Poor LoRa Gateway Range? Troubleshoot the Antenna System First

Poor LoRa gateway range is nearly always an antenna system fault — wrong band, blocked placement, a missing ground plane, or a lossy feedline — long before it is a reason to buy a higher-gain antenna.

Short answer

1. Re-test with a known-good antenna that matches your regional plan, on the shortest cable you own, outside any metal enclosure.
2. Change one thing at a time and compare RSSI, SNR and packet delivery over many packets, not one lucky uplink.
3. Fix band, clearance and feedline before you touch gain. Gain changes the shape of coverage, not the amount of margin you get for free — and regulators cap it on transmit.

This guide uses one decision rule throughout: a change is only an upgrade if it improves measured link margin at the same test points. Everything below is ordered by how often it is the real cause and how cheap it is to check.

Two terms used a lot here. RSSI is how strong the received signal is, in dBm. SNR is how far that signal sits above the background noise, in dB. They can move in opposite directions, and that difference is often the whole diagnosis.

Start With the Symptom, Not the Antenna Catalog

Never judge a change on one successful uplink. Repeat over enough packets to see whether the improvement is stable, and record the gateway, node, frequency plan, spreading factor or data rate, transmit power, antenna, cable and location each time.

1. Build a Repeatable Baseline

Before changing hardware, write down what the network already knows:

1. Gateway model, radio card and configured regional plan.
2. Node model, antenna and transmit settings.
3. Test location, distance, height, and whether line of sight exists.
4. RSSI and SNR from repeated packets, not a single message.
5. Packet counts or delivery rate over the same test window.
6. Antenna part number, stated band and gain.
7. Cable type, length, adapters, surge protector and connector interfaces.
8. Photos of the antenna, roof or cabinet, cable route and nearby metal.

Use the same node, route and time window for each comparison. If you change node position, antenna, cable and radio settings together, you will know something changed — but not what fixed it.

2. Confirm the Band Before You Compare Gain

“868 MHz” and “915 MHz” are nicknames, not interchangeable labels. The LoRa Alliance Regional Parameters define the plans as EU863-870 (common name EU868) and US902-928 (common name US915), and list the legal band per country in RP002.

Two regulatory limits also decide what your antenna is allowed to do — this is the part most range threads skip:

• Europe. ETSI EN 300 220-2 lists 863–870 MHz at 25 mW e.r.p. (about 14 dBm) with duty-cycle or polite-spectrum-access limits that differ by sub-band. “e.r.p.” is radiated power, so antenna gain counts against the ceiling.
• United States. Gateways in 902–928 MHz are typically certified under 47 CFR 15.247, where transmit antenna gain above 6 dBi must be offset dB-for-dB by reducing conducted output power.

So check three things in this order: the country’s plan, the gateway’s configured channel plan, and the antenna’s documented operating band. Only then compare gain numbers.

An antenna centred near 868 MHz should not be assumed to cover the full US915 plan. A product that documents a wider band may suit more than one plan, subject to the finished installation and local rules. For example, RFTECH’s GL868915 is documented for 868–915 MHz, 3 dBi, SMA male, 50 ohms, while the GL868F10 is centred at 868 ±10 MHz with SMA male. Those are specification examples, not a promise of a fixed distance.

If you are sanity-checking a quarter-wave reference, the free-space value is about 86.3 mm at 868 MHz and 81.9 mm at 915 MHz. That is not a finished cut length — conductor geometry, dielectric materials, matching and end effects all change the practical design. Use a validated drawing or measured data instead of trimming a radiator from the formula.

3. Give the Antenna Clearance — and a Ground Plane

A metal gateway box protects electronics and is a poor home for an internal LoRa antenna. Conductive walls block energy, and nearby panels, rails, roofs, pipes and cable bundles change the antenna’s impedance and pattern.

There is a usable number behind this. Research on antennas over metal notes that when the metal-to-antenna spacing is much less than a quarter wavelength, the reflected wave returns with a phase shift approaching 180°, which cancels the direct signal (Georgia Tech Propagation Group). At LoRa frequencies a quarter wavelength is roughly 86 mm at 868 MHz and 82 mm at 915 MHz, so treat those figures as your minimum clearance target from any large metal face.

The same wavelength drives ground-plane size. A quarter-wave or magnet-mount antenna is only half of a dipole; the metal surface under it forms the other half. As an engineering rule of thumb, aim for a conductive ground plane extending at least a quarter wavelength in every direction from the base — about 85 mm radius, roughly a 170 mm disc, at these frequencies. A thin bracket, a narrow rail or a small steel plate is not that. If an antenna is documented as ground-plane independent, use it that way; do not assume it from appearance.

Run this comparison before ordering a new model:

1. Keep the gateway and node settings unchanged.
2. Test the antenna in its current cabinet or roof position.
3. Move the same antenna to a clear position outside the cabinet, using the shortest known-good connection.
4. Keep it vertical if it is designed for vertical polarization.
5. Repeat the same packet test and compare the distribution of RSSI, SNR and delivery, not the best single result.

An external antenna often helps a metal-enclosure install because it moves the radiator outside the obstruction. It does not automatically fix anything: a long lossy cable, a mismatched connector or a bad roof-edge position can trade one fault for another. On a metal roof, avoid tucking the radiator beside a parapet, HVAC unit or large vertical steel face, and route the cable without sharp bends or crushed sections.

4. Turn dB Into Metres Before You Blame the Antenna

Most range arguments stall because everyone talks in dB and nobody converts it into distance. Free-space path loss gives you the conversion. ITU-R P.525 states it as:

Loss (dB) = 32.4 + 20 × log10(f in MHz) + 20 × log10(d in km)

At 868 MHz over 1 km that is about 91 dB. Two consequences matter for troubleshooting:

• Doubling the distance costs 6.02 dB in free space.
• So every 6 dB you find or lose is worth roughly a halving or doubling of free-space distance.

That lets you price a fault. Losing 2.7 dB in a cheap cable scales free-space distance by 10\^(−2.7/20) ≈ 0.73, about 27% shorter. In an obstructed environment, where loss grows faster than the square of distance, the same 2.7 dB costs less reach — with a path-loss exponent near 3, the factor is 10\^(−2.7/30) ≈ 0.81, about 19% shorter.

These are engineering estimates from the formula, not measured results, and they assume nothing else changed. Use them to decide what is worth chasing: a 0.6 dB cable is noise, a 3 dB adapter chain plus arrestor is a real coverage decision.

5. Calculate Cable Loss Before You Extend the Antenna

Moving an antenna out of a cabinet or above a roof is often worth it, but the feedline spends part of the link budget. Cable loss only means something when stated with cable type, frequency and length.

The Times Microwave LMR guide publishes the following attenuation at 900 MHz and +25 °C. The 5 m figures are cable-only estimates calculated as published dB per 100 ft × 5 / 30.48.

These exclude connectors, adapters, surge arrestors, mismatch, damage, water ingress and temperature effects. Feed the total into the conversion above to see what the run costs in reach, then choose the shortest practical route and get attenuation at your actual operating frequency from the cable manufacturer. Lower-loss cable preserves more margin but is thicker, heavier, less flexible and needs compatible connectors and support.

6. Gain Is a Shape, and It Is Capped on Transmit

For an omnidirectional antenna, more gain usually means a flatter vertical pattern. That helps nodes spread around the horizon and can hurt nodes above or below the antenna. A gateway high on a roof serving nearby ground-level nodes can perform worse with an 8 dBi antenna than with a broader 3 dBi pattern.

Use this as the selection rule:

• Choose a moderate, broad pattern when nodes sit at varied heights, close in, or spread through a building.
• Consider higher omnidirectional gain when nodes are mainly near the horizon and the mast is vertical with clear surroundings.
• Use a directional antenna only when the service area is deliberately concentrated in one direction and alignment can be controlled.
• Compare measured patterns and installation geometry, not dBi alone.

There is also an asymmetry almost nobody mentions. On transmit, gain sits inside your legal ceiling: an e.r.p. limit in Europe and the >6 dBi power-reduction rule in the US both mean a higher-gain antenna does not legally radiate proportionally more. On receive, gain is not regulated at all. Because LoRaWAN traffic is mostly uplink, a higher-gain gateway antenna can still improve the direction that carries most of your packets — but only if those nodes actually sit inside the narrower vertical beam. If they do not, you have bought a legal ceiling and a coverage hole at the same time.

There is no universal distance ratio for “3 dBi vs 8 dBi.” Cable loss, polarization, terrain, height, receiver sensitivity, radio settings, interference and the node antenna all stay in the link budget.

7. Check SMA and RP-SMA as Complete Interfaces

SMA and RP-SMA look almost identical from a distance. “Reverse polarity” changes the centre-contact arrangement while the outer thread stays the same, which is exactly why it gets misread.

It helps to know why RP-SMA exists at all: it is a regulatory artifact, not a performance grade. 47 CFR 15.203 requires that an intentional radiator be designed so that no antenna other than the one supplied can be used, and it prohibits a standard antenna jack. Reversing the polarity created a non-standard interface that satisfied that requirement. So “RP-SMA is for Wi-Fi, SMA is for LoRa” is a market convention, not an electrical rule, and it is not a reason to expect better or worse RF performance. Check the hardware in front of you.

Inspect both halves of every interface:

1. Outer body and thread arrangement.
2. Centre pin or socket on the gateway port.
3. Centre pin or socket on the antenna cable.
4. Whether the contacts mate without forcing.
5. Damage, looseness, contamination or a recessed centre contact.
6. Every adapter and surge protector between radio and antenna.

A correctly specified adapter can convert the interface, but it adds another connection and another possible fault. Do not assume it is lossless, or that “SMA male” alone describes the mating pair. The useful bench test is to substitute a verified short cable and antenna with the exact mating interface; if that restores performance, inspect the removed assembly separately instead of continuing to change radio settings.

While you are reading specs, keep VSWR in proportion. VSWR describes how much power reflects back from the antenna instead of radiating. A 2:1 VSWR reflects about 11% of the power, which is roughly 0.5 dB of mismatch loss — the reflection coefficient is (VSWR−1)/(VSWR+1), and mismatch loss is −10 × log10(1 − Γ²). Half a dB does not explain a tenfold range shortfall. A broken centre contact, water in the cable, or the wrong band does.

8. When RSSI Looks Bad but the Link Is Actually Fine

LoRa is designed to decode signals buried in noise. Published work puts the usable SNR range for LoRa roughly between −7.5 dB and −20 dB depending on spreading factor (MDPI), and modern concentrators are specified down to about −141 dBm sensitivity — Semtech lists that figure for the SX1303 with an SX1250 front end.

That is why RSSI alone is a bad verdict. A weak but clean signal often decodes fine, while a stronger signal sitting in a raised noise floor fails. A 10 dB rise in local noise and a 10 dB drop in received signal look very different in the two numbers together, and only one of them is an antenna problem.

If a known-good antenna, short feedline and clear mounting position do not move the baseline, widen the investigation:

• Confirm the gateway is receiving on the intended channels and regional plan.
• Compare several nodes to rule out a single-node antenna or power fault.
• Check whether loss clusters at certain times, channels or locations.
• Inspect the gateway noise floor and channel occupancy with suitable RF tools.
• Review transmit power, spreading factor or data rate, adaptive data rate behaviour and gateway logs.
• Verify power supplies, radio cards and firmware before blaming the antenna.

What to Send for Antenna or Feedline Review

Send this with an RFQ or engineering question:

• Country and exact LoRaWAN regional plan.
• Gateway and radio-card model.
• Antenna model, datasheet and required form factor.
• Gateway-port and antenna connector photos, including centre contacts.
• Cable type, length, connector ends, adapters and arrestors.
• Cabinet, mast or roof photos with dimensions and nearby metal.
• Node locations, heights and expected coverage geometry.
• Repeat-test RSSI, SNR and packet results.
• Environmental requirements and quantity.

That evidence lets a supplier review the installed RF path instead of recommending a gain number in isolation. For commercial selection, compare the outdoor LoRa gateway antennas or browse the wider LoRa antenna family. Smart meter projects with internal and external enclosure constraints should start from LoRa antennas for smart metering.

Frequently Asked Questions

What is the realistic range of LoRaWAN?

There is no single number that applies to every deployment. Range depends on the whole link budget: antenna height and pattern, terrain and buildings, feedline loss, node antenna, radio settings, interference and the packet reliability you require. Define fixed test points and measure repeated packets under the intended installation conditions.

How much range does one dB actually cost me?

In free space, 6 dB is worth about a factor of two in distance, so one dB is roughly a 12% change. In obstructed environments the same dB buys less distance because path loss grows faster with range. Treat both as estimates from the path-loss formula, not measurements.

Does cable length affect antenna reception?

Yes. Coax attenuates both transmitted and received signals, and the loss grows with length and frequency. Use the manufacturer’s attenuation figure for the named cable at your operating frequency, then add connectors, adapters and other inline components.

Is there a difference between RP-SMA and SMA?

Yes — different centre-contact arrangements with the same thread. Confirm the body, the pin or socket, and the interface on both sides. Family and gender shorthand alone is easy to misread.

Can I convert RP-SMA to SMA?

A correctly specified adapter converts the interface, but it introduces another connection. Verify both ends, inspect the centre contacts, and include the adapter in the installed loss and reliability review.

Does my LoRa antenna need a ground plane?

A quarter-wave or magnet-mount antenna does, because the metal surface acts as the missing half of the dipole. Aim for at least a quarter wavelength of conductive surface in every direction — roughly 85 mm at LoRa sub-GHz frequencies. Dipole and fibreglass designs documented as ground-plane independent do not.

Diagnose Before You Replace

The best LoRa gateway antenna is the one that covers the correct regional band, fits the real connector and feedline, produces a useful pattern for your node geometry, and works in the installed environment. Establish the short-cable, clear-location baseline first, then add the original cable, adapters and mounting position back one at a time.

If you want RFTECH to review an antenna or feedline option, send the regional plan, gateway port, cable route, installation photos and repeat-test results. We can compare documented products and tell you which specifications still need confirmation — without promising a distance the installation has not demonstrated.