Ubiquiti tuning

[HCHTech]

I'm trying to tune a new setup that is proving challenging. It's a stone house with plaster walls = wifi hell. The basement is finished, so I've got 3 APs (nano HDs), one on each floor (each with home runs to the switch, not a mesh setup). All one SSID.

To start, I've got the 2.4 radios at Medium and the 5g radios at High. Channels chosen to avoid neighbors - it's suburban, so not a crowded spectrum. My surveys show a reasonable signal in most areas that are important. I'm getting some bleed-through of the Basement AP into the first floor and vice versa because the floor is hardwood. No such problem with 1st Floor into second or vice versa because the ceiling is plaster.

The problem is that devices aren't automatically switching to the "correct" AP when the device changes floors resulting in poor performance when things move around. This seems like a textbook application for "Enable Minimum RSSI" setting. I've read through UI's KB article on RSSI, but I'm having trouble finding the best value to get the results I'm after.

First of all, the article says the allowable values are between -1 and -94, but in the controller, the only values accepted are between -67 and -90. I suppose the setting was changed and the article is just old.

The setting is basically saying that "Any device with a connection signal "worse than" (with a signal less than) the input value will be sent a de-authentication packet (be forced to disconnect and find another AP.)

The trick is find the value that is enough to encourage roaming when the device leaves the floor, but not force roaming for the lowest signal received on the original floor (i.e. farthest away from the AP on the same floor as the device). I can take a single device and figure this out with a survey. What seems impossible is to solve the equation for multiple devices at once since signal quality is very device-dependent.

I started with -67 on the first AP I was working with, but the device I was testing with kept disconnecting and reconnecting (even though it's signal strength was in the high 50's at the time.) So, does -67 really mean -57? I've bumped the setting up to (down to?) -75, which stopped the disconnect/connect loop, but only time will tell whether it has helped the main problem. I'll be back onsite next week to deliver a laptop, so can tinker a bit more.

Lastly, this multi-variable problem has to be solved independently for every AP. I'm starting to rethink the whole idea of using minimum RSSI.

Little help?

 

[HCHTech]

Paging @YeOldeStonecat - any advice on this one?

Thinking it over a bit, maybe I'm coming at it backwards. My initial thought was to set a value equal to the lowest-acceptable signal I wanted a device to have, and since I was getting between -45 and -63 on my test device depending on my location, I started with the highest number available, -67. Maybe the right answer is to come the other way, if the lowest-acceptable signal I want is, say -65, then I should start with a setting of -75.

Further, since 5G doesn't propagate very well in the first place, I think I should NOT enable minimum RSSI on the 5G radios, as those signals largely disappear when you're not on the floor of the AP. In the end, I don't care which AP a device uses, but I sold this whole upgrade on the idea that we had infinitely better control and would get much better coverage over their previous system of an old consumer router and a couple of extenders. I'd like to not have to argue this point after the fact.

 

[YeOldeStonecat]

Heya! @HCHTech
So...you're pretty correct in what you're saying.
The value that you want to put in, ....every site is different. Heck, in some sites, you'll have much different values for certain APs.

So what RSSI does, is basically the AP you set the value on, will "soft kick" a client...if that client detects an AP with a signal stronger than...<whatever you put in>. The AP basically says "Hey, if you can find someone stronger than me, get the F off of me! Get out!!!"

So say you're in school....in the english classroom. Nice strong signal of -45. Class is over. You walk 3x classes down...into history. You open your laptop back up, and it's still attached to the English room AP...at a horrible -78. Even though the room just past History...has an AP, and it's able to come in at -55. Usually clients will cling onto the first AP they associate with...using a death grip! Even if standing next to another AP on the same network/same SSID....with really good signal!

So you want to put in some settings, which encourage the clients to latch onto another AP that they determine is stronger.

YES..generally the value you input, is the level you want to encourage the AP to make the leap, if they detect stronger. Start at -75...but I usually get more to -70. I've gone into the upper -60's.

I will say though, that, a few weeks ago I did a Unifi install at an accounting firm that I took over, and I went right for -67 out of habit. And I had disconnect/reconnect issues. Much like you describe. I did use in-walls at this accounting firm, I think 5 of 'em. I bounce it back to -75. I figured perhaps an issue with recent firmware...as in the past, in some setups, I've had good luck with -67 in some stronger zones.

Here's the thing to remember...it's the prior AP that is sending the kick when the signal to the new AP is stronger than the signal back to the prior AP. It's not a setting from the upcoming AP. What is....when leaving English class....a student might head to 4x other possible classrooms...you have to pick an RSSI setting that works best for the client roaming to any of the 4x possible other APs.....now...picture doing this for a small school with 30x APs.

Signal strength is another thing. Don't be afraid to go into custom..and put try some granular settings.

So I see what you're thinking...that number for the RSSI is what it should make the jump at, the next room has a good signal of -48..so I should be able to put -50 and it will make the jump. But just start with something like -75..or maybe -73...IMO as long as something is there, it'll be an encouragement to make the leap. And make sure the prior is dialed down enough to not be strong in the cell of another AP. I always try to dial down the TX power of neighboring APs...so that within the cell of another AP they're not stonger than -65 or so.

 

[HCHTech]

Ok, so here's the thing. As far as I know, anyway, the AP giving the kick for a device to leave doesn't know that there is another AP that has a stronger signal for that particular device. Right? So because of this, WE have to make sure that when a device is in a low signal area for one AP, that it is also in a high(er) signal area for another AP. It seems to me that the controller would in fact have this information, and I can imagine a system smart enough to "instruct" devices which AP they should be connected to. However, that day is not today and the UI controllers and APs don't do that.

Ok, so that's Fact #1 - You can tell an AP when to kick a device out, but you can't tell that device where to go. As the architect of the install, you just have to make sure there is an AP with a stronger signal out there for that device to latch on to.

Fact #2 - and this is the one that makes me wonder about this setting altogether.....you can have two devices right next to each other, and because of differences in their wifi cards & antennas, have two different signal strengths. So my Laptop might have a -68, but an iPhone right next to it might have a -73. So in this case, a Minimum RSSI setting of -70 would leave the laptop alone, but kick off the iPhone. So, where does that phone go? If the Iphone has a signal strength of -68 on AP#2, then it will go there, but if it has a signal strength of -71 on AP#2, then it has no where to go (assuming you also have an minimum RSSI setting of -70 on AP#2.)

As a result, it seems to me the only workable answer ends up being the lowest-common denominator of every device in the area. You have to find the device that has the lowest signal of all of the possible devices in the environment, and tune to that, otherwise, some devices will have a good experience and some won't.

Finally, assuming you don't want to take measurements on every single device manually, you need historical signal data by device from the controller to do this job. If I go to the Clients screen, pick a client and look at properties, then select Historical Data, I get a note that "In order to access per-client statistics, client historical data collection must be enabled in Settings..Maintenance". If I go there, there is no "Enable Client Historical Data" setting, and the only setting that looks remotely like it would be related is this:



The choices for each setting are "Normal", "More", and "Debug". I've changed the Device and System logs to "More", and I will check later to see if I can get any historical data.

This stuff is making my head hurt. I guess all of this minutiae is probably handled normally by just raising your setting a point or two at a time until the client stops complaining. My problem is I want to dig until I get the "real" answer - but it's looking that that isn't really practical.

 

[YeOldeStonecat]

You're pretty much on the money...fact #1...correct. Fact #2....eh, well, yes in a room you can have clients with signal strengths that far apart. When you're doing a site survey for a school setup....you're alone, because, we'll, probably doing it in the summer, or during a holiday when the place is empty, or perhaps a few hours after school. So you do your survey...and "think" you have things pretty tight. But human bodies are like sponges for wifi signals. A classroom full of booger pickers or pimple poppers will give a pretty big difference in readings...from what you got when you were alone. So think about placement of the AP. Example...a large classroom...probably not best to use an IW-AP there along one side of the room at knee height...you want something centralized up in the ceiling.

And I think there's a bit of wishy washy with the RSSI. We're tossing around numbers like -67 or -70 or -75...but are the clients taking it that literally? Or is it more...suggestive? You and I probably shared enough notes here to say it's not that literally/exact. But it's good to have something there..as a suggestion. And use that TX power (lower) to keep APs from stepping too far deep into the cells of other APs...find that balance....adjusting TX power (usually lower), staggering channels, and employing RSSI....all those things complement each other to help make for a smooth setup.

I also find, when doing deployments in hard to penetrate setups, with lots of APs....I'll change the 5.0 channel width from the common default of 40 wide, down to 20 wide. Just a hair slower, but more stable, better penetration, and many more channels to stagger across.