This has puzzled me for ages and I’ve not thought much of it. But now that I’ve got a few AP’s in my home (3x HAP AC2) - I’m really trying to improve the ability to roam between devices, balance speed and coverage.
I use Capsman and have fine tuned the frequency usage as much as possible, ensuring that on 5GHz I’m using channels 36, 52 and 100 with Ceee. My laptop can happily connect at ~585Mb/s - and running iperf over WiFi to a device physically connected to the main HAP AC2, I get around 200Mb/s. That’s fairly acceptable for me.
But with 2.4GHz, its a slightly different story. Originally I used channels 1, 6 and 11 with the extension channel disabled to ensure there was no overlap over each AP; but in doing this, the fastest speed I could achieve was ~60Mb/s. So I changed to channel 3 and 11 with eC and forced the band to G/N (N only doesn’t seem to be an option for Mikrotik). With this, again, the fastest speed I can achieve is 60Mb/s. That is sat next to the AP, testing on a Samsung S20 5G and a laptop with an Intel Wi-Fi 6 AX201. The connection speed on my laptop shows as 300Mb/s - but the actual throughput always seems to max out around 60Mb/s.
I can’t help but wonder is there a setting I have wrong somewhere, or something that’s not right about my 2.4Ghz implementation.
300 Mbps on 2.4 GHz , yes you can. But it is not what you are thinking.
First: the speed given is the interface rate, not the data throughput. Data throughput is typical 60% of the interface rate due to the high overhead of 802.11 protocol.
Wifi is half-duplex.
But there are more conditions:
You need 40 MHz clean bandwidth. As you know there is only channel 1,6 and 11 non overlapping. Channel 3 will disturb 1 and 6 ,on 20 MHz , it will disturb all at 40 MHz bandwidth
Don’t assume the absense of other devices. Measure ! Use “Freq Usage” and “Snooper” to listen what can disturb your signal.
300 Mbps requires also 2x2 spatial streams. The client must have 2 chains also (not always the case). Multiple spatial streams is “n” only.
-The signal may not be distorted, reflected, absorbed, diffracted or the high encodong (MCS 7) will drop lower, and the 2x2 will get lost
High SNR needed, high RSSI needed. So only close to the AP.
You need short guard interval as well.
The 60% throughput range is only achievable with aggregation (A-MSDU and A-PMDU). That is n-only , not g
CAPsMAN will slow you down
Any device sending the “40MHz intolerant bit” will disable 40 MHz
15 SSID or AP’s being silent on the same channel will eat 50% of the airtime if supporting B mode, just with their beacons alone
In practical setups 40 MHz bandwidth is almost impossible today, unless you live somewhere alone in the woods.
Setting “2GHz- only- n” is no problem on Mikrotik
Using channel 3 and 11 with 40MHz introduces the devastating adjacent channel interference.
I usually get 130 Mbps interface rate with good clients on 20 MHz bandwidth
Firstly, thanks for taking the time to reply with such a detailed answer. I’m rapidly trying to learn about wireless in this capacity (my background is a engineer for large scale L3/L2 networks bgp/ospf etc.).
Its theory at this point that is interesting me more (than the practical implications of actually being able to use 40Mhz). I’m merely trying to see if its even possible to achieve more than 60Mb/s actual throughput on 2.4GHz.
From the description you’ve given, if my interface rate is 300Mb/s (ie. the rate as presented in Windows) - and I’m sat next to the AP with 2x2 MIMO client; I should in theory be able to push ~150Mb/s at best, perhaps 120Mb/s in real conditions? Is my understanding correct?
What it sounds like (again for theory and just seeing if I can make it happen once, but not leave it like this forever) is that I need to configure the 2.4GHz as N only? Yet I can’t seem to find a way to do that via CAPSMAN. Do you have any pointers for this.
It sounds increasingly like for my 200Mb/s internet connection, 5GHz is the only way to deliver this - and across a home where 5GHz won’t penetrate the walls - I’ll need dozens of APs?
Nb. You mentioned CAPSMAN will slow me down; as its only a home network, I’m using local-forwarding; so I’m assuming that statement doesn’t apply here?
Well, 60% of 300 is 180. But yes, something like this. BUT: only with a perfectly clean spectrum, angels singing and a bit of luck.
Yes, 5GHz is the way to go. One of its best features is exactly the low penetration - it keeps your neighbors interference away. Unless You live in a medieval castle, or a bunker, dozens of APs won’t be necessary. More than one, certainly. It depends upon area size and type of walls.
My “castle” is composed of foil backed insulation, plasterboard and brick walls - and foil backed insulation and doubled up plasterboard ceilings. I’m lucky if 5GHz spills through a doorway
So there’s definitely no way to ever see more than 54Mb/s on 2.4GHz with Mikrotik - or any other vendor? Even if I said my nearest neighbour was 500m away and the spectrum is clean. Feels a bit misleading that there’s so many N devices marketed at 300Mb/s when they’ll deliver a sixth of that in near-perfect conditions.
Ah, foil backed insulation is a killer - not to mention brick walls. But usually I can go through one brick wall without much problem. No foil isolation here, though.
One problem You may be facing is reflection. A brick wall will just absorb the signal - the foil will reflect it. We can do a simple test, to see if my idea about reflection is right: can you test the AP outside the house? This way we get rid of this variable.
Its raining right now, but I can do when it stops.
From a configuration perspective; is there a way to force N only via CAPSMAN. I still feel like my configuration is holding me back and that I’ve got a toggle wrong somewhere.
Just to add, I fear I’m perhaps using the wrong language. It might not be 40MHz that I’m after, but merely 20MHz + 20MHz. Ultimately, just trying to see if Mikrotik can actually do better than 54Mb/s over 2.4GHz - as that seems impossible right now and I don’t think its an environmental/spectrum/interference issue.
I’ve got a non-brand 2.4GHz N router that I can test to see.
You have all the numbers. Theoretical interface rate 300Mbps (which you get) can give 180 Mbps throughput. If all goes smoothlly and no interference.
Check your registration or status again (detail mode) for that wifi association to your client.
What is the CCQ . (You need 100% to get that 180 Mbps) It will be (much) lower than 100% when only 60 Mbps gets through. (Or the reason is not the wifi).
CCQ comes from the retransmits. You can see these by comparing the TX/ RX packets (or frames), versus the TX/ RX HW packets. (or frames)
This just to find out if wifi is the limiting factor or some other setting.
So, I tweaked some settings (setting the channel in CAPSMAN to “onlyn”) and managed to get ~120Mbps (see attached). That’s definitely much better than before; still not what I feel should be possible, but better.
Testing was performed using iperf between a WiFi client on AP2 and a wired LAN client on AP1.
44 devices on this channel 3 (40 MHz wide, so 1,2,3,4,5,6,7,8,9 as used channel)
There is no air-time fairness with Mikrotik. Without air time fairness you communicate at the speed of the slowest if there is contention for the medium (air).
This is statistics. Let say we have 2 devices competing for access to the medium. They have each 50% change to get it. (802.11 mechanisms)
If one is sending at 300Mbps and the other at 100Mbps, then they will get the same number of time-slots. Only the first one will use its timeslot 1/3 of the time, versus the second one.
This means for both the usable interface rate is 75 Mbps (300/4). With air-time fairness they each get 1/2 of the air-time, or the usable rate for the first one is 150 Mbps.
