What RF levels have been measured near 5G and other cell sites
Published surveys near cell sites keep landing far below the exposure limits. An ARPANSA survey of 50 Melbourne locations in 2022 peaked at 0.14% of the Australian limit, and a 2025 UK Health Security Agency study of 56 5G macro base stations stayed at 5% or less of the international reference level even in a worst-case calculation. In these studies the high numbers appear only right at the antenna: a worst-case estimate 1.3 m from a Belgian 5G small cell reached 68% of the ICNIRP level, and FCC agents measured up to about 300% of the US limit twelve inches in front of rooftop panels in Phoenix.
How compliance is judged in the US
The FCC's limits for the public are maximum permissible exposure values in 47 CFR 1.1310. Above 1.5 GHz the general-population limit is 1.0 mW/cm², which is 10 W/m². Between 300 and 1,500 MHz it is frequency in MHz divided by 1,500, so 0.6 mW/cm² at 900 MHz. Two details in the FCC's OET Bulletin 65 matter for anyone reading a survey:
- The limits apply to fields "spatially averaged over the body dimensions," so one hot spot at head height is not by itself a violation if the average over the body is under the limit.
- For the public, the averaging time is 30 minutes, but the bulletin notes it "is often necessary to assume continuous exposure" because nobody controls how long a passer-by stands there.
A licensee's paperwork follows 47 CFR 1.1307(b). The applicant either shows the source is exempt, prepares an evaluation against section 1.1310, or files an environmental assessment if people would be exposed above the limits. Where several transmitters share a site, every licensee whose signal exceeds 5% of the limit in the area in question shares responsibility for bringing it into compliance.
FCC or ICNIRP: which limit a percentage refers to
Four of the six measurements below come from outside the US and report results against the ICNIRP guidelines or a national standard built on them. The two sets of numbers are close but not identical, so a percentage only means something once you know which limit sits under it.
| Frequency | FCC public limit | ICNIRP public reference level | FCC limit divided by ICNIRP |
|---|---|---|---|
| 617 MHz | 4.11 W/m² | 3.09 W/m² | 1.33 |
| 900 MHz | 6.0 W/m² | 4.5 W/m² | 1.33 |
| 1,800 MHz | 10 W/m² | 9.0 W/m² | 1.11 |
| 3,500 MHz | 10 W/m² | 10 W/m² | 1.00 |
| 28 GHz | 10 W/m² | 10 W/m² | 1.00 |
The ICNIRP values use frequency divided by 200 (in W/m²) up to 2 GHz and 10 W/m² above that. The 1998 guidelines and the 2020 guidelines give the same whole-body power density values for the public at these frequencies. Below 2 GHz the FCC number is the looser of the two, so a reading equal to 1% of the ICNIRP level at 900 MHz is 0.75% of the FCC limit. At 3.5 GHz, the band of the Belgian small cells below, the two limits are the same.
The studies side by side
These are the measurements we could check against the published paper, its abstract or the FCC record itself. Each value is the highest one the authors report, and the labels say whether it was measured or extrapolated to a worst case.
Limits differ by study. Aerts, UKHSA and ARPANSA used ICNIRP-based levels; Bushberg used the FCC limit. Not shown: an FCC rooftop inspection in 2015 that found about 175% to 300% of the FCC limit twelve inches from antenna panels.
| Study | Percent of limit |
|---|---|
| Aerts 2023, Belgian 5G small cell, worst case at 1.3 m | 68% |
| UKHSA 2025, 56 UK 5G macro sites, worst case | 5% or less |
| ARPANSA 2006, 60 Australian base stations, measured | 0.2% |
| ARPANSA 2023, 50 Melbourne locations, measured | 0.14% |
| Bushberg 2025, US millimeter-wave small sites, measured | 0.0082% |
Source: Peer-reviewed papers listed in the table below.
| Study | Where and when | What was measured | Limit used | Highest | Other results |
|---|---|---|---|---|---|
| Henderson and Bangay, ARPANSA, 2006 | Five Australian cities | 60 base stations, 50 to 500 m away; CDMA800, GSM900, GSM1800 and 3G signals | ICNIRP and ARPANSA RPS3, general public | 0.2% (7.8 mW/m² from one base station) | Abstract gives no typical value |
| Henderson, Bhatt and Loughran, ARPANSA, 2023 | Melbourne, March 3 to April 27, 2022 | 50 public locations at 1.5 m height, all sources from 100 kHz to 6 GHz | ARPANSA RPS S-1, based on ICNIRP 2020 | 0.14% (2.85 mW/m²) | Median 0.0011%; 5G ranked below 4G and 3G; 26 GHz not measured |
| Calderon, Addison and Peyman, UKHSA, 2025 | 30 UK locations | 56 active 5G macro base stations in line of sight, 420 MHz to 6 GHz | ICNIRP 1998 public | 5% or less, worst-case extrapolation | 4G downlink was the largest source; 5G under 10% of the total on average |
| Aerts and colleagues, 2023 | Two 5G test sites, Belgium | Small cells in the 3.4 to 3.8 GHz band, radios about 4.5 and 5.5 m high, 0.5 to 100 m away | ICNIRP general public, 61.4 V/m | 68% worst case, 1.3 m from the antenna | People not using the network could get 5 to 30 times less at the beamforming site |
| Bushberg and Butcher, 2025 | US urban and suburban sites | Millimeter-wave small base stations, horn antenna and spectrum analyzers | FCC public limit | 0.0082% | Lowest 0.0003%; abstract does not give a site count |
| FCC Enforcement Bureau, 2015 | Phoenix office rooftop, January 14 and 15, 2015 | Twelve inches in front of carrier antenna panels | FCC general population | About 300% (298% on one panel) | Lowest about 175%; barriers broken, signs unclear |
One check on the Belgian figure. The authors computed their ratio as the square of measured field over 61.4 V/m. That reference level equals 10 W/m², which is also the FCC limit at 3.5 GHz, so the 68% would read the same against the US limit.
The units trip people up, so here is the Melbourne peak worked through. 2.85 mW/m² is 0.00285 W/m², and the paper puts that at 0.14% of the limit, which it describes as more than 700 times below it. The UKHSA ceiling of 5% means at least 20 times below. The US millimeter-wave peak of 0.0082% is roughly 12,000 times below the FCC figure. A percentage of the limit is a ratio of power, so halving the electric field strength cuts the percentage to a quarter.
Why the numbers run from 0.0082% to 68%
Distance and position explain most of it. The Belgian team measured as close as 0.5 m to radios mounted 4.5 to 5.5 m up, raised its probe to between 2.2 and 5.5 m for some readings, and reported the 68% figure for a point 1.3 m from the antenna. The Melbourne survey measured at 1.5 m above public ground, and the 2006 Australian survey stood 50 to 500 m from the base stations. Power density falls off fast with distance from an antenna: the FCC bulletin's far-field formula has power density dropping with the square of the distance.
The method matters too. A worst-case extrapolation assumes the site transmits at full capacity toward the point being measured. The UKHSA team and the Belgian team both report that kind of number, which is one reason theirs sit above the 2022 Melbourne readings, one-minute averages of whatever the network happened to be sending. The Belgian paper itself says people who are not using the network could see far less than its worst case.
Then there is what got measured. The Melbourne survey summed every source between 100 kHz and 6 GHz at each spot and still found the highest total at 0.14%. The US study looked only at millimeter-wave signals near small base stations.
What these measurements do not settle
They show compliance with today's limits at the places measured. They do not show what the limits should be. That question is still open in the US. In 2021 the D.C. Circuit sent part of the FCC's 2019 decision back, asking for a reasoned explanation that its limits protect against harms "unrelated to cancer," including long-term exposure and effects on children. On September 17, 2026, the FCC asked for fresh comment on those issues in Public Notice DA 26-997, saying the court took "no position in the scientific debate."
There are smaller gaps as well. Three of the six entries rest on published abstracts, and two of those leave out a site count or a typical value. The Melbourne survey did not include the 26 GHz millimeter-wave band, and none of these studies tracks one person's exposure over a day. The UK government's 5G health page summarizes the UKHSA work as showing exposure "well below the guidelines set by the ICNIRP," which matches the paper; the page itself gives no percentages.
Getting a reading for your own street
You have two routes. One is paperwork: ask the city or the carrier for the RF evaluation filed for the site. The other is a meter. OET Bulletin 65 notes that broadband probes can show the combined field "as a percentage of the applicable MPEs," while narrowband instruments such as spectrum analyzers look at one slice of frequency at a time, which is how you tell one transmitter from another. A reading taken a few inches from a pole-mounted radio will not match a reading on your porch, and neither one is a 30-minute body average.
Readers who want more than a published survey can write to the city or borrow a meter. Some shop at an EMF sticker store; EMFCALM phone stickers are sold in packs for several devices. This page reports measurements and leaves product questions to sellers.
If you measure, write down the distance, the height and the time of day next to the number. Without those three, a percentage of the limit tells you very little about the site.