What mesh wifi?

chattoe

Also, can you do a wired backhaul?

Mr X

You can see the UK distributors here

Ubiquiti Networks - Democratizing Professional Network Technology

Prices seem to have gone up everywhere recently but I got mine from LinTX (see link) and they had customer returns at the time, so I paid about £300 for the 4 units (all new), instead of £360 for 3.

Mr X

Yep... you can see two of my units are connected via ethernet to give max wifi capability from that mesh.
Testing these two mesh units via a wired connection, I have one connected over 20m and the other 30m cat5e cable, and I suffer minimal (if any) degradation when compared to the main wired router so both give around 940Mbs which is same as the router.

mickevh

For Ethernet usage, cable length makes no difference to performance, it's either full speed or nothing at all whether the lobes are 1m or 100m. Within Ethernet standards, max. lobe length is stated as 100m (strictly speaking, 90m for the permanent cable run plus 5m each end for patchcords.) If you do a good enough job, they can sometimes go even further.

Higher speed Ethernet (e.g. 10G) requires higher "cat" installs, (or goes less distance for a given "cat,") but 100/1000 Ethernet (which both pulse the cable at the same rate) should be fine up to 100m unless you do a spectacularly bad job of the installation - poor termination is the biggest culprit.

Bluedug

Updated on my set-up, tenda nova mw3. Bought another pack of 3 (was cheaper than two) now have 5 installed, however think I am at the range limit. We have a cabin at the bottom of the garden, placed one cube in the corner of the house closest to the cabin but signal fluctuates between good to poor at the cube in the cabin . Have tried with the wired backhaul but had very erratic speed tests, think will need to look into a mesh with longer range.

mickevh

Do you mean a wire to the cabin...? If so, that should have virtually no effect on speeds. As above, ethernet works either full speed or not at all - there's nothing in between.

If not and you're talking about a Wi-Fi link, if basic omni-directional AP's aren't working then you might consider something with separate directional antenna such as yagi's or parabolics which you can mount externally and point directly at each other with unobstructed line of sight. Some solutions offering such "point to point" (AKA "bridge" or "mesh") links can only do the PTP, (no client access,) some can do PTP   client access (and may have separate sets of antenna for each) so you may need to shop around a bit.

Getting Wi-Fi to the "bottom of the garden" using indoor AP's can be challenging because, despite placing the AP's as close as possible to the external wall, you've still shot through (typically in the UK) a cavity wall, ie two walls, then by the time whatever is left has got to the shed/cabin/summer house/whatever, it's competing with all the Wi-Fi "leaking" out from your neighbours and there's not much difference between the two.

You could perhaps try moving the AP's into windows if there are any overlooking each other, but ultimately, if that doesn't work then you need either a line of sight link with directional antenna or dig the garden up and lay in a cable (the latter being the best solution.)

Bluedug

With no Ethernet the nodes see each other but signal to them is very weak (cabin node connected to node in house and node in garage) I had hoped that by running cat 6 to the node in the cabin it would be able to utilise that to boost the strength, however that doesn’t seem to be the case. I have tested the connection in the cabin with my computer connected (direct to faceplate) and speed tests come back at 37 down, 7 up (40mbs line) and ethernet shows 1gbs. I have wired up a temporary access point in there (Netgear WAC120) but long term would prefer everything on the same “mesh” but for some reason I can’t seem to get this to work with the Tenda Nova.

mickevh

Internet based speed test sites are not very useful for testing ethernet links: The speed test effectively tests the performance of the slowest "hop" between the test device and the test server and for Internet speed tests, that is almost always your ISP link. To test local links, one is better off to use tools such as NetIO or iPerf which essentially let you create speed test server locally and take the ISP link (and everything beyond) out of the equation.

To be sure to test an gigabit link at gigabit speeds, you need to create a pathway between the NetIO/iperf server and the client that is gigabit (or better) all the way.

That said, if your computer is saying the faceplate link is running at gigabit, then it almost certainly is - gigabit generally works full speed or not at all. Using the UTP to your cabin and hanging an AP on the end of it should give you the best performance. It is doubtful the ethernet link is the problem.

This suggests something else is going wrong. I don't know the Tenda kit, but I'd want to look at the management app and see if it will tell me whether the nodes are using wired or Wi-Fi backhaul. If possible, even force the one in the cabin (if not all of them) to use wired backhaul wherever you have it.

Also, when testing, beware of "Big Wi-Fi Myth Number 2" that client devices are always "hunting for the best signal." They do not, and some clients need it to get pretty grotty before they roam. You could be sat right next to an AP, but if your client has decided to "talk" to another one further away with less good performance, you would of course see the performance from that AP rather than the one you are nearest.

Client devices, not "the system," decide which AP to "talk" to (called Associate with in Wi-Fi speak) and clients also decide if/when to roam between AP's. Even when you first power up a Wi-Fi client, it doesn't necessarily Associate with the nearest AP, so don't "just assume" that because you can "see" a particular AP close by, your client is talking to that AP.

Some Wi-Fi system have a few "tricks" up their sleeve to try and encourage client to roam more readily, but it is still ultimately up to the client to decide if/when to initiate a roaming assessment and decide whether to switch AP's or not.

Thusly, if your AP's all have the same SSID, you literally won't know which one you are talking to unless you have a special tool that tells you.

For testing purposes, one way to get around this (if the Tenda will let you) is to temporarily assign each AP a different SSID, then test them all one by one. Once you have them all working just so, then change the SSID's back to being all the same. It's the sort of thing we do when verifying/troubleshooting enterprise systems. (Or more likely, we'll wander round, give each AP a "test" SSID, check it, then rename SSID back to the same as the rest of the fleet.)

Bluedug

Thank you very much, a lot to look into! Will report back when I build up the will power to tackle it again. Happy with the AP just now which is working well.

Bl4ckGryph0n

It is a limitation of the Tenda. If you plug in a wired backhaul it is no longer a mesh network, makes sense when you think about it.

There is also not much point for going for a mesh with longer range as that doesn't exist. Well not legally as they all have to play within the legal limits. What I'd do in that situation is have a normal access point, you already have an cat 6 cable don't you? Then plug that one into the master node of the Tenda back in the house, and it then will appear that everything is on the same network, and you'll get an assignment from the same DHCP server as well.