RF safety quick reference

Use this page as a pre-job checklist when you need to go from “we have an RF source” to “we have a guideline-comparable number, with assumptions, uncertainty, and controls.”

---

1) The compliance chain (6 lines)

1. Characterize the source (bands, max credible power, duty/time behavior, antennas, beam states).
2. Define accessible scenarios (public/uncontrolled vs occupational/controlled, per location).
3. Classify region (near field vs far field, per location; note reflective boundaries).
4. Select the metric (SAR vs E/H/S vs incident/absorbed/epithelial power density vs induced metrics).
5. Select a method (screening calc, broadband survey, selective, modeling) and plan averaging + uncertainty.
6. Report + controls (results, uncertainty, limitations, boundaries/signage, procedures, change control).

---

2) Near field vs far field: quick check you can document

You need three inputs:

• frequency f
• wavelength λ = c/f (with c ≈ 3 × 10⁸ m/s)
• antenna/aperture size D (largest dimension of the radiating structure)

For electrically large apertures in free space, two common screening boundaries are:

Interpretation you can write into a report:

• If R ≫ R_far: plane-wave assumptions are more plausible (still check reflections).
• If R ≪ R_reactive: treat as strongly near-field; avoid plane-wave conversions.
• If between: radiating near field → expect spatial variability; use scan-based methods and/or modeling.

---

3) Metric selection: which “family” applies?

Use this as a category check (always confirm with your adopted framework/regulation).

Deep-dive for the >~6 GHz transition: AB. Above ~6 GHz metrics

---

4) Averaging rules (the ones that prevent most mistakes)

Rule 1: power-like quantities average linearly

If the endpoint is power density or SAR, average that quantity over time:

For constant on-state exposure, zero off-state exposure and complete periods within the averaging interval, duty cycle D gives:

Rule 2: field amplitudes generally scale with square root of D (when plane-wave-like)

If you only have far-field E and you model duty cycle:

Rule 3: never average in dB

Convert to linear units, average/combine, then convert back if needed.

Rule 4: if the metric is an area/mass average, a point peak is not the comparison quantity

If your framework defines averaging over a mass/area, your method has to compute that average (typically after you locate the maximum region).

---

5) Method escalation (broadband → selective → modeling)

Use this decision tree when you’re choosing tools.

1. Do you need band attribution or frequency-dependent criteria?

   • Yes → go frequency-selective (or a method standard that handles summation).
   • No → broadband can be appropriate for screening/mapping.

2. Is the point plausibly near-field or highly reflective?

   • Yes → avoid “convert E to S” shortcuts; use scan-based sampling, consider measuring both E and H where relevant, and consider modeling if you need internal metrics.
   • No → far-field-style E/H/S methods may be defensible.

3. Are you close to a threshold once uncertainty is considered?

   • Yes → tighten geometry control, expand spatial sampling, move from broadband to selective, and/or refine via modeling.
   • No → screening may be sufficient (as documented).

4. Does anything look “non-physical”? (flat distance scaling, stuck readings, broadband ≫ selective)

   • Yes → suspect overload/compression or method mismatch before trusting the number.

---

6) What to record (minimum, audit-friendly)

• Framework/regulatory basis + edition/revision (and why it applies).
• Clause/table pointer(s) for the limit values and multi-source summation/averaging rules you applied.
• Access classification per location (public/uncontrolled vs occupational/controlled) and the controls that make it true.
• Instrument/probe/chain model + serial + calibration date/due date; detector mode and averaging settings.
• Geometry: heights, stand-offs, distances, photos/sketches; reflective boundaries.
• Time behavior: operating state assumptions (worst-case vs typical), duty/traffic model, window length.
• Multi-source handling: inventory + summation method (broadband screening vs selective breakdown).
• Uncertainty: at least the dominant contributors (calibration, positioning/gradient, time variability).

---

7) Knowledge check

Check your understanding 1

If a source is bursty and the guideline quantity is time-averaged power density, what do you average?

Show answer

Average power density over the defined time window. In a justified plane-wave air field, this is equivalent to averaging E² and dividing by 377 Ω. Averaging E directly gives the wrong power average; averaging SAR would answer a different metric question.

Check your understanding 2

When is S = E²/377 a reasonable conversion?

Show answer

In plane-wave-like far-field conditions where the environment doesn’t strongly violate plane-wave assumptions. In the near field (including the radiating near field of large apertures/arrays), E and H can be decoupled and the conversion can be wrong.

Check your understanding 3

A broadband meter reads high but selective results disagree beyond expected uncertainty. Which measurement checks should come before attributing the difference to exposure?

Show answer

Check overload and linear operation, then whether both methods cover the same bands and use comparable time averaging and operating states. Verify selective settings and chain corrections. A discrepancy is a reason to investigate, not proof of high exposure or overload.

---

Exercises

Check your understanding 4

Exercise 1 — Region classification. An aperture has D = 0.6 m and operates at f = 3.5 GHz. Estimate R_far ≈ 2D²/λ.

Show answer

λ ≈ 3 × 10⁸/3.5 × 10⁹ ≈ 0.0857 m.
R_far ≈ 2(0.6²)/0.0857 ≈ 8.4 m.

Check your understanding 5

Constant Sₒₙ = 20 W/m² occurs during 10% of complete pulse periods, with zero between pulses. Estimate the period-averaged S bar.

Show answer

S bar ≈ D Sₒₙ = 2 W/m².