Get best of the week

Semi-Scientific Guide to Hosting a WiFi Router

WiFi - like real estate; The three main factors that influence its quality are location, location, and location.



There are almost no problems with the top floor of our test house - although, like many houses, it suffers from the terrible location of the router, far from its center.

We at Ars Technica often describe the WiFi operation scheme, write about which sets behave best. and how future standards will affect you. Today we turn to a more basic topic - we will teach you how to understand how many access points you need and where to place them.

These rules apply when it comes to a single WiFi router, a mesh set like Eero, Plume or Orbi, or access points with Ethernet backhaul support like UAP-AC from Ubiquiti or EAP from TP-Link. Unfortunately, these rules are more like recommendations, since with so many variables it’s impossible to calculate everything exactly, sitting in an armchair thousands of kilometers from your home. But if you familiarize yourself with these rules, you will at least be a little better versed in the practical aspects of what you can expect - and what you can't - from your WiFi equipment and how to get the most out of it.

Preamble


Before you start with our ten rules, let's first go over the theory of radio waves - it will help you better understand some of the rules when you understand how the power of a radio signal is measured and how it attenuates with distance and due to obstacles.


Some engineers recommend at maximum speed of -65 dBm for the maximum speed.

The graph above shows the loss curves for WiFi frequencies (horizontal is the distance from the router, red lines indicate a 2.4 GHz signal without walls, or with the addition of one or two walls , blue is the same for 5 GHz). The most important thing is to understand the units: dBm directly converted to milliwatts, only on a logarithmic decimal scale. When falling by 10 dBm, the signal power in milliwatts drops 10 times. -10 dBm is 0.1 mW, -20 dBm is 0.01 mW, and so on.

Logarithmic allows you to additively measure the signal drop, and not through multiplication. Each doubling of the distance leads to a signal drop of 6 dBm, and we clearly see this by studying the thick red curve for 2.4 GHz: for 1 m, the signal is -40 dBm, for 2 m it is -46 dBm, for 4 m it is -52 dBm .

Walls and other obstructions - including but not limited to human bodies, cabinets, furniture, appliances, will weaken the signal even more. A simple rule of thumb is -3 dBm for each wall or other significant obstacle. Thinner lines of the same color on the graph show the signal falling at the same distances when adding one or two walls (or other obstacles).

Ideally, you want to have a signal level of at least -67 dBm, but you don’t need to worry about raising it much higher than this mark - usually there is no difference in speed between the powerful -40 dBm and the weak -65 dBm, even if they are far from friend on the chart. WiFi is affected by many more factors than just signal strength; as soon as you exceed the minimum, it does not matter how much you exceed it.

In fact, a signal that is too powerful can turn out to be the same problem as too frail - many users on the forums complain about the low speed, until some smart person asks them: did you place the device right next to the access point? Move it a meter or two and try again. And, of course, the problem disappears.

Rule 1: no more than two rooms and two walls


Our first rule for placing an access point (AP) is no more than two rooms and two walls between the AP and devices. The rule is rather vague, because the rooms come in different sizes and shapes, and different houses have different composition of walls - but this is a good starting point, and it will serve you well in typical size houses and apartments with fairly modern drywall interior walls.

A typical size, at least in most of the United States, means bedrooms 3-4 meters long and living rooms 5-6 meters long along one of the walls. If we take nine meters as the average distance covering “two rooms” and add two internal walls, -3 dBm for each, our radio wave loss curve shows that the 2.4 GHz signals will feel great with a -65 dBm indicator. With 5 GHz the situation is not so good - if we need all 9 meters and 2 walls, then we will go down to -72 dBm. This is enough to establish a connection, but nothing more. In real life, a device with a -72 dBm signal at 5 GHz will see about the same throughput as a -65 dBm device at 2.4 GHz - however, formally a slower connection at 2.4 GHz will be more stable and show less delay .

Of course, all this provided that our only problems will be distance and signal attenuation. Users in rural areas and in houses with large plots of land have probably noticed this difference and understood the practical rule: “2.4 GHz is cool, but 5 GHz is a complete crap”. Urban residents or homeowners on a postage-stamp-sized plot have a completely different experience, which we will take into account in Rule 2.


If we are already starting to build mesh networks, we are preparing in full.

Rule 2: too much transmit power is bad


The signal strength at 2.4 GHz is range and effective penetration through obstacles. The downside of the 2.4 GHz signal is ... range and effective penetration through obstacles.

If two WiFi devices at a distance of “audibility” from each other transmit at the same frequency at the same time, nothing works for them: the devices for which they transmit a signal have no way to figure it out and understand which signal is intended for them . Contrary to popular belief, it doesn’t matter whether the device is on your network or not - the WiFi name and password do not matter.

To avoid this problem for the most part, any WiFi device must first listen to the broadcast before transmitting - and if any other device is already transmitting in this frequency range, then ours should shut up and wait for the end of the transfer. This does not fix the problem completely; if two devices decide to transmit at the same time, they will “collide” - and everyone will need to choose a random period of time that they will spend in anticipation before trying to transfer something again. A device that selects a smaller random number starts first - unless they choose the same random number, or some other device does not notice a respite on the air and does not decide to transmit a signal, ahead of both.

This is called a “jam”, and for most modern WiFi users, this is as big a problem as signal attenuation. The more devices you have, the more busy the network. Each of your devices may collide with another, and everyone has to respect the rules for using the ether.

If your router or AP supports this option, then reducing the power of the outgoing signal can, on the contrary, improve performance and roaming - especially if you have a mesh set or other similar scheme. 5 GHz networks usually do not need to be so attenuated, since the signal in that spectrum is attenuated quite quickly, but for 2 GHz this option can work wonders.

The last note for fans of "long-range" TD - such a TD can really give a signal stronger than usual, and finish off to a greater distance. However, it cannot force your phone or laptop to amplify the signal in response. With this imbalance, individual parts of the web page can load quickly, but in general the connection will seem unstable, since your laptop or phone will first have difficulty downloading the tens or hundreds of individual HTTP / HTTPS requests required to download each of their web pages.

Rule 3: use the spectrum wisely


In the second rule, we mentioned that all devices on the same channel compete for airtime, regardless of which network they belong to. For most people, the relationship with their neighbors is not so good that they can be persuaded to lower the transmission power - even if their router supports this function - but you can understand which channels the neighboring networks use and avoid them.

With 5 GHz, such a problem usually does not occur, but at 2.4 GHz this can be quite influential. Therefore, we recommend that most people avoid the 2.4 GHz standard. And where you can’t avoid it, use an application like inSSIDer to periodically study your radio wave environment, and try to avoid using the busiest spectrum in the area of ​​your home.

However, this, unfortunately, can be more complicated than it seems at first glance. It doesn’t matter how many SSIDs you see on a particular channel - it’s important how much airtime they actually use, and this cannot be calculated either from the number of SSIDs or from the pure signal strength of visible SSIDs. InSSIDer allows you to take another step and explore the real utilization of airtime in each channel.


The diagram from insider shows the load of each of the visible WiFi channels. In this case, almost the entire 2.4 GHz band is consumed.

In the diagram above, the entire 2.4 GHz band is mostly useless. Do not pay attention to the “empty” channels 2-5 and 7-10: 2.4 GHz equipment by default uses a channel width of 20 MHz, which in practice means that the network uses five channels (20 MHz plus half channel on each side), not one. The networks on Channel 1 actually extend from the hypothetical Channel 2 to Channel 3. Networks on channel 6 occupy channels 4 through 8, and networks on 11 occupy channels 9 through 13.


If you count the “shoulders”, up to a 2.4 GHz standard channel with a width of 20 MHz really takes a little more than four real channels 5 MHz

In 5 GHz networks, channel loading is a much smaller problem, since reducing the range and permeability of the signal means there are fewer devices to compete with. Often you can hear statements that this standard has more channels for work, but in practice this is not so if you are not involved in setting up WiFi in your enterprise where there are no competing networks. Home routers at 5 GHz are usually tuned to a channel width of 40 or 80 MHz, which means that there are really only two disjoint channels - the lower one, consisting of 36-64 channels with a width of 5 MHz, and the upper one, on channels 149-165.


Each 5 GHz network with a width of 40 MHz occupies a little more than 8 real channels with a width of 5 MHz. Each stump here symbolizes four channels with a width of 5 MHz.

In the comments, you should probably expect a discussion about these statements. Technically, you can fit four networks with a width of 40 MHz or two networks with a width of 80 MHz at the bottom of the 5 GHz band. In practice, consumer equipment operates through a stump deck with overlapping channels (for example, with a 80 MHz band centered on a 48 or 52 channel), which makes it difficult or practically impossible to achieve such spectrum efficiency in real home conditions.

Between the two standard consumer bands (in the USA) there are two more channels with the Dynamic Frequency Spectrum (DFS), but they need to be shared with devices such as commercial and military radars. Many consumer devices refuse to even try to use DFS. And even if you have a router or AP that agrees to use DFS, they must obey the most stringent requirements so as not to interfere with any radars. Users “off the beaten path” can perfectly use DFS - however, they most likely will not have problems with loading channels.

If you live near an airport, military base or port, DFS is most likely not suitable for you - and if you live outside the USA, the frequencies allowed for you may differ from what is described here (both DFS and others), depending on local laws.

Rule 4: Central Location is Best



The difference between “a router with the edge of the house” and “AP in the middle” can be critical.

Returning to the weakening of the signal, we note that the ideal place for the location of the WiFi AP is the center of the space that it needs to cover. If your home is 30 m long on one of the sides, then the router located in the middle will need to cover only 15 m in each direction, and the router on the edge (where the installers from the provider like to end the coaxial cable or DSL line) will have to cover 30 m.

The same is true for smaller rooms with a large number of APs. Remember, WiFi signals quickly fade out. Six meters - the length of a large enough living room - may be enough to ensure that the 5 GHz signal, having weakened, falls below the optimal level, if you add a couple of obstacles such as furniture or people. Which brings us to the next rule ...

Rule 5: Height - Above Human Height



Technically, the best location would be a place near the ceiling - but if that's too much, then put the AP at least at the top of the bookshelves.

The higher you can fix the AP, the better. The human body attenuates the signal by about the same amount as the inner wall - this is one of the reasons why the WiFi in your house deteriorates significantly when many friends come to the party.

By placing a TD - or router - above human height, you can avoid the need to transmit radio waves through all these annoying and signal-weakening bags of meat. Also, the signal avoids most of the furniture and household appliances - sofas, tables, ovens and cabinets.

The most ideal option would be to place the AP on the ceiling in the geometric center of the room. If this is not possible, do not worry - it will be almost as good to put it on top of the cabinet, especially if you need this AP to serve both the room where it is standing and the room on the other side of the wall.

Rule 6: divide the distances in half


Let's say some of your devices are too far from the nearest access point in order to get a good signal. You are fortunate enough to buy an extensible system, or you still have one AP from the mesh kit. Where to put it?

We observed the confusion of people in a similar situation, thinking about whether to place an additional AP closer to the first (with which it takes data) or closer to the farthest devices (to which it should transmit data). The answer is usually this: neither one nor the other. Place your AP directly in the middle between the nearest AP and the farthest client that it should serve.

The bottom line is that you are trying to save airtime by organizing the best connection possible between the long-distance devices and the new AP, and between the new AP and the one closest to it. Usually you should not give preference to one of the parties. However, do not forget rule 1: two walls, two rooms. If you cannot break the distance between the farthest clients and the main AP without violating the first rule, then place the new AP as far as the first rule allows.

If this seems too simple and logical to you, do not worry: there is one more point “only if not”, which must be taken into account. For some mesh kits, for example, Netgear's Orbi RBK-50 / RBK-53 or Plume's Superpods, the connection between the TDs has a very high throughput and works on a 4x4 scheme. Since this connection works much faster than the 2x2 or 3x3 available to clients, it may be worthwhile to reduce the quality of the communication signal between these APs so that their bandwidth is closer to that which the best of your clients can afford.

If your mesh set offers a very fast connection between APs, and you are unable to add additional APs to the scheme, then you might be better off placing the last AP closer to the clients than the previous AP. However, here you will have to experiment and study the results.

Cool thing - WiFi, isn't it?

Rule 7: Avoid Obstacles



A tightly packed bookcase is a serious obstacle to radio waves. It costs a pair of ordinary walls even with perpendicular penetration. And to cross it in length is generally useless.

If you got a particularly difficult room, there may be places where the signal simply cannot pass. Our test house had a concrete slab and several meters of dense earth, covering the line of sight between the router and the basement. We met small enterprises, just as worried that WiFi worked well in one part of the room, but it wasn’t in the other - and in the end it turned out that there was, for example, a bookcase full of books and located along the corridor , which is why several meters of processed wood weakening it were found in the signal path.

In each case, the solution is to create a workaround around the obstacle using several access points. If you have a WiFi mesh kit, use it so that the signal avoids obstacles. On one side of the obstacle, place the AP on the line of sight with the main line, so that it is visible on the other side of the obstacle and the signal does not need to go through.

With a sufficient number of APs and their careful placement, you can probably cope even with walls made of shingles and metal mesh, as they were built in the USA at the beginning of the 20th century. We saw how people successfully placed TDs in direct visibility of each other through doorways and corridors, when it would be easier to use a puncher to penetrate walls.



If too many obstacles prevent you from circumventing them from the side, above or below - see rule 8.

Rule 8: it's all about the connection between access points


Most consumers choose clean WiFi mesh sets, because it’s convenient - you don’t need to wire, just connect a bunch of access points, and let them carry out their magic there independently, without noise and dust.

It sounds comfortable, but it's actually the worst solution. Remember we talked about rules 2 and 3? These problems exist here. If your device needs to communicate with one AP, which needs to transfer data to another AP, then you already take a little more than twice as much airtime.

Okay, actually it’s not so bad - you double the use of airtime if your client is in the same place as the auxiliary AP. And since you followed rule 6 - divided the distances in half - this means that the communication quality of the main AP with the client is much better than the one organized by the client, connecting to the main AP directly. So even in the worst case scenario - when an auxiliary AP talks to a client on the same channel on which it talks to the main AP - they will be able to transmit data, consuming less airtime than if one client worked with a much longer and less quality connection.

However, it would be much better to completely avoid this problem if your APs communicate with each other at a different frequency. Two-way APs can do this by communicating with clients in the 2.4 GHz band, and between themselves at 5 GHz, or vice versa. In the real world, stubborn clients (and users) often want to connect not so optimally, in the end it turns out that there are clients at 2.4 GHz and 5 GHz, so there is no “clean” channel for internal communication.

Particularly smart sets, such as Eero, can avoid this situation by dynamically routing the intercom, minimizing congestion by transmitting in a range other than the one they are receiving, even when the ranges change. The most advanced three-way kits like Orbi RBK-50/53 or Plume Superpods can avoid this problem by using a second 5 GHz transmitter. This allows them to connect to customers either at 2.4 GHz or 5 GHz, leaving themselves an unoccupied range of 5 GHz. Orbi has a fixed and dedicated transmitter for intercom. Plume decides on the use of frequencies, depending on which version of its cloud optimizer considers the best in a particular environment).

The best option is not to use WiFi for internal communication at all. If you can lay an Ethernet cable, you must do so. It is not only faster than WiFi, it also does not suffer from problems with congestion channels. With a high network load, cheap wired APs like Ubiquiti UAP-AC-Lites or TP-Link EAP-225v3s dry even the most expensive mesh sets, if the latter are limited by internal WiFi connection. Wired intercom also solves the problem of obstacles that are opaque to radio waves - if a signal cannot be pierced through it or bypassed, then the cable stretched through it will work wonders!

Users who were unable to implement either mesh sets with WiFi or stretch Ethernet cables should consider modern equipment for transmitting signals over power lines. The results can be completely different, and depend on the quality of the wiring in the house and even on the type of connected household appliances, but in most cases the equipment of the AV2 series (AV1000 and higher) or g.hn series will be quite reliable, transmission delays will be quite low, comparable to Ethernet . Bandwidth is severely limited - in the real world, you should expect no more than 40-80 Mb / s for home conditions. If you only play games or browse the web on the Internet, then wiring data can be a much better solution than WiFi.

Having followed this path, be sure to read the instructions and take measures to encrypt communications. When testing such equipment for the first time, we accidentally built a bridge with a neighbor and reconfigured its router - it was almost the same model as ours, and the password was on it by default. "Hello, I hacked your router, I apologize" - a bad way to get to know each other, we do not recommend it.

Rule 9: usually the problems are not in bandwidth, but in delays


The good thing about the bandwidth is that it is one beautiful bright number that is easy to get by connecting to the site to check the speed or using a tool like iperf3 to communicate with the local server.

The bad thing about bandwidth is that this is a terrible way to measure both the user's impression of the network and how the WiFi network behaves under real load. Most people upset their WiFi network either when browsing the web or in games - and not when they download a large file. In both cases, the problem is not “how many megabits per second this pipe can withstand” - but “how many milliseconds it takes to complete a specific action”.

Although you can see a deterioration in the quality of a busy network by decreasing numbers of “download” speeds, this is a more complicated, confusing, and unrelated way, compared to studying application delays. Delays are a function of both simple speed and the network’s processing efficiency of traffic and airtime.

When checking WiFi networks, our favorite metric is the delay in applications, which we pretend to load a rather complex web page. More importantly, you need to measure page load in parallel with all other activity on the network. Remember the description of congestion in rules 2 and 3 - a "very fast" network with one active device can turn into a nightmare brake with many devices, or, in many cases, with one poorly connected device, which leads us to the last rule.

The conclusion of the 9th rule is that the advertised speed following the letters AC in the model is garbage. You need to trust thorough, technically competent reviewers, and not the manufacturer's speed rating on the box.

Rule 10: the speed of your WiFi network is limited by the speed of the slowest connected device



One device with a bad connection can kill the quality of communication for the entire network and all connected devices

. Unfortunately, one person trying to watch a YouTube video in a "bedroom with a worthless reception" is not only tormented by himself - his problems are overtaking others. A phone in the same room with an AP needs only about 2.5% of the available airtime to stream a video in 1080P quality at a speed of 5 Mb / s. But the phone “in a bad bedroom”, tormented by buffering and slow communication, can take 100% of the airtime of the network, and not be able to watch the same video.

Of course, streaming video occupies the incoming channel very much, and routers or APs usually refuse to transmit 100% of the time. An AP that needs to transfer a large amount of data will usually leave some airtime for other devices and request its own data, and then it will break the download time between the nearby device and the “bad bedroom” in order to try to fulfill both requests. But this still increases the window's wait time from these devices by hundreds of milliseconds, and they still have to compete with each other when opening this window.

The situation worsens if a user in a "bad bedroom" tries to upload a video, send an email or post a large photo to the social network. The router tries to leave part of the airtime to other devices - however, these restrictions do not apply to the user's phone, and he will gladly gobble up all the available airtime. Worse, the phone does not represent how much data other users requested in those brief moments when they had a window for requests. The router knows how much data needs to be delivered to each of the clients, so it can allocate time for downloading data accordingly - but all that the phone knows is that it needs to upload its data, so while everyone else is doing it, everyone else suffers. Therefore, even if you need to leave only one rule out of all this wisdom, let it be rule 10.

More articles:


All Articles