The United States has twenty five 20MHz channels in the 5GHz band and nine of them are yours without conditions. The other sixteen are DFS channels, which means the radio must listen for radar for sixty seconds before it may transmit, must stop transmitting within ten seconds if it hears any, and then cannot use that channel for thirty minutes.
The FCC also requires manufacturers to prevent the operator from switching DFS off. On top of that, the two DFS bands are the low power ones: 250mW conducted where the non-DFS bands allow 1W. So the real planning question is not which channel is best, it is how wide you make the channel, because at 20MHz you have twenty five to choose from and at 160MHz you have two, and both of those are DFS.
- Twenty five 20MHz channels in the US, sixteen of them DFS
- DFS means a sixty second check, a ten second exit and thirty minutes off the channel
- The FCC requires that DFS cannot be turned off by the operator
- The DFS bands are also capped at 250mW where the others allow 1W
- Width decides the count: six channels at 80MHz, two at 160MHz
On this page
The bandsThe four bands, and why they are not equal
The FCC divides the 5GHz band into four U-NII bands with different rules, and the rules matter more than the frequency. These are the frequency ranges as written in the regulation, and they decide which 5GHz channels are available to a wireless network.
| Band | Range | DFS required | Conducted power limit |
|---|---|---|---|
| U-NII-1 | 5.15 to 5.25GHz | No | 1W |
| U-NII-2A | 5.25 to 5.35GHz | Yes | 250mW |
| U-NII-2C | 5.47 to 5.725GHz | Yes | 250mW |
| U-NII-3 | 5.725 to 5.850GHz | No | 1W |
Read the last two columns together, because that pairing is the thing nobody mentions. The DFS bands are also the low power bands.
The FCC caps 5.25 to 5.35 and 5.47 to 5.725 at the lesser of 250mW or 11dBm plus 10 log B, while U-NII-1 and U-NII-3 permit 1W conducted. So the sixteen channels that carry the radar obligation also carry a quarter of the transmit power budget.
That is not a reason to avoid them. It is a reason to expect smaller cells there, and to stop being surprised when a client that holds onto channel 149 across the office drops off channel 100 in the same spot.
A fifth block, U-NII-4 at 5.850 to 5.895GHz, was opened more recently under stricter terms: client devices must operate under the control of an indoor access point, and clients may not connect directly to each other.
The numbersThe channel numbers, and where they come from
5GHz WiFi channels are numbered rather than counted off, and the numbering is neither arbitrary nor sequential. The center frequency is 5000MHz plus five times the channel number, so channel 36 is 5180MHz and channel 149 is 5745MHz. Because 20MHz channels sit four numbers apart, the usable list steps 36, 40, 44, 48 rather than 36, 37, 38.
| Band | 20MHz channels | Count |
|---|---|---|
| U-NII-1 | 36, 40, 44, 48 | 4 |
| U-NII-2A | 52, 56, 60, 64 | 4, all DFS |
| U-NII-2C | 100, 104, 108, 112, 116, 120, 124, 128, 132, 136, 140, 144 | 12, all DFS |
| U-NII-3 | 149, 153, 157, 161, 165 | 5 |
Twenty five in total, sixteen of them DFS, nine of them unconditional. Note that unlike the 2.4GHz band, these do not overlap. Every one of the twenty five is a clean channel, which is the actual advantage of 5GHz and the reason the wireless LAN is worth the shorter range.
WidthChannel width is the decision
This is the part that changes a design, and it is a straight trade: every doubling of channel width halves the number of 5GHz channels you have left to reuse.
| Width | Channels in the US | Non-DFS among them |
|---|---|---|
| 20MHz | 25 | 9 |
| 40MHz | 12 | 4 |
| 80MHz | 6 | 2 |
| 160MHz | 2 | Zero |
The bottom row deserves the emphasis. The two 160MHz channels sit inside DFS spectrum, so there is no way to run 160MHz in the United States without accepting the radar rules. A vendor datasheet advertising 160MHz support is not wrong; it is describing a mode whose availability depends on a radar you cannot see and cannot appeal to.
The middle rows are where real deployments live. Six channels at 80MHz sounds like plenty until you count access points: an office with twelve of them cannot give each one a private 80MHz channel without reusing four times over, and reuse at 80MHz means co-channel interference across a wider slice of spectrum than reuse at 20MHz does.
How channel bonding builds the wider channels
A wide channel is not a separate piece of spectrum. It is two, four or eight neighboring 20MHz channels bonded together, and it is named by the channel number at its center frequency. That is why a router set to 80MHz shows channel 42 or 155, numbers that are not in the 20MHz list.
| Width | Center channels available in the US | Built from |
|---|---|---|
| 40MHz | 38, 46, 54, 62, 102, 110, 118, 126, 134, 142, 151, 159 | Two 20MHz channels each |
| 80MHz | 42, 58, 106, 122, 138, 155 | Four each: 42 is 36 to 48, 155 is 149 to 161 |
| 160MHz | 50 and 114 | Eight each: 50 is 36 to 64, 114 is 100 to 128 |
Channel 50 shows the catch. Half of it is U-NII-1, but the other half is U-NII-2A, so the whole 160MHz channel follows the DFS rules.
More width raises the ceiling of a single client and lowers the capacity of a crowded room. In a dense space the right answer is usually narrower channels and more of them, which is the opposite of what the marketing on the box suggests, and no amount of prioritization recovers capacity that the channel plan gave away.
InterferenceInterference between 5GHz channels, and how to check for it
The radar rules are one source of trouble. The other is ordinary WiFi interference from neighboring wireless networks, and it comes in two kinds.
Co-channel interference. Access points and client devices on the same channel take turns, because WiFi devices listen before they transmit. Two networks sharing channel 36 do not corrupt each other so much as split the airtime, and every device that joins makes each turn come around less often.
Adjacent channel interference. At 20MHz the 5GHz channels do not overlap, so overlap appears only with bonding. An 80MHz network on channel 42 and a 40MHz network on channel 46 occupy the same frequency range, and the narrower one loses airtime to a neighbor it never chose.
Radios on neighboring channels can also bleed into each other when they sit very close together, which is why two access points on adjacent channels should not be mounted side by side.
Consumer routers that do not offer DFS channels can only pick among the nine non-DFS ones. In an apartment building or a shared office floor, that makes 36 to 48 and 149 to 165 the crowded part of the band.
Checking takes a few minutes. A WiFi analyzer app on a laptop or phone lists every network in range with its channel, width and signal strength. Look for the 5GHz channels where the strongest neighbor is weakest, and set the channel by hand or leave automatic selection on and review what it chose.
DFSWhat DFS actually does to a network
The FCC sets out the mechanism in numbers, and every one of these is a symptom somebody has filed a ticket about.
Sixty seconds of silence before it starts. A device must check for radar before it transmits on a DFS channel, and may begin only if nothing is detected within sixty seconds. That is the access point that boots and then does nothing for a minute.
Ten seconds to get off. Once radar is detected all transmissions must cease within ten seconds, of which normal traffic may continue for at most 200 milliseconds. After that, only management and control signaling is allowed while the radio vacates.
Thirty minutes of exile. A channel flagged as containing radar is subject to a non-occupancy period of at least thirty minutes, starting when the radar was detected.
And you cannot switch it off. The regulation requires that manufacturers take steps to ensure DFS functionality cannot be disabled by the operator. If a setting appears to do it, it is changing which channels are offered rather than disabling the obligation.
The detection threshold is worth knowing because it explains false positives: minus 64dBm for devices between 200mW and 1W of e.i.r.p., and minus 62dBm for lower powered ones, measured as received power averaged over one microsecond against a 0dBi antenna.
That is a very quiet signal, and airport, weather and naval radars are not the only things that can trip it.
ChoosingChoosing channels without a survey
| Situation | Reasonable starting point |
|---|---|
| Home or small office, few neighbors | Any non-DFS channel at 80MHz: 42 or 155 |
| Dense office, many access points | 20MHz or 40MHz, and use the DFS channels |
| Near an airport, port or weather radar | Non-DFS only, and expect to justify it |
| Warehouse with mobile clients | 20MHz, non-DFS, prioritize roaming over throughput |
| One client, one access point, maximum speed | 160MHz, and accept that it is a DFS channel |
The row that matters most is the second one. A dense deployment that refuses DFS is choosing between nine channels for a building that may need thirty cells, and that is a worse problem than an occasional radar event. Enable DFS, and make sure the controller knows which channels have been flagged so it does not keep trying them.
The third row is a judgment rather than a rule. Proximity to radar raises the chance of a detection, and the FCC itself recommends that anyone using these devices for critical communications should determine whether nearby government radar systems could affect their operation.
PitfallsWhere people go wrong
Choosing the widest channel available. Width helps one client in an empty room and hurts a full one. Count access points before choosing a width.
Avoiding DFS on principle. Nine channels is not enough for a dense deployment, and a radar detection costs thirty minutes on one channel rather than an outage.
Blaming the client for a DFS drop. When radar is detected the access point has ten seconds to leave, so clients get moved abruptly. That is the regulation working, not a driver fault.
Assuming a DFS channel is as strong as a non-DFS one. The DFS bands are capped at 250mW conducted where U-NII-1 and U-NII-3 allow 1W. The cell is smaller and the coverage map is not the same shape.
Trying to disable DFS. The FCC requires manufacturers to prevent it. What a setting can do is restrict the channel list, which is a different thing with different consequences.
Planning 5GHz as a copy of the 2.4GHz plan. These channels do not overlap, so the 1, 6, 11 discipline that 2.4GHz needs has no equivalent here. Every one of the twenty five is clean.
Ignoring what the clients support. Channel 165 and parts of U-NII-2C are not equally supported across older client hardware. A channel the access point can use and the client cannot is a coverage hole nobody can see from the controller.
Comparison2.4GHz, 5GHz and 6GHz, side by side
| Criterion | 2.4GHz | 5GHz | 6GHz |
|---|---|---|---|
| Non-overlapping 20MHz channels | Three | Twenty five | Far more |
| Channels free of DFS | All three | Nine | All of them |
| Range through walls | Best of the three | Moderate | Shortest |
| Interference from non-Wi-Fi | Heavy: microwaves, Bluetooth, cordless | Radar only, and only on DFS | Minimal so far |
| Client support | Universal | Universal | Recent devices only |
The middle column is the compromise, and that is why the 5GHz band carries most enterprise wireless traffic. Three clean channels at 2.4GHz is not a plan for any building with more than a few access points, and 6GHz solves the channel problem while asking every client to be new.
FAQFrequently asked questions
How many channels does 5GHz have?
Twenty five 20MHz channels in the United States, and unlike 2.4GHz none of them overlap. Sixteen of the twenty five are DFS channels and nine are not: 36, 40, 44, 48, 149, 153, 157, 161 and 165.
What is the best 5GHz channel?
For a home or a small office with few neighbors, an 80MHz channel in non-DFS spectrum, which means centered on 42 or 155. For a dense office, a narrower channel and a plan that includes DFS, because nine channels will not cover a building that needs thirty cells.
What are DFS channels?
The channels in 5.25 to 5.35GHz and 5.47 to 5.725GHz, where the FCC requires devices to detect radar and avoid sharing a channel with it. In the US that is channels 52 through 144, sixteen 20MHz channels in total.
Why does my access point stop transmitting for a minute?
Because it is on a DFS channel and performing the channel availability check. The FCC allows it to begin transmitting only if no radar is detected within sixty seconds of listening, so a minute of silence at startup is the rule working as written.
Why did all my clients drop at once?
If it happened on a DFS channel, the access point probably detected radar. Once it does, all transmissions must cease within ten seconds, with normal traffic allowed for at most 200 milliseconds of that. The clients were not disconnected by a fault, they were left behind by a legal obligation.
Can I turn DFS off?
No. The FCC requires manufacturers to take steps to ensure DFS functionality cannot be disabled by the operator. A setting that appears to disable it is restricting which channels the radio will use, which reduces your channel count rather than removing the obligation.
How long is a DFS channel unusable after radar is detected?
At least thirty minutes. The regulation calls it the non-occupancy period, and the clock starts when the radar was detected rather than when the channel was vacated.
Should I use 80MHz or 40MHz channels?
Count your access points. Eighty megahertz gives six channels in the US and two without DFS, so a dozen access points will reuse channels several times and interfere with each other. Forty gives twelve, and twenty gives twenty five. Width raises a single client's ceiling and lowers a crowded room's capacity.
Is 160MHz worth using?
Only where one client needs a very high ceiling and interference is not a factor. Both 160MHz channels in the US sit inside DFS spectrum, so there is no way to run 160MHz here without accepting the radar rules, and reuse at that width is severe.
Are DFS channels weaker?
In the United States, yes, and it is a regulatory limit rather than a hardware one. The FCC caps the two DFS bands at the lesser of 250mW or 11dBm plus 10 log B, while U-NII-1 and U-NII-3 permit 1W conducted. Expect a smaller cell on the same access point.
How are 5GHz channel numbers calculated?
The center frequency in megahertz is 5000 plus five times the channel number, so channel 36 is 5180MHz and channel 161 is 5805MHz. Twenty megahertz channels sit four numbers apart, which is why the list steps in fours rather than counting upward.
Do I still need 2.4GHz?
Usually, for range and for client compatibility. Two point four gigahertz reaches further through walls and every device supports it, but it has only three non-overlapping channels and shares its spectrum with microwaves, Bluetooth and cordless phones. Most designs run both and steer capable clients upward.
What about 6GHz?
It solves the channel shortage and the DFS problem at once, with far more non-overlapping channels and no radar obligation, at the cost of the shortest range of the three and support only on recent client hardware. It complements 5GHz rather than replacing it, in the same way 5GHz never replaced 2.4.
What are the UNII bands?
The 5 GHz band is split into UNII bands. In the United States UNII-1 holds channels 36 to 48, UNII-2A holds 52 to 64, UNII-2C holds 100 to 144, and UNII-3 holds 149 to 165. The two UNII-2 ranges require DFS, which means an access point must vacate a channel when it detects radar.
Which 5GHz WiFi channels do not overlap?
At 20 MHz width every 5 GHz channel in the standard 5GHz channel list is non overlapping: 36, 40, 44, 48 and so on up to 165. Non overlapping 5GHz channels become scarce only when you widen them, because a 40 MHz channel consumes two and an 80 MHz channel consumes four.
What 5GHz channel width should an office use?
Usually 20 or 40 MHz. A 5GHz channel width of 80 or 160 MHz looks fast on a single client and leaves too few channels to reuse across neighboring access points. WiFi channel planning in an office is mostly about reuse, so narrower channels give better results once there are several access points.
Keep readingRelated concepts
Read next · Infrastructure Quality of Service, and the Condition It Needs to Do Anything Why a crowded channel and a slow application are different problems, and what prioritization can and cannot fix. Open this next11 min- Wireless · 9 min A WLAN Is a LAN With a Radio in Front of It What the wireless LAN actually is underneath the channel plan, and where the 802.11 standards fit.
- Diagnostics · 11 min Packet Loss vs Latency, and Why They Feel the Same to Users How to tell a radio problem from a path problem once the channel plan is no longer the suspect.