A machine is being commissioned before shipment. The gateway inside the cabinet has one Ethernet WAN port left, and the manual says "uplink: RJ45". There is no fixed broadband at the site and nowhere else to pull a cable from. On commissioning day somebody takes out a phone and switches on a hotspot, but that port does not accept a wireless signal. It accepts a network cable.
The same problem has come up elsewhere. Someone wanted to share phone data with a gateway that only had an Ethernet port, and ended up with a portable router s a bridge, turning the wireless signal into one uplink Ethernet connection into that port; one reply was simpler: just buy a cellular device that takes a data card. Neither approach is wrong; the difference is how long it has to hold up and how many devices sit behind it. On a machine, the order is: first see what the equipment at the far end needs, then decide at which layer the wireless link becomes a cable. That job usually falls to an industrial cellular wifi router.
A device with one Ethernet WAN port wants an uplink, physically an RJ45, which accepts neither an SSID nor USB. The check takes two steps: look at the "uplink/WAN" line in the manual and see whether it says RJ45 or Wi-Fi, then see how it obtains an address, by DHCP, static address, or dial-up. Another class of equipment carries its own wireless client and only needs internet access, so wireless can go straight to it and no cable is needed. The class that gets missed is in between: the device has Wi-Fi, but the firmware only allows it to run as a hotspot, not as a client, so a cable is the only option. Then there is the confusion caused by a name: a port labelled WAN may only want a position that has internet access, and a router LAN port works the same, so there is no need to argue about what it is called.
The cleanest tier puts the wireless-to-cable layer inside one box: insert a card, the Ethernet port delivers Ethernet directly, and addressing, rules and internet access all belong to that one unit. The USR-G816 takes this route, with 5G SA and NSA on a Qualcomm X62 module, four Gigabit ports (one configurable as WAN or LAN, three LAN), and 1.5 kV isolation on the Ethernet ports; dual SIM, with the second slot available as an embedded eSIM; the RS232/RS485 terminal block runs at 1200 to 460800 baud, so older instruments can hang off the same unit and no separate serial device server is needed; DC 9-36 V with reverse polarity protection, a metal housing rated -35 to 75 degrees Celsius, 125 x 103 x 45 mm, DIN rail or wall mounting, and automatic failover between Ethernet, cellular and Wi-Fi. Card activation is the step that causes the most rework: IoT cards and private network cards need an APN, and leaving it blank shows up as a failure to register, which is regularly misread as no coverage at the site.
In a rented workshop, at a temporary workstation, or where a customer will only hand over one wireless account, the router joins as a client and hands the connection out through its Ethernet ports. The USR-G806w switches freely between cellular, wired and Wi-Fi internet access, with 2.4 GHz 802.11b/g/n through 2 x RP-SMA-K connectors and two 5 dBi antennas; three Fast Ethernet ports, one configurable as WAN or LAN; 104 x 102 x 28 mm, DIN rail or wall mounting, powered either through the terminal block or the DC jack, -20 to 70 degrees Celsius, hardware and software watchdogs. The cost of this tier belongs up front: the network is borrowed, the password sits with somebody else, a captive portal may need a manual click after every reboot, and the other side gives no notice before changing things, so it suits temporary and transitional use. Antennas should not be left inside a metal cabinet: log the signal level once, then move them, and the difference between the two readings is the only evidence.
Two faster options remain: an adapter that turns a phone into one uplink Ethernet connection, or simply borrowing the wireless network that is already at the site. They cover the internet leg. Traffic between the PLC, the HMI and the switch inside the cabinet is unaffected, so they suit commissioning-day acceptance, the wait for a card to be activated, and temporary workstations. They do not suit a spot left unattended all year: the phone gets carried away, system updates reboot on their own, and when a data allowance runs out the result is throttling rather than disconnection, so the link keeps showing as online while only the traffic at the back of the queue starts to time out.
The uplink takes at least one port, each device on site takes one, and running short means adding a switch: one more device, one more power feed, one more acceptance check. Whether the WAN port can be reconfigured as LAN decides how many subnets one unit can carry. The USR-G809s gives eight Gigabit copper ports and two SFP slots, two of the ports configurable as WAN or LAN plus six LAN, and one SFP shared with a copper port; dual Nano SIM, with the second as an embedded eSIM; DC 9-60 V with reverse polarity protection, 118 x 96 x 49 mm, DIN rail mounting, -25 to 75 degrees Celsius, plus an RS485/RS232 terminal block and USB and SD expansion. A point a few hundred meters away runs over SFP fiber, which also breaks the ground potential difference between the two ends.
Addressing and gateway: which unit hands out the address the device receives, and where the default gateway points; two subnets must not collide. Uplink: unplug the primary link once and time the recovery, and confirm while waiting whether data is retransmitted or suspended, and whether the PLC and HMI in the cabinet still reach each other, which shows that this leg does not depend on internet access. Signal and antennas: measure three times, cabinet door open, door closed and antennas routed outside, and record the differences in the handover document. Configuration record: export the configuration with the site number and date, and archive it together with the change list and the firmware version, so the re-test next year has a reference.
See what the far end needs first, then decide at which layer the wireless link lands, then count the ports and the power. Only then has a cellular wifi router been put in the right place.
1. What is the difference between a 5G cellular router and an ordinary Wi-Fi router?
Look for a SIM card slot. An ordinary router needs a wired broadband line first, while a cellular wifi router inserts its own card to build the uplink and then hands the network out through its ports and its wireless. One more thing: 5G means the fifth generation of cellular network, not the 5 GHz band of Wi-Fi, and the two "5G" labels in a specification sheet are regularly read as one and the same.
2. Should the choice be 4G or 5G?
Decide by whether the uplink end can carry the load, not by the advertised downlink peak. Dozens of telemetry and status points leave 4G with a margin that is easier to estimate; more than two concurrent video streams, or bulk data that has to finish inside one window, is where 5G earns its place. The USR-G816 supports both SA and NSA and falls back to 4G automatically where there is no 5G coverage, so it is worth asking which modes are available locally before selecting.
3. What matters when getting a data card?
Three things: whether the card is allowed to run in a router, whether the IMEI of the device has to be registered, and what the access point name is. IoT cards and private network cards must use the APN the carrier gives; leaving it blank shows up as a failure to register, which is most easily mistaken for "no signal here".
4. Without a public IP, can equipment in the cabinet still be maintained remotely?
Cellular lines usually receive carrier-grade private addresses, in the 10.x and 100.64.x ranges, so inbound connections from the public internet cannot arrive and port forwarding is effectively closed off. The alternative is a tunnel: the router initiates a VPN connection outward, or a device management platform opens the built-in web page. Neither requires opening a port on the customer side.
5. Is cellular internet access secure?
Security depends on the tunnel and the rules, not on the medium. The practice is to put remote access through an encrypted tunnel first, then allow traffic item by item by destination address, protocol and port, with everything else denied by default. Both the USR-G816 and the USR-G806w carry access control and IP, MAC and domain filtering, and naming a rule after the device it covers makes later review easier.
6. What is dual SIM for?
Three uses: automatic switch to the second card when the primary link fails, changing carrier without changing hardware, and, when the two cards belong to different carriers, avoiding one base station fault taking both paths down together. On the USR-G816 the second card is available as an embedded eSIM; the USR-G809s has dual Nano SIM with the second as an embedded eSIM.
7. Will data be lost at the moment the link drops?
It depends on which side handles retransmission. If it sits on the device side, the device sends its buffered data back by itself once the link returns; if it sits on the platform side, the retransmission logic has to move with it. Confirm this during commissioning: unplug the primary link once, time it, and see whether data is retransmitted or suspended during the wait.
8. How many devices can one unit carry, and how should monthly data be estimated?
Three numbers are not the same thing: the number of units online, the number of simultaneous connections, and the number of wireless clients, and the wireless side is often the one that hits its limit first. Data volume is the reporting interval times the size of one message times the number of points, plus a margin. Set the threshold at seventy to eighty percent of the plan, and have the three alerts for
