Re: PtP Relay Throughput

RouterOS Wireless Networking
1

Can you give the technical details please.

  • SXTsq-SXTsq connection is what channel/ channel width? If you get 550 Mbps data rate, it must be a channel width of 80 MHz and 2 spatial stream / SGI Short guard interval., giving an interface rate of 867 Mbps. That current information can be found in the wireless “registration” . Interesting to know is the receive signal strength, signal to noise ratio and CCQ
  • The other SXTsq-SXTsq connection is at least on a different frequency I hope, and NOT overlapping. If not , then halving the data speed is normal. The directional antenna leaks signal also backwards, even the -25 dBm lower signal leaked will make the other transmission wait.
  • Never select “auto” channel select. Make sure you have no overlap! 80 MHz will be Ceee (or eCee, eeCe , eeeC), because with XXXX you don’t know on what side the e-channels are.
  • That Omnitik AP must yet be on another 80 MHz wide channel, not overlapping at all with the other 2.
  • All Ap’s will send all broadcasts and multicasts at the basic rate. This takes a lot of airtime. Do not enable A, G or N mode. Only AC. You only want the VHT rates , not the HT rates (N) , not the 6-till-56Mbps rates (A,G), or their basic rates.
  • Use both chains for receive and transmit. This gives the needed 2x2 spatial stream. Txa/Txb is when you have only 1 antenna, this is not the case for SXTsq and Omnitik.
  • No clients on the Omnitik AP? But the AP is sending its beacons at the basic rate. If the data flow is not well managed, then it will also send the broadcasts and multicasts that are on ether2/ether3

That is 3 times 80 MHz channel width. Not easy to find that free spectrum space. You may consider reducing the Omnitik to 40 MHz, or even all to 40 MHz width, halving the throughput capacity. And if still in trouble, then exact channel match (co-channel) is much better than half overlap (adjacent channel interference). Both SXTsq at the central mast must be on different channels, if you want to relay traffic. The Omnitik might join one of the 2 SXTsq channels, but it is much better if not, as it will disturb. An AP even hidden is always transmitting its beacons !

2

Glad to help, if I can. Many decisions are in your hands, as they depend on details of the status of the links, and on the service degree you want, and on the actual usage.

So far you are in a laboratory condition. Not ideal as you say because of all the reflections you have. But it explains the clear numbers (550, 270) for a good signal, and strong co-channel interference.
Real world conditions can be worse for the connection status (interface rates), and somewhat better for the co-channel interference. (The distance between SXTsq-b and SXTsq-c can help a little bit.)

Now on your configuration. Using freq 5745-5805 , not many locations are allowed to use these frequencies (USA, Canada, Singapore, …) https://en.wikipedia.org/wiki/List_of_WLAN_channels
This is just one (1) 80 MHz band. Is there the possibility to move to the DFS channels? In the wikipedia they say 5180-5240 is even allowed outdoor in the USA !?
CLI command "/interface wireless info country-info " will tell you what you can use.
With only one 80 MHz band, you choices are limited.

-either you go for the 80 MHz bandwidth, and the setup 5745Ceee and 5805eeeC is the same band if the devices set full speed. What they decide can be seen in “registration” again.
I would add the Omnitik to the c-d frequency setting, as the modem (=most traffic) is on the a-b link. (If the omnitik cannot be set to something totally different)
This is the dynamic setup. You have co-channel interference and let it play its role. The Ceee-eeeC idea is very good for lower transmission needs, but at full load it will co-channel interfere.
The results will vary over time, as the load varies.
SXTsq-a and SXTsq-d might not receive each other well enough to do co-channel coordination. SXTsq-b and SXTsq-c will be hit by the “hidden node problem”.
This is not visible in your basement setup, as SXTsq-a and SXTsq-d can receive each other.
For the “hidden node” problem, RTS/CTS (HW protection) should be activated.
As we are in the co-channel condition with the SXTsq’s and the Omnitik-a , maybe, the signal is strong enough to just work with the Omnitik-a, without the SXtsq-b and SXTsq-c.
Using the Omnitik-a for client devices indoors is not the best condition for the signal. Be aware of the donut (thorus) shaped antenna spectrum.
My experience learns that a WAP ac (or hAP ac2, or any device with a 2-3dBi antenna emitting in all directions) is better for indoor signal than an Omnitik. (in my brick and mortar house, through wall and ceiling)

  • or you choose for the stable steady and consistent setup. Then you split the 80 MHz in 2 separate 40 MHz bands. Halving the max throughput, but consistent as both links will work at the same time in their own band. Again avoid to have the Omnitik-a in the same bands.

In all cases the hAP ac2 and the Omnitik_b should be on different frequencies from their SXTsq uplink.

The real setup will also have its own challenges. The conifer are good in absorbing the radiation, and eliminate reflections. But they can absorb much more than you wished for. Keep the “fresnel zone” in mind when making the setup design. Also know the radiation patterns. The Omnitik has the weakest signal along its vertical axis, and it has no “downtilt” pattern as far as I know. Don’t let the signal shoot far above the client devices.

A lot of “or” and “if”, as most depend on what the registration is learning, and on what you want to do where.