🌐 English (this page) · Dansk: Brugervejledning
A complete reference for ESG-SignalCreator, a Windows application that builds arbitrary I/Q waveforms, plays them on an Agilent/Keysight E4438C ESG vector signal generator, and (optionally) verifies the result on a VSA (an Agilent E4406A or a Keysight N9010A) — with an in-app Claude assistant that can drive the whole flow in natural language.
This document explains what the app does and how every feature works. For step-by-step walkthroughs, see Tutorials.md.
ESG-SignalCreator turns intent ("a 4-tone signal at 1 GHz, −10 dBm, lowest PAPR") into a real RF signal coming out of the generator, and can prove the RF matches the intent by measuring it on the analyzer. It is organised around four ideas:
- Signal personalities — pluggable sources that each produce a normalized baseband I/Q waveform (CW, Multitone, Multi-Carrier, Custom Digital Modulation, AWGN, Import I/Q).
- A deliberate pipeline — you Calculate the waveform on the PC, Download it to the generator's ARB memory, then Play it (arm the ARB + turn on RF). Each stage is explicit so you always know what is on the instrument.
- Closed-loop verification — with a VSA (E4406A or N9010A) connected to the generator's RF output, the app measures the played signal (channel power, PAPR, tone frequency, ACP, …) and compares expected vs measured.
- An assistant — an opt-in pane where Claude uses the same operations you do, through a guarded tool surface (nothing reaches the DAC or RF without your explicit approval).
Everything except live instrument I/O works offline: you can build, preview, validate, and save waveforms with no instrument attached.
The app is C#/.NET Framework 4.7.2 / WinForms. Launch the modern shell (StudioForm) normally, or pass
--classic on the command line for the original single-window UI (MainForm).
- I/Q waveform — a complex baseband signal, a stream of (I, Q) sample pairs at a sample clock (sample rate). The ESG's ARB plays these samples to reconstruct the modulated RF carrier.
- ARB (arbitrary waveform generator) — the ESG option (001/601 or 002/602) that stores and replays I/Q samples from "WFM1" volatile memory. Requires a baseband generator option.
- PAPR / crest factor — peak-to-average power ratio (dB). High PAPR (e.g. noise, OFDM-like multitone) stresses amplifiers; low-PAPR phasing (Newman) packs the same tones with smaller peaks.
- Runtime scaling — the ARB plays samples at a fraction of full scale (default 70%) to leave DAC headroom and avoid over-range clipping.
- Sample granularity / minimum samples — the ESG requires the waveform length to meet a minimum (≈60 samples) and a granularity; the validator checks this.
- Seamless loop — a played waveform loops; if the end doesn't line up with the start there is a discontinuity ("seam"). Integer-cycle lengths loop cleanly.
- VISA resource — the address string for an instrument, e.g.
TCPIP0::192.168.1.82::inst1::INSTR(LAN) orGPIB0::17::INSTR(GPIB). The app uses any installed VISA provider to open it.
- Windows with .NET Framework 4.7.2 or later.
- For live instrument control, any IVI-compliant VISA runtime installed — Keysight IO
Libraries Suite, NI-VISA, Rohde & Schwarz, Rigol, etc. The app talks to instruments through the
vendor-neutral IVI
GlobalResourceManager, so whichever provider is installed is used automatically (for TCPIP/LAN, GPIB, USB, and serial resources). - Install from the MSI on the project's Releases page (per-machine, x64; Start-menu + optional desktop shortcuts; clean uninstall). The installer enforces .NET 4.7.2 and detects a VISA runtime. See Packaging.md for build/CI details.
- On-instrument playback needs an E4438C with a baseband/ARB option (001/601 or 002/602). Verification needs a VSA (E4406A or N9010A) on the generator's RF output.
The shell is split into four regions:
| Button | What it does |
|---|---|
| Connect… | Open the connection manager and connect to the ESG (VISA resource or GPIB board/address). |
| Connect VSA… | Connect the analyzer (E4406A or N9010A, per the VSA model toggle), including the RF-path safety settings (§9). |
| Calculate | Generate the I/Q waveform from the current source + impairments (off the UI thread, with a progress bar). Updates the plots, validation and readout. No hardware. |
| Download | Encode the waveform and push it to the generator's ARB memory (WFM1). Requires a connection. |
| Play | Arm the ARB and turn RF on. |
| Stop | Disarm the ARB and turn RF off. |
| Verify | Closed-loop measure the played signal on the analyzer and show expected-vs-measured (§9). |
| Path cal… | Run the path-calibration wizard to capture cable loss + analyzer offset (§9). |
| Verify install… | Run the install self-test — a CW → AM → FM → I/Q battery measured on the analyzer (§9.7). |
| Reference | Lock the ESG and the analyzer to independent timebases or a common external 10 MHz. |
| VSA model | Toggle which analyzer the app targets — E4406A or N9010A (§9). |
| VSA Mode | Pick the analyzer measurement mode from the modes actually installed on the unit. |
| Calc → DL → Play | Run all three pipeline stages in sequence. |
| Save… / Open… | Save or load a project (.ssproj). |
The status strip at the bottom shows a status message, the Online/Offline indicator, and the connected instrument model.
Selecting a node shows its card in the centre:
- Source — pick a signal personality and edit its parameters.
- Impairments — optionally apply I/Q, AWGN, CFR and filter impairments.
- Sequence — build a sequence of waveform segments.
- Instrument settings — frequency, amplitude, RF/modulation, ARB sample clock, runtime scaling, with read-back.
- SCPI console — send raw SCPI and watch a timestamped log.
- Notifications — the validation/dependency-checker output.
- Verification — the Expected-vs-Measured table from the last Verify run.
- Assistant — the in-app Claude assistant pane.
- Three plot panes, each with a view selector: I/Q vs time, Spectrum (FFT), Constellation, CCDF, and Eye. Default views are IQ / Spectrum / Constellation; rubber-band zoom is supported.
- A results readout strip: sample count, sample clock, duration, download size (bytes), peak, RMS, PAPR, and 99% occupied bandwidth.
- A progress bar for Calculate.
- A play-state indicator (Idle / Busy / Waiting-for-trigger / Playing).
Open the Source view, choose a personality from the picker, and edit its parameters in the panel.
Each personality produces a WaveformModel (I/Q + sample rate) when you Calculate.
A single continuous-wave tone. Parameters: frequency offset from the carrier (Hz), amplitude (dBFS, 0 = full scale, negative = backed off), and starting phase (degrees). Useful as a reference and for frequency/level checks.
N equally-spaced tones. Parameters: number of tones, tone spacing (Hz, or auto), and phase strategy — Newman (minimizes PAPR), Random, or Zero (all phases aligned → high PAPR). A classic test for amplifier linearity and PAPR handling.
Several independently-placed carriers (each can be its own modulation), summed into one waveform — for multi-channel / multi-standard scenarios.
A digitally-modulated carrier. Parameters: modulation format (BPSK, QPSK, 8PSK, 16/64/256-QAM, MSK…), symbol rate (Hz), a pulse-shaping filter (RRC, RC, Gaussian, or none) with roll-off / BT (alpha), and a payload pattern (PN9/PN15/PN23, all-ones/zeros, random). Used for ACP/ACPR and modulation-quality work.
A radar-style pulse train (Signal Studio for Pulse Building, N7620A). Parameters: pulse width (s), nominal pulse repetition interval / PRI (s, ≥ pulse width), an optional raised-cosine rise/fall time (s, 0 = rectangular edges), a start delay (s), and the intra-pulse modulation:
- None (gated CW burst);
- FM formats — Linear FM chirp, Non-linear FM chirp (with a curvature 0…<1), and FM step (stepped frequency) — all using a swept bandwidth in Hz; FM/AM step use a step count;
- AM step (rising amplitude staircase);
- phase codes — BPSK and QPSK (a seedable pseudo-random code with a chip count + seed) and a Barker phase code (length 2/3/4/5/7/11/13);
- polyphase codes — Frank (order N, length N²) and P4 (configurable length).
PRI patterning selects how the gap between pulses is chosen: Fixed (the nominal PRI), Staggered (a repeating pattern of intervals), or Jittered (the nominal PRI ± a seeded random peak). Per-pulse offset tables (frequency in Hz, phase in degrees, power in dB) cycle across the train so successive pulses can be frequency-agile, phase-stepped, or power-staircased. A single pulse template is tiled to fill the waveform; a one-sample marker is emitted at each pulse start (handy as an ARB trigger/scope sync). Used for radar/EW receiver test and pulse-compression work. The remaining N7620A Option 205/206 features (custom envelope shapes, antenna-scan patterning, pattern nesting, CSV import/export, and scenario impairments) are not yet implemented (see issue #179).
Band-limited additive white Gaussian noise. Parameters: noise bandwidth (Hz), carrier-to-noise ratio (dB), and optional peak clipping. AWGN has a high crest factor (~10 dB) — a good headroom and CCDF test.
Load I/Q from a file — the app's equivalent of the Signal Studio Toolkit (N7622A) "bring your own I/Q"
workflow. Parameters: file path (you supply it), format (Auto-detect by extension, delimited
text CSV/TSV, raw interleaved little-endian Int16 .bin/.iq, Agilent/Keysight big-endian
Int16 .agt — the ESG's native ARB byte order — Agilent/Keysight big-endian 14-bit
(AgilentInt14Be; the 14-bit sample is left-justified in each 16-bit word, low 2 bits are markers),
16-bit PCM WAV, or MATLAB Level-5 .mat), source sample rate (Hz), optional I/Q swap,
and a scale multiplier. Set the format explicitly to force a byte order when an extension is
ambiguous. Lets you replay externally-captured or externally-generated (MATLAB/C++) signals.
MATLAB .mat (MAT File 5) — both uncompressed (save -v6) and the default zlib-compressed (v7)
files are read. The first suitable numeric array is used: a complex vector maps to I/Q, a real
2×N or N×2 array maps its two rows/columns to I and Q, and a real vector maps to I (Q = 0).
Marker / trigger authoring — attach a marker bit stream to the imported ARB segment: None, Start (a single marker at sample 0 — a segment trigger), Periodic (a marker every period samples), or Range (a marker block of length samples from start). The markers drive the ESG's EVENT/marker output on playback.
A dense multitone / noise-power-ratio (NPR) stimulus for amplifier and converter linearity testing (Signal Studio for Multitone Distortion, N7621B). Parameters: tone count (2 up to 4097), tone spacing (Hz), centre offset (Hz), a phase preset — Parabolic (Newman, low PAPR), Random, or Constant (aligned, high PAPR) — and an optional NPR notch (enable, width Hz, offset Hz from band centre) that clears a band of tones so you can measure intermodulation/noise falling into the notch. Composite bandwidth ≈ tone count × spacing; live PAPR is reported.
Per-tone tables. Optional per-tone magnitude (dB) and per-tone phase (degrees) tables cycle across the comb (tone k uses element k mod length), overriding the uniform power and the phase preset — use them to shape the comb or hand-tune the crest factor. Pre-distortion correction (enable) takes measured per-tone magnitude/phase error tables (from a signal analyzer, e.g. via the closed-loop Verify path) and subtracts them from the base per-tone values, pre-inverting the measured channel response so the emitted comb lands flat and IMD in the NPR notch is suppressed. Out-of-band notch-cancellation (injecting correction tones inside the notch) is not yet implemented.
A jittered clock/tone for receiver jitter-tolerance testing (Signal Studio for Jitter Injection, E4438C-SP1). Jitter is applied as timing (phase) modulation of a sinusoidal clock, so the envelope stays constant. Parameters: clock rate (Hz, the signal being jittered); periodic jitter shape (Sinusoidal / Square / Triangle / SawTooth / Exponential / Custom, or None), rate (Hz), and amplitude (UI peak-to-peak); and optional random Gaussian jitter (RMS in UI + seed). Periodic + random combine into a composite signal, and a given seed reproduces an identical sequence (digital repeatability). Amplitudes are in unit intervals (1 UI = one clock period).
Custom shape takes a one-period profile table (normalized to ±1), interpolated across the jitter period. SJ frequency sweep (enable) sweeps the sinusoidal-jitter frequency linearly or logarithmically from a start to a stop frequency across the record — the classic way to trace a jitter-tolerance mask. During a sweep the amplitude can follow a tolerance mask: a Custom mask (your frequency/UI breakpoints) or a predefined ITU-T G.8251 mask (OC-48 / OC-192 / OC-768). The predefined G.8251 corner values are approximate/representative — verify them against the standard before conformance use. Achievable-range enforcement rejects a clock or jitter frequency at/above Nyquist and, if a max jitter (UI pp) cap is set, a periodic amplitude above it.
A GSM-family carrier (Signal Studio for GSM/EDGE, N7602B) with two modulations:
- GMSK (GSM/GPRS) — the constant-envelope modulation: data bits drive an NRZ frequency pulse, Gaussian-filtered (BT, default 0.3) and integrated to continuous phase at modulation index 0.5.
- EDGE 8-PSK (v2, #186) — 3π/8-continuously-rotated 8-PSK (3 bits/symbol), pulse-shaped and peak-normalized (non-constant envelope, unlike GMSK). The pulse is a representative RRC stand-in for the linearised GMSK pulse.
Parameters: modulation, symbol rate (Hz, 270.833 ksym/s), samples per symbol, symbol count, BT (GMSK) / EDGE RRC β, Gaussian span, and data source. Sample rate = symbol rate × samples/symbol. Full burst/training-sequence framing and channel coding are not yet implemented.
A Bluetooth carrier (Signal Studio for Bluetooth, N7606B) with two modulations:
- GFSK (Basic Rate / LE) — Gaussian-filtered FSK (BT 0.5) integrated to continuous phase at a configurable modulation index (BR ≈ 0.32, LE ≈ 0.5); constant envelope.
- EDR (v2, #190) — differential π/4-DQPSK (2 Mbps,
Edr2Mbps) or 8-DPSK (3 Mbps,Edr3Mbps), pulse-shaped (RRC β 0.4) and peak-normalized; non-constant envelope. EDR keeps the 1 Msym/s symbol rate (2/3 bits per symbol).
Parameters: modulation, symbol rate (Hz; BR/LE-1M/EDR = 1 M, LE-2M = 2 M), samples per symbol, symbol count, GFSK modulation index / BT, EDR RRC β, and data source. GFSK shares the CPM engine with GMSK (GMSK is GFSK at index 0.5). The LE coded PHY, packet framing, and hopping are not yet implemented.
A W-CDMA downlink-style signal (Signal Studio for 3GPP W-CDMA FDD, N7600B). Data symbols are spread by an OVSF (Walsh) code, complex-scrambled by a PN sequence, and root-raised-cosine shaped (β = 0.22) at the 3.84 Mcps chip rate. Parameters: chip rate, samples per chip, symbol count, spreading factor (power of two) and OVSF index, modulation (QPSK…), RRC β, and scramble (enable + seed). Sample rate = chip rate × samples/ chip.
Multi-code composite (v2, #183): set code channel count > 1 to sum that many orthogonal OVSF code channels (codes 0…N-1, equal power) into a multi-code downlink — the composite is scrambled and shaped as one signal (raising the PAPR). Still representative — the named physical channels (CPICH/P-CCPCH/SCH), slot/frame structure, TFCI, transmit diversity, and HSPA channels are not yet implemented.
An HS-PDSCH-style HSPA signal (Signal Studio for 3GPP W-CDMA HSPA, E4438C-419). Same DSSS structure as §5.12 but defaulting to 16QAM on an SF-16 code — the defining HSPA feature is higher-order modulation on the high-speed shared channel. Parameters mirror W-CDMA FDD plus a modulation choice (QPSK / 16QAM / 64QAM). Multi-code (v2, #184): set code channel count > 1 to sum multiple HS-PDSCH multicodes (as HSDPA does) into the composite. Representative — HS-SCCH/ HS-DPCCH, E-DCH channels, H-ARQ, and TTI structure are not yet implemented.
A cdma2000 forward-link-style signal (Signal Studio for 3GPP2 CDMA, N7601B). Same DSSS structure as §5.12 at the 1.2288 Mcps chip rate: QPSK data spread by a Walsh code, PN-scrambled, pulse-shaped (RRC approximation of the cdma2000 filter). Parameters mirror W-CDMA FDD. Multi-code (v2, #185): set code channel count > 1 to sum several Walsh code channels into a multi-channel forward-link composite. Representative — pilot/sync/paging channels, radio configurations, and frame/PCG structure are not yet implemented, and the exact cdma2000 baseband filter is approximated by RRC.
A TD-SCDMA signal (Signal Studio for TD-SCDMA, N7612B) at the 1.28 Mcps low-chip-rate: QPSK/16QAM/64QAM data spread by an OVSF code, scrambled, RRC-shaped (β = 0.22). Parameters mirror W-CDMA FDD. Multi-code (v2, #187): set code channel count > 1 to sum several code channels within a timeslot into the composite. Representative — the 5 ms sub-frame / 7-timeslot TDD burst structure (DwPTS/UpPTS/GP), midamble codes, switching points, and HSDPA channels are not yet implemented.
A downlink OFDM signal with LTE numerology (Signal Studio for 3GPP LTE, N7624B). Uses a 15 kHz subcarrier spacing and the standard FFT size / occupied-subcarrier count for the selected channel bandwidth (1.4 / 3 / 5 / 10 / 15 / 20 MHz → 128…2048-point FFT, RB×12 subcarriers). Parameters: bandwidth, symbol count, and subcarrier/PDSCH modulation (QPSK / 16QAM / 64QAM / 256QAM). Two modes:
- Generic (default) — a plain OFDM data fill on all occupied subcarriers, for occupied-bandwidth / PAPR / spectral checks.
- Frame-structured (v2, #188) — enable frame structured to build a proper E-UTRA downlink radio-frame: a 10 ms frame of 0.5 ms slots with per-symbol cyclic prefix (Normal = 7 symbols/slot, or Extended = 6), the PSS (Zadoff-Chu) and SSS synchronisation signals on their standard symbols (central 62 subcarriers of slots 0 and 10), CRS cell-specific reference signals (antenna port 0) at the standard positions, and a PDSCH data fill on the rest — all driven by the physical cell ID (0…503). Set the subframe count (10 = one radio frame).
The frame-structured mode follows 3GPP TS 36.211 for the PSS/SSS/CRS sequences and positions; it is a representative frame, not fully conformant: single antenna port, no PBCH/PDCCH payloads or channel coding, no PDSCH scrambling. Uplink, MIMO, HARQ, and carrier aggregation remain out of scope.
The same LTE OFDM numerology and parameters as §5.16 — LTE's physical layer is common to FDD and TDD (Signal Studio for 3GPP LTE TDD, N7625B). Two modes, like FDD:
- Generic (default) — a plain downlink OFDM data fill.
- Frame-structured (v2, #189) — enable frame structured to build a proper E-UTRA TDD downlink frame (frame structure type 2): the D/S/U subframe pattern of the selected uplink-downlink configuration (0…6), the special subframe split into DwPTS / GP / UpPTS by the special-subframe configuration (0…9), TDD-positioned PSS (in the DwPTS of subframes 1 & 6) and SSS (last symbol of subframes 0 & 5), CRS, and a PDSCH fill on the downlink symbols. Uplink subframes and the GP/UpPTS are transmitted silent (the app is a downlink generator). Also set the cyclic prefix (Normal/Extended), physical cell ID (0…503), and subframe count.
Same scope caveats as FDD (§5.16): representative frame, single antenna port, no channel coding, and uplink physical channels / MIMO / HARQ / carrier aggregation remain deferred.
An 802.11 OFDM signal (Signal Studio for 802.11 WLAN, N7617B) with 11a/g/n numerology: 312.5 kHz subcarrier spacing, a 64-point (20 MHz) or 128-point (40 MHz) FFT with the standard used-subcarrier count and cyclic prefix. Parameters: bandwidth, symbol count, and subcarrier modulation (BPSK…256QAM). Two modes:
- Generic (default) — a plain OFDM data fill, for occupied-bandwidth / PAPR / spectral checks.
- Frame-structured (v2, #191, 20 MHz only) — enable frame structured to build a representative 802.11a/g PPDU: an optional L-LTF training preamble (two long training symbols + double guard interval) followed by data OFDM symbols that carry the four pilot subcarriers (±7, ±21) with the standard polarity, using a selectable guard interval (Long 0.8 µs / Short 0.4 µs, the 802.11n short-GI option).
The L-LTF sequence and pilot positions/polarity follow IEEE 802.11. It's a representative PPDU — the L-STF/L-SIG fields, channel coding/interleaving, MAC framing, MIMO, and 80/160 MHz (11ac/ax, which exceed the ESG) remain deferred.
A fixed-WiMAX (IEEE 802.16-2004) 256-FFT OFDM signal (Signal Studio for 802.16-2004, N7613A) with 200 used subcarriers. Parameters: channel bandwidth (Hz; sample rate ≈ bandwidth × 8/7), cyclic-prefix ratio (1/4, 1/8, 1/16, 1/32), symbol count, and modulation (BPSK…64QAM). Two modes:
- Generic (default) — a plain OFDM fill of the 200 used subcarriers.
- Frame-structured (v2, #192) — enable frame structured for the exact 256-FFT map with the eight pilot subcarriers (±13, ±38, ±63, ±88; BPSK, modulated by the standard 802.16 pilot PRBS), and an optional preamble symbol (energy on every 4th subcarrier → four identical quarter-symbol repetitions in time, the WiMAX-preamble hallmark).
Pilot positions and the pilot PRBS follow IEEE 802.16-2004. Representative frame — the exact two-symbol long preamble, FCH, DL/UL-MAP, DCD/UCD, RS-CC channel coding, and MAC framing remain deferred.
A mobile-WiMAX (IEEE 802.16e) scalable-OFDMA signal (Signal Studio for 802.16 WiMAX, N7615B) at the fixed 10.9375 kHz subcarrier spacing, with a selectable FFT size (128 / 512 / 1024 / 2048 for ≈ 1.25 / 5 / 10 / 20 MHz) so the sample rate scales with FFT. Parameters: FFT size, cyclic-prefix ratio, symbol count, and modulation (QPSK…64QAM). Two modes:
- Generic (default) — plain OFDM (no OFDMA subchannel permutation).
- Frame-structured (v2, #193) — enable frame structured for a DL-OFDMA-style frame: an optional preamble symbol (a BPSK PN carried on every 3rd used subcarrier — the OFDMA-preamble spacing) and data symbols with a DL-PUSC pilot pattern (two pilots per 14-subcarrier cluster, at cluster positions 4 and 8).
The preamble's every-3rd-subcarrier layout and the 14-subcarrier DL-PUSC cluster follow the 802.16e structure, but this is a representative frame — the exact preamble PN per IDcell, the full PUSC/FUSC/AMC subchannel permutation, FCH/DL-MAP/UL-MAP, MIMO, and CTC/CC coding remain deferred.
The DAB COFDM signal underlying Terrestrial-DMB (Signal Studio for T-DMB, N7616B). Keeps a 2.048 MHz signal bandwidth across all four transmission modes (I/II/III/IV), which set the FFT size (2048 / 512 / 256 / 1024), active carriers, and guard interval. Parameters: mode, symbol count, data source. Two modes:
- Generic (default) — a plain-QPSK OFDM fill (DQPSK approximated by QPSK).
- Frame-structured (v2, #195) — enable frame structured for a DAB transmission frame: the synchronisation channel (a null symbol of silence for frame sync, then a phase-reference symbol carrying a known per-carrier phase) followed by differentially-encoded DQPSK data symbols (each carrier's phase accumulates the QPSK data delta from the previous symbol).
The null + phase-reference symbols and the differential (DQPSK) encoding follow the DAB frame structure (ETSI EN 300 401). Representative — the exact phase-reference table, the FIC/MSC multiplex, the TII, and convolutional coding remain deferred.
A DVB-T COFDM signal for an 8 MHz channel (Signal Studio for Digital Video, N7623B). Parameters: transmission mode (2K = 2048-FFT / 8K = 8192-FFT), guard-interval ratio (1/4…1/32), symbol count, and modulation (QPSK / 16QAM / 64QAM). Sample rate is the DVB-T elementary rate (64/7 MHz). Two modes:
- Generic (default) — a plain OFDM data fill of the active carriers.
- Frame-structured (v2, #196) — enable frame structured to insert the standard DVB-T scattered
pilots: on each symbol the carriers where
k mod 12 == 3·(l mod 4)carry boosted (4/3) BPSK pilots, so the pilot pattern shifts by 3 subcarriers every symbol and repeats every 4 symbols. Pilot values come from the DVB-T reference PRBS (X¹¹ + X² + 1).
The scattered-pilot positions, 4/3 boosting, and reference PRBS follow ETSI EN 300 744. Representative — the continual/TPS carriers, PRBS energy dispersal, RS/convolutional coding, and MPEG-TS framing remain deferred, and the other digital-video standards (ISDB-T, ATSC 8VSB, DVB-C/S QAM, DTMB) are not implemented.
An analog FM broadcast signal (Signal Studio for Broadcast Radio, N7611B). The baseband multiplex is an audio test tone (mono), optionally with a 19 kHz stereo pilot and a 38 kHz DSB-SC stereo subcarrier, frequency-modulated onto the carrier. Parameters: audio-tone frequency, stereo on/off, peak deviation (75 kHz), sample rate, and length.
RDS (v2, #194): enable RDS to add the 57 kHz data subcarrier — a 1187.5 bps biphase (Manchester) data stream (differentially-encoded PRBS), DSB-SC on 57 kHz (3× the 19 kHz pilot), with a configurable RDS deviation (~2 kHz of the 75 kHz total). Representative — a single test tone rather than program audio, real RDS group content (PS/PTY/RT/…), pre-emphasis, and SCA are not implemented, and the digital broadcast formats (DAB/DAB+ → see §5.21 T-DMB; XM/HD Radio) are covered elsewhere or deferred.
The Impairments view applies optional, independently-toggled effects to the calculated waveform (in order) so you can model real-world imperfections or test correction:
- I/Q impairments — gain imbalance, quadrature (phase) error, and DC offset. (A gain imbalance produces a measurable image tone — see the verification tutorials.)
- AWGN — add noise to the signal at a set level.
- CFR (crest-factor reduction) — reduce PAPR (peak windowing/clipping) to test or emulate CFR.
- Filter — apply an additional FIR filter.
Each has a checkbox to enable it and a property grid to edit its settings. Impairments are applied during Calculate, after the source produces its baseband I/Q.
The deliberate three-stage flow keeps you in control of what reaches the hardware:
- Calculate — builds the I/Q off the UI thread (progress bar), applies any enabled impairments, then refreshes the three plots, runs validation (§8), and updates the readout. No hardware is touched, so this is always safe.
- Download — turns the ARB off, encodes the I/Q to interleaved 16-bit, two's-complement, big-endian samples, frames them as an IEEE-488.2 definite-length block, and writes them to the generator's volatile WFM1 memory. Requires a connection.
- Play / Stop — Play selects + arms the ARB segment (at the sample clock, with runtime scaling) and turns RF on, showing the play-state indicator; Stop disarms and turns RF off.
Calc → DL → Play runs all three in one click. The download size and target are shown in the readout / notifications.
After every Calculate the dependency checker reviews the waveform against the connected (or default) target capability profile and lists findings in Notifications with severities (Info / Warning / Error). It checks:
- Minimum samples and granularity (the ARB's length rules).
- Memory cap — does the waveform fit the installed baseband option's sample memory?
- Sample-clock and carrier limits — within the instrument's range?
- DAC over-range — would the samples clip at the chosen scaling?
- Loop seam — will the waveform loop seamlessly (integer-cycle length)? A discontinuity is a warning.
Errors should be resolved before downloading; the verification path and the assistant both re-run this checker as a safety gate before any hardware action.
With a VSA on the generator's RF output, the app becomes a closed-loop generate → measure → compare system. The analyzer only ever receives RF. Two analyzers are supported: the Agilent E4406A and the Keysight N9010A (EXA).
Choosing the analyzer. The VSA model toggle on the action bar (next to Connect VSA…)
selects which analyzer the app targets — E4406A or N9010A — and is remembered between sessions.
The selection drives the connect dialog's title, its default interface and address hint (the E4406A
defaults to GPIB, e.g. GPIB0::17::INSTR; the N9010A defaults to LAN/USB, e.g.
TCPIP0::<ip>::hislip0::INSTR), the per-model input-damage default, and the identity check —
connecting an instrument that doesn't match the selected model is refused. The N9010A is validated
against the Keysight X-Series manuals; the E4406A path is additionally hardware-validated. Confirm the
N9010A's max safe input against its data sheet before driving power (see below).
The N9010A runs periodic auto-alignments that can pause a measurement for seconds. The app waits these out using a Service-Request (SRQ) notification rather than a fixed timeout, so a measurement that coincides with an alignment completes normally instead of failing spuriously.
Connect VSA… opens the VSA connection form, which includes the RF-path safety settings:
- Armed — turn this on when the analyzer is physically on the ESG output and must be protected.
- Analyzer max safe input (dBm) — the damage threshold, seeded from the selected model (E4406A type-N input ≈ +35 dBm, default gate +30 dBm; N9010A +30 dBm / 1 W max safe input per its data sheet, 5989-6529EN). Override it for your unit.
- Path loss (dB) — any inline pad/attenuator between the ESG and the analyzer.
When armed, the power safety gate blocks any commanded ESG power that would put more than the safe level at the analyzer input (accounting for path loss). This gate guards both the manual UI and the assistant.
Verify measures the played signal and compares it to expectations, populating the Verification table (metric, expected, measured, Δ, tolerance, pass/fail, with a summary):
- Channel power vs the commanded level minus path loss.
- PAPR vs the value computed from the generated I/Q.
- Tone frequency (for a single tone) vs carrier + offset.
Path cal… drives a clean unmodulated carrier at a known level, measures it on the analyzer, and records commanded − measured as the inline path loss — applied to both the safety gate and Verify so subsequent runs are self-consistent. RF is returned off when done.
The Reference menu sets the ESG and the analyzer to independent internal timebases or to a common 10 MHz external reference (a house reference or the ESG's 10 MHz OUT cabled to the analyzer) for clean frequency comparisons. It reports the resulting source of each instrument.
The VSA Mode menu lists the measurement modes actually installed on the unit (read live from the
analyzer's :INSTrument:CATalog?): always Basic, plus any option-gated communications-standard
personalities (GSM, EDGE, cdmaOne, cdma2000, 1xEV-DO, W-CDMA, NADC, PDC, iDEN). Selecting one switches
the analyzer's mode; an uninstalled mode is refused with a message that lists what is installed
(see §9.8).
Under the hood the app provides typed VSA measurements (also exposed to the assistant, §10): Channel Power, ACP/ACPR, CCDF / PAPR (Power Statistics), Spectrum marker (tone frequency/power, occupied BW), Waveform (time-domain peak/mean/peak-to-mean), and Power-vs-Time with a configurable power mask (pass/fail over time windows) for bursted signals.
Verify install… runs a short, guided generate → play → measure → compare battery that proves the whole install and configuration works end-to-end, across signal types rather than one. It synthesizes four signals as ARB I/Q and plays each through the ESG, measuring each on the connected analyzer:
- CW — an unmodulated tone; checks channel power, PAPR (≈ 0 dB), and tone frequency (carrier + offset).
- AM — 50% at 100 kHz; the elevated PAPR fingerprints the amplitude path.
- FM — 500 kHz deviation at 100 kHz; the constant-envelope PAPR (≈ 0 dB) fingerprints the frequency path.
- I/Q multitone — a 4-tone Newman signal; channel power + PAPR for the full complex path.
Results appear in the Verification view (expected vs measured per step) with an overall PASS/FAIL. It needs a baseband-capable ESG and a connected analyzer; the input-damage safety gate is enforced before any RF, and RF is returned off when done. AM/FM are verified via power/PAPR (not analog demod). On a FAIL, a troubleshooting dialog lists each failed check's likely cause and ordered fixes (e.g. excessive AM → over-driven ESG or analyzer mis-read → lower level / check ARB scaling / re-run Path cal…).
To run the same battery by hand — reading the analyzer yourself, with every UI control, value and expected reading spelled out and a standalone VSA settings checklist — follow the Manual Verification Procedure.
The app binds to what the connected unit actually reports, not to a fixed model configuration, so it
never offers a personality the hardware can't run or accepts a setting the instrument would silently
reject. On connect it reads *IDN?, *OPT?, and the live ? MAX/MIN limits and reconciles them against
the static profile (the effective profile). Two tiers:
- Core (always present). Functions available on any supported unit regardless of options:
- ESG (E4438C): ARB waveform download & playback, frequency/amplitude/RF-output control, reference locking. (Baseband ARB itself requires an installed baseband generator option — see below.)
- Analyzer (E4406A): Basic measurement mode. (N9010A): SA and IQ Analyzer modes. Channel Power, CCDF/PAPR, Spectrum marker, and Waveform measurements run in these core modes.
- Option-gated (present only if the unit reports the option).
- ESG: baseband generator / ARB memory depth — the reconciled sample count and sample-clock
ceilings reflect only the baseband options
*OPT?actually reports, and the download path reads back*OPC?+:SYSTem:ERRor?so a rejected waveform is surfaced, not assumed loaded. - Analyzer: communications-standard personalities (GSM, EDGE, cdmaOne, cdma2000, 1xEV-DO, W-CDMA, NADC, PDC, iDEN). These appear in the VSA Mode menu only when installed.
- ESG: baseband generator / ARB memory depth — the reconciled sample count and sample-clock
ceilings reflect only the baseband options
If you select a mode that isn't installed, the app refuses it with a clear message (naming the installed modes) rather than relying on a silent instrument-side rejection. Connecting a model the app doesn't support (anything other than the selected E4406A/N9010A analyzer or the E4438C ESG) is refused at connect.
The Assistant view is an opt-in pane where Claude drives the app in natural language through a tool surface — never synthetic clicks or hidden SCPI. It is off until you enable it and provide an API key.
- A transcript of your messages and Claude's replies (streamed live).
- An input box with Send and Stop (Stop cancels the in-flight turn).
- A settings strip: Enable assistant (master switch), Auto-approve hardware, Allow raw SCPI, and Set API key….
- Inline confirmation cards appear in the transcript whenever Claude wants to do something that touches the instrument — with the action, its parameters, and Approve / Decline.
- Read (run freely):
get_app_state,list_personalities,get_current_config,get_validation_results,get_results_readout. - Configure (project/PC state only):
set_source_personality,configure_cw,configure_multitone,configure_custom_modulation,configure_awgn,configure_import_iq,select_plot_view,set_project,calculate_waveform. - Hardware (each behind confirmation):
connect_instrument,disconnect_instrument,download_waveform,play_rf,stop_rf,set_instrument_settings. - Measure / verify (read the analyzer):
get_vsa_state,measure_channel_power,measure_acp,measure_ccdf,measure_spectrum_peak,measure_waveform,verify_signal. - Gated (opt-in):
send_raw_scpi— an advanced escape hatch, disabled until you tick Allow raw SCPI, and always confirmed per call.
Enforced in the dispatcher, not the prompt:
- read / configure run with no prompt; hardware / destructive require an Approve/Decline
card. Auto-approve hardware can skip the prompt for ordinary hardware tools, but
play_rf,connect_instrument, andsend_raw_scpialways confirm (RF emission, bus takeover, raw commands). - A pre-execution validation gate re-runs the dependency checker before
download_waveform/play_rfand refuses on a hard validation error — even if you approved. - Commanded power goes through the input-damage safety gate (§9.1).
- Instruction-source boundary: anything returned from a tool (a file's contents, an instrument response) is treated as data, never commands — Claude won't act on instructions hidden in tool output.
- The API key is stored encrypted with Windows DPAPI (per-user) and never written to projects, logs, or the request body. Bulk data (e.g. raw I/Q arrays) is minimized before being sent.
Reads issued in one turn run concurrently; configure/hardware steps stay serialized in order; long conversations are compacted automatically.
- Sequence view — assemble multiple waveform segments into a sequence (table + script views), with subsequences and batch compile, for multi-segment playback.
- Projects — Save… / Open… persist the active source + settings as a
*.ssprojJSON file. - Exports — export the waveform as raw ARB bytes, CSV, or a SCPI script; and use built-in test-model presets as starting points.
- Markers — ARB markers are supported for triggering/segmentation.
The SCPI console sends raw SCPI to the connected instrument and shows a timestamped request/response
log — useful for ad-hoc commands and debugging. (The assistant's equivalent, send_raw_scpi, is gated
and confirmed; see §10.)
ESG-SignalCreator.HilHarness.exe is a console runner for automated hardware tests (CI / bench
regression), separate from the GUI:
# ESG-only: connect, *IDN?/*OPT?, download a CW, arm the ARB, read back (RF off, safe)
ESG-SignalCreator.HilHarness.exe "TCPIP0::192.168.1.82::inst1::INSTR"
# Closed-loop battery across signal types + a frequency sweep, with a JSON report
ESG-SignalCreator.HilHarness.exe --vsa GPIB0::17::INSTR --all --dwell-seconds 3 --json report.json
# A single signal type, or a flatness power sweep
ESG-SignalCreator.HilHarness.exe --vsa --signal multitone
ESG-SignalCreator.HilHarness.exe --vsa --flatness
It enforces the input-damage safety gate, keeps the analyzer running during dwell so the front panel tracks live, exits non-zero on failure, and can emit a machine-readable JSON report.
- "Connect the ESG first" / Offline — connect via Connect… before Download/Play/Verify.
- Open fails / resource not found — confirm a VISA runtime is installed and the resource string is correct (try the SCPI console's discovery, or the connection manager's Find).
- Download/Play disabled — you need a calculated waveform and a connection; check Notifications for validation errors.
- Validation errors (memory/min-samples/over-range) — adjust length, sample clock, or scaling; the message names the limit. Hardware actions are refused while a hard error stands.
- Verify fails on power — run Path cal… so the path loss is captured, and confirm tolerances in the verification profile.
- Assistant says it's disabled / no key — tick Enable assistant and Set API key….
- Assistant won't touch hardware — that's by design; approve the inline card. If a hardware action is refused despite approval, a validation error is blocking it — fix it (see Notifications).
- CI release queued forever — the release workflow needs a self-hosted Windows runner with a VISA provider installed (see Packaging.md).
- The analyzer only ever receives RF. Arm the RF-path safety and set the max safe input + path loss before driving power into it; the gate then blocks unsafe levels.
- Play turns RF on. Use Stop to turn it off. The assistant always confirms
play_rf. - Treat raw SCPI (console or
send_raw_scpi) as the advanced, fully-logged escape hatch — it can do anything the instrument allows.
See Tutorials.md for hands-on, build-up walkthroughs of every feature.