The port is the expensive one, and the damage spreads. Start the threads by hand and square, tighten with the wrench rather than by feel, and turn the nut and not the cable body. A cross-threaded or over-torqued port is a bench repair on an eighty-five thousand dollar instrument, and the same damage transfers to the next connector it mates with — which on this rig is a calibration module. Never leave a cable hanging off a port either: side load at the connector is cable stress, and it is the slow version of the same failure. Cap what you are not using, and touch a grounded surface before you touch a center pin.

The horn is dropped, not worn out. Two hands, support its weight while you mate it, and never hang it from its own connector.

The cable is kinked, not worn out. Respect the bend radius, do not coil it tight, do not lay it where somebody will step on it, and do not pull it straight out of a tight bend.

The absorber is priced by the square foot. The pyramids shed if you brush them and a crushed tip is a permanently worse quiet zone, so nobody leans on the walls and nothing gets rested on the floor pyramids. There is no repairing a tip; there is only buying more absorber.

One more that costs nothing to avoid: nothing that transmits goes into an analyzer port. The port is an input, and it is not built to be driven.

The single most useful habit this lab can leave you with is the order of operations: calibrate, verify, then measure. An unverified calibration is an unmeasured antenna, and the difference costs about ninety seconds to check and an afternoon to discover afterwards.

The foam is not optional equipment — it is what lets you take your hands off the antenna while you read the screen, and Lesson 9 explained why that matters. Standing the antenna in the chamber enforces the rule for you: with the door shut you cannot be holding the element and reading the trace at the same time, which is exactly the mistake this lab is trying to train out of you.

A torque wrench is worth using if the bench has one: an under-tightened SMA connector is a slow, intermittent way to lose a calibration.

The service will even run with no positioner attached at all, and says so — the scan button reads no positioner and everything that turns the tower is grayed out, while the Sweep tab keeps working. That is the shape of today’s measurement: one port, one sweep, nothing moving.

The default span is wide enough for a horn and wrong for your element: a 101-point sweep across a gigahertz is a 10 MHz grid, and a thin dipole’s whole VSWR \(\le 2\) band can be 40 MHz. Four points across a resonance will not find its minimum and cannot give you a crossing frequency worth quoting. Points also set how finely you can place the marker, because the marker steps from one measured frequency to the next and not between them. This is the one setting where the default will quietly cost you the measurement, so change it before you calibrate, not after.

The wizard asks which plane you calibrated because nothing in the data afterwards can tell. Lesson 9 made that argument on paper; the dropdown is the argument made unavoidable, and the answer is written into the record — and into the header of every file you export — so that a measurement read weeks later still says what it was referenced to. Pick the one-port option and the checklist rewrites itself to match: one cable end to unmate, one port of the module to mate, and nobody sent into the chamber to disconnect an antenna for no reason.

The cal pill in the strip along the bottom is the other half: it reads cal, uncal, or cal? when the instrument will not say. An uncalibrated sweep is valid data and is never blocked, but it should not be possible to take one without noticing.

Verifying against a load is the step that catches everything else — a cable swapped after the calibration, a loose connector, a sweep changed without meaning to. It works because the load is a standard the calibration did not use to define itself, so a good answer is evidence rather than arithmetic.

A capture is not a scan and is not written to runs/. It lives in the browser until you export it, which is deliberate: the service ships complex S-parameters and nothing else, and every derived number you are about to read — dB, impedance, VSWR, return loss — is computed from them on the page. The payload stays a measurement rather than a measurement plus somebody’s arithmetic.

If you do ask for several parameters, STOP declines to start the next one rather than interrupting the sweep in flight. A single sweep is one exchange with the instrument and there is nowhere to interrupt it.

This is Lesson 9’s “four names for one number” with the instrument doing the conversion instead of you:

with \(Z_0 = 50\ \Omega\), which on this instrument is a fact and not a setting. Two readings are worth recognizing on sight. The readout says open where \(\Gamma\) reaches +1, because the impedance there is unbounded rather than merely large. And it says VSWR ∞ whenever \(\vert \Gamma \vert \ge 1\): a passive antenna cannot reflect more power than it receives, so that is noise, or a calibration that no longer matches the sweep, and quoting a number from it would dress the problem up as a measurement.

Read the resonance twice, once as the dip in dB and once as the reactance passing through zero — the real-axis crossing on the chart. The two readings disagree when something is wrong with the reference plane, and that disagreement is the cheapest diagnostic you have.

On the log-magnitude trace that bar sits at −9.5 dB, so the magnitude plot finds the crossings quickly and the marker confirms them exactly. Interpolate between the two bracketing points rather than quoting whichever one happens to be closer; with 401 points across ±30% of a 900 MHz resonance the grid is about 1.3 MHz, and a 44 MHz bandwidth quoted to the nearest grid point is carrying 3% of avoidable error.

Two details in that header are there because of how easily a file lies. A capture of a single reflection parameter is written as .s1p rather than an .s2p with the other three columns zeroed — a zero-filled .s2p claims a through path of exactly zero and an infinitely reflective second port, and a reader will plot those numbers quite happily. And if you do capture a partial set, the header names what was not measured and written as zero. The Touchstone column order for a two-port file is \(S_{11}\), \(S_{21}\), \(S_{12}\), \(S_{22}\), which is not the order most people expect and is the usual source of transposed data.

Rename each file as you save it — l10_freespace.s1p, l10_hand.s1p, l10_bench.s1p. The export stamps the clock into the filename, which keeps three captures apart and tells you nothing about which is which a week later. Three files with the sweep unchanged between them are directly comparable, which is the whole reason for not touching the VNA panel once you have calibrated.

In detail:

1. An annotated \(\vert S_{11}\vert\) plot. Resonance marked, both VSWR \(= 2\) crossings marked, axes labeled with units.

2. A results table: \(f_0\), \(Z\) at resonance, VSWR \(\le 2\) bandwidth in MHz and in percent, for all three perturbation configurations.

3. A Smith chart with the marker on the resonance, and one sentence identifying that point as the real-axis crossing.

4. A paragraph on the perturbation results. What moved, in which direction, by how much, and why. Connect at least one observation to the near-field argument from Lesson 9.

Hand in the three exported files with it, and state the reference plane you calibrated at. A plot with no markers and no units is a screenshot rather than a measurement, and the paragraph carries the largest share of the grade.

The midterm project, due 2 Oct, asks you to build a dipole, tune it, measure it, and defend the numbers you report about it. The antenna itself is not in question — everyone builds the same one — so what you defend is the measurement. Tuning is this lab: capture, read the sign of the reactance, trim, capture again. Calibration and Smith-chart fluency are the rate-limiting skills on that project, so practice both here, where the cost of a mistake is an afternoon rather than a week.