SIGVIEW includes various analysis modes for AI signal reporting function, organized into two categories: general-purpose modes that work well with any type of signal, and domain-specific modes optimized for particular signal types such as vibration, audio, electrical power, and biomedical signals.


The list of modes is loaded from a configuration file and may be extended in future versions. Advanced users can also add custom modes by editing the configuration file (see "Extending the Analysis Modes" at the end of this topic).

General Modes

General Signal Description


A comprehensive overview covering all aspects of the signal: amplitude, spectral content, temporal behavior, and quality.


When to use: As your default starting point when you want an all-around understanding of what the signal contains. Good for documentation and reporting when no specific analysis focus is needed.


What you get: A structured description organized into sections covering signal overview, amplitude and distribution characteristics, spectral character including dominant frequency and harmonics, temporal behavior and stationarity, signal quality metrics, and two to three key observations highlighting the most notable characteristics. Typically 300 to 500 words.

Signal Quality Assessment


A focused diagnostic evaluation that rates the signal's overall quality and identifies specific issues with noise, distortion, dynamic range, and temporal stability.


When to use: When you need to determine whether a recording is suitable for further analysis, or when you suspect quality issues with a measurement.


What you get: An overall quality rating from Excellent to Unusable, detailed assessment of noise level based on signal-to-noise ratio, distortion evaluation using THD and THD+N, dynamic range analysis, temporal stability check, stationarity assessment, and two to four actionable recommendations tailored to the signal type. Typically 250 to 400 words.

Spectral Analysis Summary


A deep dive into the frequency-domain characteristics of the signal, including dominant frequencies, harmonic structure, octave band profiles, periodicity, and frequency stability over time.


When to use: When frequency content is your primary concern. This is the right mode for identifying tonal components, analyzing harmonic relationships, assessing effective bandwidth, understanding spectral energy distribution, or checking whether the dominant frequency is stable or varying over time.


What you get: Dominant frequency identification with source interpretation (musical note, motor RPM, or other), harmonic decay analysis with implications for the source type, spectral distribution interpretation using centroid and flatness, octave band energy profile, bandwidth assessment relative to Nyquist frequency, periodicity assessment from autocorrelation, and frequency stability analysis from the dominant frequency over time data. Typically 300 to 500 words.

Signal Classification


Attempts to identify what type of signal this is based purely on its measured characteristics, without relying on file name or metadata.


When to use: When you have an unknown or unlabeled signal and want to identify its likely source. Also useful for verifying that a signal matches its expected type, or for automated categorization of large signal collections.


What you get: A primary classification with confidence level (such as "machine vibration from rotating equipment, High confidence"), supporting evidence organized by metric category, one to two alternative classifications with explanations of why they are less likely, detailed signal characteristics regardless of classification, and notes on which metrics are most and least conclusive. Typically 300 to 500 words.

Anomaly Detection


Scans all metrics systematically for unusual values that may indicate problems with the signal, the recording equipment, or the source being measured.


When to use: As a quality check before detailed analysis, or when something seems unusual about a signal but you cannot pinpoint the issue. Also useful as a routine check in automated measurement workflows.


What you get: A list of detected anomalies, where each entry includes the specific metric that is unusual, its measured value, the expected range for that signal type, two to three possible causes, a severity rating of Low, Medium, or High, and a suggested action. The analysis checks for clipping and saturation, DC offset, noise issues, spectral anomalies, harmonic irregularities, temporal instabilities, Time-FFT anomalies, and stationarity concerns. If no anomalies are found, this is stated explicitly. Typically 200 to 500 words depending on the number of findings.

Executive Summary


A brief, non-technical summary designed for audiences who need to understand the key facts without signal processing expertise.


When to use: For inclusion in reports, presentations, or emails where the reader is a manager, client, or other non-specialist. Also useful when you need a quick, high-level overview of a signal.


What you get: A one-sentence summary capturing the essential nature of the signal, four to six key facts expressed in plain language without jargon, and a short suitability assessment. Maximum 200 words, written in a professional tone suitable for formal documentation.

Signal Comparison (usable only from Control Window)


Compares two or more signals side by side, highlighting differences and similarities across all metric categories.


When to use: When comparing before and after recordings, different operating conditions, different measurement channels, different equipment, or different time periods. Available from the Control Window when multiple signals are selected, or via manual usage with the clipboard prompt.


What you get: A structured comparison covering amplitude and dynamics, spectral differences described in plain language, quality comparison, temporal behavior differences, statistical profile comparison, a summary table of the eight to ten most important metrics side by side, and a conclusion summarizing the most significant differences. Typically 400 to 600 words.

Statistical Summary


A rigorous statistical characterization suitable for research papers, technical reports, or formal data documentation.


When to use: When you need publishable statistical descriptions, want to formally assess distribution properties, or need to quantitatively document stationarity and periodicity. Appropriate for academic or regulatory reporting contexts.


What you get: A dataset description, a descriptive statistics table covering central tendency, dispersion, shape, and extremes, distribution analysis with Gaussian fit assessment, frequency-domain statistics including Wiener entropy, temporal stationarity quantification using coefficient of variation and range ratio, and a periodicity assessment from autocorrelation. Typically 400 to 600 words with formatted tables.


Domain-Specific Modes

Predictive Maintenance Assessment


Assesses machine health based on vibration or operational data and identifies potential developing faults.


When to use: For vibration signals from accelerometers or velocity transducers mounted on rotating or reciprocating machinery. Best used as part of a regular condition monitoring program to detect developing problems before they cause failures.


What you get: A machine health rating of Good, Acceptable, Warning, or Critical with justification. Rotational speed estimate in RPM derived from the dominant frequency. Vibration level assessment including crest factor interpretation. Fault indicator analysis checking for imbalance, misalignment, bearing defects, gear mesh problems, mechanical looseness, resonance, and cavitation. Temporal trend analysis showing whether vibration levels are stable, increasing, or intermittent. Three to five maintenance recommendations with urgency levels (routine, schedule soon, or immediate action). Typically 300 to 500 words.


If the signal does not appear to be vibration data, the AI will note this and provide a general assessment instead.

Vibration Analysis


Detailed vibration engineering analysis following industry practices, with individual frequency component identification and severity assessment.


When to use: For in-depth vibration engineering work when you need to identify each significant frequency component and its likely mechanical source, assess harmonic relationships in detail, evaluate broadband versus narrowband energy distribution, and rate severity against standards such as ISO 10816.


What you get: A measurement summary with key parameters. A list of identified frequency components, each with its frequency in Hz and RPM, likely mechanical source (1x rotational, 2x misalignment, bearing frequency, gear mesh, structural resonance, and others), amplitude relative to the dominant component, and whether it belongs to a harmonic series. Harmonic decay rate analysis with interpretation. Broadband versus narrowband energy characterization using spectral flatness. Time-domain characteristics including crest factor, kurtosis, and skewness interpretation. Stationarity assessment. Severity rating with reference to applicable standards. Typically 400 to 600 words.


If the signal does not appear to be vibration data, the AI will note this and provide a general assessment instead.

Audio Signal Analysis


Analysis tailored for audio recordings, using audio-appropriate terminology and evaluation criteria.


When to use: For speech recordings, music, environmental sound recordings, test tones, or any audio content. Especially useful for assessing recording quality and identifying the characteristics of the audio content.


What you get: Audio content type identification (speech, music, ambient sound, or test tone) with confidence level. Frequency content described in audio terms: bass content from 20 to 250 Hz, midrange from 250 to 4000 Hz, treble and presence from 4 to 8 kHz, and high frequencies above 8 kHz. Spectral centroid interpreted as perceived brightness. Dynamic profile including crest factor interpretation for audio (compressed, typical, or very dynamic). Recording quality assessment checking for clipping, DC offset, bandwidth limitations, and noise. Pitch and tonality analysis with musical note identification when applicable. Temporal character description in audio terms (sustained, decaying, rhythmic, speech-like). Typically 300 to 500 words.


If the signal does not appear to be audio, the AI will note this and provide a general assessment instead.

Electrical / Power Signal Analysis


Analysis focused on power quality, harmonics, and grid-related characteristics for voltage or current measurements.


When to use: For signals from power system measurements, including voltage or current waveforms from the electrical grid, generators, motors, inverters, or power electronics equipment.


What you get: Fundamental frequency identification (50 Hz or 60 Hz system). Harmonic analysis with each harmonic mapped to typical load types: third harmonic from non-linear loads, fifth and seventh from six-pulse rectifiers, even harmonics indicating asymmetry. THD interpreted against IEEE 519 power quality standards. Waveform quality assessment including crest factor deviation from the ideal sine wave value of 1.414. Frequency stability evaluation for grid-connected systems. Noise and interference identification from unexpected octave band energy. Temporal RMS behavior interpreted as load changes, voltage sags, swells, or transient events. Overall power quality assessment with recommendations. Typically 300 to 500 words.


If the signal does not appear to be an electrical power signal, the AI will note this and provide a general assessment instead.

Biomedical Signal Analysis


Analysis tailored for physiological signals with medical terminology and clinically relevant interpretations.


When to use: For ECG, EEG, EMG, PPG, respiratory, or other physiological recordings. Useful for initial signal characterization and quality assessment before detailed clinical analysis.


What you get: Signal type identification (ECG, EEG, EMG, PPG, or respiratory) with confidence level and reasoning based on frequency range, periodicity, and statistical properties. Rate estimation: heart rate in beats per minute for ECG-like signals, breathing rate for respiratory signals, derived from autocorrelation periodicity. Rate variability assessment from dominant frequency over time statistics. Waveform morphology interpretation (skewness, kurtosis, and crest factor in the context of physiological waveforms). Frequency band analysis mapped to physiological ranges: below 0.5 Hz for baseline wander and motion artifacts, 0.5 to 4 Hz for cardiac and respiratory fundamentals, 4 to 30 Hz for higher harmonics and EEG bands, above 30 Hz for muscle artifacts and power line interference. Signal quality assessment for clinical use including SNR adequacy, artifact evidence, and stationarity. Processing and clinical recommendations. Typically 300 to 500 words.


Important: AI signal analysis is not a substitute for clinical diagnosis. The biomedical analysis mode provides signal characterization to support, not replace, professional medical interpretation.


If the signal does not appear to be biomedical in nature, the AI will note this and provide a general assessment instead.

Communication Signal Analysis


An RF-focused analysis that identifies the likely modulation type (AM, FM, FSK, PSK, QAM, OFDM, etc.), estimates carrier frequency and occupied bandwidth, and assesses signal quality from a communications perspective.


When to use: When the signal is (or may be) a captured communication waveform and you want to understand what kind of modulation is present, where the carrier sits, how much bandwidth it occupies, and whether the signal is clean enough to demodulate. Useful for spectrum monitoring, interference investigations, characterizing unknown transmissions, and sanity-checking recordings of known RF systems. Works best with broadband captures that include a carrier and its sidebands; for already-demodulated baseband signals, the mode will recognize this and adjust its analysis accordingly.


What you get: Sanity checks (is this actually a communication signal, is the carrier below Nyquist, is it passband or baseband), carrier frequency and occupied bandwidth estimates, a ranked list of plausible modulation types with confidence levels and the specific evidence supporting each, estimates of modulation parameters where derivable (AM modulation index, FM frequency deviation, FSK frequency shift, approximate symbol rate), RF quality assessment covering SNR, spurs, harmonics, spectral symmetry, compression and noise floor character, optional matching to known standards (AM/FM broadcast, cellular, Wi-Fi, Bluetooth, LoRa, DVB, etc.) with frequency and bandwidth comparisons, and a clear statement of what cannot be determined from aggregate metrics alone — notably that PSK and QAM cannot be reliably distinguished without I/Q phase data. Ends with recommended follow-up analyses in Sigview (spectrogram for burst signals, zoom-FFT on peak clusters, IQ capture for constellation analysis). Typically 400 to 700 words.

Extending the Analysis Modes


The analysis modes are defined in the file sigview_analysis_modes.json, located in SIGVIEW's Application Data folder. Advanced users can edit this file to modify existing prompts or add entirely new analysis modes. Changes take effect the next time SIGVIEW starts, and new modes automatically appear in the Analyze Signal and Copy Analysis Prompt to Clipboard submenus.


Each mode entry in the file contains four fields: id (a unique identifier used internally, with no spaces), label (the name shown in menus), description (a short tooltip description), and system_prompt (the full prompt text sent to the AI).


When creating custom prompts, keep in mind that the AI receives only the numerical metrics from the JSON report. It does not have access to the raw signal waveform. Effective prompts reference the specific metric names and description fields that appear in the report, and provide clear interpretation guidelines for the expected signal type.