Solutions

Antenna height beats transmit power at 470 MHz

Twelve decibels separate a 2 m mast from an 8 m one. The two-ray arithmetic behind 470 MHz coverage, and the range where extra height stops paying.

A meter box on the wall of a factory reads -104 dBm at 3.9 km from the gateway. It works in dry weather and drops packets after rain. The obvious fix in the meeting is a 5 W amplifier. The less obvious fix is the antenna, which is bolted to a cabinet 1.6 m above a concrete apron, at the same height as a row of parked forklifts.

Both fixes add decibels. Only one is cheap, legal and permanent. What follows is the arithmetic behind that choice, and the places where the arithmetic stops working — because on a 470–510 MHz link, height is not a vague best practice. It is a term in the path loss equation with a coefficient you can compute.

Height enters the link budget twice

Two rays leave the transmitter. One goes straight to the receiver; the other bounces off the ground somewhere in between and arrives late. At 490 MHz the second ray is not a rounding error. The phase difference between them is

dphi = 4 * pi * h1 * h2 / (lambda * d)

with both heights in metres, distance in metres and lambda = 0.612 m at 490 MHz. At short range the two rays add and cancel as distance changes — that is the ripple you see when you walk a receiver away from a mast. Beyond a few hundred metres they are almost perfectly out of phase, so what is left is not the sum of two strong signals but the small difference between them. That difference falls off as 1/d², not 1/d, and the path loss becomes

PL = 40 * log10(d) - 20 * log10(h1 * h2)

Two consequences follow directly. First, the slope is 40 dB per decade instead of 20, so ground-level sub-GHz links lose range much faster than a free-space calculation suggests. Second, height appears as a simple 20·log term: doubling the height of either antenna buys 6 dB, quadrupling buys 12 dB, regardless of distance. This is the classic two-ray ground reflection result, and it holds while the ground is the dominant reflector and the distance is well past the breakpoint.

That 6 dB per doubling is the whole argument in one number. A 6 dB gain from an amplifier means quadrupling transmit power: 26 dBm becomes 32 dBm, 400 mW becomes 1.6 W, and in most jurisdictions that is outside the type approval for the band, as our 470–510 MHz rules walkthrough sets out. Six decibels from a longer pole costs a pole.

Antenna height

Gain vs an 8 m mast

Transmit power needed to match it

EIRP at that power

1 m (cabinet top)

+18.1 dB

44 dBm

25 W

2 m (wall bracket)

+12.0 dB

38 dBm

6.3 W

3 m (short pole)

+8.5 dB

34.5 dBm

2.8 W

5 m (eaves)

+4.1 dB

30.1 dBm

1.0 W

8 m (roof mast)

0 dB

26 dBm

0.4 W

The right-hand column is the part that ends the discussion. Every row assumes the same receiver at 1.5 m, 26 dBm at the transmitter and 2 dBi at both ends. Matching an 8 m mast with an amplifier on a 1 m bracket means 25 W in a band where the equipment is licensed at a fraction of that.

The breakpoint is 62 metres, not 600

The 40 dB/decade slope does not apply everywhere. Close in, the direct ray wins and the link behaves like free space until the two rays are far enough out of phase to cancel. The crossover is at

d_bp = 4 * pi * h1 * h2 / lambda

which at 490 MHz gives 62 m for a 2 m mast, 154 m for 5 m and 246 m for 8 m. Inside that radius, raising the antenna changes almost nothing about the median level. Outside it, every metre counts.

Path loss at 490 MHz for 2, 5 and 8 m masts against a 1.5 m receiver, with the free-space curve and the breakpoints marked

For scale: free-space loss at 1 km is 86.2 dB. At 5 km, the two-ray model puts a 2 m mast at 138.4 dB, a 5 m mast at 130.5 dB and an 8 m mast at 126.4 dB. That is 52 dB more than free space for the 2 m case — the difference between a link that closes with 20 dB of margin and one that does not close at all, from geometry alone.

What the hardware actually contributes

The radio in this calculation is a plain narrowband transceiver, not a spread-spectrum link with processing gain to hide behind. Our TP5500 transparent RS485 bridge and the TP5510 concentrator share the same RF front end, so the same numbers apply to both.

Parameter

TP5500

TP5510

Operating frequency

470–510 MHz

470–510 MHz

Channel bandwidth

≤200 kHz per carrier

≤200 kHz per carrier

Symbol rate

2.4 / 19.2 / 76.8 kbps

2.4 / 19.2 / 76.8 kbps

Transmit power

26 ± 1 dBm

26 ± 1 dBm

Receiver sensitivity

−116 dBm at 19.2 kbps

−116 dBm at 19.2 kbps

Channels

1

1

Network role

Point-to-point, transparent

Broadcast and directed, up to 100 terminals

Interfaces

RS485, RS232, DC 12 V, SMA female

RS485, RS232, DC 12 V, SMA female

Supply

DC 12 V (5–28 V)

DC 12 V (5–28 V)

Protection and mounting

IP65, wall or DN35 rail, 116 × 85 × 41 mm

IP65, wall or DN35 rail, 116 × 85 × 41 mm

Temperature

−40 °C to +70 °C

−40 °C to +70 °C

One sanity check before trusting any of it: thermal noise in 200 kHz is −174 + 10·log₁₀(200,000) = −121 dBm. A −116 dBm sensitivity therefore implies about 5 dB of receiver noise figure, which is exactly what a competent narrowband front end should show. If a datasheet claims −116 dBm at 200 kHz with a 2 dB noise figure, one of the two numbers is wrong.

With 26 dBm and 2 dBi at each end, the budget is 146 dB. Run it through the model and a 2 m mast reaches 7.7 km to a 1.5 m receiver, 5 m reaches 12.2 km, 8 m reaches 15.5 km. The radio horizon for the same three cases, using 4.12·(√h₁ + √h₂) for a 4/3 Earth, is 10.9, 14.3 and 16.7 km. Height buys kilometres until the horizon, then stops.

Range allowed by 146 dB of link budget versus antenna height, against the radio horizon

Five ways height makes things worse

Nulls at short range. Inside the breakpoint the two rays interfere, and the nulls sit at predictable distances: 9.8 m, 4.9 m and 3.3 m for a 2 m mast; 39 m, 20 m and 13 m for an 8 m mast. A meter room 20 m from the gateway can measure 10 dB worse after the mast is raised. This is the most common "the upgrade broke it" report in sub-GHz deployments, and it is geometry, not a faulty radio.

Received power versus distance from 4 to 80 m for 2 m and 8 m masts, showing how raising the antenna moves the interference nulls outwards

The horizon is a wall. Free-space loss alone would let 146 dB span 972 km. The Earth allows 11 km from a 2 m mast to a 1.5 m meter. No amplifier changes that number; only height does, and only as the square root.

Diffraction is not a door. A single roof edge 5.6 m above the line of sight costs about 8.8 dB at 4 km, computed with the single knife-edge method in ITU-R P.526. Raising the mast so the edge sits 3.7 m above the line recovers only 0.9 dB of that — diffraction loss changes slowly with clearance once the path is already blocked. In a street canyon you are not dealing with one edge but six, and each one is roughly another 6 dB.

The Fresnel zone is wide. At 470 MHz the first zone has a 12.4 m radius at 1 km and 39 m at 10 km, against 5.6 m and 17.7 m at 2.4 GHz (ITU-R P.530 is the reference treatment of clearance and multipath on these paths). That width is why 470 MHz shrugs off a fence post and dies behind a treeline: a mast in the middle of the path costs more than a hill at the end of it.

Hardware quietly eats the gain. Three metres of RG-58 at 490 MHz is about 0.9 dB; a connector that took water is 5–10 dB, and it fails slowly enough to look like seasonal fading. An antenna clamped directly to a metal mast detunes, and a high-gain whip tilted by 10° moves its pattern off the horizon.

One link, computed to the last decibel

Take a 4.2 km path from a rooftop mast to a meter at 1.5 m, with a roof edge 8 m high at 1.6 km from the mast. Fresnel radius at the edge is 24.6 m, so the edge sits deep inside the first zone whatever the mast height. Working through the plane-earth loss, the knife-edge loss and the remaining margin:

Mast height

Line of sight at the edge

Edge above line

Knife-edge loss

Received power

Margin over −116 dBm

3 m

2.43 m

5.57 m

8.8 dB

−110.7 dBm

5.3 dB

6 m

4.29 m

3.71 m

7.9 dB

−103.7 dBm

12.3 dB

8 m

5.52 m

2.48 m

7.3 dB

−100.6 dBm

15.4 dB

Three metres of extra mast converts a 5.3 dB margin — a link that fails in rain — into 12.3 dB, with the same radio and the same antenna. The same result from an amplifier needs 7 dB more power, 33 dBm, plus the regulatory work that goes with it. Note also that going from 6 m to 8 m buys only 3.1 dB: the model has a logarithm in it, and the first metres are the valuable ones.

This is the same budget that decides how far 470 MHz actually reaches; the difference here is that the antenna height is a variable rather than an assumption.

What a field check should record

The model tells you where to put the mast. It does not tell you what your site will do, because the reflection coefficient depends on what the ground is made of. Six steps, in this order:

  1. Measure at three distances — 100 m, 1 km and 2 km if the site allows — with the antenna at the height you plan to install, not on a temporary tripod at 1 m.
  2. Take 50 samples per point and record the median. Instantaneous RSSI on a moving vehicle swings 6–10 dB over a few metres, and a single reading is not evidence.
  3. Record the geometry with each reading: mast height, receiver height, distance, and the height and distance of the worst obstruction between them. A number without geometry cannot be compared to anything later.
  4. Note what the ground is along the reflection path. Wet asphalt, standing water and metal roofing behave very differently from crops and gravel, and the difference shows up as 5–10 dB.
  5. Compare the measurement with the prediction before adding power. If the site is 15 dB worse than the model, you have an obstruction or a broken connector, and no amount of dB will fix it.
  6. Re-measure one to two months after installation. Connector degradation is gradual, and a 5–10 dB drop is the usual signature of water ingress rather than a change in propagation.

The model itself is eleven lines, and worth having in the survey notebook:

import math
lam = 299.792458 / 490.0          # 490 MHz

def path_loss(d, h1, h2=1.5):
    fspl = 20 * math.log10(4 * math.pi * d / lam)
    ground = 40 * math.log10(d) - 20 * math.log10(h1 * h2)
    return max(fspl, ground)      # crossover at d = 4*pi*h1*h2/lam

for h in (2, 5, 8):
    print(h, round(path_loss(4200, h), 1))

It is a model, not a measurement, and it is optimistic on rough ground: the 40 dB/decade slope assumes a flat, reflective surface, and real terrain sits between 30 and 40. That is also what the published field data shows. A 2025 test of 915 MHz LoRa nodes at 1.5, 5 and 10 m over 100 m to 1 km found the 1.5 m installation below −115 dBm at the far end while the 10 m installation stayed above −95 dBm — a spread of at least 20 dB, against 16.5 dB predicted by 20·log₁₀(10/1.5) for the same height ratio. The measurement report is worth reading for the shape of the curves rather than the absolute values, since its band, power and terrain are not yours.

FAQ

Does antenna gain do the same job as height?

Partly. Gain is bought by narrowing the vertical pattern, so an 8 dBi whip compared with a 2 dBi one can add 6 dB along the horizon — the same 6 dB as doubling the mast height. The difference is what happens to the rest of the pattern: the gain antenna has less coverage above and below it, which matters when meters sit 2 m below the antenna on the same building. Height also lifts the antenna out of the clutter, which no pattern can do.

Can I just raise the transmit power instead?

You can add decibels, but not cheaply and not always legally. Six dB means quadrupling power, and 26 dBm is already close to what the band allows for this class of equipment. Power also fails against the horizon and against obstruction, which are the two limits that actually bind at multi-kilometre range. And in a star network the uplink is usually the weak direction: the concentrator can transmit at 26 dBm, the battery meter cannot.

How high is enough?

Enough to clear the worst obstruction by roughly 60% of the first Fresnel radius, and enough that the path is not grazing the ground for most of its length. In practice that lands at 5–10 m for a rooftop mast on a 2–5 km link. Past that, the horizon and the link budget take over and extra height returns less than 6 dB per doubling.

Why did RSSI get worse after we raised the mast?

Because you moved the nulls. Under 80 m the two rays interfere, and the null spacing scales with height — an 8 m mast puts its first null at 39 m where a 2 m mast had one at 9.8 m. Move the antenna 2–3 m sideways, or lower it, and re-measure before concluding the radio is faulty.

A 2 m mast and a 25 W amplifier produce the same number on a link budget. Only one of them is still there in five years, and only one of them can be installed on a Tuesday afternoon.

More in Solutions

  • Solutions

    How do drones transmit images?

    The real-time transmission of images and video from an unmanned aerial vehicle (UAV) to a ground station is a cornerstone technology that enables applications…