That question draws the boundary line rather neatly. In an office, weak coverage is answered by adding one more AP. In a yard, on a workshop floor or inside an outdoor enclosure, weak coverage usually comes with no second power outlet and no second cable run, so the only option left is to send the available RF energy where it is needed. One specification becomes two different problems depending on where it is read. What decides whether an industrial wireless router is the right purchase is whether the site sits outside the assumptions of a standard IT room.
The first threshold is not on the datasheet. It is the location. A climate-controlled, dust-filtered room that people walk through daily is what commercial equipment is designed for. A plant exterior wall, the top of a workshop cabinet, a high shelf in a warehouse, or an unattended pump station and solar site fall outside that envelope from the start.
The way to settle it is to draw the mounting position once: which stretch of rail or which cabinet panel, whether a breaker or a variable frequency drive sits alongside, which side the cables arrive from, and where the antennas exit. Metal enclosure walls absorb signal, so once antennas are routed outside the cabinet, feed cable length and hole position have to be fixed together, and moving them later means stripping the cabinet again. Leave clearance from heat-generating modules, and let the bend radius of the conductors decide how much room is kept above and below. Once this part is settled, size, mounting method and antenna form filter out about half the candidates in one pass. A model such as the USR-G806w, at 104 x 102 x 28 mm, mountable on rail or wall, with external antennas that can be led outside the cabinet, suits the small points where only a dozen centimeters of rail remain.
The operating temperature on a datasheet describes the air around the device, not the air inside the enclosure. A sealed outdoor cabinet under direct sun runs considerably hotter inside than outside. A ceiling void peaks at midday under radiant heat from the roof with no ventilation. The device then adds its own heat on top of both.
The way to read the number: measure two points during the hottest stretch of the year, air temperature at intake height and chassis surface temperature. Compare the combined figure against the stated ceiling and keep twenty degrees Celsius of margin or more. When measurements run close to the ceiling, relocation comes before a change of model; metal housing, fanless conduction cooling and wide-temperature components all rank after that step. The USR-G816 is rated from -35 to 75 degrees Celsius, pairs a metal housing with industrial-grade components and thermal design, and carries RS232 and RS485 terminals that connect legacy instruments directly, saving a serial device server and one acceptance test. Rail or wall mounting both work.
Site power is commonly 24 or 48 V DC, and it dips at the instant motors start or stop. That makes the wide-input range more informative than the nominal voltage, and reverse-polarity protection on the terminals removes the cost of the one miswired connection. Where conditions allow, feed two supplies from different circuits.
As the number of devices in a station grows, power has to be counted together with port count: does each device get its own Ethernet port, or does a switch have to be added. The USR-G809s provides two ports configurable as WAN or LAN, six LAN ports and two SFP fiber ports, with wide input DC 9-60 V and reverse-polarity protection on an industrial terminal block, and it mounts on a rail. With one subnet per port, the gateways all land on the same device, and the cabinet does without an extra box, an extra supply and an extra round of re-inspection.
New sites, remote stations and assets that travel with vehicles are three places where cellular is the primary link rather than a backup. Construction runs in weeks while a SIM can be activated in minutes, and pulling dedicated cable carries costs and follow-on maintenance out of proportion. Two things tend to be skipped when the card is ordered. IoT cards and VPDN private-network cards have to be settled at order time, and a card inserted without the APN filled in simply never registers, which is regularly misread as absent signal. An antenna left inside a metal cabinet wastes the card; log the signal strength once before relocating anything.
On this point the USR-G816 offers 5G SA and NSA on a Qualcomm X62 module, dual SIM with the second slot available as an embedded eSIM, so changing carrier does not mean changing hardware. It fails over automatically between Ethernet, cellular and Wi-Fi, and commissioning should include unplugging the primary link once with a stopwatch to confirm whether field traffic is retransmitting during the switch or simply stopped.
At unattended sites, how soon a fault is noticed often costs more than the device itself. Three things matter: whether an alarm fires after the drop, who receives it, and whether it can be handled without travelling to the site.
The USR-G806w provides daily and monthly traffic histograms plus a live session view, and pairs them with three alarm classes, device offline, weak signal and allowance exceeded, delivered by email and SMS. Thresholds set at seventy or eighty percent of the plan mean an overage is known before the invoice arrives. A remote management platform also allows parameters to be changed, reboots issued and firmware upgraded in batches, so editing an allowlist once does not require dispatching a vehicle. Hitting one of the five conditions is enough to justify specifying industrial-grade equipment; hitting two or more and still quoting commercial hardware tends to convert the saved difference into site visits.
What deserves asking at the end of a selection is not how high one datasheet line reads, but whether these things come with a way to handle them: whether the unit fits the available rail, whether margin can be measured on the hottest day of the year, whether it can reach the network on its own, and whether anyone is notified when the link drops. The difference between an industrial wireless router and commercial equipment comes down to these things that do not happen in an office.
1. How is a 5G cellular router different from an ordinary Wi-Fi router?
Look for a SIM slot. An ordinary Wi-Fi router shares broadband that already reaches the premises, while a 5G cellular router uses its own SIM to join a carrier network and then shares that connection across its Ethernet ports and wireless, combining modem, routing and switching in one box. Another common mix-up is 5G against 5 GHz: the first is the fifth generation of cellular networks, the second is a Wi-Fi frequency band. The names differ by a single character and the technologies have nothing in common. Industrial models add wide input voltage, wide temperature range, rail mounting and alarms on top of that.
2. Is 5G worth it when 4G is enough?
Judge by upstream traffic rather than downlink speed. PLC polling and energy reporting, which move tens to hundreds of bytes every few minutes, are served well by 4G and coverage is usually more stable. Upstream becomes the bottleneck with several camera feeds, bulk program downloads or pulling historical footage remotely, and that is where 5G earns its place through lower latency and more upstream headroom. Where 5G coverage drops out, the device falls back to 4G automatically, so the choice should follow upstream demand at the busiest hour rather than the peak figure on the box.
3. What should be checked when ordering a SIM card?
Carrier data plans and IoT plans differ sharply over the APN, whether the card is bound to a device, and whether a public address is issued, so private-network and VPDN arrangements have to be settled at the ordering stage. The classic symptom of an unfilled APN is a registration that never completes, easily mistaken for absent signal. Estimate monthly usage before activation as reporting interval times message size times number of sites times 1.5, and set thresholds at seventy or eighty percent of the plan so an overage is not discovered at month end.
4. Can a site device still be reached without a public IP?
Cellular links usually deliver private or carrier-grade NAT addresses, so unsolicited inbound traffic from the public internet never arrives and port forwarding is effectively off the table. The workable approach is to have the device originate an encrypted tunnel outward, or to use the matching remote management platform, so that no inbound port has to be opened on the public side. Acceptance testing can run the other way round: scan the site address from the public side and every target port should come back closed.
5. Is cellular access less secure than a wired line?
Only the hop from site egress to carrier changes. The path inside the cabinet, from Ethernet port to PLC and supervision host, is untouched, and none of the long-standing internal-network issues go away. What the medium genuinely adds comes in three parts: the SIM is a credential that can be removed, so it belongs behind a cabinet lock and in an inventory; the APN decides whether traffic leaves through a public or a dedicated gateway; and there is one more bill that can be consumed, which a threshold alarm should watch.
6. What is dual SIM for?
Two cards can sit on different carriers, which avoids single-coverage dead spots and allows a carrier change at plan expiry or signal degradation without replacing hardware. Making the second one an embedded eSIM removes even the trip to open the cabinet. Switching triggers on link availability rather than signal bars, so commissioning should include unplugging the primary link once with a stopwatch, and confirming whether field traffic retransmits during that wait or simply stops.
7. Is data lost while the link is down?
That depends on which side handles retransmission, and it is worth asking during selection. Handled on the device side, data is buffered locally and filled in once the link returns. Handled on the platform side, how complete that stretch of data is depends on when service resumed. Combined with the three alarm classes (offline, weak signal, allowance exceeded) over email and SMS, and a remote management platform that changes parameters, reboots and upgrades firmware from a distance, most issues do not wait for the next site visit.
8. How many devices can one unit carry, and how is data usage estimated?
Device count comes down to two things: the terminal limit on the Ethernet ports and on wireless, and whether heavy lateral traffic exists inside the station. With many devices on site, look at port count and whether one subnet per port is possible; a remote point carrying three or four IO channels does not need a high port count. Traffic is estimated from upstream services. Daily and monthly histograms show the trend, while the live session view shows which terminals are connected now and where they connect to. The two together are what localise which device is consuming the data.
