Tom's Tech Corner: The Evening Wifi Slowdown, Fully Accounted For

Dear Tom,

Every night around 7 o'clock, our wifi gets slow — the videos buffer, the tablet lags, all of it. It's fine in the morning. What's going on?

— Denise from Tulsa, Oklahoma

Denise, thank you for this question, which I confess delighted me the moment I read it, because it touches on one of the great unresolved elegances of the modern home: the router. The router, which is of course just a stateful non-deterministic packet oracle sitting quietly in your hallway, is not merely a box that "sends the internet." It is, at minimum, a continuously operating decision engine, and to understand why it slows down at 7 p.m. sharp we have to build the whole apparatus up from first principles. I'll try to keep this brief, though of course the full derivation belongs in Appendix B, which I have not included here for space but which you should absolutely picture existing.

Let's begin, as we must, with the electromagnetic spectrum — not casually, but properly, assuming of course standard atmospheric conditions and a lambda in the low centimeter range, which is simply what your router is emitting whether or not anyone in the house has ever thought about it. Your wifi signal is a radio wave, and radio waves, as you'll recall from the general theory, propagate outward in a sphere until they strike something — a wall, a refrigerator, a neighbor's competing sphere — at which point they do what physicists call "interfering," which is a bit like two conversations happening in the same kitchen, except the kitchen is made of electrons and neither conversation knows the other exists. At 7 p.m., Denise, every house on your block is having this same electron kitchen conversation simultaneously, because that is when America, as a civilizing force, sits down to stream television. This is not a metaphor. This is contention, in the formal sense, and it is the first of at least four unrelated causes we'll need to walk through before I can, in good conscience, call this question answered.

The second cause is thermal. This surprises people, but it shouldn't, once you consider that your router is a small computer performing several hundred million operations per second, and computers, as we know from the foundational work on entropy, cannot process information without generating waste heat — this is simply the Second Law doing what it does, and no amount of hoping will exempt a $40 plastic router from thermodynamics. By 7 p.m., your router has been quietly computing since 6 a.m., accumulating heat the way a bathtub accumulates water, and its internal processor — I want to be precise here — begins what's called thermal throttling, a defensive deceleration that, incidentally, is the exact same phenomenon that governs deep-sea cable-laying ships, which brings us to a point I think is worth pausing on.

The undersea cable is, in a sense, the router's spiritual ancestor, and I don't think you can fully appreciate the wifi signal in your kitchen without appreciating the 1858 transatlantic telegraph cable, which failed after three weeks because nobody had yet worked out signal attenuation over distance — a problem your router re-solves, silently, every single evening, which I think is rather moving if you sit with it. The shipping container, similarly, revolutionized the transport of physical goods in the 1950s by standardizing the unit of transfer, which is precisely what a "packet" is doing in your wifi network right now: it's a shipping container for data, standardized so that your router doesn't have to reinvent the loading dock every time your daughter's tablet requests a cartoon. I mention this not as trivia but as necessary background — you cannot understand modern packet-switched networking without understanding twentieth-century containerized shipping, and frankly I think more people should know this.

The third cause is what network engineers call the "noisy neighbor problem," though "problem" undersells it — it's really a statistical inevitability. Assuming, as we should, a Poisson distribution of streaming start-times across your neighborhood, the probability that all fourteen households on your cul-de-sac begin simultaneous video sessions clusters overwhelmingly around the dinner hour, for reasons that are more sociological than technical, though the sociology and the technology converge beautifully at exactly the moment your screen freezes. See Figure 3 for the full distribution curve, which plots household streaming onset against a standard evening, and which I wish I could show you here, though the shape, if you're curious, resembles a bell that has been sat on.

The fourth cause, and I promise we are nearly at the end of the necessary apparatus, is what's called channel saturation in the 2.4 gigahertz band, which your router shares — this is the part people never expect — with your microwave oven, your Bluetooth speaker, your baby monitor if you have one, and, in a delightful historical footnote, the exact same frequency range originally reserved by international treaty for industrial, scientific, and medical equipment, which is why wifi has always had to sort of politely share a hallway with your popcorn. At 7 p.m., when the microwave runs and the Bluetooth speaker plays and the router is doing all of the above simultaneously, you are experiencing what I'd call a fully saturated spectral commons, a phrase I rather like and may start using at dinner parties.

So: electromagnetic contention, thermal throttling, statistical neighborhood clustering, and spectral commons saturation. Four independent, rigorously grounded causes, all converging, with a kind of grim mathematical beauty, at exactly 7 p.m. in your home in Tulsa. I want to stress that this is not a malfunction. This is your network behaving exactly as the underlying physics, thermodynamics, statistics, and regulatory history predict it should. If anything, Denise, the remarkable thing isn't that your wifi slows down at 7 — it's that it works at all, given everything arrayed against it.

I recognize I haven't told you what to do about it, but that's because there isn't really a "do" here so much as an "understand" — once you see the full shape of the problem, the way I've just laid out, you'll find the slowdown reads less like an inconvenience and more like a kind of nightly proof, replaying itself, that the whole invisible edifice is still standing. I find that reassuring. I hope you will too.

Glad I could clear that up!

— Tom