AiMesh Adventures including Stock v Merlin on node

[Morris]

Part 1:

Adding a second as a node. I factory reset an old and proven RT-AX86U which had Merlin code and add it as a node. Within minutes I notice some of my IOT devices dropping and connecting again and again. I recall seeing others with similar complaints about IOT devices dropping. I turn off the node and they stay connected. I turn it back on and they start dropping. I connect my phone to the node and speed is terrible. I bring the node near the router and ethernet connect it. Connect my phone to it and communication is great. As signal was reported great in the nodes original position, I did not think it was a WiFi signal issue yet tried a much closer position and still bad speed with WiFi 5-Ghz backhaul. I load Asus firmware on the RT-AX86U node with the router RT-AX86U on Merlin. Place the node where it original was tried and it works great! There appears to be an issue with Merlin 386.5_2 on a RT-AX86U running as a node and RT-AX86U running the same version of Merlin. Asus 3.0.0.4.386.46061 works fine on the same node with Merlin 386.5_2 on the router.

Part 2:
I've been running on a single RT-AX86U since my node died mid fall. The node was used to cover my back yard as the router covers our home. With warmer weather we have been using our outdoors areas again and the outdoor WiFi has been sub par. I purchased a pair of XD6 connected them up and all was good though one of the two XD6 took a few tries to bring up to firmware rev. The code's checksum was good and after the upgrades everything ran great for 12 hours. Then the router (RT-AX86U) started crashing about every 15 minutes. As things had been stable to 12 hours I suspected the router and replaced it with a brand new RT-AX86U with Merlin. Crashes continued leaving me scratching my head as the router had been running great with the past few revs of Merlin. I talked with my wife, a very experienced computer tech and she asked "What changed?". I replied, "I added the XD6 nodes yet it ran great for 12 hours. You are asking the correct question so I'm powering them down." 30 minutes later no crashes and I complemented my wife. I waited 3 hours and then suspected the XD6 with the difficult firmware upgrade. I powered on the other XD6 and everything continued to run fine. 3 hours later I powered up the questionable XD6 node and 12 minutes later the router rebooted. I wondered if the node has a bad firmware load so reset to factory, went through a setup and loaded the firmware via an ethernet connection. It seemed good. Factory reset, joined to AiMesh and instant router crash. I'm returning the XD6 pair.

AiMesh seems rather sensitive and I'm surprised with it's age. I plan to add a single XD6 as a second node with one RT-AX86U as node and another as a node. This provides a quick switch for our main router should it develop an issue.

 

[Tech9]

AiMesh seems rather sensitive and I'm surprised with it's age.

What I found from experience for stable AiMesh setup:

- stock Asuswrt firmware
- the same model routers running the same firmware
- wired backhaul
- no USB attached storage used as NAS

If you change one of the above, it may or may not work.

 

[Ranger802004]

Part 1:

Adding a second as a node. I factory reset an old and proven RT-AX86U which had Merlin code and add it as a node. Within minutes I notice some of my IOT devices dropping and connecting again and again. I recall seeing others with similar complaints about IOT devices dropping. I turn off the node and they stay connected. I turn it back on and they start dropping. I connect my phone to the node and speed is terrible. I bring the node near the router and ethernet connect it. Connect my phone to it and communication is great. As signal was reported great in the nodes original position, I did not think it was a WiFi signal issue yet tried a much closer position and still bad speed with WiFi 5-Ghz backhaul. I load Asus firmware on the RT-AX86U node with the router RT-AX86U on Merlin. Place the node where it original was tried and it works great! There appears to be an issue with Merlin 386.5_2 on a RT-AX86U running as a node and RT-AX86U running the same version of Merlin. Asus 3.0.0.4.386.46061 works fine on the same node with Merlin 386.5_2 on the router.

Part 2:
I've been running on a single RT-AX86U since my node died mid fall. The node was used to cover my back yard as the router covers our home. With warmer weather we have been using our outdoors areas again and the outdoor WiFi has been sub par. I purchased a pair of XD6 connected them up and all was good though one of the two XD6 took a few tries to bring up to firmware rev. The code's checksum was good and after the upgrades everything ran great for 12 hours. Then the router (RT-AX86U) started crashing about every 15 minutes. As things had been stable to 12 hours I suspected the router and replaced it with a brand new RT-AX86U with Merlin. Crashes continued leaving me scratching my head as the router had been running great with the past few revs of Merlin. I talked with my wife, a very experienced computer tech and she asked "What changed?". I replied, "I added the XD6 nodes yet it ran great for 12 hours. You are asking the correct question so I'm powering them down." 30 minutes later no crashes and I complemented my wife. I waited 3 hours and then suspected the XD6 with the difficult firmware upgrade. I powered on the other XD6 and everything continued to run fine. 3 hours later I powered up the questionable XD6 node and 12 minutes later the router rebooted. I wondered if the node has a bad firmware load so reset to factory, went through a setup and loaded the firmware via an ethernet connection. It seemed good. Factory reset, joined to AiMesh and instant router crash. I'm returning the XD6 pair.

AiMesh seems rather sensitive and I'm surprised with it's age. I plan to add a single XD6 as a second node with one RT-AX86U as node and another as a node. This provides a quick switch for our main router should it develop an issue.

I read a similar article not too long ago where someone had a similar issue with their nodes running Merlin and it had something to do with the CPU clock speed so they wrote a script to step it down and it made the nodes stable.

 

[Morris]

What I found from experience for stable AiMesh setup:

- stock Asuswrt firmware
- the same model routers running the same firmware
- wired backhaul
- no USB attached storage used as NAS

If you change one of the above, it may or may not work.

The XD6 + RT-AX86U as nodes with RT-AX86U have been stable. I suspect what you describe is related to my findings related to mixed firmware versions. WiFi backhaul works fine with the expected reduced performance and wired is faster.

 

[Morris]

I read a similar article not too long ago where someone had a similar issue with their nodes running Merlin and it had something to do with the CPU clock speed so they wrote a script to step it down and it made the nodes stable.

That's interesting. Possibly the code is using a timing loop rather than clock and interrupts.

 

[Morris]

The XD6 + RT-AX86U as nodes with RT-AX86U have been stable. I suspect what you describe is related to my findings related to mixed firmware versions. WiFi backhaul works fine with the expected reduced performance and wired is faster.

Another thought related to same router and Asus firmware. Regression testing is tricky and there are a lot of models. Even the most basic testing would test two of the same routers. I'm not confident they do very extensive testing. The members of this forum may test more than they do.

 

[Morris]

I'm up for a couple of days and have learned quite a bit. The XD6 is a gem and works perfectly on the latest firmware with a WiFi backhaul
The RT-AX86u running Merlin is a lovely router, I wish I could say the same for it as a node running Merlin or Asus firmware. The 2.4Ghz band dose not work properly, I had to disable the radio and the radio works fine when it's used as a router. Using WiFi backhaul works for hours and then hosts connected to it start to have communication problems. I've connected it via ethernet and powerline and it seems stable though my powerline connection is slower than I'd like. I'll observe longer before running a cable.

I've run for a few years using 5Ghz/160 with auto channel and of cause DFS enabled. Also used auto channel on 2.4Ghz. For now while trouble shooting the communications I've gone to 5Ghz/80 fixed channel and a fixed channel for 2.4Ghz. This prevents reconverting when an aircraft flies over. I'll consider 5Ghz/160 with DFS fixed and auto once proven stable for a few days. I'm happy to be stable and will attempt to speed things up slowly.

 

[jata]

@Morris - thanks for sharing this info and experience with Aimesh and stock v. Merlin.

I have a similar set of adventures and have settled on AX86S main router (merlin) with 2x AX58U as nodes (stock) and everything is working well - even my Apple homepod mini's are behaving finally. The nodes are wifi backhaul but I find this is working better that with ethernet over powerline.

One thing in your post above caught my eye - the 5Ghz/160 setup. I have my network at 5Ghz/80 fixed + 2.4Ghz/20 fixed. I was thinking about switching to 5Ghz/160 as i'd like to increase the node connection bandwidth. Have you found 5Ghz/160 stable and faster in real world testing? - More about LAN (NAS access) than WAN as my internet is only 100/40.

 

[Morris]

@Morris - thanks for sharing this info and experience with Aimesh and stock v. Merlin.

I have a similar set of adventures and have settled on AX86S main router (merlin) with 2x AX58U as nodes (stock) and everything is working well - even my Apple homepod mini's are behaving finally. The nodes are wifi backhaul but I find this is working better that with ethernet over powerline.

One thing in your post above caught my eye - the 5Ghz/160 setup. I have my network at 5Ghz/80 fixed + 2.4Ghz/20 fixed. I was thinking about switching to 5Ghz/160 as i'd like to increase the node connection bandwidth. Have you found 5Ghz/160 stable and faster in real world testing? - More about LAN (NAS access) than WAN as my internet is only 100/40.

I live near an airport. Later in the day I went to CH 36 160Mhz width. It worked great once the network converged and this took way too long. A few hours later aircraft started flying over my home and on the first sound of a jet, the mesh fell apart. I then went to 36/80 and even that did not work well as there appears to be over compensation for the need to power down on CH 36 when aircraft weather radar is received. I went to 149/80 and the network immediately converged.

The XD6 had no problems on 36/160, it was the RT-AX86U node that appears to have DFS issues. As a router the RT-AX86U did not misbehave on DFS channels and properly reconfigured per US FCC requirements. User a wireless monitor, I observed that the RT-AX86U sometimes announces 160Mhz wide channels when the router is not set that way.

I'm going to try a temporary ethernet run to each node to see how that works as you report issues with that setup. I appreciate the heads up. I'll report my findings when I do this later

 

[jata]

All good. I have zero issues with proper/real ethernet from the nodes but it's not possible at my house (without a lot of work).

I find powerline ethernet to work ok but it is slower than wifi backhaul on the nodes in my experience. Maybe newer/better powerline adapters would be better but I doubt it.

I have very few aircraft flying over so I have switched to 5Ghz/160 fixed + 2.4Ghz/20 fixed and I will see how it goes and report back.

one question - do you see the nodes wifi on the list of devices in the wireless log in system log on main router? I do. I'm thinking that the 5Ghz bandwidth reported would be the same/similar as for backhaul performance. Is this a reasonable assumption?

 

[Morris]

Yes, the nodes show up in the Merlin Wireless Log. I hope 5Ghz/160 fixed works for you. It will take me 30 minutes to an hour to run the ethernet cable.

 

[heysoundude]

The members of this forum may test more than they do.

Of course- why pay for something when you can get it for free, and have your intended users pay for the privilege of doing it by buying the hardware?
I'm certain this plays no small part in Asus' relationship w/ RMerlin.

 

[Morris]

Of course- why pay for something when you can get it for free, and have your intended users pay for the privilege of doing it by buying the hardware?
I'm certain this plays no small part in Asus' relationship w/ RMerlin.

Any problem caught during QA will prevent hours of first, second and third level support and prevent the need to grow the support team. Third level support which interrupts developers can be a huge burden on an organization. The RMerlin Asus relationship is an interesting one and it's clear that they value his input as much as he there's.

 

[uwish]

After upgrading my Router to the 6000, I moved all my AX86 and tried to set up in AiMesh mode using Merlin on all of them. I never got it to work, at all. Even when the nodes were physically connected to the router. I had to revert back to latest stock firmware. I know some people have had success creating the AiMesh on stock then flashing to Merlin, but the network is unstable, with poor speeds etc.

I still have one device that for some reason is on or off the network at random times. Some of my other IOT did the same thing until I removed auto in the channel assignments. Once I fixed a specific channel, they are all behaving better. But my Nest still drops occasionally and at random times. Both my nodes are ethernet backhauled and I get fantastic speeds anywhere in the house with my 1Gig shaw connection. The average is 650 down on any AX device.

 

[jata]

I agree that ethernet backhaul will always be preferred so @uwish i'm jealous of your setup.

I have tested using iperf3 the speed between my nodes and a RPI4 connected directly to my main router using a new CAT6 cable. Here are my results and I don't think I can improve them using wifi backhaul - but happy if others can help.

Both nodes are reporting connection 'great' in the GUI and wifi log reporting 5GHz/80 node connection 648 / 1080 Mbps with -66 dB

Nodes are running latest stock (3.0.0.4.386_48908)

• baseline: macbook pro 2014 ethernet to main router (AX86S) to RPI4 = 90MB/s (750Mb/s) - so max/saturating the 1GB network
• wifi baseline: macbook pro 2014 wifi 5Ghz to main router (AX86S) to RPI4 = 90MB/s (750Mb/s) - so max/saturating the 1GB network
• wifi node1: macbook pro 2014 wifi 5Ghz to node (AX58U) to RPI4 = 30MB/s (250Mb/s)
• wifi node2: macbook pro 2014 wifi 5Ghz to node (AX3000) to RPI4 = 30MB/s (250Mb/s)

So the node to router wifi backhaul on both nodes is poor I think. Would hope for around 60+MB/s.

Have tried using 5GHz/160 but I found that performed worse and I have switched back to 5Ghz/80

Anyone else have similar (or preferably better) results for node performance/speed?

 

[Morris]

I agree that ethernet backhaul will always be preferred so @uwish i'm jealous of your setup.

I have tested using iperf3 the speed between my nodes and a RPI4 connected directly to my main router using a new CAT6 cable. Here are my results and I don't think I can improve them using wifi backhaul - but happy if others can help.

Both nodes are reporting connection 'great' in the GUI and wifi log reporting 5GHz/80 node connection 648 / 1080 Mbps with -66 dB

Nodes are running latest stock (3.0.0.4.386_48908)

• baseline: macbook pro 2014 ethernet to main router (AX86S) to RPI4 = 90MB/s (750Mb/s) - so max/saturating the 1GB network
• wifi baseline: macbook pro 2014 wifi 5Ghz to main router (AX86S) to RPI4 = 90MB/s (750Mb/s) - so max/saturating the 1GB network
• wifi node1: macbook pro 2014 wifi 5Ghz to node (AX58U) to RPI4 = 30MB/s (250Mb/s)
• wifi node2: macbook pro 2014 wifi 5Ghz to node (AX3000) to RPI4 = 30MB/s (250Mb/s)

So the node to router wifi backhaul on both nodes is poor I think. Would hope for around 60+MB/s.

Have tried using 5GHz/160 but I found that performed worse and I have switched back to 5Ghz/80

Anyone else have similar (or preferably better) results for node performance/speed?

I'm not surprised by your findings. When I was running with a single node, my testing showed a very small improvement going from 80Mhz to 160 Mhz channels for my laptops that have the typical two antenna setup. Just over 1-Gb throughput at 160 and just under at 80. Nothing to get excited about yet if you are transferring large files or a lot of them 160 might save a little time. It might matter more if your network is busy and you have multi stream capable devices. Tri-Band with dedicated backhaul 160 channel will be the best yet a Tri-Band router that runs Merlin or for that matter any is rather pricy. I have to widen the holes that my cables go floor to floor and then I'll be running ethernet backhaul.

 

[Morris]

I started wiring today and now have a ethernet connection from the XD6 node to the RT-AX86U node. As the XD6 is very well behaved as a node it is providing the wireless backhaul from the RT-AX86U router. This is a reasonably fast setup and worked immediately. I'll started drilling the holes for the link from router to RT-AX86U node yet did not finish that run. It's coming in the near future.

 

[jata]

Thanks @Morris for keeping this thread alive. I don't really 'need' my network to be faster but it is a bummer seeing the difference of node vs main router on wifi backhaul with everything setup optimally(ish).

My simple question is why are the nodes maxing at 30MB/s with a 5Ghz 'great' connection?

Lets say the backhaul link is 800Mb/s and actual throughput is 60% of that - this should still deliver 60MB/s

 

[Morris]

Thanks @Morris for keeping this thread alive. I don't really 'need' my network to be faster but it is a bummer seeing the difference of node vs main router on wifi backhaul with everything setup optimally(ish).

My simple question is why are the nodes maxing at 30MB/s with a 5Ghz 'great' connection?

Lets say the backhaul link is 800Mb/s and actual throughput is 60% of that - this should still deliver 60MB/s

Not on dual band as the node must receive the frame and then transmit from buffer. This cuts the throughput to a maximum of half. This is assuming that the frame is not stripped of packets and then packets reframed to transmit which adds a tiny bit of additional overhead. To avoid the buffering and retransmit delay, use of a wired backhaul or a different radio is required. If you want a pure wireless solution Tri-Band routers are the best solution and of them the 6E routers may avoid congestion in your area.

 

[jata]

I see. Thanks.

I already have 4 asus routers in the house so not planning to buy any more

Looks like I have to be happy with what I have or run some cables to link the nodes