4 ms·

Really good whitepaper on WiFi 6 (802.11ax): https://www.cisco.com/c/dam/en/us/products/collateral/wireless/white-paper-c11-740788.pdf https://www.cisco.com/c/d

by mrb 8y ago

Really good whitepaper on WiFi 6 (802.11ax): https://www.cisco.com/c/dam/en/us/products/collateral/wireless/white-paper-c11-740788.pdf https://www.cisco.com/c/dam/en/us/products/collateral/wirele... Also: http://www.ni.com/white-paper/53150/en/ http://www.ni.com/white-paper/53150/en/ Main differences with 802.11ac: - denser subcarriers (or "tones") spacing; they are spaced by 78.125 kHz instead of 312.5 kHz, for example a 80 MHz channel now allows 980 data subcarriers, up from 234. That's a 4.19× improvement (4× higher density doesn't translate in exactly a 4× improvement because 11ac had a bit more "pilot" or "null" subcarriers not used for data.) - new 1024QAM mode (encodes 10 bits per symbol, up from 8 bits with 256QAM). That's a 1.25× improvement. - downside is that the symbol length had to be increased from 3.2 µs to 12.8 µs, and the symbol interval too from 0.4 µs to 0.8 µs. That's a 3.78× reduction in efficiency. So the maximum data rate for a single stream on a 80 MHz channel increased by 4.19×1.25/3.78 = 1.39× between WiFi 5 and WiFi 6 (you can confirm it with published max data rate: 802.11ac = 433 Mbit/s, 802.11ax = 600 Mbit/s; and 600/433 = 1.39). But the best feature of WiFi 6 is that different subcarriers can be concurrently used by different users, what they dubbed OFDMA (https://en.wikipedia.org/wiki/Orthogonal_frequency-division_multiple_access https://en.wikipedia.org/wiki/Orthogonal_frequency-division_...). This means that during the same 12.8 µs timeslot, even on a small channel like 20 MHz, you can have 9 concurrent users each assigned 26 subcarriers and each transmitting 26 different symbols (total of 234 symbols transmitted concurrently.) Whereas with WiFi 5, all subcarriers of the 20 MHz channel have to be used by the same user.

zaroth 8y ago
Great information! I wonder how the subcarrier allotments are orchestrated, particularly under hidden neighbor conditions. Presumably the AP has to orchestrate it based on QoS settings, but the client-OS will need upgrades throughout the stack to signal how much bandwidth should be allotted/reserved at different times. Sounds complicated and tricky, and not likely to work very well for devices where bandwidth requirements are extremely bursty and unpredictable.
ricardobeat 8y ago
Due to the increased efficiency, the speed increase is supposed to be much higher. At CES they showed up to 11Gbits: https://www.zdnet.com/article/d-link-asus-tout-802-11ax-wi-fi-routers-but-youll-have-to-wait-until-later-in-2018/ https://www.zdnet.com/article/d-link-asus-tout-802-11ax-wi-f...
- zamadatix 8y ago
  That's just the marketing, AC would have been 7 gbit/s if measured in the same ludicrous 8x8 160 MHz setup (though technically AC only specifies up to 4x4 it doesn't matter, clients are almost exclusively 2x2, even high end laptops). Even in a full greenfield ax deployment it'll be a rare sight to see your phy rate pop above 1 gbit/s half duplex as a client. Like the parent comment said the actual felt speed gains in a real world situation are going to be from OFDMA allowing concurrent use at slower speeds.
  - cptskippy 8y ago
    I always took these numbers to be like a Switch's fabric speed. Not the speed an individual client would receive.
  - rsynnott 8y ago
    MacBook Pros have had 3x3 for a while now (except for the non-touchbar 13” post-2016, which is really more of a MacBook Air internally anyway).
    - zamadatix 8y ago
      Yep, it's about the one client I ever see be 3x3. They use a specially customized version of the 2014 BCM943602 to do it. Don't think I've ever seen a 4x4 client pop into our logs that wasn't an AP to AP bridge but I know there are a handful of adapters made which support it. Interestingly Intel gave up on making 3x3 cards a couple of years ago but will sell you a 160 MHz 2x2 ¯\_(ツ)_/¯
- mrb 8y ago
  Well this D-Link router shown at CES is "tri-band" (a misnomer, it should be called "tri-channel") so "11 Gbit/s" is just the sum of the maximum data rate achievable in 3 different channels (one channel in the 2.4 GHz band, plus two channels in the 5 GHz band), where each channels is as wide as possible (40 MHz at 2.4 GHz, and 160 MHz at 5 GHz), and where each channels uses 4 MIMO streams (ie. 4 antennas transmitting concurrently.) For comparison, the number I quoted in my parent post is just for 1 channel at 80 MHz with 1 MIMO stream. D-Link can claim "11 Gbit/s" because: - first channel at 2.4 GHz is 40 MHz wide (468 data subcarriers) where 802.11ax can operate at 286.8 Mbit/s, multiplied by 4 streams = 1147 Mbit/s - second channel at 5 GHz is 160 MHz wide (1960 data subcarriers) where 802.11ax can operate at 1201.0 Mbit/s, multiplied by 4 streams = 4804 Mbit/s - third channel also provides 4804 Mbit/s 1147 + 4804 + 4804 = 10755 Mbit/s which D-Link's marketing team rounds to "11 Gbit/s." It goes without saying that a typical client WiFi device (phone, laptop) will never reach 11 Gbit/s. For starters it will only use 1 of the 3 channels (4804 Mbit/s maximum) and most devices are 2×2 thus capable of 2 streams (2402 Mbit/s maximum.) Even if they are 3×3 you'd be lucky in real-world conditions to get about one and a half times or twice the bandwidth of one stream.
gammateam 8y ago
all I want to know is the maximum speed file transfer I can expect? with unlimited pipes from my ISP, and a fat 802.11ax router, what can I theoretically expect on speedtest.net I feel like your analysis neglected using multiple bands or channels or whatevers at once, and how the routers all can do double, tri quad bands already and might do more for this technology so we talking 3 Gbit/s?
- bdamm 8y ago
  How many other users are in your local area, and their use of the common medium, also matters. So it's hard to really state what boost you can expect in your own home. My read on this new standard is that it's built with conflict reduction in mind.

by mrb 8y ago

zaroth 8y ago

Great information! I wonder how the subcarrier allotments are orchestrated, particularly under hidden neighbor conditions. Presumably the AP has to orchestrate it based on QoS settings, but the client-OS will need upgrades throughout the stack to signal how much bandwidth should be allotted/reserved at different times. Sounds complicated and tricky, and not likely to work very well for devices where bandwidth requirements are extremely bursty and unpredictable.

ricardobeat 8y ago

Due to the increased efficiency, the speed increase is supposed to be much higher. At CES they showed up to 11Gbits: https://www.zdnet.com/article/d-link-asus-tout-802-11ax-wi-fi-routers-but-youll-have-to-wait-until-later-in-2018/ https://www.zdnet.com/article/d-link-asus-tout-802-11ax-wi-f...

zamadatix 8y ago
That's just the marketing, AC would have been 7 gbit/s if measured in the same ludicrous 8x8 160 MHz setup (though technically AC only specifies up to 4x4 it doesn't matter, clients are almost exclusively 2x2, even high end laptops). Even in a full greenfield ax deployment it'll be a rare sight to see your phy rate pop above 1 gbit/s half duplex as a client. Like the parent comment said the actual felt speed gains in a real world situation are going to be from OFDMA allowing concurrent use at slower speeds.
- cptskippy 8y ago
  I always took these numbers to be like a Switch's fabric speed. Not the speed an individual client would receive.
- rsynnott 8y ago
  MacBook Pros have had 3x3 for a while now (except for the non-touchbar 13” post-2016, which is really more of a MacBook Air internally anyway).
  - zamadatix 8y ago
    Yep, it's about the one client I ever see be 3x3. They use a specially customized version of the 2014 BCM943602 to do it. Don't think I've ever seen a 4x4 client pop into our logs that wasn't an AP to AP bridge but I know there are a handful of adapters made which support it. Interestingly Intel gave up on making 3x3 cards a couple of years ago but will sell you a 160 MHz 2x2 ¯\_(ツ)_/¯
mrb 8y ago
Well this D-Link router shown at CES is "tri-band" (a misnomer, it should be called "tri-channel") so "11 Gbit/s" is just the sum of the maximum data rate achievable in 3 different channels (one channel in the 2.4 GHz band, plus two channels in the 5 GHz band), where each channels is as wide as possible (40 MHz at 2.4 GHz, and 160 MHz at 5 GHz), and where each channels uses 4 MIMO streams (ie. 4 antennas transmitting concurrently.) For comparison, the number I quoted in my parent post is just for 1 channel at 80 MHz with 1 MIMO stream. D-Link can claim "11 Gbit/s" because: - first channel at 2.4 GHz is 40 MHz wide (468 data subcarriers) where 802.11ax can operate at 286.8 Mbit/s, multiplied by 4 streams = 1147 Mbit/s - second channel at 5 GHz is 160 MHz wide (1960 data subcarriers) where 802.11ax can operate at 1201.0 Mbit/s, multiplied by 4 streams = 4804 Mbit/s - third channel also provides 4804 Mbit/s 1147 + 4804 + 4804 = 10755 Mbit/s which D-Link's marketing team rounds to "11 Gbit/s." It goes without saying that a typical client WiFi device (phone, laptop) will never reach 11 Gbit/s. For starters it will only use 1 of the 3 channels (4804 Mbit/s maximum) and most devices are 2×2 thus capable of 2 streams (2402 Mbit/s maximum.) Even if they are 3×3 you'd be lucky in real-world conditions to get about one and a half times or twice the bandwidth of one stream.

gammateam 8y ago

all I want to know is the maximum speed file transfer I can expect? with unlimited pipes from my ISP, and a fat 802.11ax router, what can I theoretically expect on speedtest.net I feel like your analysis neglected using multiple bands or channels or whatevers at once, and how the routers all can do double, tri quad bands already and might do more for this technology so we talking 3 Gbit/s?

bdamm 8y ago
How many other users are in your local area, and their use of the common medium, also matters. So it's hard to really state what boost you can expect in your own home. My read on this new standard is that it's built with conflict reduction in mind.