rigexpert-swift

RigXSwift — a native macOS application and Swift package for RigExpert antenna analyzers.

Not affiliated with, endorsed by, or supported by Rig Expert Ukraine Ltd. "RigExpert" and "AntScope" are their names, used here only to say which hardware this works with.

Not a port of AntScope2. No source is shared with it — the AA serial protocol, the analyzer capability table and the OSL calibration algebra were read out of its MIT-licensed source and reimplemented. The result is 5,000 lines of Swift against 72,000 of Qt C++, and an app bundle of 950 KB against the official build's 85 MB.

Everything below was measured with a real AA-30.ZERO against physical references.

What it does

• Connects over a serial port and sweeps
• Plots SWR, resistance and reactance, a hand-drawn Smith chart, and a time-domain view of the cable, with a hover cursor
• Applies an open/short/load calibration captured from the command line
• Measures a coaxial cable's length, velocity factor and loss from the sweep alone
• Detects when the reactance crossings belong to the feedline rather than the antenna, and measures the feedline instead of misreporting it as a resonance
• Shades the amateur bands over the SWR and R/X plots, and reports the best SWR in each one — Italy, IARU Region 1, 2 or 3, since the bands are not the same width everywhere
• Zooms to a band with one click, and shows where the CW, digital and phone segments of it are, with a dashed rule at every division
• Counts the samples the analyzer failed to measure, rather than hiding them
• Opens saved Touchstone .s1p files and overlays them for comparison, from this or any other instrument
• Exports Touchstone .s1p

Does it work

Verified against known references, not asserted:

measurement	reference	result
50 Ω load, uncalibrated	50 Ω	R 47.4–48.2 Ω, SWR 1.04–1.06
50 Ω load, calibrated	50 Ω	R 49.2–50.0 Ω, SWR 1.00–1.02
RG-58 length, 2.97 m by tape	2.970 m	2.962 m (velocity factor 0.662 vs 0.66 nominal)
RG-174 velocity factor	0.66 solid PE	0.66, allowing 2 cm for connectors
RG-174 loss, 1–30 MHz	≈0.1 dB/m datasheet	0.126 dB/m
Feedline length, 1.66 m of cable	1.66 m + connectors	1.747 m (8.7 cm over three connector interfaces)
Dipole resonance, 2.04 m tip to tip	143/L = 70.1 MHz	70.33 MHz, 0.33%
Feedline end by TDR	1.747 m, known two other ways	1.72 m, resolution ±0.47 m

System repeatability, from two sweeps taken without touching anything: ±120 kHz on a 67 MHz resonance, about 1% on |Γ|. Those two sweeps, and the four that follow them through a controlled experiment on feedline length and surroundings, are in Examples/dipole-experiment.

Getting started

./make_app.sh                       # builds build/RigXSwift.app
./test.sh                           # 88 tests

The command line client needs no UI and is the quickest way to check a connection:

swift build
.build/debug/rigx-probe demo                    # against a simulated analyzer
.build/debug/rigx-probe ports
.build/debug/rigx-probe identify /dev/cu.usbserial-3130 --baud 115200
.build/debug/rigx-probe sweep /dev/cu.usbserial-3130 1 30 200 --out antenna.s1p
.build/debug/rigx-probe calibrate /dev/cu.usbserial-3130 0.1 30 501 --out cal.json
.build/debug/rigx-probe cable Examples/rg174-1.01m-shorted.s1p --length 1.01

make_app.sh builds the bundle around the executable SwiftPM produces; there is no .xcodeproj. Xcode still opens Package.swift directly. When xcode-select points at the Command Line Tools the script borrows Xcode's toolchain through DEVELOPER_DIR, so it needs no sudo.

Structure

module	contents
RigXCore	Frequency, Impedance, Reflection, Sweep, DeviceProfile, Calibration, BandPlan, cable and sweep analysis
RigXIO	Touchstone .s1p
RigXTransport	the AA protocol, AnalyzerSession, SerialChannel, and a simulated analyzer
RigXSwiftApp	the SwiftUI app
rigx-probe	command line client

Reflection is the hub: SWR, return loss, ρ, phase and the Smith coordinate are all functions of Γ, so they derive from one type rather than being recomputed in six places.

Frequencies are a type, never a bare Double. The original passes numbers whose unit depends on the layer — megahertz on the wire, kilohertz in the device table, whatever the header says in a Touchstone file — and that is a whole class of bug that simply cannot occur here.

ByteChannel is a protocol, which is why the entire session can be tested with no hardware attached: SimulatedAnalyzerChannel answers the real command vocabulary from memory, in chunks that deliberately fall in the middle of lines. A network bridge would be another implementation and nothing above it would change.

Hardware

The AA-30.ZERO has no USB of its own: it needs 5 V, ground and a 3.3 V UART at 38400 baud. Arduino/RigXBridge carries those bytes to the Mac. Stack the shield on an Arduino UNO R3, upload, and the analyzer appears as an ordinary serial port. The Mac side of the bridge runs at 115200, which is why --baud 115200 appears in the examples above.

Other Arduino boards are not covered: the analyzer's UART lands on D4 and D7, and whether a board can receive on those pins depends on its core rather than on its speed.

Known limitations

• Italian and English only, switchable from the globe in the toolbar
• The app loads a calibration but cannot capture one — that flow is on the command line
• The Smith chart shows the cursor but cannot set it; use the SWR or R/X view
• TDR is marked experimental: the transform is validated but new
• Comparison overlays appear on the SWR chart; the other views draw the selected trace
• The app's views have no tests. Everything below them does

Author

Andrea Pede, IZ0TWS.

Provenance

The RigExpert AA protocol, the analyzer capability table and the OSL correction were derived by reading AntScope2, which is MIT-licensed; its copyright notice is retained in LICENSE. The three .s1p fixtures under Tests/RigXIOTests/Fixtures are that project's own files, used here only to exercise the parser — they are not usable calibration standards, all three read close to an open.

Reading that source also turned up bugs worth knowing about if you use the original: parallel reactance is wrong on the calibration-corrected path, the Touchstone parser rejects the specification's own lowercase kHz, and an unguarded Windows registry write leaves a stray file inside the macOS app bundle.