Skip to content

VNA frequency-domain measurement ​

A vector network analyzer (VNA) tests a device at selected frequencies and measures both the magnitude and phase of its reflected and transmitted signals. The result is a complex S-parameter at each frequency. The VNA time-domain reflection note uses frequency measurements to calculate a synthetic impulse response; this note stays in the frequency domain and follows the VNA inductor S-parameters REPL example. Keysight's frequency-domain overview describes direct VNA sweeps of a device under test.

What the test receivers measure ​

For a two-port device driven at port 1, let a1 be its incident wave, b1 its reflected wave, and b2 its transmitted wave. Match port 2 so that no wave arrives from that side. Then:

S11(f)=b1(f)a1(f).S21(f)=b2(f)a1(f).

The first index names the responding port; the second names the driven port. These are complex ratios, so a dB plot shows only their magnitudes. The phase remains available in the complex readings. See S-parameters for the novice for the wave definitions and an inductor calculation. Rohde & Schwarz's S-parameter primer also explains reflection, transmission, magnitude, and phase.

Choose frequencies directly ​

The inductor example analyzes a 5 nH series inductor from 1 to 3 GHz:

js
var g = nP.global;
g.fList = g.fGen(1e9, 3e9, 101);

var tx = nP.mod({
  mode: 'VNA-frequency',
  fList: g.fList
});

fGen() includes both endpoints. The 101 values give 100 intervals, each 20 MHz wide. VNA-frequency accepts any nonempty, strictly increasing list of nonnegative frequencies; equal spacing is not required. mod() puts a unit complex reference at each listed frequency and makes the circuit's fList match. The points represent independent tones in a sweep, not FFT bins of a sampled waveform. There is no tList, sample rate, Nyquist limit, or inverse FFT in this mode. A physical VNA applies frequencies in sequence; nP calculates their independent steady-state responses.

Connect the ideal measurement paths ​

The REPL circuit first uses a default Coupler180() as a matched equal splitter. Its port 4 feeds reference receiver R, while port 2 drives the input coupler. The input coupler's port 2 drives the inductor and port 3 feeds reflection receiver A. The output coupler's port 4 feeds transmission receiver B. Unused ports are matched.

text
source -> splitter -> input coupler -> 5 nH inductor -> output coupler
              |            |                                   |
              R            A                                   B

The nP.nodal() result exposes four ports: source input, R, A, and B. Its three demodulators select external output ports 2, 3, and 4. Every unused coupler port is matched.

Inductor between the VNA input and output couplers

VNA source and three receiver signal paths

Compare all three receivers ​

Each default coupler has an equal-power split. Its through-wave voltage factor t and coupled-wave voltage factor k are both 1/2. The splitter puts t at R and sends t toward the input coupler. The reflected signal then passes through the input coupler's through and reverse-coupled paths. The transmitted signal passes through the input coupler's through path and the output coupler's coupled path. Relative to R, either test receiver therefore sees the same product:

tk=12.

For the unit source used by mod(), the raw complex receiver voltages are:

R(f)=12V.A(f)=122S11(f)V.B(f)=122S21(f)V.

The receiver settings are R gain 1, A gain 2, and B gain 2. A and B each specify referenceReceiver: referenceRx, so the receiver divides by R's complex phasor before applying its settable gain. Thus 2A/R=S_{11} and 2B/R=S_{21}. The gain defaults to 1, including for a direct mod-to-demod connection. nP does not infer these settings from the VNA mode or inspect arbitrary fixture paths.

There is no attenuator in any of the three VNA REPL circuits. The three ideal couplers split power without dissipating it. A 3 dB pad in front of R would be a different fixture: it would lower R's raw voltage by 10−3/20. In that hypothetical circuit, setting R's gain to 103/20 while retaining A and B gains of 2 would recover the same ratios. The receiver settings describe the actual paths; VNA-frequency and VNA-time do not assign gains on their own.

js
var referenceRx = nP.demod(circuit, {
  outputPort: 2,
  referenceGain: 1,
  noise: false
});
var reflectionRx = nP.demod(circuit, {
  outputPort: 3,
  referenceReceiver: referenceRx,
  referenceGain: 2,
  noise: false
});
var transmissionRx = nP.demod(circuit, {
  outputPort: 4,
  referenceReceiver: referenceRx,
  referenceGain: 2,
  noise: false
});

var s11Table = reflectionRx.out('response dB');
var s21Table = transmissionRx.out('response dB');

nP.lineChart({inputTable: [s11Table], title: 'Measured S11', metricPrefix: 'giga'});
nP.lineChart({inputTable: [s21Table], title: 'Measured S21', metricPrefix: 'giga'});

For the complete circuit construction, run VNA inductor S-parameters. The receiver also offers out('response magnitude'), out('response phase') in degrees, out('response real'), and out('response imaginary') as numeric tables. Its out('spectrum') retains the raw receiver phasors, and out('frequencies') returns the swept hertz values.

Check one point at 2 GHz ​

The series inductor's impedance is Z=j2πfL. At 2 GHz and 5 nH, Z=j62.831853Ω. With Z0=50Ω, the ideal inductor model gives:

S11=Z2Z0+Z.S21=2Z02Z0+Z.

The receiver values and the direct nP.L(5e-9) S-parameters agree:

Quantity at 2 GHzReflection pathTransmission path
Raw receiver voltageS11/(22) VS21/(22) V
After dividing by R and applying gain 20.283043+j0.4504770.716957−j0.450477
Corresponding inductor parameterS11S21
Magnitude0.5320180.846733
Magnitude in dB−5.481473 dB−1.445070 dB
Phase+57.858°−32.142°

The raw complex values retain the signs of their imaginary parts. Division by R removes the source/splitter level; the factor of two removes the coupler sampling factor. The direct inductor table comes from l1.out('s11dB', 's21dB'); it provides an independent check of the two test paths.

Relationship to the time-domain mode ​

QuestionVNA-frequencyVNA-time
Starting gridSupplied fListPower-of-two tList starting at zero
FrequenciesSelected directly; may be unevenGenerated DC-to-Nyquist FFT grid
Receiver resultComplex tone response at each frequencySynthetic time-domain waveform
Inverse FFTNoneUsed by demod().out('waveform')
ExampleThree receivers measuring a 5 nH inductorSame receiver fixture with a 5 nH inductor, or R/A locating cable faults

The VNA inductor impulse response example uses this same three-receiver circuit with VNA-time. Its 4,096-point tList generates an equally spaced DC-to-8 GHz grid; A and B are divided by R and corrected by the same settable gain of 2 before the inverse FFT. Their plotted outputs are the normalized reflection and transmission impulse responses. The time-domain reflection note explains that grid and the three plots.

Here the direct 1-to-3 GHz sweep reports frequency-domain S-parameters. Its 20 MHz point spacing does not create a time record or a displayed travel time. The ideal couplers have no finite bandwidth or mismatch. VNA-frequency supports deterministic demod() reception with noise: false; it does not simulate receiver noise.