If you don’t keep up with #WIFIQ on Twitter, you should. It is a question-of-the-day series started by Lee Badman (@wirednot) and always sparks worthwhile discussions amongst numerous WLAN professionals. Here is the question from Thursday July 16, 2015. Clicking the image will open the discussion in a new tab.
I would respond by saying that it is not necessarily a bad design, depending on the requirements. Classic IT response, right? “It depends.” Hallway APs certainly make it hard to add capacity without introducing massive amounts of co-channel interference. A hallway DESIGN may end up working just as well as having access points in rooms given proper channel planning, attenuation measurements, and validation. Many tweeps pointed out that sometimes APs in the hallway is a customer requirement. Verticals such as hospitality and healthcare may not allow access to guest rooms or surgical suites for many reasons. I think nine times out of ten, hallway APs are never part of an actual design; they are thrown up willy-nilly by folks who just don’t know any better. All you really need to say is “Put your RF where your clients are!”
If you haven’t read my last post “Fixing Hallway-Fi, Vol.1” please check it out. By the way, the TM on the title is just a joke. There is also an excellent comment from Keith Parsons (@KeithRParsons). The post describes the steps I took to optimize a hallway “design” in a single story school building.
One of the points made in the #WIFIQ discussion was that you need to consider all three dimensions in a multi-floor hallway network. Indeed, RF energy propagates left, right, forwards, backwards, up AND down. The first school I optimized is a two-story building. I did the work on it back in March, but didn’t document it as well as I should have. Luckily I have the survey files handy, so I was able to take some screenshots and hopefully put together a useful post. I also have utilized Keith’s handy tips from his comment on my previous post, so the visualizations should be easier to understand.
For any signal strength images, my ranges are as follows:
- Want: -72 dBm or greater (shades of green, yellow and orange)
- Don’t Want: -73 dBm to -85 dBm (grey)
- Don’t Care: less than -85 dBm (white)
For channel overlap images, the colors indicating the number of overlapping access points are:
- Green: none
- Yellow: two
- Red: three
There are seven Juniper WLA522 access points in this network, so in 2.4 GHz there were never more than three overlapping APs; at least RingMaster got that right. All were set to auto-channel, and here are the default (aka maximum) power levels for various channels for that model. You can see they were all over the map.
Without further ado, the initial signal strength maps for the first floor (4 APs) and second floor (3 APs). You can also see the channel plan selected by RingMaster. I’ve placed them so the second floor is above the first – should help to visualize the building. I forgot to include the AP names on the map. On the second floor, left to right, is RES5-RES6-RES7. The first floor, left to right, is RES1-RES2-RES3-RES4.
Once again, high signal levels everywhere and a ratio of 1 AP per 2 classrooms. I’ll go through the capacity plan later on. Thank goodness these APs are all mounted horizontally on the ceiling!
Now, lets think about how a hallway design in a two-story building is different than in a one-story. A single floor building (such as the example in my previous post linked above) may have wings, creating horizontal separation between the hallways and thus introducing attenuation from walls and free space loss. Imagine a building shaped like an upper-case T. A client sitting in the intersection of the hallway will hear each and every access point in the building. However, the access points at the ends of the hallways may enjoy a teeny tiny bit of isolation from each other thanks to the walls and free space in the direct paths between them; there may be four walls (two of which are exterior) and 100 feet between them. In a two-story building, you lose the “hallway design aids” of corners, wings and free space loss, and end up with your access points separated by 10-15 feet of vertical space and maybe 10-20 dB of attenuation from floor materials, depending on your construction. If your EIRP is 20 dBm, an optimistic estimate is that your access points are hearing each other at -40 dBm or higher. Have a look at these next maps, which show the signal strength on each floor from the access points on the other floor.
You can see that the second floor access points actually provide coverage for the entire building, excluding the gymnasium. The two greenest shades indicate signal levels of -50 dBm or greater. We can attribute the weaker coverage on the floor above to the (assumed) downtilt of the internal omnidirectional antennas. I say assumed because I cannot for the life of me find a spec sheet with antenna patterns for the Juniper WLA 522, even in their support portal. You might now be thinking, “Why not just turn off all the main floor access points and be done with it?” Certainly that is an option, albeit one that I failed to think of at the time. Again, capacity plan coming later. Here are the channel overlap maps. Remember, green=1, yellow=2, red=3 overlapping APs.
Here are the pop-up lists of overlapping signal strengths from key areas on each floor. Quite a bit of contention going on, even in 5 GHz because 40 MHz channels were enabled.
With the devices I have to support, I need to maintain 2.4 GHz coverage throughout the building. Lets have a look at the capacity plan for this school now. Here is the input and output portions of the Capacity Planner from Revolution Wi-Fi (@revolutionwifi).
The number of devices at this K-8 school is low – roughly 4 laptops (many of which are 2.4 GHz only), 2 iPads, and an Apple TV in each classroom. (Edit: we just bought 20 new laptops for this school, so the count went up slightly). With no high school students, the number of personal devices is limited to teachers and thirty or so older students. You can see from the capacity planner results that my current number of seven APs is overkill for capacity, but I’ll work with it. I know that the iPads and Apple TVs will go to 5 GHz given the chance, and I’m going to assume that half of the personal devices (phones) will as well. I probably should have changed the band steering ratio in the planner to indicate that…
Here are the steps I took to optimize this WLAN:
- Set channel width in 5 GHz to 20 MHz to eliminate channel reuse.
- Turn off 2.4 GHz radio in RES1, RES3 and RES6.
- Disable Juniper’s load balancing feature because it’s crap and we aren’t dense enough to need it.
- Disable low rates and set beacon rate to 24 Mbps (does not reduce CCI due to preamble).
- Deploy this static channel and power plan:
- RES1: 2.4 GHz off, Ch.36 @ 10 dBm (max)
- RES2: Ch.1 @ 14 dBm (max), Ch.157 @ 15 dBm
- RES3: 2.4 GHz off, Ch.165 @ 15 dBm
- RES4: Ch.11 @ 7 dBm, Ch.40 @ 10 dBm (max)
- RES5: Ch.11 @ 12 dBm (max), Ch.48 @ 15 dBm
- RES6: 2.4 GHz off, Ch.153 @ 15 dBm
- RES7: Ch.6 @ 15 dBm, Ch.161 @ 15 dBm
Here are the resulting visualizations from the validation survey.
I haven’t wasted space with the channel overlap maps for 5 GHz, because there isn’t any – with seven radios on 20 MHz channels, there is no channel reuse. There is some overlap in 2.4 GHz. I had to reuse channel 11 on the access points at the opposite ends of the building. I used channel 11 because it has the lowest maximum transmit power (12 dBm) for the WLA522. RES4 on the main floor is transmitting only high enough to cover the gymnasium, but it still results in co-channel interference with RES5 on the second floor. The CCI is limited to the hallways, but that doesn’t make it any better. It still affects the network by creating contention for AP transmission airtime, and may cause retries on client transmissions. I plan to move RES4 into the gym with a patch antenna eventually. The gym coverage is the main reason I can’t use only the three APs on the second floor.
Andrew von Nagy recently wrote a series of tweets about knowing which CCI actually matters – are the overlapping channels the same as the primary access point for the area your client is in? If not, you probably can ignore it. Having APs in the hallway really increases the size of that area, and this effect is multiplied on a multi-story building. I think the only areas where there may be concerns are the middle of the upstairs hallway and in the library and room 303. Clients in these areas may be connected to RES5 but will hear RES4 only 5-10 dB lower. The other area all have radios on other channels with stronger signals to connect to. We all know that clients always connect where you expect them to…
The main point I wanted to make in this post is that while you may think of hallways on different floors as being separate physical spaces, depending on your building’s construction they may have less attenuation between them than two rooms.
Thanks for reading and commenting!