There is a section in both the AP903 and AP802 manuals that says:-

For your reference, we make a real mesh environment test and get the following record.
(Use VigorAP APP to do internet speed test with different hops mesh node.)
Internet Download Speed (for root and hop1 ~ hop3):
iPad connects to Root : 80Mbps
iPad connects to hop1 Node : 49Mbps (Uplink RSSI : -55dBm)
etc

However, since they used the Internet as their Speed Test source, it's hard to know whether to expect 49/80 of normal/wired speed - or 49Mbps as the actual answer. The speed tester built into Draytek's 'Wireless APP' (left-hand icon, top-right of screen), uses various .ZA & .RO addresses - which don't exactly inspire confidence!

So, I set up a test rig, took a few measurements - and learned a few things along the way!

Test software was the Android "WiFi Speed Test" application (pro) from the Google Play store , with the Test Server running on a PC under Windows 10. This PC was connected to an AP903 LAN Port. The app. ran on a Huawei Mate 20 Pro. The test used a 200MB transfer, using 4K packets.

I only investigated the 5GHz band.

The AP903 was positioned on the 2nd floor, at one end of the building and was set to 'Auto 20/40/80 MHz' mode. It chose channel 36, 80 Mhz (i.e 34-50)
It discovered and auto-provisioned the AP802 over wireless - it chose channel 44, 40MHz mode (i.e. 42-50). This overlapped that chosen by the AP903.

Taking a speed test at the intended location of the AP802 (about 5m diagonally below the AP903, on the 1st floor), gave an average speed of 222Mbps down/326Mbps up.

With the AP802 positioned here, the Huawei was moved to the ground floor (another 5m diagonally).

The recorded speeds (averaged over 3 runs) were 59Mbps down / 67Mbps up.

The AP802 was now switched to wired-uplink mode (the location had been chosen to facilitate this) and the tests repeated.

The recorded speeds (averaged over 3 runs) were now 146Mbps down / 127Mbps up.

I still don't know if this is a fixed overhead - but at least I know the answer is not 49Mbps :wink:

From other tests I did, I found that selecting an 80MHz channel vs a 40MHz channel, does indeed double the throughput. When in Mesh Mode, the AP802 writes the message:-

Code:

Don't set 80M, Force VHT_BW 20/40MHz on mesh

to Syslog and will not use 80MHz.

This may be a requirement to make 'wireless meshing' work - but it also does it in wired-uplink mode, thus immediately halving the potential throughput.


With the AP802 set to "AP" mode, Channel 52, 80MHz selected (so no overlap with AP903), it achieved speeds comparable to those recorded for the AP903.
For now, I've abandoned Mesh Mode and set the AP903 to "AP" mode as well

AP Assisted Roaming appears to still work, but with no SYSlog confirmation.

A somewhat more comprehensive test than my fiddling so far....

But from my previously reported "mesh" stress testing - i.e. switching off one of the mesh AP903s and see what happens - I too have come to wonder whether the Draytek "mesh" system is really worth the bother and expense. It just doesn't seem to work as a "mesh" as I understand it. As for having some devices with a dedicated 5GHz backhaul and others without, what's that all about?

I've been looking at the Syslogs from the AP903s and I can't see anything like your "force 20/40" . Is that something which only happens when the AP starts up in mesh mode?

Piste Basher wrote:
I've been looking at the Syslogs from the AP903s and I can't see anything like your "force 20/40" . Is that something which only happens when the AP starts up in mesh mode?

Yes, only in Mesh mode; it's just part of the immense 'splurge' that is output as the device boots.

Do you see the same 20/40 mode being forced on your AP903 Mesh Nodes (including the wired ones) :?:
A throughput of 59Mbps is only acceptable, if you're only connecting to the Internet - and you have a typically naff FTTC connection to it. (It seems to be halved from its potential maximum, by having 80MHz disabled and then (more than) halved again if it's being used as the Backhaul.)

I too have come to wonder whether the Draytek "mesh" system is really worth the bother and expense. It just doesn't seem to work as a "mesh" as I understand it. As for having some devices with a dedicated 5GHz backhaul and others without, what's that all about?

I think the actual Mesh function is quite a 'static' thing; you can probably see the topology, by studying the various uplink/downlink nodes reported by each node. I don't think it goes searching at the time of failure. What you would really need to do, I suppose, is set all your AP903's to 'wireless' Mesh mode for a proper test.

From what I can make out, once 'wired-uplink' is selected, all you get is the ability to make a change on the Root Node and then push it out to the clients. Useful for the initial setup I suppose, but after that, well ... (there's only the 'Object' list that changes much in my setup, and it's a minor convenience not having to update it twice, when I add a new entry).

An interesting piece of the jigsaw, is this opinion on the use of 5GHz...
https://www.cablefree.net/wirelesstechnology/unlicensed-wireless-links/using-the-5ghz-band-in-the-uk/

They wrote: I will cover this in more detail in part 2 of the blog, but the bottom line is for indoor use 2.4GHz should be your weapon of choice unless the 2.4GHz spectrum is too crowded or your building is open plan.

I've not yet spotted "Part 2 of the blog" ...

Something else, that occurred to me ...

I'll bet that when running in wireless mode, better throughput can be had by using WDS or Universal Repeater mode, rather than as a Mesh-node. They all half the ultimate bandwidth, but in the case of the first two, that can be half the 80MHz throughput, rather than half the bandwidth of the enforced 40Mhz mode as a Mesh-node.

I'll try it, when I get the inspiration :wink:

Just a quick response to one point - as you may recall I was running the 2926ac as root, with 3 wired nodes and 1 wireless (in the garage). Last week I thought I'd try having all nodes wireless in order to see if they formed a "self healing" mesh as advertised, since the wired/wireless setup did not appear to do so. So I dismantled the mesh and started from scratch, searching for and adopting nodes from the 2926ac. No problem with the first one, in an adjacent bedroom. I could also find and adopt the AP in the garage. But no matter how many times I searched I could not find the APs in the Kitchen or the Lounge, even though they are physically closer than the one in the garage and, by using the AP discovery feature, I could "see" them both - and each with a reported signal strength higher than that of the one in the garage

That is a mystery to me.

So at present I have two wired nodes and two wireless - bedroom connected to router directly over 5GHz, garage connected to either kitchen or lounge over 5GHz, kitchen and lounge wired back to router.

Haven't had the energy to fiddle further as yet.....