IN15 · Where does the cable end?

Vector Network Analyzer

How can a network analyzer separate a device from its cables? A network analyzer compares outgoing RF waves with the incident wave, keeping both amplitude and phase. Known standards reveal the instrument and cable errors at a chosen reference plane. Reflection and transmission corrected to that plane describe the connected device; moving the plane requires a model of the intervening line.

Follow incident and reflected RF waves, calibrate two ports and move a known reference plane before retaining complex reflection and transmission.

Where does the device begin?

Preview reflection $-12.4\,\mathrm{dB}$

Preview transmission $-0.313\,\mathrm{dB}$

Preview transmission phase $-78.2^\circ$

Withheld check error $0.0324\,\%$

DeviceSource directionFrequency cursor

f/\mathrm{MHz}

100

Each calibration cable

\ell_c/\mathrm m

0.25

Each known port extension

\delta/\mathrm m

0.05

Filter capacitance

C/\mathrm{pF}

40

Drag the device carriage to change the actual cable. Gold marks calibration; pink marks the reported plane.

Standards, receiver and model

Declared kit definitionsEach known spacer

\ell_s/\mathrm m

0.05

Load resistance

R/\Omega

100

Instrument error factor

q_e

1

Receiver quadrature noise bound

n_r

0.000001

Nominal available source power

P_0/\mathrm W

0.001

Acquisition seed

s_0

188

Both ports measure open, short and load. A known through fixes forward and reverse tracking. A separate load and series resistor test the result without entering the fit. Changing a cable, receiver or declared kit requires fresh standards.

Real fifty-ohm RMS waves; positive complex time convention. Matched, nondispersive lossy TEM lines. This reduced two-port model assumes zero switch terms and isolation. Frequency is assumed calibrated. Relative wave calibration does not establish absolute power.

Cable centreline lengths use the model scale; radii and component sizes are enlarged. Moving wave rings show phase and direction at schematic speed and amplitude. The known spacer bounds port extension; no extrapolation outside the calibrated frequency grid.

Model assumptions

A real 50-ohm RMS traveling-wave model uses coherent receiver ratios, passive reciprocal error boxes and matched TEM cables. SOL standards at both ports plus a known through calibrate a reduced two-port model, including source and load mismatch. Isolation and switch terms are zero; this is not a commercial full SOLT uncertainty model. Separate check standards never enter the fit. Receiver ratios do not establish absolute power. Frequency is assumed calibrated; noise is bounded complex receiver noise. Known port extension removes only the specified matched spacer. No dispersive, connector-mode, temperature or nonlinear effects are modeled.

A saved sweep, at its saved plane

$f/\mathrm{MHz}\quad;\quad20\log_{10}|S_{ij}|/\mathrm{dB}$

Turquoise: corrected reflection; violet: corrected transmission. Amber: raw receiver ratios at the instrument, before correction. Dashed grey: separate ideal device reference at the saved reporting plane. No noisy or nonpassive result is clipped into the unit circle.

Raw waves, calibration and check standards

Kept sweeps
Record	Device	Calibration	Check error	Valid

Raw receiver channels
Channel	Real part	Imaginary part

Measured standard wave ratios
Channel	Real part	Imaginary part

Coefficients fitted from saved standards
Channel	Real part	Imaginary part

JSON and CSV retain raw complex incident/outgoing waves for both excitations, standards, coefficient snapshots, exact geometry, processing plane, bounded noise, seed and version. Touchstone uses real fifty-ohm reference and the declared RI column order; invalid sweeps cannot be exported as corrected Touchstone.

Start with the essentials

Focus question: How can a network analyzer separate a device from its cables?
One-sentence intuition: Known standards define a reference plane; a separate check tests whether the correction predicts something it did not fit.

Core mathematical model

Real-reference RMS power waves

\begin{aligned}a&=\frac{V+Z_0 I}{2\sqrt{Z_0}},\qquad b=\frac{V-Z_0 I}{2\sqrt{Z_0}}\\P&=|a|^2-|b|^2,\qquad Z_0=50\,\Omega\end{aligned}

Current points into each device port. Voltage and current are RMS phasors with positive complex time convention. Net power is incident wave magnitude squared minus outgoing wave magnitude squared. Receiver ratios alone do not calibrate absolute watts.

A reflection travels through a line twice

\Gamma_{\mathrm{plane}}=\Gamma_{\mathrm{device}}e^{-2\gamma\ell},\qquad\gamma=\alpha+\mathrm i\frac{2\pi f}{v}

A known matched spacer adds an outward and return path to reflection. Transmission accumulates the sum of the two one-way spacer lengths. Extension removes that specified propagation; it cannot guess an unknown connector.

Common difficulties

A matched load has no resolved reflection phase

Typical misconceptionEvery reflection must have a meaningful angle.

Better mental modelAt a reflection null, noise determines the angle. The phase is withheld when the wave ratio falls below the stated floor.

One-port correction is not full two-port calibration

Typical misconceptionCorrecting two reflections also corrects transmission.

Better mental modelThe two receivers need a through relation, and both mismatch feedback paths enter the matrix inverse. This reduced model explicitly omits switch terms and isolation.

Run the experiment

01
Follow both directions
Compare the filter and through presets, then reverse the source.
What to observe: Incident and reflected markers follow the physical coaxial paths. The stored sweep contains both excitation directions.
02
Move the cable end
Drag the device carriage to change cable length, then acquire fresh standards.
What to observe: The old calibration becomes incompatible when the actual cable changes. Old saved records keep their original geometry.
03
Move only the reporting plane
Choose the reference-plane preset and increase the extension.
What to observe: The pink rings move along the known spacer. Corrected phase changes by round-trip reflection or one-way transmission propagation.
04
Test the calibration
Compare the defined and wrong-kit presets. Inspect the withheld checks and export a sweep.
What to observe: Open and short parasitics matter. The independent load and series-resistor checks never contribute to the calibration fit.

Vector Network Analyzer

A saved sweep, at its saved plane

Choose the boundary before naming the device

Core mathematical model

Real-reference RMS power waves

A reflection travels through a line twice

Common difficulties

A matched load has no resolved reflection phase

One-port correction is not full two-port calibration

Run the experiment

Follow both directions

Move the cable end

Move only the reporting plane

Test the calibration

Related experiments and articles

How physics instruments measure