Two numbers move in opposite directions
Channel width is the clearest example in home networking of a setting that improves one number by damaging another. Doubling the width doubles the spectrum a single transmission occupies, which roughly doubles the peak rate. It also halves how many independent channels exist in the same band, which doubles how many networks have to share each one.
Wireless is half duplex and shared. Only one device on a channel transmits at a time, and everyone within earshot waits their turn. So the throughput a network actually achieves is its peak rate multiplied by the share of airtime it gets, and the share of airtime falls as more networks land on the same channel. The calculator above shows both terms, because looking at either one alone gives the wrong answer.
The arithmetic behind the disappointment
Work through what happens as you widen the channel in a building where every channel already has someone on it. The width doubles, so the peak rate index doubles. The number of channels halves, so the number of networks sharing each channel doubles, so the airtime share halves. Multiply the two and you get exactly what you started with.
That is the result the table above is showing when several rows come out identical. In a fully contended environment, total airtime is conserved and channel width redistributes it rather than creating it. Widening is not a small gain, it is no gain, and the average column says so plainly.
What is not conserved is reliability. A wider channel overlaps more neighbouring transmissions, so it collides more often, and every collision means a retransmission that consumes airtime twice. It is also more likely to encounter interference somewhere across its span, and a channel is only as clean as its dirtiest part. So in a dense building the wide channel is the same on average and worse in the tail, which is exactly where users notice: the call that stutters, the stream that rebuffers once an hour.
When wide is genuinely the right answer
None of the above applies to an uncontended band. A detached house with few neighbours on a band with plenty of channels has more channels than networks, every network gets essentially all of its own airtime, and the peak rate is the whole story. There the widest width your equipment and your regulator permit is the right choice, and it is not close.
| Situation | Networks vs channels | Sensible width |
|---|---|---|
| Detached, quiet band | Far fewer networks than channels | As wide as permitted |
| Suburban, moderate band | Comparable | Middle widths, and measure |
| Dense block of flats | Many more networks than channels | Narrow, and add access points instead |
| Any building, low band | Very few genuinely independent channels | The narrowest option |
The crossover the calculator reports is the point where the number of networks equals the number of channels at that width. Below it, widening pays. Above it, widening stops paying, and continues not to pay however much further you go.
Counting neighbours honestly
A scan taken standing next to the router is the wrong scan. Take it at the place where the problem is, because that is where the client is deciding whether to transmit, and the set of networks audible there is different. Networks that are strong enough to make your radio defer are the ones that cost you airtime; a network barely detectable at the noise floor costs almost nothing, which is why the calculator lets you discount the weak ones rather than counting every name in the list.
Also count the networks rather than the access points, since one physical box frequently advertises several names, and each one costs a little airtime on its own for management traffic even when nobody is connected to it. And count per band. The set of neighbours on the low band and the high band are usually very different, which is one of the arguments for moving what you can to the band with fewer of them.
What to do when the channel plan has run out of room
The situation this page most often reveals is one where no channel setting is going to help, because the band is full whatever you do with it. At that point the useful moves stop being about frequency and start being about geometry and about traffic.
Lowering transmit power sounds backwards and is often right: a smaller cell hears fewer neighbours and is heard by fewer of them, and if the clients are close enough it loses nothing. More access points closer to the clients does the same thing from the other direction, and the access point coverage calculator works out how many and how far apart. Anything that can take a cable should, which removes its airtime from the pool entirely, and the ethernet drop calculator prices that in cable and jacks. And it is always worth confirming that the wireless is the actual bottleneck before rearranging it, which means checking demand against the connection in the household bandwidth calculator and running the wired versus wireless comparison in the Wi-Fi improvement guide first.
Questions people ask
Should I use 20, 40, 80 or 160 MHz?
It depends entirely on how many other networks share the band where you are, which is why this page asks rather than recommending. With far fewer networks than channels, take the widest width your equipment and regulator allow, because peak rate is the whole story and nothing is competing for the air. With more networks than channels, the wider option gives back on airtime exactly what it gained on rate, and loses on reliability because it overlaps more neighbours. The tipping point is where the network count reaches the channel count at that width, and the calculator reports it directly.
Why is my wide channel slower than the narrow one was?
Because throughput is rate multiplied by airtime share, and you improved the first term while damaging the second. A wide channel spans more spectrum, so it overlaps more of your neighbours, so it has to wait for more of them and collides with more of them. Every collision costs a retransmission, which consumes the airtime twice. In a dense building this can leave the wide channel genuinely worse rather than merely equal, and the effect shows up as inconsistency rather than as a lower speed test result, because a speed test that happens to run during a quiet moment looks fine.
How many channels do I actually have?
Take it from a scan on your own equipment in your own location, because availability is a regulatory matter that differs by country and by band and changes over time. Nothing on this page asserts what you are permitted to use. What is general rather than jurisdictional: the lower band has very few genuinely non-overlapping channels because the channel numbers overlap each other heavily, while the higher bands have considerably more. Some channels in some places carry additional conditions on their use. Your equipment knows what it is allowed to do where it is; a web page does not.
Is it better to change channel or add an access point?
Changing channel helps only when some channel is meaningfully quieter than the one you are on, which in a dense building is rarely true and rarely stays true, since your neighbours equipment is also choosing automatically and moving. Adding an access point helps in a way that does not depend on anyone else, because it puts the client closer to a radio, which raises the rate, which means each transfer occupies the air for less time, which leaves more airtime for everyone. Faster transmissions are shorter transmissions, and in a contended environment shortness is the resource that matters.
Do my own access points interfere with each other?
Yes, whenever there are more of them than there are non-overlapping channels at the width you have chosen, which the calculator flags. Two of your own radios on the same channel are contending exactly as a neighbour would. The mitigation is to reuse channels only between radios that are far enough apart not to hear each other well, which is the same reuse pattern any cellular system uses, and to consider a narrower width specifically so that each radio can have a channel to itself. A narrower channel that nobody else is on frequently beats a wider one being shared with your own equipment.