#1941: Why You Can't Zigbee-Wi-Fi Your House

The "mesh" promise fails when you hit the coordinator bottleneck. Here's why multiple hubs don't work like Wi-Fi.

0:000:00
Episode Details
Episode ID
MWP-2097
Published
Duration
24:39
Audio
Direct link
Pipeline
V5
TTS Engine
chatterbox-regular
Script Writing Agent
Gemini 3 Flash

AI-Generated Content: This podcast is created using AI personas. Please verify any important information independently.

The promise of Zigbee is seductive: a self-healing mesh network where every device makes the whole system stronger. But in practice, many smart home enthusiasts hit a wall where adding more devices makes their network slower, not better. The problem isn't the protocol itself—it's the fundamental architecture of how Zigbee networks are built and managed.

The Coordinator Bottleneck
At the heart of every Zigbee network sits a single coordinator—the "brain" that manages the entire system. While modern coordinators like the SMLIGHT SLZB-06 use powerful chips like the TI CC2652P, they still face hard limits. A coordinator can typically handle only 32-50 direct connections before it must rely on routers to offload traffic. The total network capacity for these chips caps out around 200 devices, and even high-end models using newer EFR32 chips might only push to 300-400 devices.

But here's the catch: just because a chip can hold 400 addresses in its routing table doesn't mean it can process 400 devices simultaneously. The coordinator must store routing information, manage encryption keys, and handle incoming packets in real-time. When it runs out of processing capacity, you start seeing "Adapter Busy" errors and latency spikes—your kitchen lights might take two seconds to turn on after a motion sensor triggers.

The Router Reality
Many users misunderstand what makes a mesh network actually work. Battery-powered devices like motion sensors and door contacts are almost always "End Devices"—they don't repeat signals because that would drain their batteries in days. Only mains-powered devices like smart plugs and wired switches act as routers, forming the backbone of your mesh.

This creates a critical vulnerability: if you use smart bulbs as routers and someone flips the physical wall switch off, that entire section of your mesh disappears. Sensors that were routing through that bulb suddenly become stranded, frantically searching for new parents and flooding the network with "Where am I?" packets. This "routing storm" can bring a coordinator to its knees.

The Hidden Killer: Chatter
Even if you have enough routers, the real limit often comes from device "chattiness." Zigbee operates at a mere 250 kilobits per second shared across the 2.4 GHz band. A single power-monitoring plug that reports energy usage every half-second can consume more bandwidth than a dozen temperature sensors checking in every ten minutes. When multiple high-frequency devices compete for airtime, packet collisions occur, and the coordinator spends all its time requesting retransmissions rather than processing new commands.

The Wi-Fi Comparison Fallacy
Perhaps the most common misconception is treating Zigbee like Wi-Fi. In Wi-Fi, you can add multiple access points that all connect to the same router, and your devices roam between them seamlessly. Zigbee doesn't work this way because of its fundamental architecture.

Every Zigbee network has a unique PAN ID (Personal Area Network Identifier), and by protocol definition, there can be only one coordinator per PAN ID. The coordinator holds the encryption keys and acts as the "Trust Center"—the source of truth for the entire network. If you plug in a second coordinator, it starts an entirely separate network with different keys and a different PAN ID. Your devices on Network A cannot use routers on Network B. They're two completely separate silos.

The Professional Solution
The correct way to scale beyond 200 devices is to run multiple independent Zigbee networks on different channels—one on Channel 11, another on Channel 25, for example. Both networks can report to the same Home Assistant or MQTT broker, which acts as a translator between them. This allows you to create automations where a button on Network A controls a light on Network B, but it requires careful management. You need to remember which device belongs to which hub, and adding a new sensor requires deciding which network it should join.

This approach works, but it's not the seamless "corporate campus" experience many users imagine. It's more like managing two separate countries that happen to have diplomatic relations.

Key Takeaways

  • Zigbee networks are fundamentally limited by their coordinator's processing capacity and routing table size
  • Only mains-powered devices act as routers; battery devices are endpoints that don't repeat signals
  • High-frequency reporting devices can saturate Zigbee's limited bandwidth, causing latency and packet loss
  • Multiple coordinators create separate networks; they cannot share load like Wi-Fi access points
  • Professional scaling requires running independent networks on different channels and managing them as separate entities

The bottom line: Zigbee is powerful but has real physical and architectural limits. Understanding these constraints helps you design a network that actually works rather than one that sounds good on paper.

Downloads

Episode Audio

Download the full episode as an MP3 file

Download MP3
Transcript (TXT)

Plain text transcript file

Transcript (PDF)

Formatted PDF with styling

#1941: Why You Can't Zigbee-Wi-Fi Your House

Corn
You ever buy a piece of tech because the marketing promised you a self-healing, infinite web of connectivity, only to find yourself staring at a smart bulb that won't turn on while you're standing three feet away from it? It is the classic smart home betrayal. You’re holding your phone, you’re pressing the virtual button, and nothing. It feels like the digital equivalent of being ghosted by your own living room.
Herman
Herman Poppleberry here, and Corn, you are touching on the entire "mesh" industrial complex. The marketing for Zigbee and Z-Wave makes it sound like magic—this idea that because it is a mesh, it is inherently invincible. People treat it like a digital sourdough starter where you just keep adding flour and it grows forever, but in reality, it is a very finite resource with some hard physical and architectural ceilings. It’s like a highway system; just because you can add more lanes doesn't mean the interchange at the center can handle ten thousand cars a minute. Eventually, the toll booth—which is your coordinator—becomes the bottleneck. Think about a crowded cocktail party. If five people are talking, everyone hears everyone. If five hundred people are in that same room, even if they are all "repeating" what they hear, the noise floor rises so high that the person standing next to you might as well be on Mars.
Corn
Well, today’s prompt from Daniel is about exactly that. He is asking about Zigbee coordinators and the actual limits of scaling these networks. He specifically mentions the SMLIGHT series, which we both know is the current darling of the power-user community, and he wants to know when these things actually start to sweat. Plus, the million-dollar question: can we treat multiple coordinators like Wi-Fi Access Points to blanket a mansion in Zigbee goodness? Daniel is essentially asking if he can build a "corporate campus" level of connectivity in a residential setting.
Herman
It is a phenomenal prompt because it hits that intersection of "it sounds simple on the box" and "it is incredibly complex once you have a hundred devices." By the way, today's episode is powered by Google Gemini 1.5 Flash, which is fitting since we are talking about high-performance coordination and processing at the edge.
Corn
Let’s start with the "mesh" part. The marketing says every device makes the network stronger. Daniel calls this out as having a "big caveat." So, Herman, why isn't my battery-powered motion sensor making the network stronger for my door lock?
Herman
This is the fundamental misunderstanding of Zigbee roles. In a Zigbee network, you have three types of players: the Coordinator, which is the brain; the Routers, which are the backbone; and the End Devices, which are the leaves. The caveat Daniel is talking about is that battery-powered devices—your sensors, your buttons, your contact triggers—are almost always End Devices. They do not repeat signals. They are selfish. They wake up, whisper their data to a parent, and go back to sleep to save battery. If they acted as routers, they'd be dead in three days because they'd have to keep their radios listening 100% of the time to catch and pass on messages from their neighbors. Imagine a bucket brigade where half the people are taking a nap; the water isn't moving very far.
Corn
So if I have a house full of fifty battery sensors and one hub, I don't actually have a mesh. I have a very crowded star topology. It’s basically fifty people trying to shout at one librarian at the same time.
Herman
Precisely. You only get a mesh if you have Routers, which are typically mains-powered devices like smart plugs, wired switches, or some bulbs. But even then, there is a "bad neighbor" problem. If you use certain smart bulbs as routers—especially older ones like the famous Sengleds or even some early Hue bulbs—and someone flips the physical wall switch off, that part of your mesh just... dies. It is like a bridge that disappears whenever someone turns off the lights on the deck. All the sensors that were "talking" through that bulb are suddenly stranded. They start frantically searching for a new parent, which drains their battery and floods the airwaves with "Where am I?" packets. This is what we call a "routing storm." It’s the digital equivalent of a GPS recalibrating while you’re driving sixty miles an hour through a tunnel.
Corn
That is why people buy those dedicated IKEA or Aeotec repeaters, right? They are just "always-on" routers that don't have a light bulb attached to them that might get switched off by a confused guest. They’re like the permanent infrastructure of the city.
Herman
Right. They are pure infrastructure. But even with good routers, you hit the Coordinator bottleneck. Daniel mentioned the SMLIGHT series, specifically things like the SLZB-06. These use the Texas Instruments CC2652P chip, which is a beast compared to the old CC2531 sticks we used to use, but it still has limits. Think of the CC2531 like a 1990s pager, and the SMLIGHT like a modern smartphone. It’s better, but you still can’t run a whole corporation off one phone.
Corn
But why is that? If the SMLIGHT has a powerful antenna and a fast processor, why can't it just handle everything? Is it a memory issue or a radio frequency issue?
Herman
It’s both. Every device in a Zigbee network needs an entry in the "Routing Table" and the "Neighbor Table" of the coordinator. These tables are stored in RAM on the Zigbee chip itself. If you have 200 devices, the coordinator has to remember the best path to reach every single one of them. If that table fills up, the coordinator literally forgets how to talk to your back door sensor. It has to drop an old entry to make room for a new one. It’s like trying to memorize a phone book; eventually, names start falling out of your head.
Corn
Okay, let's get into the numbers. If I’m running an SLZB-06, which is a high-end consumer coordinator, how many devices are we talking about before the wheels fall off? Is it fifty? Five hundred? Daniel seems worried that even with high-end gear, there’s a "glass ceiling" he’s going to hit.
Herman
It is less about the total number and more about the "Direct Child" limit versus the "Network" limit. A coordinator like the SLZB-06 can usually only talk to about thirty-two to fifty devices directly. If you want more than that, you must have routers to offload those connections. Now, the total network capacity for that specific chip is technically around two hundred devices. If you step up to the newer SLZB-06M or the Ultra models using the EFR32 or newer chips, you might push three hundred or four hundred. But—and this is a huge but—just because the chip can hold 400 addresses in its table doesn't mean it can process the traffic of 400 devices simultaneously.
Corn
Two hundred sounds like a lot until you realize a modern smart home has a sensor on every window, every door, motion in every room, leak sensors under every sink, and then you add the lights. It adds up fast. I mean, a three-bedroom house could easily hit 150 devices if you’re thorough. But what does "overwhelmed" actually look like? Does the coordinator just catch fire?
Herman
I wish it were that dramatic. It’s actually much more annoying. It starts with latency. You walk into the kitchen, the motion sensor triggers, and the lights come on... two seconds later. You check the logs in Zigbee2MQTT or Home Assistant, and you see "Adapter Busy" or "MAC No ACK" errors. Basically, the coordinator is trying to manage the routing tables, handle the security keys, and process incoming packets, and it just runs out of air. It’s like a waiter trying to take orders for 50 tables at once; eventually, he just stands in the middle of the floor staring into space because he can't decide who to help first.
Corn
Is it just the count, or is it the "chattiness" of the devices? Because I imagine a power-monitoring plug that reports every half-second is a lot heavier than a temperature sensor that checks in once every ten minutes.
Herman
That is the hidden killer. If you have fifty plugs reporting energy usage constantly—showing you every single watt your fridge or TV is pulling—you are saturating the bandwidth. Zigbee is a low-bandwidth protocol. We are talking two hundred and fifty kilobits per second shared across the entire 2.4 gigahertz channel. Compare that to Wi-Fi 6, which is gigabits. When you have high-frequency reporting, you get packet collisions. The coordinator spends all its time asking for retransmissions. It’s like trying to have a conversation in a room where everyone is shouting their bank balance every five seconds. You can’t hear the person trying to tell you the house is on fire.
Corn
So, the "just add more repeaters" advice might actually make things worse if you’re already at the bandwidth limit? I’ve seen people on forums say "My network is slow, so I added five more smart plugs," and then they wonder why it got even slower.
Herman
It can. Every hop a packet takes through a repeater adds latency and increases the chance of a collision. If a signal has to jump through four bulbs to get to the coordinator, that is four times the airtime used for a single message. It is why you want a "flat" mesh where most things are only one or two hops away from the coordinator, rather than a long chain. If you have a long, skinny house and your coordinator is at one end, and a signal has to hop through six devices to reach the other end, that’s six opportunities for that packet to be dropped or delayed. Think of it like a game of telephone; the more people in the line, the more likely the message is going to be "The cat is on the mat" by the time it reaches the end, instead of "The stove is on fire."
Corn
Okay, so let's tackle Daniel's big question. The Wi-Fi analogy. In Wi-Fi, if my coverage is bad, I add another Access Point. They all talk to the same router, and my phone roams between them. If I'm in the garage, I'm on the Garage AP; if I'm in the bedroom, I'm on the Bedroom AP. Can I just stick three SMLIGHT coordinators around my house and have one giant, happy Zigbee family?
Herman
Short answer: No. Long answer: Absolutely not, and trying will make you go bald from stress.
Corn
(Laughs) Tell us how you really feel, Herman. Why not? It seems so logical. If I have an Ethernet-connected SMLIGHT in the basement and another one in the attic, why can't they just share the load?
Herman
Because of how the Zigbee protocol is architected. A Zigbee network is defined by a unique ID called a PAN ID (Personal Area Network Identifier). By protocol definition, there can be only one Coordinator per PAN ID. The Coordinator is the "Trust Center." It holds the encryption keys. It decides who is allowed to join. It’s the "Source of Truth" for the entire network map. You cannot have two "captains" on the same ship. If you plug in a second coordinator, it starts its own, entirely separate network. It will have a different PAN ID, different encryption keys, and it will essentially be a stranger to the first network.
Corn
But wait, could I run two separate networks and just have them both feed into the same Home Assistant or MQTT broker? Like, "Upstairs Network" and "Downstairs Network"?
Herman
You can, and that is actually the professional way to scale beyond the two hundred device limit. You run Network A on Zigbee Channel 11 and Network B on Channel 25. They are invisible to each other. Your devices on Network A cannot use routers on Network B. They are two different silos. But since they both report to the same "brain" like Home Assistant, you can still create an automation where a button on Network A turns on a light on Network B. Home Assistant acts as the translator between two different foreign countries.
Corn
That sounds like a management nightmare for the average user, though. You have to remember which device is paired to which hub. If I buy a new sensor, I have to think, "Okay, am I in the kitchen? That’s Hub B." It’s not "roaming" like Wi-Fi.
Herman
There is no roaming in Zigbee. Once a device is paired to a coordinator, it is married to it. It stores that specific coordinator's long address. If you move it to the other side of the house where a different coordinator has a stronger signal, it won't care. It won't "see" the other coordinator as a valid parent. It will keep trying to talk to its original coordinator until it dies of exhaustion or you manually reset it and pair it to the new one. This is why the Wi-Fi comparison falls apart. Wi-Fi is designed for mobile clients; Zigbee was designed for static industrial sensors.
Corn
So, if I have a massive property, like a farm or a mansion, and I can't do the "multiple AP" thing easily, what is the alternative? Daniel mentions using a coordinator in "Router Mode." Is that the secret sauce? Can I turn one of those powerful SMLIGHTs into a super-repeater?
Herman
That is a very clever workaround. You can take a high-gain device like an SLZB-06, flash it with Router firmware instead of Coordinator firmware, and place it in a far-off garage. It connects back to your main coordinator over the Zigbee mesh, but because it has a massive antenna and a powerful radio, it acts as a "super-router" for all the little sensors in that garage. It gives you that "hub" feel without breaking the "one coordinator" rule. It basically acts as a massive megaphone for the main coordinator.
Corn
But it’s still limited by that one main coordinator’s processing power and the 2.4 gigahertz interference, right? You’re still funneling everything back to that single brain.
Herman
Yes. You are still sharing the same narrow pipe. This is why channel planning is so critical. Zigbee lives in the same 2.4 gigahertz neighborhood as Wi-Fi and Bluetooth. There are only sixteen Zigbee channels, and only a few of them—like 15, 20, and 25—usually fall in the "gaps" between standard Wi-Fi channels 1, 6, and 11. If you scale up to 200 devices and you don't plan your channels, you aren't just fighting your coordinator's CPU; you are fighting your neighbor's Netflix stream. If your Wi-Fi is on Channel 6 and your Zigbee is on Channel 17, they are literally overlapping and screaming over each other.
Corn
I love that we’re talking about "channel planning" for light bulbs. It really shows how far we've come from just screwing in a piece of glass and wire. It’s like being a radio frequency engineer just to make sure the bathroom light turns on. But okay, if I’m an enthusiast and I’m hitting that 150-device mark, and things are getting twitchy—maybe a light takes three clicks to respond—what is my move? Do I start pruning devices, or do I go to the "silo" method?
Herman
The first move is an audit. Look for the "chatties." This is the step most people skip. Go into your Zigbee2MQTT dashboard and look at the "Last Seen" and the message frequency. Do you really need your washing machine plug to report voltage changes every two seconds? Probably not. You only need to know when the cycle is done. If you increase the reporting threshold—say, only report if the wattage changes by more than 10 watts—you can cut your network traffic by 80% overnight. That clears up airtime for the stuff that matters, like motion sensors.
Corn
That makes a lot of sense. It’s about signal-to-noise ratio. But what about the physical environment? I’ve heard people say that even the material of your walls can dictate when you hit that "glass ceiling" Daniel is worried about.
Herman
Oh, absolutely. If you live in a modern house with drywall and wood studs, Zigbee is fairly happy. But if you live in an old European home with two-foot-thick stone walls, or a loft with reinforced concrete and rebar, your mesh is going to struggle. Rebar acts like a Faraday cage. In those environments, you might hit the "scaling limit" at only 40 devices because the coordinator is constantly re-routing packets that can't penetrate the walls. You end up with a very "deep" mesh with five or six hops, which, as we discussed, is a recipe for instability.
Corn
So in that case, Daniel’s idea of multiple coordinators—the "silo" method—might actually be a necessity rather than a luxury. If you can’t get a signal through the wall, you put a dedicated "embassy" in that room connected via Ethernet.
Herman
Ethernet is your best friend when physics is your enemy. If you have a detached garage or a basement bunker, don't try to mesh your way there. Run a Cat6 cable and put a second PoE SMLIGHT in that room. It’s a separate Zigbee network, but Home Assistant makes it feel like one.
Corn
Let’s talk about the hardware for a second. Daniel mentioned the SMLIGHT series. Why are these so popular compared to, say, a $15 USB stick plugged into a Raspberry Pi? Is it just the antenna, or is there more under the hood?
Herman
It’s the delivery method. A standard USB stick has to be physically plugged into your server. If your server is in a metal rack in the basement, or tucked behind a desk, your Zigbee signal is already starting at a disadvantage. Metal and concrete are Zigbee killers. The SMLIGHT units are Power-over-Ethernet (PoE). You can put them anywhere you have an Ethernet jack. You can put them in the center of your house, on the ceiling, or in a hallway, and they talk to your server over your rock-solid Ethernet backhaul. It decouples the "radio" from the "computer."
Corn
That seems like a massive win for scaling. It lets you put the "ears" of the network where people actually live, not where the noisy fans and the server rack are. It’s like having a central microphone in the middle of the room instead of trying to listen through a keyhole from the basement. And they have those fancy web interfaces too, right?
Herman
They do. You can log directly into the SMLIGHT’s IP address and see the health of the Zigbee chip itself, independent of your automation software. It tells you the chip temperature, the RAM usage, and even has a built-in radio frequency signal analyzer. For someone like Daniel who is worried about hitting the ceiling, that data is gold. You can actually see if the chip is "sweating" before the lights stop working.
Corn
So, looking forward, do you think this is where Thread and Matter change the game? Because Daniel’s frustration with the "one coordinator" limit seems like a fundamental flaw in Zigbee’s aging design. Thread is essentially "Zigbee but with IP addresses," right? Does it solve the "one coordinator" problem?
Herman
Thread is a huge step forward because it is designed to be multi-hub from day one. In a Thread network, you have things called "Border Routers." If you have an Apple TV, a HomePod, and a Google Nest hub, they all act as Border Routers for the same Thread network. It is much more like the Wi-Fi mesh dream Daniel was asking about. If one hub fails, the others take over seamlessly. There is no "single point of failure" brain like there is in Zigbee. It’s a more democratic system.
Corn
So Zigbee is the aging rock star who is great at a small club but struggles with a stadium tour, while Thread is the new kid designed for the arena?
Herman
That is a great way to put it. Zigbee is incredibly mature, it is cheap, and the device selection is massive. You can buy a Zigbee sensor for five bucks. But it’s a "single-point-of-failure" architecture by design. Thread fixes the "captain" problem by allowing multiple leaders on the same team. However, Thread is still in its awkward teenage years. The reliability isn't quite there yet, and the "Border Router" handoffs can still be a bit flaky depending on which brand of hub you're using. We’ve all seen the "Matter" horror stories where a Google hub and an Apple hub fight over who gets to control the light bulb.
Corn
But we're in 2024, and I still have a drawer full of Zigbee sensors that work perfectly fine. I'm not replacing them all for Thread just yet. It feels like Zigbee still has a lot of life left if you know how to treat it. I mean, if it’s not broken, why fix it? Especially when Zigbee devices are so much cheaper right now.
Herman
Nor should you replace them! A well-tuned Zigbee network is still faster and more reliable for most people than an early-stage Thread setup. It just requires that "Junior Admin" mindset we talked about. Don't trust the "self-healing" marketing blindly. Be intentional about your routers. If you have a dead spot, don't just throw a light bulb at it—throw a dedicated, high-quality repeater. Keep your coordinator in a central location, ideally away from your Wi-Fi router, and don't be afraid to split the network if you're turning your house into a literal computer.
Corn
I think the practical takeaway for Daniel and the listeners is: check your device count. If you're under a hundred, you're probably fine with one good SMLIGHT coordinator and a few decent plugs as routers. If you're approaching two hundred, or if you see that "Adapter Busy" error, start looking at that second coordinator on a different channel. And for the love of all that is holy, stop using your smart bulbs as the primary backbone of your security system. If a guest turns off the lamp, your front door sensor shouldn't go offline.
Herman
(Laughs) Yes! Buy a five-dollar dedicated repeater, plug it into an outlet that isn't controlled by a switch, and save yourself the headache when someone accidentally hits the light switch. It’s the cheapest insurance you can buy for your sanity. And remember, "more devices" doesn't always mean "better mesh." Sometimes, "more devices" just means "more noise."
Corn
This was a deep one. I feel like I need to go home and re-map my entire mesh now. I noticed my kitchen motion sensor took a beat longer than usual this morning, and now I’m wondering if my toaster is being too "chatty." Or maybe my neighbor just got a new Wi-Fi router that’s stomping all over my Zigbee Channel 11.
Herman
I already did mine this morning. I found out my smart blinds were reporting their position every time the sun moved a millimeter. Every single percent of movement was a packet. I changed it to report only every five percent, and my motion sensors are back down to twelve milliseconds of latency. It’s beautiful, Corn. The lights come on before I even realize I’ve walked into the room. It feels like living in the future again instead of living in a laggy video game.
Corn
Of course you did. You probably have a dashboard for your dashboard. I bet you have a real-time graph of your Zigbee packet success rate projected on your bedroom wall.
Herman
Don't give me ideas, Corn. I have a spare projector and a lot of empty wall space.
Corn
Well, thanks to Daniel for the prompt. This was a great dive into the "mesh" of it all and a good reality check on the marketing versus the physics. It’s easy to get caught up in the hype of "infinite scaling," but at the end of the day, we’re still dealing with radio waves in a very crowded 2.4GHz spectrum.
Herman
Before we sign off, a big thanks to our producer, Hilbert Flumingtop, for keeping the gears turning behind the scenes and making sure our own "network" stays up. He’s the real coordinator of this operation.
Corn
And a huge thanks to Modal for providing the GPU credits that power the AI behind this show. Without them, we’d be doing this with a pen and paper, and nobody wants to hear that. It would be a very quiet podcast.
Herman
This has been My Weird Prompts. If you enjoyed this dive into Zigbee physics and network architecture, leave us a review on Apple Podcasts or Spotify—it really helps other nerds find the show. We love reading the reviews, even the ones that tell us we’re too geeky.
Corn
You can find all our episodes, the transcripts, and the RSS feed at myweirdprompts dot com. See you in the next one, hopefully with zero latency and a perfectly balanced mesh.
Herman
See ya. Keep those packets flowing.

This episode was generated with AI assistance. Hosts Herman and Corn are AI personalities.