[Prompt]
custom: Let's talk about wifi planning. The rookie way to roll out a Wi-Fi network is to use what the ISP pr | Hosts: corn, herman

[Response]
Corn: Daniel sent us this one — he wants to talk about Wi-Fi planning, and he's basically framed it as two ways to get it wrong. The first is the hope-and-pray method: you plug in whatever plastic box your ISP gave you, stick it in a corner, and assume it'll cover the whole house. Spoiler — it doesn't. The second is the overcorrection: you buy a three-pack mesh system, scatter nodes everywhere, and accidentally turn your network into a traffic jam of backhaul channels fighting each other. The smarter way, he says, is to approach it like an architect — actually plan your coverage, take real measurements, and only then buy hardware.

Herman: The timing on this is perfect, because we're now at the point where Wi-Fi seven gear is actually on shelves — and the cost of guessing wrong has never been higher.

Corn: Explain that part. Why does Wi-Fi seven make bad planning more expensive?

Herman: Wider channels, more bands, more backhaul options — all of which sound great until you realize they're also more ways to accidentally sabotage yourself. Wi-Fi seven introduces three hundred twenty megahertz channels. That's enormous bandwidth, but it also means the six gigahertz band goes from having about seven non-overlapping channels at one sixty megahertz down to just three. So if you're the type who buys a mesh system and scatters four nodes around a twelve hundred square foot apartment, those nodes are now fighting over an even smaller pool of clean airspace.

Corn: The margin for error shrinks as the capability expands. That's a neat little trap.

Herman: It's the classic "more rope to hang yourself with" situation. And here's the thing — most people don't even know they're in the trap. They see "Wi-Fi seven" on the box, they see "mesh" on the box, they assume the technology will figure it out. But the physics doesn't care about the marketing.

Corn: The router can't out-negotiate a concrete wall.

Herman: And that's really the heart of what Daniel's getting at — there's this whole discipline of Wi-Fi planning that treats your house like a coverage problem to be modeled and measured, not a guessing game. You map the space, you account for what the walls are made of, you figure out where signal actually goes and where it doesn't, and only then do you decide how many access points you need and where they go.

Corn: Which is the exact opposite of how ninety-nine percent of people do it. Most folks buy the hardware first, then wander around the house holding their phone up like a dowsing rod, trying to find the one spot where Netflix doesn't buffer.

Herman: The dowsing rod approach. I've seen it. I've done it. I've been that person in my own home before I knew better.

Corn: We've all been that person. The question is whether we stay that person in the Wi-Fi seven era, when the stakes are higher and the interference landscape is more crowded.

Herman: That's what makes this worth digging into. Because the tools to do this properly have gotten genuinely accessible. You don't need to be a network engineer with a two thousand dollar Ekahau license anymore. There are apps on your phone right now that can give you eighty percent of the insight you need. The gap between "I guess this is fine" and "I know exactly what my network is doing" has never been smaller.

Corn: Let's break down what actually goes wrong when you don't plan — and why the fix isn't just buying more hardware.

Herman: There are really two distinct failure modes here, and they're almost opposites of each other. The first one is what I'd call the "hope and pray" approach — and this is probably eighty percent of households. You get the router from your ISP, you plug it in wherever the cable comes into the house, and you assume it'll cover everything. It doesn't. The signal hits a concrete wall or a floor and drops twenty decibels, and suddenly your bedroom is a dead zone, your garage may as well be on the moon, and you're also getting hammered by your neighbor's router blasting away on the exact same channel.

Corn: Co-channel interference from the guy next door who also plugged his ISP router into the corner and called it a day. It's a tragedy of the commons played out in two point four gigahertz.

Herman: And then there's the second failure pattern, which is the overcorrection. Someone reads a blog post about mesh systems, decides more is always better, and drops four or five nodes into a twelve hundred square foot apartment. Each node is transmitting backhaul traffic to the others, and if they're too close together, those backhaul transmissions start competing with the actual client traffic your devices are trying to send. You've spent four hundred dollars to make your network worse.

Corn: Under-provisioning gives you dead zones, and over-provisioning gives you a network that's technically alive but functionally useless. Both come from the same root problem — you never measured anything.

Herman: And that's the mindset shift Daniel's pointing toward. Wi-Fi planning isn't really about signal strength in isolation. It's about understanding how radio waves propagate through your specific space — what the walls are made of, where the interference sources are, what your actual client density looks like. Signal strength, or RSSI, is just one variable. You also need to care about signal-to-noise ratio, channel utilization, packet error rates. A strong signal on a crowded channel can perform worse than a moderate signal on a clean one.

Corn: The architect's approach isn't "how many bars do I have," it's "what is the actual usable throughput at every point where I'm going to sit down with a device.

Herman: And you model that before you spend a dollar on hardware. You map the floor plan, you account for attenuation through walls and floors, you figure out where the dead zones actually are, and then you decide how many access points you need and where they go. You validate with real measurements afterward. It's the difference between guessing and knowing.

Corn: Guessing got us through the Wi-Fi five and six eras because the penalties were smaller. Fewer bands, simpler channel layouts, less crowding. But with Wi-Fi seven and those three twenty megahertz channels you mentioned, guessing wrong doesn't just leave you with a slow corner of the living room — it can tank the whole network.

Herman: The tools to do this properly have quietly gotten good enough that there's no real excuse anymore. But we'll get to those.

Herman: To understand how to plan properly, we have to look under the hood at how Wi-Fi signals actually behave in a real building. And the tools that do this well — Ekahau, NetSpot — they're not just drawing pretty heatmaps. They're running physics models.

Corn: You're saying my ISP router doesn't do this.

Herman: Your ISP router is a hopeful little box with an antenna and a dream. These planning tools use something called RF propagation modeling. At its core, it's ray-tracing — the same idea used in computer graphics, except instead of tracing light rays bouncing off surfaces, you're tracing radio waves passing through walls, floors, furniture. The tool needs to know what those obstacles are made of, because different materials attenuate the signal differently.

Corn: Attenuate meaning weaken.

Herman: And the numbers are surprisingly specific. The ITU indoor propagation model — that's the International Telecommunication Union, they maintain the standard reference — gives you per-material loss figures. Drywall, about three decibels. Glass, around five. A concrete wall? That's a factor of thirty-two in terms of actual power reduction.

Corn: One concrete wall turns a perfectly good five gigahertz connection into something that barely works.

Herman: This is exactly what Ekahau's predictive heatmaps show. You import a floor plan, you assign building materials to each wall — this one's drywall, this one's concrete, this one's glass — then you place virtual access points on the map and run the simulation. The tool calculates expected signal strength at every point in the space, accounting for every wall the signal has to punch through. There's a classic example from their documentation: you place an AP, everything looks fine, then the heatmap hits a concrete wall and you see an immediate twenty decibel cliff. The planned five gigahertz connection drops straight to two point four gigahertz fallback on the other side of that wall.

Corn: Two point four is the slow lane — more range, but crowded and capped on throughput.

Herman: And here's where it gets more interesting. Ekahau doesn't just show you signal strength. That heatmap is actually calculating expected data rates at every point. Not bars — megabits per second. Because their whole philosophy, and they've written about this extensively, is that coverage alone doesn't guarantee performance.

Corn: Coverage versus performance.

Herman: Coverage is RSSI — received signal strength indicator. How loud the router is at your device. Performance is whether you can actually stream video or load a webpage. And the gap between those two things is where most home networks die. You need to also care about SNR — signal-to-noise ratio — which is how much louder your router is than the background noise and interference. You need to care about channel utilization, which is what percentage of airtime on a given channel is already occupied by other devices and neighboring networks. You need to care about packet error rates — how many frames are arriving corrupted and need to be retransmitted.

Corn: You could have a screaming-loud signal at minus fifty dBm, but if the noise floor is also high, or the channel is saturated, or you're getting a ten percent packet error rate, that minus fifty is meaningless.

Herman: And NetSpot's survey mode makes this brutally visible. They have a signal-to-noise ratio view where you can walk your space and see, point by point, what the actual SNR is. There's a scenario that shows up all the time: you're standing in a spot with full bars, RSSI looks great, but NetSpot reveals you've got adjacent-channel interference bleeding over from a neighbor's router parked on channel six. Your SNR is five dB. That's terrible. Meanwhile, a spot two rooms over with a weaker signal but twenty-five dB of SNR will perform dramatically better.

Corn: The bars on your phone are lying to you constantly.

Herman: The bars are a single number trying to summarize a multi-variable equation. It's like judging a car by how loud the engine is.

Corn: Which brings us to the other half of this — the coverage versus capacity tradeoff. You mentioned it earlier, but walk me through the physics of what happens when too many clients pile onto one access point.

Herman: Wi-Fi is a shared medium. Only one device can transmit at a time on a given channel. So if you've got a single AP in a room showing minus fifty dBm everywhere — great coverage — but thirty clients are all trying to talk to it simultaneously, they're taking turns. Each device gets a fraction of the available airtime. Throughput per device collapses. It doesn't matter that the signal is strong. The channel is the bottleneck.

Corn: Proper planning means deciding upfront whether a given zone needs blanket coverage or high-density capacity.

Herman: That's the distinction. Blanket coverage is for IoT sensors, smart thermostats, things that send tiny packets occasionally and just need to stay connected. You can stretch one AP across a lot of square footage for that. High-density capacity is for a living room where someone's streaming four K video, someone else is gaming, and a third person is on a video call — all simultaneously. That zone might need its own AP even if the signal from the next room technically reaches it. Because "technically reaches" and "actually works under load" are two completely different things.

Corn: The planning tools let you model both scenarios before you drill a single hole in the wall.

Herman: That's the whole point. You place virtual APs, you set your client density assumptions, you run the predictive model, and you see where the bottlenecks are going to be before they exist. NetSpot's workflow is a bit more accessible for home users — you upload a map image, you trace the walls, you drop APs onto it, and it generates the heatmap overlay. Their home tier is forty-nine dollars a year, which includes both the predictive mode and the survey mode for walking the space afterward. Ekahau is the professional-grade tool — their license runs north of two thousand dollars, but it gives you packet-level metrics, full protocol analysis, and integration with their purpose-built Sidekick scanner. Different tools for different needs, same underlying physics.

Corn: The predictive model is step one. But you said something important a minute ago — you validate with real measurements afterward. The model is a hypothesis. You still have to test it.

Herman: That's where the over-provisioning trap gets especially ugly, because it's not just a predictive modeling problem — it's something you can actually measure and watch happen in real time.

Corn: Walk me through the physics of what happens when someone puts mesh nodes too close together. You mentioned backhaul interference earlier, but what's actually colliding with what?

Herman: Think of it this way. In a mesh system, every node has two jobs. It has to talk to your devices — that's the client-facing traffic — and it has to talk to the other nodes to pass that traffic back to the main router. That second job is the backhaul. If you place nodes fifteen feet apart, they're screaming backhaul traffic at each other on the same channel, and your phone trying to load a webpage is now competing with a node-to-node conversation happening at the exact same frequency.

Corn: The mesh is essentially shouting over its own conversation.

Herman: Wi-Fi seven makes this worse. Those three hundred twenty megahertz channels I mentioned — they're spectacular for throughput, but they also mean the six gigahertz band now has only three non-overlapping channels. If your mesh backhaul is on six gigahertz and your client traffic is also on six gigahertz, and you've got nodes close enough that their coverage circles overlap heavily, you've just created a self-jamming network.

Corn: The wider the channel, the fewer lanes available, and the more carefully you have to plan who's driving in which lane.

Herman: And here's a concrete case that shows up all the time. Fifteen hundred square foot house, someone buys a three-node Wi-Fi six E mesh system. They unbox it, put one node in the living room, one in the hallway, one in the bedroom — ends up being about fifteen feet between each node in a rough triangle. Looks reasonable on paper. But the backhaul on six gigahertz is saturating the band, and clients on five gigahertz are seeing fifty percent throughput loss. Not because the hardware is bad — because the geometry is bad.

Corn: What's the fix? Just move them further apart?

Herman: That's what the measurement tells you. Same person runs a NetSpot survey, walks the house with their phone, takes readings in every room at seated height — not floor level, that's important, because Wi-Fi behaves differently at the height where your devices actually are. They discover that the far end of the house, about forty feet from the nearest node, is actually still getting a usable signal. So they move one node out there. Now the backhaul has enough physical separation that the signals aren't stomping on each other, and client throughput doubles.

Corn: From moving one box forty feet.

Herman: From measuring instead of guessing. And the survey itself is not complicated. You walk the space systematically — every room, at the height where you actually use devices, which is seated height, roughly three to four feet off the ground. You note every spot where signal drops below minus seventy dBm, because that's the typical threshold where reliable video streaming starts to fall apart. Netflix buffering, Zoom freezing — that's the minus seventy cliff.

Corn: You don't need a two thousand dollar Ekahau license to do this.

Herman: You absolutely don't. NetSpot's home tier is forty-nine dollars a year, and it gives you both the predictive heatmapping and the survey mode. On Android, WiFi Analyzer is free and gives you RSSI per channel, channel utilization, a real-time graph of signal strength as you walk. Is it as precise as an Ekahau Sidekick? The Sidekick has calibrated antennas — it can measure packet-level metrics like retries and CRC errors, which a phone's Wi-Fi chip simply doesn't expose. A phone gives you RSSI and channel noise, but not packet loss.

Corn: For someone trying to fix their home network, phone-level data is probably enough.

Herman: More than enough. It's eighty percent of the insight for zero percent of the cost. The key is being systematic about it. Don't just wave your phone around. Walk to each spot where you actually sit down with a device, hold the phone at seated height, wait a few seconds for the reading to stabilize, note the number. If you're below minus seventy, that's where you need another node — or where you need to reposition one you already have.

Corn: The workflow is: predictive model first to figure out roughly where things should go, then walk the space with a phone to validate, then adjust. Not buy a mesh system and hope the app's auto-placement wizard knows what your walls are made of.

Herman: That auto-placement wizard, by the way, has no idea. It's optimizing for signal strength between nodes, not for what happens to your client traffic when the backhaul saturates. It's a perfect little machine for generating the over-provisioning failure pattern we've been talking about.

Corn: Which brings us back to the practical question. What does someone actually do with all this?

Corn: First thing — before you buy anything, draw a floor plan. Doesn't have to be architectural blueprints. Just a rough sketch of your space, walls marked, rooms labeled. Then grab a free tool — WiFi Analyzer on Android, or NetSpot's free tier — and take readings at the spots where you'd consider putting an access point. What you're hunting for is anywhere the signal drops below minus seventy dBm. That's your "needs another node" line. Below that, video streaming gets shaky, calls drop, the whole thing degrades.

Herman: The floor plan matters because it forces you to think about what's between point A and point B. That wall you've been ignoring for five years? If it's concrete, it's eating fifteen decibels. Your floor plan tells you that before you even take a reading.

Corn: Second thing — if you're using a mesh system, give the nodes some breathing room. Thirty to forty feet apart in typical residential construction. Drywall, wood studs — that's the standard. Closer than thirty feet, you're paying for backhaul interference whether you realize it or not. And if your mesh supports a dedicated backhaul band — six gigahertz or, even better, wired Ethernet — use it. Take the backhaul conversation off the client channels entirely.

Herman: Wired backhaul is the cheat code. If you can run an Ethernet cable between nodes, you've just eliminated the entire self-interference problem. The mesh becomes a set of coordinated access points instead of a radio shouting match.

Corn: Third — stop looking at bars. Your phone's bar display is doing public relations for your Wi-Fi, not reporting the truth. What you actually want is SNR. A signal at minus fifty dBm with five dB of SNR is garbage. A signal at minus seventy-five with twenty-five dB of SNR will run circles around it. Use a tool that shows SNR, not just RSSI. WiFi Analyzer does this. NetSpot does this. It's the difference between feeling good about your network and actually having one that works.

Herman: All three of these are weekend projects. Not "call a consultant" projects, not "learn RF engineering" projects. Draw the floor plan, walk the space with your phone, check the numbers, move a node or two. That's it. The gap between guessing and knowing is about two hours of your time and possibly zero dollars.

Corn: Which is a much better deal than four hundred dollars on a mesh system that's accidentally jamming itself.

Herman: We've got the weekend plan. But here's the open question I keep turning over — as Wi-Fi seven becomes the default and the six gigahertz band fills up, does this kind of manual planning become mandatory for everyone, or do the mesh systems finally get smart enough to do it themselves?

Corn: That's the bet the mesh companies are making, right? "Self-optimizing" — the box figures out channel assignments and backhaul routing so you don't have to. And to be fair, they've gotten better. But "better" and "good enough to handle three non-overlapping channels with four nodes in a small apartment" are different things.

Herman: This is where the AI-driven planning tools get interesting. Ekahau has their AI Pro feature now — you upload a photo of your floor plan, and it automatically suggests AP placements based on the geometry and materials it detects. NetSpot's got their own AI heatmap that does something similar. The machine learning is doing the propagation modeling for you.

Corn: The expert-level planning becomes accessible, but it's still planning — you're just outsourcing the thinking to a model instead of doing it yourself. Which is different from the mesh system claiming it'll figure everything out on the fly.

Herman: A predictive model that says "put the node here" before you buy is fundamentally different from a mesh system that tries to fix bad placement after the fact. One prevents the problem, the other manages the symptoms.

Corn: I think that's the real through-line. The tools are getting smarter, but the physics isn't changing. Concrete still eats fifteen decibels whether an AI points it out or you do.

Herman: Now — Hilbert's daily fun fact.

Hilbert: In the nineteen thirties, researchers studying sharks off the Azores discovered that the ampullae of Lorenzini — the electroreceptive organs in a shark's snout — can detect electric fields as weak as five billionths of a volt per centimeter, and a single hammerhead may carry over three thousand of these sensory pores across its head.

Corn: That's unsettling.

Herman: So to land this — the question isn't really whether you should plan your Wi-Fi. It's whether, five years from now, you'll even have to think about it, or whether the planning just happens invisibly before you ever open a box. And I suspect the answer is somewhere in the middle — the tools will get better, but the person who measures will always have a better network than the person who trusts the wizard.

Corn: This has been My Weird Prompts. Thanks to our producer Hilbert Flumingtop. If this episode saved you from buying a mesh system you didn't need, rate us five stars and tell a friend who's still using their ISP's router.

Herman: We'll be back next week.