Latest / Logosofia no Dia a Dia / Wi-Fi Settings That Actually Matter
Transcript
- [Prompt] custom: That has elevated wireless internet to an art form. It's definitely ubiquiti. I was poking around th | Hosts: corn, herman [Response] Corn: Daniel's been poking around his Ubiquiti controller again. He's got a single Wi-Fi 7 access point and he's staring at this interface that shows him things he didn't know existed — channel scanners, power sliders, band spacing toggles. And the thing that struck him is how this one little AP, with the right controller behind it, could scale up to run an airport. But all that granular control raises a question: what do these settings actually do, and why should someone with one AP or five care about them? Herman: The timing here is genuinely interesting. Ubiquiti's self-hosted controller just became their recommended path for self-hosters, which means a whole wave of enthusiasts are seeing these settings for the first time. You go from a TP-Link Deco where everything's hidden behind a friendly app that says "everything's fine" — to a screen full of RF scans and dBm sliders. It's like going from driving an automatic to sitting in a cockpit. Corn: The Deco's version of optimization is basically a green checkmark and a reassuring pat on the head. Herman: And that's fine for what it is. But the Ubiquiti approach says: here's the actual radio environment you're living in, here's what's interfering with it, and here are the levers. Now you decide. Which is both empowering and completely overwhelming if you don't know what those levers do. Corn: That's what we're going to unpack. Daniel's question is essentially: I've got this powerful controller, I can see channel optimization, transmission power, band spacing — what do these parameters mean, why isn't maxing everything out the answer, and how do you actually configure this intelligently whether you're running one AP or a whole fleet? Herman: The core tension here is that Wi-Fi is a shared, contested medium. It's not like a water pipe where more pressure means more flow. It's more like trying to have a conversation in a crowded room. Shouting louder doesn't always help. Sometimes it just makes everyone else shout louder too. Corn: The physics of shouting in a crowded room. That's basically the episode. Herman: That's the episode. And that interface — the one that shows you things you didn't know mattered — is exactly where we're going to start. Herman: Let's talk about what Ubiquiti actually is, because the ecosystem matters for understanding why these settings exist. UniFi isn't a consumer mesh system with a pretty app bolted on. It's enterprise networking gear that happens to be priced and packaged so enthusiasts can buy it. Same controller software that runs a university campus runs Daniel's single AP. Corn: Which is why the interface feels like it was designed by someone who assumed you'd read the manual. Because in the enterprise world, you have. Herman: That's the tension. The self-hosted controller path they're now recommending puts you in charge of your own network operations center, essentially. You're running the same software stack — the controller handles device adoption, firmware updates, configuration pushes, RF environment monitoring. It's a full management plane. The difference between this and a Deco or an Eero isn't that the physics are different — radio waves are radio waves. It's that you can see the physics. Corn: The Deco is making the same decisions under the hood — picking channels, adjusting power — it just doesn't tell you what it chose or why. Which means when something goes wrong, you're troubleshooting a black box. Herman: And Ubiquiti surfaces three parameters that form the core of what separates prosumer Wi-Fi from the plug-and-pray approach. Channel optimization — that's the AP scanning the spectrum, detecting what networks and noise sources are out there, and picking the cleanest channel. Transmission power control — how loud each AP shouts. And band spacing, or channel width — how much spectrum you're claiming for each transmission. Those three things interact. You can't optimize one in isolation. Corn: The thing Daniel noticed, which is the right thing to notice, is that the controller offers to do this automatically in the background. There's a toggle for that. Which raises the obvious question: if it can optimize itself, why do I need to understand any of this? Herman: Because "automatic" doesn't mean "optimal for your specific situation." The auto-optimizer is making decisions based on what it can measure at a given moment. It doesn't know your floor plan. It doesn't know which rooms actually need coverage and which are storage closets. It doesn't know that your neighbor fires up a ham radio setup every Thursday evening. The settings are there because the RF environment is dynamic and personal — and understanding what the levers do means you can make better choices than a generic algorithm. Corn: The central question isn't "what should I set these to." It's "what's happening in my airspace, and how do I work with it rather than against it. Herman: Which is a fundamentally different mindset from the consumer approach, where the answer is always "buy another node" or "move it closer." This is about understanding the medium. And that's where we should start — with channel optimization, because it's the feature most people discover first and the one that reveals the most about what's actually happening in your radio environment. Herman: The channel optimizer works by briefly turning each radio into a listener instead of a broadcaster. Your AP has separate radios for 2.4, 5, and 6 gigahertz. When a scan kicks off, it takes one radio offline for maybe a hundred milliseconds — short enough you won't notice a drop — and during that window it's doing what's called a clear channel assessment. It sweeps across every channel in the band, measuring energy levels, decoding beacon frames from neighboring networks, and logging what it finds. Corn: It's not just checking signal strength. It's reading the actual network identifiers, figuring out who's on which channel and how loudly they're talking. Herman: It builds an interference map. It knows "there's a network called SmithFamily on channel six at minus sixty-five dBm, another one on channel six at minus seventy-two, and channel eleven has one weak signal at minus eighty." Then it runs a scoring algorithm — UniFi uses a proprietary one but the principle is the same across enterprise gear — and picks the channel with the lowest aggregate interference score. The automatic option in the UniFi controller lets this happen on a schedule, typically in the middle of the night when nobody's streaming anything. Corn: The trade-off is that brief service blip. Which is fine at three in the morning and less fine during a Zoom call. Herman: That's why the scheduling exists. The deeper thing here, and this is what most people miss, is that channel selection often matters more than how much power you're pushing. I've seen real-world scenarios where a channel with a minus seventy dBm signal and zero contention delivers better throughput than minus fifty dBm with three other APs screaming on the same channel. Corn: Because co-channel interference isn't just background noise. It's another device actively competing for airtime. Herman: That's the key. Wi-Fi is a polite protocol — it listens before it talks. If another AP is on your channel, your AP has to wait its turn. Every frame your neighbor transmits is a frame your AP can't send. So you can have a beautiful strong signal and terrible performance because you're sharing the channel with three other networks. The minus seventy dBm channel with no neighbors? Your AP never has to wait. It just transmits. Corn: The optimizer isn't hunting for the strongest signal. It's hunting for the emptiest room. Herman: That's why the 2.4 gigahertz band is such a nightmare. Only three non-overlapping channels — one, six, and eleven — and in an apartment building you might see twenty SSIDs crowding those three slots. The optimizer has to pick the least bad option. Compare that to 6 gigahertz on Wi-Fi 7, where you've got fifty-nine non-overlapping twenty-megahertz channels. The optimizer's job is dramatically easier because there's actual space to work with. Corn: 6 gigahertz has its own problem. Shorter range, worse at going through walls. So the channel might be pristine, but if the signal can't reach your kitchen, it doesn't matter how clean it is. Herman: Which is exactly why channel optimization becomes more critical, not less, on 6 gigahertz. 4 you can sometimes get away with cranking power to punch through walls — it's a blunt instrument, but it works. On 6 gigahertz, the physics doesn't give you that option. The attenuation through drywall is significantly higher. So you have to be smarter about channel selection because you can't compensate with brute force. Corn: The automatic optimizer is useful, but it has a blind spot: it only sees what's happening at scan time. Daniel's neighbor's Deco system picked its channel once during setup and never re-evaluated. The UniFi controller can rescan, but only when you tell it to or on schedule. In between scans, the RF environment can shift. Herman: That's the contrast with consumer mesh that really matters. A TP-Link Deco does a single channel selection at boot and sticks with it. Six months later, three new neighbors have moved in with their own networks, and suddenly your once-clean channel is a traffic jam. The Deco won't tell you and won't fix it. The UniFi controller, especially with the self-hosted path giving you historical scan data, lets you see the degradation happening and do something about it. That visibility is the whole game. Corn: Channel selection is half the battle. The other half — and this is where things get counterintuitive — is transmission power and how it interacts with channel width. Herman: Most people's instinct, and I get it, is to crank power to max. Thirty dBm on 2.4 gigahertz, the full legal limit. More power equals better coverage, right? The problem is Wi-Fi is a two-way conversation. Your AP can blast at thirty dBm, but your phone? Your phone's transmitter is maybe fifteen dBm on a good day. So the AP is shouting loud enough that the phone can hear it from the backyard, but the phone can't shout back. Corn: The asymmetric link budget problem. The AP sounds like it's right next to you, but your client is whispering. Herman: And the phone sees full bars because it's measuring the AP's signal, not its own ability to reach the AP. So it stays connected, tries to transmit, fails, retries, fails again. Meanwhile the AP is wasting airtime on retransmissions. You get the illusion of great coverage and the reality of terrible performance. Corn: That's before we even talk about what max power does to your neighbors — including your own other APs. Herman: In a multi-AP setup, three APs all at high power are essentially three people shouting at each other from different corners of the same room. They can all hear each other, so they all have to wait their turn. And your clients, which are the ones actually trying to have a conversation, can't get a word in edgewise. Corn: The instinct to turn it up is exactly backwards in most cases. Herman: It's one of those rare cases where the intuitive thing is the wrong thing. And this is where channel width enters the picture and makes everything more complicated. Wider channels — eighty megahertz, a hundred sixty megahertz, and on Wi-Fi 7 you can go up to three hundred twenty megahertz on the 6 gigahertz band — wider channels are like claiming more lanes on the highway. Higher throughput, more data per second. But you're also claiming a huge chunk of spectrum, and in a dense area, you're overlapping with way more neighbors. Corn: Running a hundred sixty megahertz channel at max power in an apartment building is basically the Wi-Fi equivalent of driving a semi truck through a parking garage. Herman: That's exactly what it is. You're blasting a wide signal across channels that your neighbors are trying to use, creating contention not just on one channel but across a whole block of them. And remember, Wi-Fi is polite — your AP sees activity on any part of that wide channel and it waits. So you've taken a huge slice of spectrum and you're deferring to everyone on it. Corn: Which means the theoretical throughput of a hundred sixty megahertz channel is great on paper, but in a dense environment, you might actually get better real-world performance on forty or eighty megahertz because you're not constantly waiting. Herman: This is where the interaction between power and channel width becomes the thing you have to think about. They're not independent sliders. High power with wide channels is the worst combination in density. Low power with narrow channels is the most polite. And somewhere in between is usually the sweet spot for your specific environment. Corn: For Daniel's single AP setup, what's the practical starting point? Herman: Medium power, eighty megahertz channels on 5 gigahertz, and let the channel optimizer pick the cleanest spot. That gives you good throughput without being a bad neighbor to yourself. If you have coverage dead zones, the answer usually isn't cranking power — it's adding a second AP at low or medium power. One loud AP creates that asymmetric link problem we talked about. Two quieter APs give you real coverage. Corn: In a multi-AP house — say three APs in twenty-five hundred square feet — the mistake people make is setting all three to high power and wondering why their devices won't roam. Herman: Because the client makes roaming decisions based on signal strength, not throughput. If all three APs are blasting at high power, your phone sees strong signals everywhere and thinks "I'm fine right here, thanks." It sticks to the first AP it connected to even when you've walked to the other end of the house and a different AP would give you much better performance. Drop those APs to medium or low, and suddenly there are real signal gradients — the phone notices when it's leaving one AP's range and entering another's. Corn: The bane of multi-AP setups. Herman: You can see it in the UniFi controller. The client history shows you exactly which AP a device is connected to, and you'll spot the laptop that's been clinging to the basement AP while sitting in the upstairs office. Dropping power fixes that. I've seen setups where going from high to medium improved overall throughput by twenty to thirty percent just because clients started roaming properly. Corn: The self-hosted controller Daniel's running makes this diagnostic process actually possible. You're not guessing — you've got historical RF scan data, you can see interference patterns at different times of day. Herman: This is the real advantage of the self-hosted path Ubiquiti is now recommending. You can run scans at eight in the morning when everyone's at work, then again at eight in the evening when every neighbor's network is lit up and their kids are streaming. The interference profile is completely different. With historical data, you can schedule your channel optimization for the worst-case scenario — or you can manually tune for the time of day that matters most to you. Corn: You're not just optimizing for right now. You're optimizing for Tuesday night when the whole neighborhood is online. Herman: And that's something a Deco or an Eero simply cannot do. They're making decisions in a black box, once, with no visibility and no memory. The UniFi controller gives you a time-lapse of your RF environment. That's the difference between reacting to problems and understanding them. Corn: If you're sitting there thinking, okay, what do I actually do with all this? Let's turn it into a practical checklist. Four things you can do right now in your controller. Herman: First one: run a channel scan, but don't just run it once. Run it at different times of day. Morning when the neighborhood's quiet, evening when everyone's streaming, maybe a weekend afternoon when you've got the highest density of active networks. The interference profile shifts dramatically across the day, and the self-hosted controller Daniel's using gives you that historical data — you can actually see the patterns. Corn: Then schedule your automatic optimization for whatever window is off-peak in your house. Three in the morning, four in the morning — whenever nobody's awake and nobody's running a backup job. The scan takes a radio offline for a fraction of a second, but you still don't want it happening during a late-night gaming session. Herman: Second: for a single AP setup like Daniel's, start with medium power and eighty megahertz channels on 5 gigahertz. That's your baseline. Walk around with a device and actually measure throughput, not just bars. If you find a dead zone, resist the urge to crank power. The better move is almost always adding a second AP at low or medium power rather than trying to make one AP cover the whole house by shouting. Corn: One loud AP creates that asymmetric link problem — full bars, no throughput. Two quieter APs give you real coverage where clients can actually talk back. Herman: That brings us to the third one, which is for multi-AP setups. Lower power is almost always better. Set each AP to low or medium. Let the channel optimizer assign non-overlapping channels, or do it manually if you know your environment well enough. Then verify that clients are actually roaming. The UniFi controller's client history is your friend here — it shows you which AP each device is connected to, and you can spot the sticky ones that won't let go. Corn: A laptop clinging to the garage AP while it's sitting in the bedroom. You'll see it immediately. Herman: If you spot that, dropping power further on the AP it won't release is usually the fix. You're creating a cleaner signal gradient so the client notices it's time to move. Corn: Fourth one, and this is the one people overlook: don't ignore 2.It's slower, yes. But it punches through walls better than 5 or 6 gigahertz ever will. That's where your IoT devices live — smart plugs, thermostats, door sensors. Keep the channel width locked at twenty megahertz on 2.That's the only width where the three non-overlapping channels — one, six, eleven — actually stay non-overlapping. Go wider and you're creating cross-channel interference that degrades everything on the band. Herman: Those IoT devices are chatty in ways you don't notice until the band gets clogged. A temperature sensor transmitting a tiny payload every thirty seconds isn't a bandwidth hog, but it's claiming airtime. On a clean twenty-megahertz channel, that's fine. On a forty-megahertz channel overlapping with two neighbors, it's just more noise in an already noisy room. Corn: Those four steps — scan at multiple times, medium power with eighty megahertz for a single AP, low power with verified roaming for multi-AP, and twenty megahertz on 2.4 for IoT — that'll get you ninety percent of the way to an optimized network without touching anything more exotic. Herman: The beautiful thing is none of this requires being a network engineer. It's just acknowledging that your radio environment is real, it's specific to your house, and it changes. The controller gives you the visibility. These four steps give you the strategy. Corn: There's a bigger question lurking behind all of this. Corn: The question is whether all this visibility and control is a transitional phase. Wi-Fi 7 brings three hundred twenty megahertz channels and a whole new band, but the 6 gigahertz spectrum won't stay empty forever. As it fills up, does manual optimization become unsustainable? Are we headed for a world where the complexity forces us back to black boxes — just smarter ones? Herman: I think about this a lot. Ubiquiti is already layering machine learning into their optimization engine. The controller doesn't just scan and pick — it's starting to learn patterns, to anticipate interference before it happens. The trajectory seems pretty clear: AI-driven automatic optimization that makes these settings invisible again, but this time with actual intelligence behind the curtain instead of a one-time channel pick at boot. Corn: We're in this weird middle period where the tools are powerful enough to expose the knobs, but not quite smart enough to turn them perfectly on their own. Herman: The question is whether the broader market follows Ubiquiti into that middle period or skips straight past it. The Deco approach — total abstraction — works for most people. But the self-hosted controller path Ubiquiti is now recommending suggests there's a real and growing audience that wants the knobs. Enthusiasts who'd rather understand the physics than trust a green checkmark. Corn: The line between prosumer and enterprise is blurring, and Ubiquiti's betting that blurring continues. The same controller software running an airport runs Daniel's single AP. That's not an accident — it's an argument about who wants control and why. Herman: I think the argument holds. Not because everyone needs to become a network engineer, but because understanding these parameters isn't about certification-level knowledge. It's about recognizing that Wi-Fi is a physical medium. It has walls. It has neighbors. It has times of day when the air gets crowded. The best setup works with those constraints rather than pretending they don't exist. Corn: That's the thing Daniel's controller revealed, right? Not a set of intimidating engineering dials, but a map of the actual environment he's living in. The settings are just how you respond to what the map shows you. Herman: Once you've seen the map, you can't unsee it. You stop thinking about Wi-Fi as magic and start thinking about it as a shared space. That's the real shift. Everything else is just sliders. Corn: Now: Hilbert's daily fun fact. Hilbert: The dual number in Slovene — a grammatical form used specifically for exactly two of something — survives as a rare linguistic fossil from the early medieval period. The surname "Dvojnik," still found in Slovenian records, literally means "the one who is two" and may have originated as a name for twins. Corn: A name that means "the one who is two." That's either profound or deeply confusing. Herman: I'm going to be turning that one over for a while. Corn: This has been My Weird Prompts. Thanks to our producer, Hilbert Flumingtop. If you enjoyed this, do us a favor and leave a review wherever you listen — it helps. We'll be back with another one soon.