Move a Coax Stub Notch by Doubling Its Length

A coaxial cable can suppress a frequency even when its far end has nothing connected. Attach one end as a branch on a tee and leave the other end open. The branch is called a stub. Compare two cable lengths to see how the first notch moves and why the longer cable passes the signal near the shorter cable's notch.

Build the tee fixture

A Tektronix AFG1022 feeds the straight path of a Pomona 4896 BNC tee and an external Tektronix 50 ohm feed-through load at a RIGOL DHO804. One open-ended 10 or 20 foot cable attaches to the side port. The calculated graph shows a notch at about 15.5 megahertz for 10 feet and 7.8 megahertz for 20 feet.
The two branch cables are alternatives. The plot uses a lossless line model. Its two deep dips are calculated. Full-size PNG · Editable SVG

Before connecting any input or output cables, connect the DHO804’s rear Ground Terminal to earth ground with the supplied ground cable, following the DHO800 User Guide. Its Type-C power input does not ground the instrument through the power cord. Use the supplied power adapter and keep the AFG1022’s protective earth connected. Use one Tektronix AFG1022 function generator and a RIGOL DHO804 oscilloscope. Begin with the generator output off. A Bayonet Neill–Concelman (BNC) coaxial cable has a center conductor for the signal and an outer shield for the return. The cable shields share the signal reference and do not replace the oscilloscope’s required earth lead. Work only with the generator's small signal.

  1. Use a short Pomona 2249-C-36 50 Ω RG58C/U cable from AFG1022 channel 1 output to the in-line female port of a Pomona 4896 BNC tee. The straight path on this tee has one male and one female port. The perpendicular port is female.
  2. Plug the tee's male port into the female input of a Tektronix 011-0049-02 50 Ω feed-through terminator. Plug the male output of the terminator directly into RIGOL channel 1. The DHO804's analog input is nominally 1 MΩ, with a small parallel capacitance. The external terminator provides the actual 50 Ω load from center to shell while passing the signal to the scope. Keep it attached for every run.
  3. Use the side port first with no cable, then with a Pomona 2249-C-120 (10 ft or 3.048 m), then with a 2249-C-240 (20 ft or 6.096 m). Both are male-to-male 50 Ω RG58C/U assemblies. In each stub run leave the cable's far connector open with its center and shell separated. Place it on a nonconductive support away from metal and keep both connectors stationary during readings. Support the tee and cable so their weight does not strain the oscilloscope input.

On the AFG1022 set channel 1 to sine, zero offset and 1.00 V peak to peak with the 50 Ω load display setting. The 50 Ω setting scales the displayed amplitude for a matched load. The generator's source impedance remains about 50 Ω. A 1 V peak-to-peak sine across 50 Ω dissipates about 2.5 mW. The specified feed-through is rated 2 W average. The generator documentation describes short-circuit tolerance and output shutdown when overloaded. If it disables the output, stop and check the connections. Turn output off before each cable swap.

Predict where each open stub will pull the signal down

A cable's characteristic impedance, Z0, relates the voltage and current of a traveling wave. The Pomona data give nominal Z0 = 50 Ω and a capacitance per unit length, C′, of 32.2 pF/ft for cable type C. In a low-loss line, the propagation delay per foot is approximately Z0C′ = 1.61 ns/ft. The first quarter-wave frequency is one divided by four times the one-way delay through the cable. This estimates 15.53 MHz for 10 ft and 7.76 MHz for 20 ft. Cable length tolerances and connector geometry make these predictions approximate.

At a quarter wavelength, a wave reflected from the open end travels back with the phase that makes the voltage at the stub input very small. The ideal open stub is reactive, so the notch is associated with reflection back toward the source. At half a wavelength the input looks open again. At the short stub's first notch frequency, the long stub is approximately a half wavelength. Replacing the short stub with the long one should therefore recover much of the received voltage at that same frequency.

For an ideal lossless stub of length ℓ, write its electrical length as θ = 2πfℓ/v, with v the propagation speed. Its input impedance is −jZ0 cot(θ), where j denotes the imaginary unit. The corresponding shunt admittance is j tan(θ)/Z0. With 50 Ω at the source and 50 Ω at the load, a voltage divider then gives the ideal normalized magnitude |2/[2 + j tan(θ)]|. Connector parasitics, cable loss and the scope input capacitance affect the depth of a real notch. Use the model for the frequency and recovery pattern.

Measure four sweeps on the same frequencies

Set the RIGOL channel to direct-current (DC) coupling and 1× readout for the direct cable. Use normal acquisition with the channel bandwidth limit off. Set the CH1 edge trigger level near 0 V in Auto sweep mode. Set the horizontal scale so several cycles are visible and verify an actual sample rate of at least 500 MSa/s (million samples per second) at each frequency. That gives about 26 samples per cycle or more in these bands. Adjust vertical range to show several divisions without clipping. At a deep notch, a stopped display or a measurement held from a previous frequency can give the wrong value. Confirm each acquisition is fresh and the waveform still looks like a clean sine. After a generator change, let the view settle and clear old measurement statistics before reading peak-to-peak voltage (Vpp). Repeat a reading if it wanders appreciably.

Repeat the complete grid for each mode
ModeTee side portPurpose
baseEmptyInitial reference
short10 ft stubHigher-frequency notch
long20 ft stubLower notch and recovery
restoredEmpty againCheck fixture stability

For each mode record frequencies from 6.0 to 10.0 MHz and from 13.0 to 19.0 MHz, every 0.5 MHz in each band. There are 22 frequencies and 88 readings in the blank worksheet. Keep the source amplitude, load setting, feed cable and terminator fixed. Save a screen near each dip for your own evidence. Use the same vertical scale for corresponding measurements whenever it resolves both signals. When a deeply attenuated trace needs a finer scale, record the actual Vpp in volts.

For this fixture, record only a readable signal of at least 15 mV peak to peak. If the trace reaches the acquisition noise floor or the scope cannot confirm a fresh sine at the stated frequency, leave that measurement unresolved. A displayed zero or a blank measurement is insufficient to claim the ideal zero in the model.

Normalize each dip and test recovery

For each frequency compare the two no-stub readings. The checker requires them to be within 5% of the initial reading and individually between 0.75 and 1.25 Vpp. It divides each stub reading by the average of the two references at that frequency. This reduces sensitivity to the frequency response of the fixed generator, cable and receiving path.

Run python3 stub_sweep.py observations.csv. The file header is mode,frequency_mhz,vpp. For each expected band the checker finds the lowest sampled point, checks that it has at least 50% amplitude suppression and requires the two points 1 MHz away to rise at least 0.08 in normalized amplitude. It requires the sampled minima to lie near a 2:1 frequency ratio and the long stub to recover to at least 0.75 of the reference at the short stub's minimum. This reports sampled points without fitting an exact resonance frequency from a coarse grid. The recovery check compares amplitudes and does not measure the cable's phase directly.

Exit 0 means the observations support both dips and the recovery. Exit 1 means the data are incomplete or do not establish the intended comparison. Exit 2 reports a file or parsing error. If a dip is missing, check the terminator and the actual lengths, the far end of each cable and both no-stub sweeps. If the lowest sample is displaced, a new fine sweep can help explain the result. The example and automated tests are synthetic. No hardware measurements are included.

Challenge

A 10 ft open stub has a first notch near 15.5 MHz. If a 40 ft stub of the same type replaces it, where would its first notch move? How will it behave near 7.75 MHz?

Answer

The first notch moves down by a factor of four to about 3.88 MHz. Near 7.75 MHz the cable is one half wavelength long and its input should look approximately open. The waveform should recover there in the ideal circuit. The first notch of that 40 ft cable falls outside the delivered test grid.

Complete Python analyzer

Download the standalone analyzer or use the experiment archive for the diagram, worksheet, synthetic sweep and tests.

Complete source for stub_sweep.py
#!/usr/bin/env python3
"""Compare two open shunt-stub sweeps with before and after no-stub controls."""

import argparse
import csv
import math
from pathlib import Path
import sys
from typing import NamedTuple


HEADER = ("mode", "frequency_mhz", "vpp")
MODES = "base", "short", "long", "restored"
LOW = tuple(n/2 for n in range(12, 21))
HIGH = tuple(n/2 for n in range(26, 39))
GRID = LOW + HIGH
ROUNDING = 1e-12


class EvidenceError(ValueError):
    """The observations do not establish the intended notch comparison."""


class Notch(NamedTuple):
    mhz: float
    ratio: float
    left_ratio: float
    right_ratio: float


def require(ok, label, message):
    if not ok:
        raise EvidenceError(f"{label}: {message}")


def parse(rows):
    rows = iter(rows)
    require(tuple(next(rows, [])) == HEADER, "CSV", "use the exact worksheet header")
    measurements = {mode: {} for mode in MODES}
    for line, row in enumerate(rows, 2):
        label = f"line {line}"
        require(len(row) == 3, label, "expected mode, frequency_mhz and vpp")
        mode, raw_f, raw_v = row
        require(mode in measurements, label, "unknown mode")
        try:
            f, volts = float(raw_f), float(raw_v)
        except ValueError as error:
            raise EvidenceError(f"{label}: use numeric MHz and volts") from error
        require(f in GRID, label, "frequency is outside the fixed half-megahertz grid")
        require(math.isfinite(volts) and .015 <= volts <= 1.25,
                label, "voltage outside the readable 0.015–1.25 Vpp screen")
        require(f not in measurements[mode], label, "duplicate mode/frequency")
        measurements[mode][f] = volts
    for mode, readings in measurements.items():
        missing = sorted(set(GRID) - readings.keys())
        require(not missing, mode, f"missing grid frequencies: {missing}")
    return measurements


def notch(ratios, mode, band, limits):
    # Ascending frequency resolves exact ties by reporting the lower sample.
    f = min(band, key=ratios.__getitem__)
    low, high = limits
    require(low <= f <= high, mode, f"minimum lies outside the {low}–{high} MHz window")
    floor = ratios[f]
    require(floor <= .50 + ROUNDING, mode, "notch depth is less than the 0.50 ratio screen")
    left, right = ratios[f-1], ratios[f+1]
    require(min(left, right) - floor >= .08 - ROUNDING,
            mode, "the adjacent ±1 MHz observations do not bracket the dip")
    return Notch(f, floor, left, right)


def compare(measurements):
    reference = {}
    for f in GRID:
        before, after = measurements["base"][f], measurements["restored"][f]
        label = f"{f:g} MHz"
        require(.75 <= before <= 1.25 and .75 <= after <= 1.25,
                label, "no-stub reference amplitude outside 0.75–1.25 Vpp")
        require(abs(after-before) <= .05 * before + ROUNDING,
                label, "restored amplitude differs from before by more than 5%")
        reference[f] = (before + after) / 2
    short, long = ({f: measurements[mode][f] / reference[f] for f in GRID}
                   for mode in ("short", "long"))
    a = notch(short, "short", HIGH, (14, 18))
    b = notch(long, "long", LOW, (7, 9))
    ratio = a.mhz / b.mhz
    require(1.7 - ROUNDING <= ratio <= 2.3 + ROUNDING, "comparison", "minima are not near a 2:1 frequency ratio")
    require(long[a.mhz] >= .75 - ROUNDING, "comparison", "long stub did not recover near the short-stub notch")
    return {"short": a, "long": b, "frequency_ratio": ratio,
            "long_at_short": long[a.mhz], "short_at_long": short[b.mhz]}


def main():
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument("observations")
    args = parser.parse_args()
    try:
        with Path(args.observations).open(newline="", encoding="utf-8-sig") as stream:
            result = compare(parse(csv.reader(stream, strict=True)))
    except EvidenceError as error:
        print(f"UNRESOLVED: {error}", file=sys.stderr)
        return 1
    except (OSError, UnicodeError, csv.Error) as error:
        print(f"ERROR: {error}", file=sys.stderr)
        return 2
    for mode in ("short", "long"):
        v = result[mode]
        print(f"{mode}: lowest sampled point {v.mhz:.1f} MHz, ratio {v.ratio:.3f}, "
              f"shoulders {v.left_ratio:.3f}/{v.right_ratio:.3f}")
    print(f"Short/long sampled frequency ratio: {result['frequency_ratio']:.3f}")
    print(f"Long-stub recovery near predicted half-wave: {result['long_at_short']:.3f}")
    print(f"Short-stub ratio at long-stub minimum: {result['short_at_long']:.3f}")
    return 0


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

Related experiments and primary sources