SLZB-06 vs a USB Coordinator on Ethernet: Placement Fixes the Stick Doesn’t

The USB coordinator is a good radio in a stupid place. It hangs off the back of a NUC, next to an NVMe enclosure, on a three-inch cable, inside a metal desk. LQI looks like a crime scene. The forum says buy an SLZB-06. The SLZB is a good radio you can put on a bookshelf with Ethernet. That is the product: placement, not magic silicon. If you mount the SLZB in the same rack closet as the USB stick used to live, you bought a PoE tax and kept the problem.

SLZB-06 versus a USB dongle on a long USB run or a networked serial is a placement and failure-domain argument. I have run Sonoff-style sticks on five-meter extensions and I have run an SLZB on PoE in a hallway. The hallway won. The chip family was not why.

What the SLZB actually changes

You move the 2.4 GHz radio off the USB 3 hash and off the host that reboots for Docker. Z2M or ZHA talks TCP to a radio that can sit where the mesh is, not where the compute is. If the stick was next to an SSD, start with USB 3 and Zigbee before you buy Ethernet. PoE means one cable. The web UI lets you flash and peek without crawling behind a PC. Those are real.

You do not get more routers. You do not get a new channel plan. You do not get immunity from a lamp-switch parent. You get a coordinator you can put in the geometric center of the apartment instead of the office corner. If your mesh already has routers and the only crime was USB 3, an extension cable is cheaper. If you cannot run USB that far without a flaky repeater, Ethernet wins.

When a USB stick plus a cable is enough

The host sits near the house center. You can use a quality USB 2 extension, ferrite, and distance from SSDs. The host does not reboot daily. You already have Z2M working. Buying SLZB to “upgrade firmware culture” is a vibe. Measure LQI at the desk after the extension. If the bedroom is still sad, you need routers or a radio in the bedroom’s direction, not a receipt.

USB also fails in ways Ethernet does not: device node flips after a reboot, pass-through into a VM that ate the stick, a hub that drops the serial. If your HA is a VM and the stick keeps vanishing from the guest, a networked coordinator is an integration fix, not only an RF fix. That is a valid SLZB reason that is still about placement of the serial endpoint, not about dBm.

When Ethernet placement is the whole point

The computer lives in a closet. The house lives everywhere else. A USB extension will not reach the hallway without becoming a project. PoE to a high shelf in the corridor is the project that matches the mesh. I put the SLZB where I would put a Wi-Fi AP if I only had one: visible, central, not in a rack.

Power independence is the other win. The NAS can die. The radio can stay up if PoE comes from a switch that is still on. Z2M on the NAS will still be down unless you split that too. An SLZB does not keep automations alive by itself. It keeps the mesh from pairing to a ghost if you recover Z2M quickly. If Z2M is down for an hour, bindings still work; HA automations do not. Know which failure you are buying around.

New failure modes you just purchased

The network is now in the radio path. A guest VLAN, an isolation flag, or a switch that blocks mDNS will make the SLZB disappear. Give it a static DHCP lease on the same LAN as Z2M, or a VLAN you route on purpose. Do not put the coordinator on the IoT VLAN “for security” and then wonder why HA cannot open the socket. The coordinator is infrastructure. Treat it like the AP.

Firmware and the web UI are another surface. Flash on a calendar, not on a whim after a forum scare. Take a backup of the Zigbee network before you flash, the same as a USB stick. An SLZB does not excuse skipping the backup. It makes you feel modern while you skip it.

Two coordinators on one channel if you “just test” the SLZB without migrating is how you split a mesh. Migrate or re-pair. Do not run USB and Ethernet radios as roommates.

Latency over Wi-Fi to an SLZB you powered with a cheap injector and then backhauled on the same 2.4 band is a joke I have seen in the wild. If you do not have Ethernet in the hallway, run Ethernet. A Wi-Fi-connected coordinator on 2.4 next to Zigbee is the USB 3 problem wearing a different hat. PoE is the point.

Migration without romance

Backup Z2M or ZHA. Move the network to the new radio using the documented path for your stack, or re-pair if the backup story is messy. Expect a night of devices that picked bad parents. Walk the house with a sensor. The SLZB will not reparent for you just because it is higher on the wall. It will hear better. Parents are still a mesh problem.

Keep the USB stick as a labeled spare with the last-known firmware. When the SLZB’s switch port dies, you will want a stick and an extension, not a shopping session.

Z2M and ZHA over the network

Zigbee2MQTT speaking to a remote adapter is a well-trodden path. You will set a host and port instead of a tty. That is easy. What is not easy is assuming the socket is as reliable as USB. A flapping switch port looks like a mesh storm. Watch the adapter connection in Z2M logs separately from device LQI. If the socket drops when the vacuum hits a cable, you have a physical Ethernet problem wearing a Zigbee costume.

ZHA can use networked coordinators too, with more opinions about which stacks are supported this month. Read the current docs for your exact SLZB firmware mode. Ember versus other firmwares is a support matrix, not a vibe. I pin the mode that Z2M or ZHA already documents and I do not dual-boot personalities weekly.

Security people will want TLS to the radio. Fine if you set it up. A coordinator on the LAN with an open serial TCP port is a new toy for anything that can reach that VLAN. That is another reason not to drop it on the IoT network next to the cheapest camera. Infrastructure VLAN, static lease, firewall to the HA host only if you are that person. Most households just need “not guest.”

Temperature and cheap injectors: an SLZB in an attic or a sealed plastic box will thermal-throttle or reboot. Hallway shelf, open air, not a south-facing window. PoE midspan bricks fail like all bricks. Label the port on the switch so you do not disable PoE during a “cleanup.” I have taken a mesh down with a VLAN prune that included the coordinator port. The USB stick would have survived that prune. Both designs have a string you can pull. Know which string.

A buying rule

Buy SLZB-class Ethernet if the compute cannot sit near the mesh and you can run a cable to a better room. Buy a USB stick and a long USB 2 cable if the host is already in a decent place and your only crime was USB 3. Buy routers either way if you have twenty batteries and two bulbs.

I would not buy Ethernet to fix a channel overlap with the apartment Wi-Fi. I would not buy Ethernet to fix ZHA versus Z2M. I would buy Ethernet to put the coordinator where a human stands, not where the GPU lives.

The close

The SLZB-06 is a placement device. A USB coordinator on a long, clean USB 2 run is also a placement device. Silicon is secondary. If the radio stays in the rack, you will write the same post next year with a new SKU in the title. Put the coordinator in the open, keep one radio, back up the network, and spend the rest of the money on always-on routers.

Placement fixes what the stick cannot. Ethernet is how you place it when USB cannot reach. That is the whole comparison. Everything else is a case and a web UI.

If you remember one test: pair a device in the worst room before you buy, with the USB stick on a long cable in the hallway, held in your hand. If that works and the closet fails, you do not have a chip problem. You have a map problem. The SLZB is a permanent version of your hand in the hallway. Buy it when you are tired of holding your hand there. Skip it when a cable would have done.