Wi-Fi 7 Multi Link Operation (MLO) discussion

[thiggins]

This discussion was cluttering up the ASUS GT-BE89 thread, so I moved them here.

 

[RTWZ80]

I think the key differentiator being touted for Wifi 7 vs Wifi 6/E is Multi-Link Operation (MLO) where your Wifi 7 router and Wifi 7 client can use up to two bands concurrently or quickly switch between two bands to maximise throughput and compensate for the 6Ghz band's relatively short range.

Multi-link Operation (MLO) gives WiFi 7 routers and devices an undeniable advantage over last-gen tech - Edge Up

WiFi 7 marks the debut of Multi-Link Operation (MLO), a technology that lets clients and access points use more than one band simultaneously.

edgeup.asus.com

 

[tgl]

I think the key differentiator being touted for Wifi 7 vs Wifi 6/E is Multi-Link Operation (MLO) where your Wifi 7 router and Wifi 7 client can use up to two bands concurrently or quickly switch between two bands to maximise throughput and compensate for the 6Ghz band's relatively short range.

Seems like the same old story to me: the manufacturers tout behavior that you will only reach if you live in the sticks with no nearby wifi-using neighbors. Where I live, reality is that the 2.4GHz band is completely saturated and 5GHz is getting there (at least if you can't use DFS channels, which again is typical in urban locations). I believe that the 6GHz band will make life better. I don't believe that multiband operation is going to add much on top of that.

 

[RTWZ80]

Seems like the same old story to me: the manufacturers tout behavior that you will only reach if you live in the sticks with no nearby wifi-using neighbors. Where I live, reality is that the 2.4GHz band is completely saturated and 5GHz is getting there (at least if you can't use DFS channels, which again is typical in urban locations). I believe that the 6GHz band will make life better. I don't believe that multiband operation is going to add much on top of that.

I didn't say that the 6Ghz band wouldn't be a game changer. MLO supposedly will allow the user to switch between the 5Ghz band which has greater range and the 6Ghz band which has higher throughput. So both can be used concurrently to overcome the shorter range encountered on the 6Ghz band as well as the increasingly crowded 5Ghz band depending on network conditions. Of course, this is all in theory until we see more Wifi 7 clients hit the market and allow users to test if that's a reality.

Currently using the RT-BE96U and am getting up to 1600-1900 Mbps u/l and d/l wireless speeds on the 5Ghz band even behind closed doors in rooms. The 6Ghz band while also speedy, roughly 1300 Mbps u/l and d/l, sees a sharp drop-off in speed to below 200 Mbps once I close the door to each room near the main router or AiMesh node. These were tested on Wifi 6 (DFS) and Wifi 6E clients though. I'm still waiting for decent Wifi 7 clients to hit the market before testing them for myself.

 

[Jansen3]

I think the key differentiator being touted for Wifi 7 vs Wifi 6/E is Multi-Link Operation (MLO) where your Wifi 7 router and Wifi 7 client can use up to two bands concurrently or quickly switch between two bands to maximise throughput and compensate for the 6Ghz band's relatively short range.

Multi-link Operation (MLO) gives WiFi 7 routers and devices an undeniable advantage over last-gen tech - Edge Up

WiFi 7 marks the debut of Multi-Link Operation (MLO), a technology that lets clients and access points use more than one band simultaneously.

edgeup.asus.com

This is why I’m excited about Wifi 7. MLO is a big game changer. The stuff coming out of the 802.11BE TG is pretty interesting if you want to get deeper into it.

There’s actually no limit on implementation, just client support. You could technically have quad-band support, but most will be tri band for obvious reasons.

 

[Qbcd]

No, your posts haven't made anything clearer. What you're stating is not how things work. But, let's drop it, sure.

Ok, so for the record you are saying that MLO or CA in cellular networks is not adding or aggregating several channels together? You're so adamant about this being wrong that I am curious what you think the actual correct description of these technologies is, in general terms. I mean it's literally called "multi-link operation", I feel like it's pretty self-explanatory, but I could be missing something.

 

[Jansen3]

Quad-band will be really useful for wireless backhaul I think, or segmenting your home network e.g. separate bands for family use, IoT devices etc.

I was referring specifically to quad band MLO, which will require 4 radios and a whole lot of processing power. Qualcomm really sees it as the best way to lower latency, more so than to boost bandwidth.

But most MLO implementions will be tri band since you already have those radios. It’ll mostly be firmware that unlocks the potential, don’t think there’s too much on the hardware side to optimize.

Of course the client side will be the most important component. I believe Apple will be going heavily into Wifi 7 for MLO and constant connectivity. This is what everyone was hoping for with Smart Connect, but a much cleaner implementation.

 

[glens]

I haven't investigated but it just occurred to me how slick it could be if MLO is configurable for an association with one radio transmit and the other receive, for pseudo-effective full duplex operation. Is this in the deck of cards currently being shuffled?

 

[Jansen3]

I haven't investigated but it just occurred to me how slick it could be if MLO is configurable for an association with one radio transmit and the other receive, for pseudo-effective full duplex operation. Is this in the deck of cards currently being shuffled?

From what I recall, MLO has multiple operating modes and one of these is Simultaneous Transmit and Receive (STR). It allows for using 1 radio (ie 5 GHz) for upload and 1 radio (ie 6 GHz) for download as a valid combo, which is effectively a full duplex connection as you postulated.

 

[Jansen3]

This is from TP-Link’s website, but I believe it’s written into the 802.11BE draft docs.

 

[RTWZ80]

I was referring specifically to quad band MLO, which will require 4 radios and a whole lot of processing power. Qualcomm really sees it as the best way to lower latency, more so than to boost bandwidth.

But most MLO implementions will be tri band since you already have those radios. It’ll mostly be firmware that unlocks the potential, don’t think there’s too much on the hardware side to optimize.

Of course the client side will be the most important component. I believe Apple will be going heavily into Wifi 7 for MLO and constant connectivity. This is what everyone was hoping for with Smart Connect, but a much cleaner implementation.

It'll be interesting to see how MLO will improve the throughput performance on quad band routers given that the current configuration (2.4G, 5G1, 5G2, 6G) ends up splitting the 5G band in half and makes it harder to be able to consistently access the DFS channels. With a number of the Wifi 7 quad band routers including the GT-98/Pro switching things up (2.4G, 5G, 6G1, 6G2), they will likely derive higher aggregated throughput from Enhanced Multi-Link Single Radio Modes as per MediaTek's plug here:

 

[glens]

From what I recall, MLO has multiple operating modes and one of these is Simultaneous Transmit and Receive (STR). It allows for using 1 radio (ie 5 GHz) for upload and 1 radio (ie 6 GHz) for download as a valid combo, which is effectively a full duplex connection as you postulated.

Thank you. Of course it will require discrete antennae per radio for fullest effect. Can only hope...

Ought to bite a notable chunk out of latency as well.

 

[RTWZ80]

6 GHz is terrible unless you're in the same room. Lower legal power limit than the higher DFS 5 GHz channels and higher frequency, for AiMesh wireless backhaul it's useless, and for clients too 5 GHz is better. If you're not near an airport, the DFS channels above 100 are excellent with 30 dbm legal power limit, which if both devices can utilise especially, makes a huge difference. Furthermore, you can't use 3rd party antennas on any device with 6 GHz, so again any sort of point-to-point type use case is hindered.

But yes if you have hardwired APs in every room, go nuts with 320 MHz channels hopefully without interference.

That's true, I'm getting higher speeds and range on the 5Ghz (DFS) bands vs the 6Ghz bands. That said, MLO is supposed to combine the use of both 5Ghz and 6Ghz bands to give you a 'best of both worlds' throughput and usage scenario. However, as the feature has yet to be released on routers or Wifi 7 clients (PCs) there's no way to know how well it'll work until we can test it ourselves.

 

[Qbcd]

That's true, I'm getting higher speeds and range on the 5Ghz (DFS) bands vs the 6Ghz bands. That said, MLO is supposed to combine the use of both 5Ghz and 6Ghz bands to give you a 'best of both worlds' throughput and usage scenario. However, as the feature has yet to be released on routers or Wifi 7 clients (PCs) there's no way to know how well it'll work until we can test it ourselves.

But the way I see it, it'll contaminate my interference-free high-power DFS channel with high-latency congested 2.4 GHz and weak-signal 6 GHz... Plus it's more overhead and potential for glitches. I don't need super high speeds, I'd rather have the stability and performance of my selected channel. So I don't know, I mean carrier aggregation makes a lot of sense for cellular networks, but for Wi-Fi it seems gimmicky. I'm sure there will be a way to disable it, I just wonder how that will play with AiMesh and Smart Connect.

Seems that most of the new features of Wi-Fi 7 are only useful if you're in the same room as the AP. 320 MHz channels - useless with lower signal strength, even 160 MHz is hardly useful, you're mostly just raising the noise floor and introducing more potential for interference. 4K QAM - nice, but again unless you're in the same room you won't get that. MLO - gimmicky, see above. 16 spatial streams - won't actually be used. Preamble puncturing - ok, this one sounds pretty good, but let's see if it actually makes a difference.

And so if you wanna stay on 80 MHz channels and not use MLO and can't get the SNR for 4K QAM, then you're basically getting the same performance as Wi-Fi 6.

 

[Jansen3]

But the way I see it, it'll contaminate my interference-free high-power DFS channel with high-latency congested 2.4 GHz and weak-signal 6 GHz... Plus it's more overhead and potential for glitches. I don't need super high speeds, I'd rather have the stability and performance of my selected channel. So I don't know, I mean carrier aggregation makes a lot of sense for cellular networks, but for Wi-Fi it seems gimmicky. I'm sure there will be a way to disable it, I just wonder how that will play with AiMesh and Smart Connect.

This doesn’t make any sense at all. Signals on different frequencies will not “contaminate” your wifi.

Smart Connect sucks. MLO is a far better implementation of it. You get additional frequencies that will lower latency, with additional protection from random signal loss that occurs.

It looks like Asus is using the most powerful processor available in a router for their Wifi 7 generation. TP-Link has been intentionally coy about their CPU specs. This and the additional RAM will certainly go a long way towards stability and performance.

There are many reasons why so many companies are backing Wifi 7 and it will see rapid adoption. Even Apple, which has traditionally been slow with Wifi adoption, will be adopting it quickly. They even changed their plans to use their own wifi chips and extended their contract with Broadcom because they won’t be able to implement Wifi 7 on their own.

 

[Qbcd]

With 5GHz operating within a hair's width away from 6GHz, I don't see how 6GHz bands will be impacted much (for range).

2.4GHz cannot 'contaminate' the 6GHz band. And the processors are significantly more powerful than anything we have today, so I don't expect a slowdown there either.

There are more improvements to WiFi 7 than a mere doubling of the bandwidth. On a properly built (hardware-wise) and fully fleshed out (firmware-wise) model, the differences should be significant.

6 GHz has a much lower legal power limit compared to channels 100-128. And it can be about 1 GHz higher than channels 100-128, which does make a difference as well. Before you say that high power is useless because your devices can't talk back to the AP 1) It's still very much useful for download, 2) If you're connecting 2 devices that both support the higher limit (e.g. 2 routers in AiMesh), then it does make a substantial difference. I've tested 24 dbm vs 30 dbm links and the latter is much better, how could it not be.

On the "contamination" point - of course it does, if you're aggregating 2.4 GHz and 6 GHz onto my Channel 100, then you're introducing all the latency and signal strength issues of those links into the aggregated connection. Maybe MLO is smart enough to use the high-performance band as an anchor and manage the rest intelligently, but I don't trust it, much rather not introduce the extra variables. I prefer stability over speed. And moreover, MLO probably requires the same power limit on all bands, so now my Channel 100 has to go down to 24 dbm, which kills its main advantage. It'll need to be tested ultimately, but I don't see how it can be as good as using a single channel.

Smart Connect sucks

I need to use Smart Connect in my deployment, and of course I can't count on clients supporting MLO for at least another 10 years. But they may not be mutually exclusive.

 

[Jansen3]

On the "contamination" point - of course it does, if you're aggregating 2.4 GHz and 6 GHz onto my Channel 100, then you're introducing all the latency and signal strength issues of those links into the aggregated connection. Maybe MLO is smart enough to use the high-performance band as an anchor and manage the rest intelligently, but I don't trust it, much rather not introduce the extra variables. I prefer stability over speed. And moreover, MLO probably requires the same power limit on all bands, so now my Channel 100 has to go down to 24 dbm, which kills its main advantage. It'll need to be tested ultimately, but I don't see how it can be as good as using a single channel.

I need to use Smart Connect in my deployment, and of course I can't count on clients supporting MLO for at least another 10 years. But they may not be mutually exclusive.

The terminology you’re using is a bit weird, but I think I get your concerns.

It depends on firmware and which MLO mode is used, but the way to think of it is all three bands are connected simultaneously, which gives you much greater reliability than Smart Connect, which switches between 2.4 and 5 GHz.

Also note that MLO is being built specifically around latency concerns. Rollover and interference mitigation were cited in the 802.11 TG for MLO.

 

[Qbcd]

The terminology you’re using is a bit weird, but I think I get your concerns.

It depends on firmware and which MLO mode is used, but the way to think of it is all three bands are connected simultaneously, which gives you much greater reliability than Smart Connect, which switches between 2.4 and 5 GHz.

Also note that MLO is being built specifically around latency concerns. Rollover and interference mitigation were cited in the 802.11 TG for MLO.

Yeah they're not mutually exclusive in theory, you can have a client connect to an SSID corresponding to a single channel as normal, and then if the client supports MLO and it's enabled, you can aggregate additional channels from the same band or other bands, with the main one you're connected to acting as "anchor", similar to how 5G NSA works. Well I hope you're right about the potential performance benefits, will have to see what the reviews say, but I'm still skeptical and of course I don't want to compromise on power limit.

Carrier aggregation in LTE-A/5G at least doesn't come with any performance benefits, other than bandwidth of course, as far as I know. In fact it's better to be connected to just 1 channel than several to reduce overhead, this is why carriers (at least in Europe) try to acquire contiguous spectrum for 5G, preferably 100 MHz blocks in band n78, because it's better to connect customers to one big channel than aggregate several.

Another thing is, I wonder if you will be able to customize which bands/channels you'd like to aggregate and at what signal strength, so you can exclude 2.4 GHz for instance. Knowing how glitchy Asus' channel management is in general, I don't have too much confidence in their implementation of MLO.

 

[L&LD]

6 GHz has a much lower legal power limit compared to channels 100-128. And it can be about 1 GHz higher than channels 100-128, which does make a difference as well. Before you say that high power is useless because your devices can't talk back to the AP 1) It's still very much useful for download, 2) If you're connecting 2 devices that both support the higher limit (e.g. 2 routers in AiMesh), then it does make a substantial difference. I've tested 24 dbm vs 30 dbm links and the latter is much better, how could it not be.

On the "contamination" point - of course it does, if you're aggregating 2.4 GHz and 6 GHz onto my Channel 100, then you're introducing all the latency and signal strength issues of those links into the aggregated connection. Maybe MLO is smart enough to use the high-performance band as an anchor and manage the rest intelligently, but I don't trust it, much rather not introduce the extra variables. I prefer stability over speed. And moreover, MLO probably requires the same power limit on all bands, so now my Channel 100 has to go down to 24 dbm, which kills its main advantage. It'll need to be tested ultimately, but I don't see how it can be as good as using a single channel.

I need to use Smart Connect in my deployment, and of course I can't count on clients supporting MLO for at least another 10 years. But they may not be mutually exclusive.

It is impossible to aggregate 2.4GHz and 6GHz on Channel 100. You're misunderstanding how this works. There is no contamination as you're defining it. But, depending on hardware and firmware implementations, I agree that one model may be better than another at doing MLO. But that is par for the course. Not all Class 7 routers will be worth buying or considering.

As I don't use channels 100 - 128 (I use the higher channels), the power is the same in my case. I've never said higher power was useless. The router's sensitivity/noise rejection must also match though to be a true range increase.

~5.9GHz to ~6.9GHz is ~17%. I'm sure antennae design can 'easily' account for that on a properly designed and spec'd router.

MLO will be supported by new WiFi 7 clients as soon as they are released after WiFi 7 is certified. It won't be in 2033 when that feature is available to be used.

 

[Qbcd]

It is impossible to aggregate 2.4GHz and 6GHz on Channel 100.

That's not what I said, I meant adding channels from those bands to Channel 100.

As I don't use channels 100 - 128 (I use the higher channels), the power is the same in my case. I've never said higher power was useless. The router's sensitivity/noise rejection must also match though to be a true range increase.

I know you didn't say it, but I get preemptively defensive about it because there's always at least one guy who says higher power is useless because the client can't shout as loud as the AP, and I'm tired of hearing that, sorry if that was rude.

I forget what the limits are in the US, but in the UK and Europe Channels 100-140 have a 30 dbm or 1W limit, so that's 6 dbm higher than channels 36-64 for instance, it's really noticeable. And being a higher DFS channel, it's almost always clear, so no interference. Once you get used to that performance which is almost on the level of a wired connection, adding highly congested lower-power channels to it that are going to have noticeably higher latency, jitter, bufferbloat and late or lost packets, is like... no way. I don't really understand how MLO is going to mitigate that, but we'll see.

That being said, this is just my use case. I need APs to reach across 2 walls sometimes. But if you have APs in every room and are not near an airport, and you have access to say 160 MHz of completely clear 5 GHz and 160 or 320 MHz of completely clear 6 GHz, aggregating them would totally make sense and you can push insane speeds.

And hey if MLO can be disabled, then it doesn't hurt to have the option. I'm not hating on it, just saying it's not strictly better to have it enabled.

Sorry for the long posts, I'm in a rambly mood.

MLO will be supported by new WiFi 7 clients as soon as they are released after WiFi 7 is certified. It won't be in 2033 when that feature is available to be used.

Again, that's not what I meant, but I don't feel like typing anymore.