← all posts

Why I Start at 40 MHz for Enterprise 6 GHz

September 21, 2026 · by Brent Leekley · ~7 min read

Why I start at 40 MHz for enterprise 6 GHz. Three reasons in a row: the clients are mobile-first with one or two spatial streams needing a few megabits, AFC notches on Cisco's Standard-Power implementation take a whole wide channel when they clip any part of it, and college football stadiums running outdoor AFC reportedly chose 40 MHz. Caption: in 6 GHz the width goes first, because the width you choose constrains the mode you can use.

Third post in the 6 GHz power mode series. Posts one and two covered the LPI versus Standard Power decision and the AFC machinery behind Standard Power. This one covers channel width, which is the spectrum column of the power, spectrum, and operations trade. The mode decision and the width decision are really the same conversation, and in 6 GHz the width goes first, because the width you choose constrains the mode you can use.

What are your clients actually doing?

Your clients are doing less than the spec sheet assumes, and that is the whole argument in one line.

Spec sheets and launch decks talk about 160 and 320 MHz channels, and the math on those slides adds every radio together into one headline number. The clients paying the bill for your design do not look like that. The devices arriving on enterprise 6 GHz networks are mobile-first: phones and tablets with one or two spatial streams, running video streaming, messaging, and browser traffic that needs a few megabits per second. A 40 MHz channel with a good MCS rate (modulation and coding scheme, which sets how many bits fit in each transmission) covers that with room to spare.

Battery is part of it too. Sustaining an ultra-wide channel costs a mobile radio power and heat, and handset radios are built to prioritize battery life over peak channel width. So even when you hand a phone a 160 MHz channel, the phone gets a vote on how hard it works, and it usually votes for its own battery. Wide channels mostly benefit a small population: laptops doing sustained, large transfers like backups or big file syncs, traffic that would do just as well over a wired connection. That is not the population filling an auditorium.

Buying width you cannot spend is the same mistake as buying a three-quarter-ton truck for a ten-mile commute. It is not that the truck is bad. It is that you paid for a capability you will never put to work, and in a channel plan you pay for it in reuse.

Why do clients already prefer 6 GHz?

Because the operating systems decided for you, and that turns out to do a lot of design work on your behalf.

Apple documents a 6 GHz preference in its auto-join band selection, and Android's network selection is more involved and leaves more to the OEM. The panelists on the Tech Field Day podcast that prompted this series tested band selection across their own iOS, Android, Windows, and Mac devices and reported a consistent operating-system-level preference for 6 GHz over 5 GHz, even at equal channel width. Treat that as an observation of the devices they tested rather than a guarantee for every OS version and OEM. Keith Parsons described a roaming test across a large site, sampling every ten seconds for around twenty minutes, in which a 6 GHz-capable client never once dropped back to 5 GHz; I have not listened for the exact timestamp, so go to the episode for the full setup.

That preference pays you twice. Every capable client that moves to 6 GHz stops consuming 5 GHz airtime, and the legacy devices stuck on 5 GHz get a faster network without you touching anything on their band. It is the rare change that improves a band you did not work on. And because the preference holds at equal width, you do not need ultra-wide channels to collect it. What you need is 6 GHz cells the clients can actually reach, which loops straight back to the talk-back and power mode discussion from post one.

Client band preference. Across iOS, Android, Windows and Mac the panel reports a consistent operating-system-level preference for 6 GHz over 5 GHz, and it holds even when both bands offer the same channel width. In a roaming test across a large site, sampling every ten seconds for around twenty minutes, a 6 GHz-capable client never once dropped back to 5 GHz. Every capable client that moves off 5 GHz gives that airtime back to the legacy devices stuck there.
The band preference holds at equal width, so you do not have to buy clients with bandwidth.

How channel width interacts with each power mode

In the US, LPI's power spectral density rule adds 3 dB of allowed transmit power each time width doubles, up to the band's EIRP cap (equivalent isotropically radiated power), so an AP that spends the allowance can hold its SNR (signal-to-noise ratio) and wide channels genuinely work. That is a US/FCC number and it assumes the transmitter actually uses the allowed power. It is not how every regulatory domain sets LPI. In the EU, Commission Implementing Decision (EU) 2025/913 caps LPI at 23 dBm mean EIRP with a 10 dBm/MHz density limit across 5945 to 6425 MHz. Because 10 dBm/MHz across 20 MHz already works out to 23 dBm, a 20 MHz channel reaches the total cap on its own, and there is no density headroom left to spend on a wider one. The doubling argument below is a US argument. If your design goal is maximum per-client throughput, and you are working under the US rules described above, LPI with wide channels (80, 160, or Wi-Fi 7's 320 MHz) is the highest-throughput configuration Wi-Fi currently offers, and I would not talk anyone out of it when that is the actual goal.

Standard Power is where width gets expensive. The AFC exclusion masks from post two can remove chunks of the band, and because a traditional 802.11 channel has to be one unbroken run of spectrum, and because Cisco excludes static preamble puncturing when Standard Power and AFC are both in use, a notch anywhere inside a wide channel's block removes the entire channel on that platform today.

Think about parking. A lot with a few pillars scattered through it still parks plenty of cars, because a car can go almost anywhere the pillars are not. Put one of those same pillars in the middle of a bus bay and the bus does not fit at all, and the pillar did not get any bigger. That is a notch against a 40 MHz plan versus a 160 MHz plan. The obstruction is the same size in both cases, and the bill is completely different.

How much you lose depends on the incumbents around your site. A clean grant can support 160 MHz, and Cisco documents 160 MHz Standard Power operation, so this is not a prohibition. You are not entirely flying blind, either: Cambium and Federated Wireless both offer tools to check likely spectrum availability before you deploy. A preliminary result is not a final authorization, and the grant you actually get can still differ, so treat it as planning input, not a guarantee. Still, a wide-channel Standard Power plan is a bet on the incumbent map at every site you deploy, and a notch you did not plan for takes out a channel you could not afford to lose. Standard Power at 20 or 40 MHz needs a much smaller contiguous block to route around any given notch, which is why it keeps a deep channel pool and healthy reuse across a wider range of grants. Verify the usable channel count at your design's minimum EIRP for each site; a narrow plan improves your odds, it does not guarantee them.

So the modes pair naturally with widths. LPI pairs with wide channels and buys throughput. Standard Power pairs with 40 MHz and buys client talk-back. If you have a phone at the edge of a large room, a 40 MHz Standard Power cell will serve it better than an 80 MHz LPI cell, because the wider cell does the phone no good if the AP cannot hear the phone answer.

U.S./FCC rules, and this assumes the transmitter uses the allowed power. How channel width affects signal to noise ratio in each band. In 5 GHz U-NII-1 through U-NII-3 a total power cap does the limiting, so doubling the width adds 3 dB of noise on the same power and SNR drops with every doubling. In 6 GHz Low Power Indoor the binding limit is power density, so the rules allow 3 dB more power each time the width doubles, up to the band's 30 dBm EIRP cap, and SNR holds flat from 20 MHz through 160 MHz. The EU sets LPI differently: a 23 dBm total EIRP cap and a 10 dBm/MHz density limit mean roughly 20 MHz already hits the total cap, so this doubling argument does not carry over there.
U.S./FCC, assuming the transmitter uses the allowed power. The 5 GHz reflex of narrowing a channel to protect the cell edge does not carry over to 6 GHz LPI here, but check your own regulatory domain.

The evidence from high-density deployments

Outdoor stadium deployments have been running AFC longer than indoor ones, so they are the closest thing we have to operational history, and the panel on the Tech Field Day podcast pointed to college football stadiums as their working example. Those networks reportedly run 40 MHz channels under AFC, at high power, with massive reuse, and hold up under some of the densest client loads Wi-Fi sees.

That is good evidence for that venue, not proof every enterprise should make the same choice. It cuts against wide channels being a free win for a stadium's crowd and channel plan, operating under real AFC grants. Context still matters: HPE's own large-public-venue RF planning guidance recommends 20 MHz for its scenarios, a different number for a similar problem, which is a reminder to size the width to your venue rather than borrow one stadium's number.

There is also a portability argument that shows up on the invoice eventually. The US got the full 1,200 MHz; much of the world did not, and 80 MHz plans get difficult in regions with less spectrum. A 40 MHz design built in the US carries to most international sites with far less rework, which is worth real money to anyone running a global standard.

The two modes pair naturally with two channel widths, in the US. Low Power Indoor pairs with wide channels of 80, 160 or 320 MHz and buys throughput, because the power density rule holds SNR. Standard Power pairs with 20 or 40 MHz and buys client talk-back, because a deep channel pool survives a wider range of AFC grants.
Each mode has a width it works best at, and they are not the same width.

Where do I come down?

I start at 40 MHz for dense enterprise deployments with many lightly active clients, in either mode, then validate capacity, latency, and coverage against the actual client mix before calling it done. Wider channels stay on the table when the demand and the channel plan justify them.

Go wider when you can point to the client population and the application traffic that needs it, and your mode supports it. LPI keeps wide-channel SNR by rule in the US, so it is the mode that can most reliably carry an 80 or 160 MHz design there. Standard Power benefits from a clean grant at every site, a condition you can check for in advance with a planning tool but cannot fully confirm until the real grant comes back.

If you are going to Standard Power for talk-back reasons, planning 20s and 40s from the start means the AFC masks matter less wherever they land, and a plan built that way survives more grants than one banking on a clean one. The last post in this series covers how to see all of this on a Catalyst 9800 over NETCONF, and how to monitor the AFC pieces so the database never surprises you.

The rest of the series

  1. Choosing a 6 GHz Power Mode: Low Power Indoor or Standard Power.
  2. What AFC Requires From a Standard Power Access Point.
  3. Why I Start at 40 MHz for Enterprise 6 GHz (this post).
  4. Configuring and Monitoring 6 GHz Standard Power on a Catalyst 9800, Programmatically.

← all posts