2

TLDR: Why do DIY access points suck, and what is big networking hiding from me?

I'm going through another set of network upgrades, and 802.11 ax adaptors on the client side are really attractive for my wireless needs for systems with no wireless adaptor. Intel's AX200s are lovely low cost ax capable units that seem really interesting. 802.11ax capable routers however, are pretty pricy, costing more than my x86 capable router or many mesh networks.

I've typically used proper APs for wifi, but its the one part of the network I have no control over.I've played around with DIY routers a lot (since I apparently make most consumer routers cry) . While ram and CPU constrained, a consumer router, even a cheap,cheerful or geriatric one is going to run circles around say a top of the line laptop, with the latest wireless gear running hostapd.My setup hardly reaches half the range of my AP even with external antennas.

Now, its pretty clear that the secret sauce isn't in CPU (a raspberry pi will run circles around it) and a good chunk of modern routers and their firmware run linux, and even hostapd.

Picking a router I've used in the past - the RT-N56U

A quick look at the sadly defunct wikidevi and info depot shows the hardware it uses- it uses the Ralink RT3092 (which also appears in a mini PCIe formfactor) for 5ghz with 2 antennae and Ralink RT3662F (from the SOC) for 2.4 ghz with 3. - there's nothing particularly magical about either one of these - other than having seperate radios and the sheer total number of antennas (and many desktop wireless cards have 3 or more). That said, I've seen them go up to 6.

There's also mentions of a power amplifier for each band.

On the other hand the Turris Omnia seems to use entirely off the shelf mini PCIe cards and no amplifiers.

Assuming I wanted to build a better DIY AP - which of these would I be looking at - dedicated cards per band and/or better/more antenna, power amplifiers (can I even legally bolt one into a DIY setup?) or are there other factors in play?

Improve this question
2
  • In most jurisdictions you can legally bolt amplifiers onto your output as long as the radiated power does not exceed the limit. This is generally not helpful though as it increases both signal AND NOISE.
    – davidgo
    Nov 23, 2019 at 23:30
  • More and better antennas are definitely a thing - but TANSTAFL - ie if an antenna has greater gain in horizontal plain it will compensate by less in a vertical plain
    – davidgo
    Nov 23, 2019 at 23:35

1 Answer 1

Reset to default
1

For maximum performance Wi-Fi, you want each radio chain to have a high quality, high power, PA (power amplifier - the standard term for the transmit amp) and a high quality LNA (low-noise amp: the receive amplifier). You want each radio chain to have its own well-optimized single-band antenna. You want to de-correlate your antennas to maximize your MIMO benefits. You want to design and position your antennas to have the appropriate coverage pattern for how your AP is likely to be positioned and used. You want to tune your transmit power per antenna and per MCS to ensure you always hit the regulatory limits (without exceeding them) for every possible transmission.

You want to keep your radios cool. Heat is a source of RF noise. Read up on "black body radiation" if this is news to you.

You want to keep your radios EMI-shielded from the rest of the system.

For the rest of the system, you mainly need to make sure it's not a bottleneck. If you're trying to do it in an embedded device you've got to make sure you have enough CPU power to keep up with the CPU demands of the Wi-Fi driver and software, and Cake or FQ-CoDel SQM, and NAT (if enabled) and Gigabit Ethernet (helps to have lots of hardware offloads in your GigE NIC).

Improve this answer
5
  • Funny thing is "the rest of the system" probably is overpowered - though I've run the base software on a raspberry pi clone, this is a core i3-4010U based firewall box. I wonder how much of the rest is doable on a consumer system or a NUC class machine... And of course, the last part of the first paragraph I suspect comes into the domain of "Well Equipped, specialised EE lab"
    – Journeyman Geek ♦
    Nov 24, 2019 at 3:53
  • I suspect many APs actually have hardware acceleration at least for WLAN-Ethernet bridging, while on general purpose systems that just goes through the OS.
    – u1686_grawity
    Nov 24, 2019 at 4:00
  • That's vaguely why I focused on wireless - planning on popping a AX200 into that thing to see how it goes, I can probably get/swap the antenna in theory. For the low cost of base level PC level AX gear, the AP side of it really pricy :/. I don't really think I have too much flexibility short of building a full fat matx or full atx router.
    – Journeyman Geek ♦
    Nov 24, 2019 at 4:07
  • @grawity WLAN to LAN bridging is usually the software bridge in the OS. LAN switching is a 5-port switch chip. NAT (WAN to [W]LAN) is sometimes hardware accelerated though, especially if your SOC is less than 1GHz single core.
    – Spiff
    Nov 24, 2019 at 5:22
  • @Spiff: Software bridge on Linux can still be offloaded to hardware, at least I'm sure I've seen it for some switch-chip models that Linux natively supports, and I wouldn't be surprised if a SoC could do it for WLAN as well.
    – u1686_grawity
    Nov 24, 2019 at 5:37

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .