GPS sync, TDMA, kids in 3rd grade class, what do they all have in common? They are all designed to make someone take their turn. Imagine what a classroom would be like if every kid spoke at the same time (yea, I’m sure some teachers think they do anyway). Synchronization from an AP was a great idea to start with and, in some environments, is both necessary and works pretty well. In other environments, mainly pretty much almost every one we deal with, good old WiFi mode works better.
Synchronization solves two problems, crappy sector antennas that leak more than a 75 Harley and users that stomp on top of each other like parents at a Walmart at Christmas trying to get the AP’s attention (Hidden Node and density issues). Then there is the Mother whose kid can say Mommy 32,000 times in the span of 27 seconds while hanging upside down on their leg. In this scenario, this would be the torrent users. But what if you could simply turn around so that kid is behind you immediately and becomes invisible and inaudible. If I could patent a shirt or something that did that, I’d be rich (I tried to get the Get Smart Control Cone of Silence, but Amazon was out of them). As I’m not, the second-best thing I could do is simply tell the kids to take their turn which is what the wireless manufacturers did (I tried the spray bottle but the only ones it affected were the cats).
Now if both parents facing back to back were talking to 2 different kids simultaneously and speaking at the exact same time, then we would assume that the kids would hear us (yea, wishful thinking with most kids). This is the problem synchronization basically solves. I’m oversimplifying it but you get the idea. Two sector antennas with clients 180 degrees out of phase have theoretically have sufficient s/n ratios so that the clients really only “see” the data from the one sector pointing at them. If you have 4 sector antennas, then they can all be transmitting at the same time using only 2 channels. If you have 6 sector antennas, you need 3 channels and so on and so forth. But guess what, if 2 antennas are back to back on the same pole 180 degrees out of phase with insane f/b ratios, it doesn’t matter if you have polling. And if your radios have sufficient filtering to avoid bleedover from the other APs close to you or you have sufficient spectrum to move them far, far, away from each other in the spectrum, with antennas that have, yes, you guessed it, insanely low f/b or side lobe rejection, it doesn’t matter. And, yes, one more time, if the antennas are spread out on a building or separated by sufficient distance which is just a few feet and they are on different channels, then, and say it with me, IT DOESN’T MATTER if you have polling.
Sounds pretty straightforward so what is the problem? Well, there are a couple. If you have 100Mbps of capacity on a sector and you use some type of fixed-frame synchronization protocol, you will lose 25-50% of your download capacity because it has to be split between upload capacity and to keep all the timing the same. The reality in our industry today is that 95% or higher of the traffic is downloaded, not uploaded so you are wasting a whole lot of good airtime. If you need the capacity, then you either have to make it up with wider spectrum (harder to get a good s/n ratio which we will get into later) or settle for lower speeds with your clients. Finding contiguous spectrum is hard sometimes when you are moving from 20MHz to 40MHz or higher so having a higher efficiency in band is very important.
WiFi though was built with the idea that every kid should have the opportunity to talk whenever they want. Not every kid will get heard the first time but if you don’t try to put 60 kids in the room with teacher, then they might have good odds their message gets through. Of course, if the little angels are guaranteed to shut up 98% of the time (I can dream), then there is a great chance they get heard by teacher most of the time, which is the reality of the relationships today between the AP and the CPE. And yes, you are playing the odds but when 1-2% of your total bandwidth is your upload capacity, slap down my chips and call me Amarillo Slim.
Video streaming has changed the dynamic and need for polling for most users. Now throw in the ground-breaking RF Elements horns which have a F/B ratio higher than my batting average and the flexibility and size to keep the Homeowners Associations off my back. The only thing that is even close to matching it in a sector format and performance today are Mimosa Networks new sector antennas but they still require significant pole space if you need several on the same pole. It’s also only available in 42 degree beam patterns whereas the RF Element horns are available in everything from 11-80 degrees (+-3dB). The Mimosa sector though, has more gain and can support 4×4 MIMO so 2 tools for 2 different applications.
So back to the point of how all this works together to make polling really not that necessary any more. It’s also how using the RF Element antennas gives you the flexibility to adapt both the budget and coverage to maximize your AP or MicroPop deployments. We deploy both Ubiquiti and Mimosa equipment meaning our APs are 2×2 or 4×4. Because most of these locations are limited to a single pole that also has to handle backhaul equipment, vertical pole space per antenna is critical. Sector antennas simply take up too much vertical space for high-density MicroPops, especially if you need 4-8 of them and the homeowner trims their grass with scissors.
Now throw in the better filtering on the AC radios with the Mimosa radios and the Prisms versus some of our past 802.11n products, and you don’t have to worry as much about adjacent channel noise. With the fact that 5GHz has up to 540MHz of spectrum to work with, there is a lot of room to put some space between channels on the same AP. A side note, the Mimosa radios have an additional 5MHz of spectrum in 5160-5340MHz at the ends of this spectrum bank with a 20MHz channel and an additional 10MHz in the middle compared to the Ubiquiti Prisms. I’m not sure about Cambium ePMP but it might be limited even further on the edge spectrum. Either way, it’s good information to have if you are using the strategy of not using GPS and using WiFi modes on your APs or you need all the spectrum you can find due to interference. This also explains some of what we have seen in the field spectrum scans.
Ahh, now back to polling and high interference. Here is a fact, a packet either gets through or it doesn’t. Yes, you didn’t expect such deep thoughts from me but periodically I surprise you. If it doesn’t, it needs to get retransmitted. Whether that happens in the same time allotted or in the next time around, could be the difference between 8ms latency and 80ms latency. If the interference is you, then you could use that information to rework your polling scheme. If it’s not you, then you are pretty much screwed and you are better off letting the 802.11 PHY Layer do battle to the death in trying to get the packets through if you want the lowest latency. Keep in mind that UDP packets don’t transmit back, thus lowering the amount up upload capacity you need and reducing interference. I can already feel some people cringing while reading this. But I’m willing to defend this position because most of my areas are higher interference and this is where the horns or Mimosa’s new sector antennas come into play.
Time to put all of this together and what you might do if you have different scenarios. Let’s start with a single pole scenario. In that case, we usually start with an Omni until we get a sufficient number of customers on the system to justify a higher Capex on the deployment. In our case, most of these areas are going Mimosa. A little A5-360 or A5c with N5-360 on the pole can handle customers up to a couple miles away in some cases (we have a lot of inventory of both). After we get enough customers and we know where they are located, we start parsing it down into 2 A5cs with horns covering all the customers. That might range from four 30-degree horns, wider pattern horns, or in some cases, horns to cover users in a few directions but if there is a mass of users in one direction, possible an MTI 4×4 panel or now, Mimosa 4×4 panel. This again though, depends on how much pole space we have to work with. The MTI antennas are smaller. Keep in mind that you want to try to get a Mimosa AP to a 4×4 configuration instead of 2×2 covering 2 different zones but sometimes it’s just not cost-effective for ethe revenue.
With MicroPop coverage, we also use the horns for sectional coverage which eliminates the need for 360 degree coverage. We might have one MicroPop cover 90 degrees, another strategically placed near it to cover 180 degrees but not overlapping the first Micropop coverage zone, etc… Think Tetris or that crazy 2000 piece puzzle that was missing a piece at the end (which is the guy that will surely call us and we can’t hit from some reason right after we think we are done). In this case, we can keep noise down to a minimum and if we know where competitors towers are (and we do), we can use that to our maximum advantage. If you are the guy defending the top of the hill and you are surrounded and you aren’t mobile or agile, that limits your deployment strategy greatly. Micropops give us more flexibility to play pin the coverage on the customer with the beam patterns.
Most of our Ubiquiti deployments today are on buildings or towers where we can put up as many poles as necessary. In this case, we have been historically working with the widest pattern antenna that keeps the noise within certain levels. Since we always start out with an unknown number of customers in some direction we usually don’t know of yet, the goal is to keep the Capex down until we start discerning where we need to focus things. We also may start out with 10MHz channels for the first handful of users. Basically, we are trading off cost versus beam pattern versus throughput. A Ubiquiti Prism with an RF Horn is very inexpensive so the reality is that you really don’t care up to a point whether you have 5 or 50 users on an AP because it pays for itself in a month or two and could let you charge a higher ARPU. Even if you had to put up 12 of them or more, that still less than $5K. 50 customers cover that in a couple of months and that’s a worse case scenario. We play other games with this type of setup also.
There are many other variations on this as density increases, the distance difference between closer users and farther users varies more, or you are trying to penetrate more vegetation. You also have to analyze the areas and keep rechecking to make sure things don’t change but our areas are very dynamic, many areas don’t have that problem. But in all of these cases that we use with the horn antennas, the lack of side lobes and high F/B ratios have us pretty much getting rid of polling protocols and gaining back a lot more bandwidth capacity. Of course this doesn’t work for everyone. But it’s working for us pretty well and we are in some very high density environments with as many as 2000 homes in an area of roughly 1.5×1.5 miles. In the other extreme, we have customers out 10’s of miles but the strategy still applies.
Bottom line, only use polling when needed. Test both options if you can, depending on what equipment you are running. Polling is most effective when you have the same interference on both sides so you can find clean spectrum to work with. But if you don’t, good old 802.11 PHY Layers were built from the ground up to get packets through, pretty effectively if I might add, for when the spectrum is as dynamic and unpredictable as a pack of puppies at feeding time.