition. Solder directly between the HIGH and P copper pads. |
| B1 and adapter | Short rigid copper link bridging R1 pads, with firm low-resistance fasteners. Two short rigid tinned copper posts fit the fixture jaws. Keep component leads short, support them away from the seals and use no solderless breadboard. |
| Meter and fixture | Keysight E4980BL-032 (20 Hz to 300 kHz) and direct 16047A Kelvin fixture with 16047A-701 shorting plate. Connect the four BNC connectors directly; make each force and sense contact touch the respective post. |
| DC check and discharge | Floating Fluke 115 on the 6 V DC range, then a separate 1 kΩ, 0.25 W discharge resistor. Use the meter’s COM at LOW. |
Make the copper layout small enough to fit between the jaws, and mark where the jaw edges touch. Put both B1 endpoints immediately beside R1 so the bypass loop stays short. Retain the same board orientation, post contact depth and lead geometry. If the short strap or capacitor requires moving the jaws, repeat the correction and the entire set. C1 is out of circuit for the resistor-only and plate controls.
Before changing a strap, link or capacitor connection, set the LCR bias to 0 V and stop measuring. Disconnect the adapter from the source or keep the source in its inactive state, discharge C1 through 1 kΩ, and verify less than 50 mV across C1. With 100 µF and 1 kΩ, the nominal time constant is 0.10 s. After the meter confirms discharge, remove the discharge resistor. Never fit a shorting plate or a replacement link across a charged capacitor. The only powered source at the fixture is the LCR meter.
short state without editing the correction. That is a contact/zero repeatability screen. The same plate does not give an independent traceable absolute zero. Keysight gives a typical residual model of about 20 nH and 1 mΩ for the supplied plate; 20 nH is about 12.6 mΩ at 100 kHz before correction. Keep the geometry and the correction fixed through all accepted states.reference. Leave B1 open, remove C1 and fit the short P-to-LOW link in its place. Bias stays off. The reference includes its short link and contacts, so it is a comparator for the increment. Record both components across all four frequencies. Remove the temporary link before reinstalling C1.base_before, then discharge and open B1 for series. Reapply and verify the same bias after the configuration change. Discharge, reclose B1, and repeat as base_after. At each frequency, wait for stable readings before recording. Do not improve a bad point by choosing only the most favorable reading. Rerun the full state if its contacts, correction or temperature changed.The positive bias keeps the capacitor polarized under the small AC signal. At the worksheet’s maximum current and maximum 120 Hz reactance, the capacitive AC swing is roughly √2 × 1.2 mA × 17 Ω ≈ 29 mV peak. It is small compared with the verified positive DC bias. R1 AC dissipation at 1.2 mA RMS is only about 0.32 µW. These are circuit estimates, and the circuit has no external ripple-current or load source.
Fill the blank twenty-row worksheet and run python3 esr_compare.py observations.csv. Input numbers are signed in volts, ohms and milliamperes, with no engineering suffix. Exit 0 means all four points support the increment; 3 means valid input with at least one unresolved point; 2 indicates a malformed or incomplete worksheet. JSON includes the measured vectors, magnitude and a reason for every failed screen.
max(20 mΩ, 1% × max(|Xbefore|, |Xafter|)) in reactance. The reactance screen is looser where the ideal magnitude is large.The two-subtraction span shows how much the observed increment changes when you choose either baseline. It does not describe the instrument’s absolute accuracy, temperature uncertainty, or probability of future readings. A supported point is evidence for a real increment in this fixture. An unresolved point can come from contact drift, a weak excitation reading or residual inductance. Compare the cause reported and repeat the corresponding controls before interpreting the capacitor.
If you need a magnitude for comparison, calculate it from each recorded R and X. At high frequency, a small positive X can be real and near-resonant behavior can make inferred capacitance unstable. The worksheet makes no claim about resonance between its four sample frequencies.
Download the Python analyzer or get the schematic, worksheet and synthetic host tests. Python uses its standard library. The supplied tests exercise software behavior and contain no bench data.
#!/usr/bin/env python3
"""Compare R-X observations of a capacitor with and without a series resistor."""
import argparse
import csv
import json
import math
from pathlib import Path
FREQS = (120, 1000, 10000, 100000)
CAP_STATES = ("base_before", "series", "base_after")
STATES = ("short", "reference", *CAP_STATES)
X_LIMITS = {120: (-17.0, -10.0), 1000: (-2.1, -1.2),
10000: (-0.24, -0.09), 100000: (-0.08, 0.10)}
HEADER = ("state", "frequency_hz", "dc_v", "r_ohm", "x_ohm", "ac_ma")
EXPECTED = {(state, f) for state in STATES for f in FREQS}
EPS = 1e-10 # Round-off slack for exact inclusive boundaries.
def bounded(value, low, high):
return low - EPS <= value <= high + EPS
def read_rows(stream):
reader = csv.reader(stream, strict=True)
if tuple(next(reader, ())) != HEADER:
raise ValueError("Expected header: " + ",".join(HEADER))
rows = {}
for row in reader:
if len(row) != len(HEADER):
raise ValueError(f"Wrong field count at line {reader.line_num}")
state, f_raw, *numbers = row
try:
key = state, int(f_raw)
values = tuple(float(value) for value in numbers)
except ValueError as exc:
raise ValueError(f"Invalid number at line {reader.line_num}") from exc
if key not in EXPECTED or key in rows:
raise ValueError(f"Unknown or duplicate state/frequency at line {reader.line_num}")
if not all(math.isfinite(v) and abs(v) <= 10000 for v in values):
raise ValueError(f"Nonfinite or unphysical input at line {reader.line_num}")
rows[key] = values
if set(rows) != EXPECTED:
raise ValueError(f"Missing observations: {sorted(EXPECTED - rows.keys())}")
return rows
def displayed(r, x):
return {"r_ohm": r, "x_ohm": x, "magnitude_ohm": math.hypot(r, x)}
def compare_frequency(rows, f):
reasons = []
for state in STATES:
dc, _, _, ac = rows[state, f]
cap_connected = state in CAP_STATES
dc_low, dc_high = (1.40, 1.60) if cap_connected else (-0.05, 0.05)
if not bounded(dc, dc_low, dc_high):
reasons.append(f"{state} DC bias")
if not bounded(ac, 0.80, 1.20):
reasons.append(f"{state} AC current")
dc_levels = [rows[state, f][0] for state in CAP_STATES]
ac_levels = [rows[state, f][3] for state in CAP_STATES]
if max(dc_levels) - min(dc_levels) > 0.050 + EPS:
reasons.append("capacitor states DC bias mismatch")
if max(ac_levels) - min(ac_levels) > 0.100 + EPS:
reasons.append("capacitor states AC current mismatch")
_, short_r, short_x, _ = rows["short", f]
_, ref_r, ref_x, _ = rows["reference", f]
_, before_r, before_x, _ = rows["base_before", f]
_, series_r, series_x, _ = rows["series", f]
_, after_r, after_x, _ = rows["base_after", f]
if math.hypot(short_r, short_x) > 0.020 + EPS:
reasons.append("short residual")
if not bounded(ref_r, 0.180, 0.260) or abs(ref_x) > 0.020 + EPS:
reasons.append("resistor reference")
if not all(bounded(r, 0, 1.0) for r in (before_r, series_r, after_r)):
reasons.append("capacitor fixture real resistance")
if not all(bounded(x, *X_LIMITS[f]) for x in (before_x, series_x, after_x)):
reasons.append("capacitor reactance")
x_tolerance = max(0.020, 0.01 * max(abs(before_x), abs(after_x)))
if abs(before_r - after_r) > 0.020 + EPS:
reasons.append("baseline real drift")
if abs(before_x - after_x) > x_tolerance + EPS:
reasons.append("baseline reactance drift")
delta_r = sorted((series_r - before_r, series_r - after_r))
delta_x = sorted((series_x - before_x, series_x - after_x))
if delta_r[0] < 0.100 - EPS or any(abs(d - ref_r) > 0.030 + EPS for d in delta_r):
reasons.append("real increment does not match reference")
if any(abs(d - ref_x) > x_tolerance + EPS for d in delta_x):
reasons.append("reactance does not match reference")
return {
"frequency_hz": f,
"supported": not reasons,
"reasons": reasons,
"short_magnitude_ohm": math.hypot(short_r, short_x),
"reference": displayed(ref_r, ref_x),
"before": displayed(before_r, before_x),
"series": displayed(series_r, series_x),
"after": displayed(after_r, after_x),
"delta_r_ohm": delta_r,
"delta_x_ohm": delta_x,
"reactance_screen_ohm": x_tolerance,
}
def analyze(rows):
results = [compare_frequency(rows, f) for f in FREQS]
return {"all_supported": all(row["supported"] for row in results), "frequencies": results}
def main():
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("observations", type=Path)
args = parser.parse_args()
try:
if args.observations.stat().st_size > 16384:
raise ValueError("Input exceeds 16 KiB")
with args.observations.open(encoding="utf-8-sig", newline="") as stream:
result = analyze(read_rows(stream))
except (OSError, ValueError, csv.Error) as exc:
parser.exit(2, f"Input error: {exc}\n")
print(json.dumps(result, indent=2, allow_nan=False))
return 0 if result["all_supported"] else 3
if __name__ == "__main__":
raise SystemExit(main())
At a fixed 1 mA RMS, a 1.0 Ω real component dissipates 1.0 µW. An ideal −j1.0 Ω reactance has the same magnitude and dissipates zero average power. With a capacitor fixture the real part is the component used for P = IRMS²R.
The RC lesson derives the ideal transient relationship. Here the same capacitance is examined under small-signal AC, with the dissipative component separated from reactance.