Spectrum Analysis of Random RF Samples — City of Seattle, Washington

1. What “spectrum analysis” means in an urban RF context

In field RF work, spectrum analysis typically means estimating power spectral density (PSD) and related time–frequency products (e.g., spectrograms) from received electromagnetic energy, then computing derived metrics such as occupancy (duty-cycle / fraction of time above a threshold), peak lists, and channel utilization. Large-scale “spectrum observatory” systems implement this by repeatedly sweeping a receiver across wide frequency ranges, storing PSD samples, and deriving min/max/mean and occupancy views over time windows.¹

Microsoft’s Spectrum Observatory documentation describes a station architecture based on a wideband receiver (RFeye) measuring PSD from 30 MHz to 6 GHz, completing a full scan about every 3 seconds, and producing derived occupancy and power-density visualizations for specific locations including Seattle, WA (historical node) as one of the measurement sites.¹

2. A defensible “random sampling” design for Seattle

“Random samples within Seattle” should be interpreted as a probability-based spatial sampling plan inside the City of Seattle boundary, with explicit time-of-day coverage to capture diurnal variation (commuter peaks, business hours, nighttime quiet periods).

2.1 Recommended sample plan (practical)

• Sampling frame: City of Seattle polygon; exclude water if desired (or include shoreline/port if maritime RF is in scope).
• Design: Stratified random sampling by land-use class (downtown/commercial, residential, industrial/port, parks/hills) to reduce variance.
• Sample size: 20–60 sites is typical for a “city snapshot,” depending on equipment/time; repeat a subset weekly for trend detection.
• Time coverage: At each site, collect at least 2 time blocks (e.g., midday and evening) across multiple days.
• Measurement dwell: Wideband sweeps for 10–30 minutes per site per time block; add targeted narrowband captures where needed.

2.2 Instrumentation notes (urban constraints)

• Receiver: Spectrum analyzer or SDR + calibrated front end; avoid overload in downtown by using attenuation and pre-selection when available.
• Antennas: Discone (wideband VHF/UHF), plus band-optimized antennas for 700–2600 MHz and 2.4/5/6 GHz.
• Compliance: Limit work to energy measurements; do not attempt to demodulate or recover content. (Many signals are protected communications.)

3. Analysis workflow used for each sample

3.1 Wideband PSD estimation

For each sweep segment, compute PSD using FFT-based estimates (commonly Welch averaging for stability), and store PSD versus frequency and time. A “spectrum observatory” approach explicitly stores relative PSD samples and then aggregates them into min/max/mean and occupancy products across hours, days, and months.¹

3.2 Occupancy (energy-detection) metric

A standard approach is energy detection: mark a frequency bin “occupied” when PSD exceeds a threshold (e.g., noise-floor estimate + X dB). Occupancy is the fraction of time bins where the condition is true. The threshold choice materially affects results; Microsoft’s Spectrum Observatory documentation notes that measured values are intended to be relative and that occupancy inference requires careful study.¹

3.3 Contextual band interpretation

For Seattle, high-level “what you will usually see” in RF power surveys includes: broadcast FM (88–108 MHz), aeronautical VHF (108–137 MHz), marine VHF near the waterfront (channels including 156.8 MHz for channel 16), NOAA Weather Radio in the 162.4–162.55 MHz range, public-safety and other land-mobile allocations in UHF, cellular blocks in multiple bands, and dense unlicensed activity at 2.4 GHz and 5 GHz (and increasingly 6 GHz). Rules for common unlicensed regimes are defined in FCC Part 15 (e.g., §15.247 and §15.407).⁴⁵⁶⁷

4. Examples (illustrative) from “random” Seattle sample points

The table below shows example outputs you would produce from a Seattle campaign. Values are illustrative placeholders showing the structure of reporting (bands, observed peaks, and inferred occupancy classes). Replace with your measured outputs.

4.1 Example plot products you should generate for each sample

• PSD chart: mean PSD versus frequency with peak markers and labeled allocations.
• Spectrogram: waterfall (time vs frequency) for bands of interest (e.g., 2.4 GHz and 5 GHz).
• Occupancy chart: occupancy (%) by channel/band under a documented threshold policy.
• Peak list: top-N emissions with estimated center frequency, bandwidth class, and persistence.

5. How to interpret results responsibly

• Locality: RF is highly location-dependent (buildings, terrain, antenna height). Random sampling reduces bias but does not eliminate it.
• Time variance: Occupancy changes minute-to-minute; report both instantaneous and aggregated metrics.
• Threshold sensitivity: Document the occupancy threshold method; run sensitivity tests (e.g., +3, +6, +10 dB above noise-floor) to show robustness.
• Regulatory context: When interpreting dense unlicensed bands, note that permitted operation is defined by FCC rules (Part 15), including §15.247 (2.4 GHz ISM and 5.8 GHz ISM) and §15.407 (U‑NII).⁴⁵