Wi-Fi through walls: does 5 GHz really lose to 2.4 GHz?

Three bands, the same power out, antennas of the same gain at both ends. Add walls and change what they are made of. In this model 5 GHz starts below 2.4 GHz before any wall is in the way.

A 2020 home-network tip sheet from the US Federal Communications Commission gives the usual advice in two halves. The 5 GHz band is the faster one. The 2.4 GHz band generally covers more ground.

The second half is about reach, and reach is a subtraction. Power out, minus what distance takes, minus what each wall takes. The tip sheet does not say how many decibels each of those is.

This page does that subtraction. The controls below carry distance, wall count and wall material into one equation, worked for every band at once. Which band comes out on top is an output here, not an assumption.

The wall decides the direction, the count only how far

How many walls are in the way counts the walls. What the walls are made of picks one of five rows from a published table of building materials.

Start with no walls at all. The model already puts 5 GHz 7.07 dB below 2.4 GHz. That gap comes from the two centre frequencies under one rule of this model: every antenna has the same gain in both bands. The gap does not move with distance. So 5 GHz does not fall behind at some wall count. It starts behind.

Now add concrete. The model puts one 100 mm wall at 8.07 dB for 2.4 GHz. For 5 GHz it puts the same wall at 14.03 dB. Each wall then widens the gap by a further 5.96 dB. At every concrete thickness the slider allows, the two bands only move apart.

That is also why thicker is not always lossier. At 5 GHz the model puts a 5.5 mm pane of glass at 3.34 dB. An 11 mm pane comes out at 0.38 dB.

So glass can run the other way. At 11 mm the model puts each pane at 3.26 dB for 2.4 GHz and 0.38 dB for 5 GHz. Every pane closes the gap by 2.89 dB. After three panes the gap card flips to 5 GHz above 2.4 GHz and reads 1.59 dB.

At 12.5 mm, plasterboard closes it too, and barely. One sheet costs 1.09 dB at 2.4 GHz and 1.01 dB at 5 GHz, so each sheet narrows the gap by 0.08 dB. Eight sheets leave it at 6.43 dB, within a decibel of the empty room. A 100 mm brick wall widens it by 0.43 dB.

In this model the material and its thickness decide which way the gap moves, and the count decides how far. The chart above shows it. Here is how to read it.

One concrete wall, counted by hand

Load the Through one concrete wall preset. Leave the power and threshold sliders where they start: 20 dBm out, −82 dBm threshold. Then follow 2.4 GHz to the end.

Distance first. Ten metres is 0.01 km, so 20·log₁₀(0.01) = −40.0000, and 20·log₁₀(2437) = 67.7371. Add 32.4478 and the path gives 60.18 dB. Leave the distance in metres with this constant and the answer comes out 60 dB too large. It still looks plausible.

The wall next. At 2437 MHz the table gives η′ = 5.24. And σ = 0.0462 × 2.437^0.7822 = 0.0927 S/m. So η = 5.24 − j·0.684. The slab formula returns 8.07 dB at 100 mm.

Add the two and Total path loss reads 68.26 dB. The two rounded parts add to 68.25; the model rounds once, at the end. Take 68.26 dB from 20 dBm and the signal card reads −48.3 dBm. That is 33.7 dB above the threshold.

Now read the 5 GHz row, with nothing else changed. At 5500 MHz the free-space loss rises to 67.26 dB. The same wall rises to 14.03 dB. That is 81.29 dB in total, and the card reads −61.3 dBm.

The two readings differ by 13.03 dB. The figure above reaches the same number a second way, from the two frequencies plus the one wall. Two routes, one answer. Both of them rest on five rules.

Five rules, and the first is about antennas

Every figure the model produces follows these five. Drop any one and those figures stop meaning what they say.

1. Every band leaves the router with the same radiated power, and every antenna is isotropic, with the same gain in every band.
2. One straight path through empty space. No other path reaches the device, bounced or bent.
3. Each wall is one flat slab of a single material. The wave arrives square on.
4. Wall losses add. Two walls cost twice what one wall costs.
5. This is signal, never signal against noise.

No simulation stands behind the model's figures. Each one is a closed form, evaluated directly.

The material coefficients. ITU-R P.2040-4 (September 2025), § 3, Table 3, f in GHz. The fitted range sits beside each row.

Every b in these five rows is zero, so the rows report no trend in η′ with frequency. The band dependence sits in σ and in the wavelength alone. The model refuses a band outside a material's fitted range rather than stretch the fit.

The slab. Equations (43b) and (44) of the same recommendation, one layer, square-on arrival. To stand in for a stud partition, count its two sheets as two slabs; the cavity between them is not modelled. The default thicknesses are slider settings, not citations. Rule one is the one the next section takes apart.

A real router need not keep the first rule

Equal power and equal gain are rules here, not facts. A real router need not drive the two bands with the same power, or through antennas with the same gain. In the United States even the power limits differ between parts of the 5 GHz band. Say 5 GHz had to draw level through two 100 mm concrete walls. The model needs 18.99 dB more in the higher band, about 80 times the power.

Everything that makes an indoor signal indoor. The distance term assumes one clear path through empty space. Indoors, floors, ceilings and metal reflect, and the equation has no term for that. Antennas are isotropic, so pattern, orientation and the hand holding the phone are absent. So is interference. A crowded band looks the same here as an empty one, at 2.4 GHz and at 5 GHz alike.

The walls. Adding decibels per wall treats the walls as independent; the reflections bouncing between two walls are left out. Steel, foil-backed insulation, pipes and metal studs are not among this page's five materials. A slanted path crosses more material than a square-on one.

The spread inside each row. The table's rows are curve fits to more than ninety published material characteristics, gathered from eight sources, and the recommendation calls them examples. The About section below sets this slab beside two other published models, and for a concrete wall those two differ by almost 5 dB at 2.4 GHz and by more than 12 dB near 5 GHz. The formula under the distance term is older than Wi-Fi.

The formula underneath, written down in 1946

Harald Friis set out the formula under the distance term in 1946, in three pages of the Proceedings of the IRE. The first IEEE 802.11 wireless LAN standard followed in 1997, half a century later.

Friis wrote it with areas, not decibels. The received power is the transmitted power times the two antennas' effective areas, divided by the wavelength squared and the distance squared. Give both antennas the area of an isotropic antenna, λ²/4π, then take logarithms, and the distance term on this page falls out.

That substitution is where this page's 7.07 dB comes from. Hold both antennas' areas fixed instead and the same formula puts the higher band 7.07 dB above the lower one. So the answer to the title question depends on the antennas before it depends on any wall.

About this page

The test file computes the distance loss two ways: once from the unit constant, once from the wavelength. It computes the slab transmission three ways: the closed form, the first four thousand terms of the internal-reflection series summed directly, and an optical characteristic matrix. They agree to eight decimal places in every case the file runs. How I check things.

Checked against two other published models. 3GPP TR 38.901, Table 7.4.3-1, models concrete as 5 + 4f dB, with f in GHz. At 2437 MHz that gives 14.75 dB, and this slab reaches 14.75 dB at 200 mm. At 5500 MHz the same two come out at 27.00 dB against 26.47 dB. ITU-R P.1238-13 has a table of floor penetration factors, Table 5, whose apartment entries are footnoted "per concrete wall": 10 dB at 2.4 GHz and 13 dB at 5.2 GHz. This slab reaches 9.90 dB at 130 mm and 2.4 GHz, and 13.34 dB at 100 mm and 5.2 GHz. The three sources give different losses for a concrete wall, and none of them names one thickness for it.

Centre frequencies. 2437, 5500 and 6525 MHz, picked rather than derived. Another channel in the same band shifts the distance term by 20·log₁₀ of the ratio of the two frequencies.

Sources.

• Harald T. Friis, "A Note on a Simple Transmission Formula," Proceedings of the IRE 34(5), 1946, pp. 254–256, doi:10.1109/JRPROC.1946.234568.
• Recommendation ITU-R P.2040-4 (09/2025), § 3 and Table 3; equations (43b) and (44), itu.int/rec/R-REC-P.2040.
• Recommendation ITU-R P.525-5 (11/2024), equation (6), itu.int/rec/R-REC-P.525.
• Recommendation ITU-R P.1238-13 (09/2025), Table 5, itu.int/rec/R-REC-P.1238.
• 3GPP TR 38.901 version 16.1.0 (Release 16), § 7.4.3.1 and Table 7.4.3-1, published by ETSI as TR 138 901 V16.1.0 (2020-11), etsi.org.
• Federal Communications Commission, "Home Network Tips for the Coronavirus Pandemic," 26 March 2020, docs.fcc.gov.
• Power limits by part of the 5 GHz band in the United States: 47 CFR § 15.407(a), ecfr.gov.
• "IEEE Standard for Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications," IEEE 802.11-1997, doi:10.1109/IEEESTD.1997.85951.
• The threshold slider has no source and claims none — it is a line you place.