Monday, April 11, 2011

Chapter 23 - I Can See Clearly Now The Rain has Gone!

WISPs are starting to get a little recognition with Obama’s new initiative to provide more funding and loans for Rural Broadband. He announced a new initiative to take back more bandwidth below 1GHz and auction it off. Then he came up with the brilliant idea to use those funds to pay for a real public safety infrastructure. The timing of that was about a week after Chapter 21. I’m not saying Obama is a big fan of “Tales from the Towers” and that he would plagiarize other people ideas like extending the Patriot Act or keeping Guantanamo Bay open, but, okay I am saying… Everybody has to have their fantasies.
Then the Obama administration announced they are using leftover funds and probably some other money the fed just printed up out of thin air yesterday, to extend loans to bandwidth providers. Apparently the several billions already spent didn’t end up providing a lot of bandwidth to the end users (big surprise), but funded lots of fiber projects, bandwidth for government entities, and data collection consultants (lots of value there). So they created a new program to try and help the WISP operators again.
The sad part is that it’s the wrong thing to do. Throwing money to one company who can fill out forms better than their competitor screws up the free market in that area. Nature abhors a vacuum and a WISP operator can now profitably serve areas with as little as 10 users. If users are willing to pay, someone will set up shop to make money. The government doesn’t need to give money to companies to provide service to these areas. For example, giving money to Hughenet, SkyBlue, Echostar, and Spacenet to lower the cost of satellite internet was just plain stupid and totally against free-market principals. Local WISP operators who can provide better service or higher bandwidth now have to lower prices to compete. That means less revenue for those operators to expand their service. Of course, once a user experiences real internet service other than satellite, they typically won’t go back, regardless of price.
For example, Triad Wireless operates an area that has 54 total homes over about 15 square miles. Most of the users were on satellite for $60 per month. That meant those users had no real-time action game playing, no Netflix, no video-conferencing, no Skype, no remote backup software, etc… Due to the remoteness of the area, the only landline option was a T-1. We brought the circuit to one centrally localized house, extended the system with wireless, and 10 users signed up with an install fee of $250 which wasn’t far off from the satellite provider (it’s also a really long drive from our office). Even though the bandwidth wasn’t ideal, it opened up a lot more options than those users had with satellite. For example, one of the residents is a teacher who had to drive 20 miles into town to find a faster internet to teach their class twice a week. His video conferencing simply didn’t work over satellite.
As word of mouth spread on the difference in performance between a T-1 shared between 10+ people (19 at last count) and the satellite systems they had, most of the satellite users in the area dropped their contracts. The increased user count allowed us to expand the system to 3Mbps at a cost 33% higher than satellite. Most of the users would rather pay $80 per month for up to 3Mbps of real bandwidth with no monthly caps than pay $60 per month to SkyBlue for limited capabilities. Some of the users opted to stay with the 1.5Mbps peak. The users know the bandwidth is limited and if someone is hogging it for a long time, we throttle them back for a while. When things slow down, we open the pipes back up. It’s worked for 2 years like a charm.
That doesn’t mean everyone can afford to pay $80 a month. This area has very large acreage properties in a very remote area 25 miles from a small city. In other areas where it’s economically possible, Triad Wireless plans on rolling out a $10 or less monthly service under the “Education Everywhere” program for lower income students. It just depends on the environment but it demonstrates that it’s pretty easy for a WISP to get into a lot of nooks and crannies not addressed by cable or DSL providers.
Satellite companies typically charge for installation and a certain level of access fee to make a profit. I have no problem with that. However, if the government steps in and gives the satellite companies money to eliminate the install fee and subsidize the monthly fee to offer $40 per month service, that gives them an unfair advantage over a local WISP operator. I’m not saying that I can’t compete in a fair market, I just can’t compete when the government is taking taxes they collect from my company and then shovel it into the bank account of my competitor to be used against me. I would rather pay the Mafia since I they would guarantee I wouldn’t have any competitors muscling in on my business. I would also get better protection, and at least I know the rules going in. Losing a kneecap is better than watching your business and economic viability collapse because some feel-good politician with absolutely no business experience 2000 miles away just sold you out to a big contributor, but enough about our useless federal government that isn’t competent enough to balance the books on a lemonade stand.
More spectrum simply needs to be opened up for free that can use equipment that doesn’t take the Manhattan project to design. Regulate it like 3.65GHz but don’t let any cellular companies use it. If a company owns spectrum in an area, then they shouldn’t be allowed in. If Jim Wright can be allowed to keep Southwest Airlines out of Texas (sure, politicians can’t be bought), then the FCC should be able to keep Verizon, Sprint, and AT&T out of the new band. If those big companies are allowed in this band, smaller operators might just as well start setting up shoe polish stands. The most efficient companies in terms of speed of deployment, lowest cost, and best customer support are the independent WISP operators. The closer the owner/operator is to the customer, the better the service.
Of course, the telecommunications group donating the least amount of money to our “we never met a donation that didn’t influence our vote” politicians are also the WISP operators. From what I’ve seen of GE’s tax bill, several congressman and senators are probably sitting on “I love Jeffrey Immelt” tattoos. I do know that a bunch of Democrats added “I love Netflix to their other cheek recently”. I’m not holding out much hope that they will do the right or even the smart thing for the small business owners who employ 80% of the people in this country. I’m also not holding much hope for my invite from Obama to ride all over the world in Air Force One.
Fortunately for the rural market, several new technologies may start removing the chokehold that local fiber and wired incumbents have in the U.S. For example, if you want local loop fiber in many states, although you can find a middleman, you probably still end up on one of the two local incumbents in the state. If you are in Mexico or Canada, you end up on one, count them, one local loop fiber provider. Last I checked, that’s usually not the best deal for the consumer but it makes it possible for the richest man in the world to keep his place on top in Mexico. At the same time, the government wants to help WISPS which are usually a handful of employees, by asking them to fill out paperwork that takes 2 attorneys, a CPA, 5 grant writers, 3 RF and network engineers, a market research expert, and two psychics. So how do the smaller WISPs continue to compete in what is clearly a difficult business environment caused by government rules?
We talked about what the cost of bandwidth was in a data center. In some areas, it’s as little as $1-$2 per MB (probably shared somewhere) or the highest pricing I have seen for a Tier 1 provider is around $5-$6 per MB. If the connection point includes local loop, the cost is probably $20 per MB or higher. Since the profit of an ISP is the reselling of that bandwidth, it’s important that cost stay low. That means we start at the data center whenever possible. I recognize this isn’t an option for everyone, but for this model, we are banking on the idea there is a reachable data center.
Let’s assume that your state has 2 incumbent carriers for local loop and you remember what happened to Covad when they built their company on a single source supplier model. I just don’t trust a local loop provider that also happens to be a competitor. Most companies have the ethics to not lower your system performance to damage your company. But think about this, do you want to build your company on the premise that it’s easier to ask for forgiveness than permission and let the lawyers handle it after you are out of business? The local loop provider could also easily mess with your pricing model and either undercharge you or provide better performance since they know your numbers.
I also wouldn’t bet my survival on a rogue IT guy buried inside a big company whose job may be on the line when their revenues go down. There are too many ways to slightly screw up a network without anyone being aware if or how it’s done. Telmex is the experts at that. Just try to make a Skype or VoIP call across the Mexico border to a Telmex business DSL client to avoid paying international phone rates. It sounds like Obama speaking without a teleprompter under 10’ of water. You can’t build a house on a shaky foundation.
So if we don’t’ want to use local loop, what’s option 2. We build our own backhaul to connect to the big boys. If your coverage area is within a couple hundred miles of a major city, wireless backhaul can now move 1Gbps for 20 miles or more with no compression. Compression is useless with video streaming since the packets compress as well as a Sumo wrestler in a Southwest airlines coach seat. The days of 100 users sharing a T-1 circuit are over. To be competitive today you need lots of bandwidth. You need to plan on moving big chunks of data long ways early on even if you don’t need a lot to begin with. Video streaming will be the major type of traffic over the next few years. Historically, this would take multiple radios bonded together or multiple hops.
Let’s run the numbers and see how this works out for a really big pipe. 1Gbps with local loop charges probably cost no less than $15K per month. That’s $180,000 per year. 1Gbps pipes are the sweet spot with Tier 1 providers in terms of cost per megabyte. If you can use the new hardware from Bridgewave or Exalt, who now have products that can move 1Gbps over a single pair of radios for about $25-$30K, to go 20 miles, then you have passed the first test. Assume that there are some obstruction or terrain issues between the data center where the fiber feed is available and that you need an additional tower in the middle.
Going over the Capex to go 20 miles, the fixed cost is two radio pairs costing $40,000K installed worst case, assuming an outside installer for a total of $80,000. The monthly costs are $7K at the data center and probably about $500 per month on the tower co-location fee. Toss in the $15K-$20K you are going to have to pay for the survey and licensing of those frequencies, and that puts the first year cost at $190,000. It takes approximately 12-13 months to break even over paying a local loop carrier. However, year two you are now saving over $100,000 or enough Moola to put up a parallel link to handle the growth of your WISP operation since your service now has awesome bandwidth.
The only key difference between the products is how far you can go. The problem with the Bridgewave is that it’s only available in 18GHz and 23GHz frequencies. This is fine for the 10 miles hops and mild climates. Go further than that and the antennas will be big enough to saucer 5 kids down snow hill or you will be singing “Don’t let the rain come down” (http://www.youtube.com/watch?v=xrThyvv4sko) every time it gets cloudy. Keep in mind that when you use dishes that Shaqille O’Neil can hide behind in higher frequencies, tower load and sway also now becomes an issue. In this case, it’s easier to put an 8’ dish on a building at 10 feet and a 4’ dish on a tower rather than a 6’ dish in both places. The 4’ dish also has a wider beam pattern allowing for a little more tower movement. The sizes of the dishes are generalized but you get the idea.
Exalt has the option to use 6GHz and 11GHz which means 10 miles could become 20+ miles. Curvature of the earth starts to come into play in flat terrain areas at those distances if nobody left you a pair of convenient mountains to work off of. Even if you keep the hops to 20 miles, that now means 6-8 hops gets you border to border coverage for the entire state of Indiana from Indianapolis. Since I doubt you get $15 per MB pricing all over the state even if it was available, these radios now just technically and financially opened up very big pipes for pretty much most of the country.
Based on the amount of interest I’m seeing from several venture capital groups I’ve talked to over the last few weeks, I’m sure I’m not the only guy that figured this out. Going back to some numbers we discussed in the past, 3600 users with that type of bandwidth generates $100,000 to $150,000 per month in revenue. We are spending $200,000 for our backhaul and that cost drops by $100,000 the second year. The revenue stream is a minimum of $1,200,000 per year when the system is built out. It doesn’t take an accounting genius to figure out that certain WISP models can be a highly profitable endeavor. However, if we can figure that out here, imagine how much number crunching is going on at the big cellular companies. They are promising enough bandwidth to your phone to download a Peterbuilt. Unfortunately when you try it, you find out that if you want the trailer, it will cost another $10 and you are only allowed 1 gallon of gas to run the whole thing for 30 days. Either that or they throw up a 25mph speed limit after you driver for an hour. The end result is that the federal government will let the multi-billion dollar cellular companies charge for the use of bandwidth on a per MB basis but the mom and pop WISPS have to give it away for free. I’m sensing a double standard here.

Wednesday, March 2, 2011

Chapter 22 - Is Law Enforcement the Red–Headed Step Child of the Broadband Movement?


When Wireless Broadband was on everyone’s lips 8 years ago, we all thought we would be able to use our laptops everywhere. About the same time that Earthlink and Metro-Fi realized you can’t make “Free” pay off for their investors, 3G started moving in to fill the void. Then YouTube came along, pushed 3G to the ground and said, “I spit in the face of your puny bandwidth (insert Austrian Accent Here).” 3G then said, “Oh yea, my big brother, 4G, is coming and he will take care of you. You will be sorry.” YouTube said, “BTW, meet my cousins, Hulu and NetFlix”. 4G took one look at these guys and said, “I’ll fight you guys but you have to have one hand tied behind your back, both legs tied together, and we are only going to fight for 10 minutes. After that I win and you have to go home”.
In the meantime, the Big Land Barons who wanted to expand their holdings went to the Lords of the FCC and whispered in their ears, “Our lands aren’t big enough and we have rocks and trees in our way. If you kick out the squatters that are on some of your prime land, we will buy it from you and enrich your coffers.” So it was that the Lords of the FCC created The Great Plan. They took The Great Plan to the Council of Kings and said, “We need you to pass a law to kick out the old squatters. We will tell them they are doing the right thing for the country and then resell their land for a fat purse. We will then give some of the land to new squatters for free who will also develop the land and make our Kingdom better.”
When the Sheriffs heard of The Great Plan, they said unto the Lords of the FCC, “We have worked our lands to death and need new lands. The Bandits are smarter, faster, and meaner. We need new lands all across the Kingdom and in every village so Sheriffs are closer and can work together to stop the bandits.” In the Great Plan, the Lords of the FCC decided to appease the Council of Kings and not ask them to use the coin from the Great Sale to pay for the Huts and Barns and lookout towers with 360 degree coverage and an 8” spyglass mounted in the middle. Oh no, they were more clever than that. The thought they could fool the Big Land Barons into paying for the not only the land, but for the huts and barns and lookout towers with 360 degree coverage and an 8” spyglass mounted in the middle (I’m a huge Arlo Guthrie fan). The Lords of the FCC said unto the Sheriffs, “we will grant you new lands in order for you to protect the peasants from the bandits.” Thus, the Lords of the FCC were happy with their new plan and went forth to preach it. And so it was written, and so it was done.
To the peasants, the Lords of the FCC preached, “We are confiscating ill-gotten lands given away by the former Lords of the FCC in order to give you better services and enhance your cities and huts”. Prosperity will grow across thy land and all will be connected to the Great Message.” And the peasants and other squatters did not object for it was not their land being confiscated.
The FCC executed The Great Plan. They confiscated the lands and sold them off to the new Land Barons. The new Land Barons offered the Lords of the FCC a great amount of coin, thus ensuring that the Lesser Land Barons weren’t going to be able to buy the new prime land. At the same time, the FCC told the Sheriff’s, “We have decided that your land is too valuable to just give to you so we are going to sell it to the Big Land Barons. However, we are going to make the Big Land Barons build you houses and barns and lookout towers with 360 degree coverage and an 8” spyglass mounted in the middle.”
When the Land Barons read the scrolls written by the Lords of the FCC on how to use the new lands designated for the Sheriffs, they went to the Lords of the FCC and said, “We do not want that land. It is too costly and cumbersome to build the houses and barns and lookout towers with 360 degree coverage and an 8” spyglass mounted in the middle.” They then paid the Lords of the FCC a great amount of coin for the land they had already bought. The Lords of the FCC quickly left the Great Hall so as to not have to explain to the Sheriffs why their plan had failed and the Sheriffs still had no new land.
After that, many Sheriffs sent many riders to the Lords of the FCC requesting the new land. The Lords of the FCC spent long seasons pontificating on what to do about the land they confiscated for the Sheriffs. After a very long time, they came up with The Little Plan. The Lords of the FCC decided in the Little Plan that they would grant the land directly to the Sheriffs. They said, “We will give you the lands, but we will tell you what you may build on the land. You may only build huts and barns and lookout towers with 360 degree coverage and an 8” spyglass mounted in the middle as to our specification.” The Sheriffs said, “We have no coin to build huts and barns and lookout towers with 360 degree coverage and an 8” spyglass mounted in the middle. Where is the coin from the Land Barons for sale of the other Land?” The Lords of the FCC said, “It has been confiscated by the Council of Kings and redistributed to friends of the Great Messiah and the Council of Kings. There is no coin for you. You have to find your coin from somewhere else to follow the Llittle Plan.” And that my dear friend, brings us to today.
Basically State and Municipal Law Enforcement sort of got left out in the cold when it came to the new 700MHz frequencies. It’s not that they didn’t eventually get the bandwidth they needed; they were told that they must deploy the most expensive technology available to actually use it. Instead of using the money from the auction to build out this infrastructure, each municipality must fund their own infrastructure. Of course, considering that 3 different agencies in the Justice Department haven’t been able to build a simple radio together after wasting hundreds of millions of dollars with no end or product in sight, it’s probably too much to ask them to build a nationwide wireless system. This edict was also given at a time when government budgets are getting hammered from top to bottom. Since public safety has no money to deploy even two tin cans and a kite string, I don’t think we will be seeing high-speed bandwidth for mobile public safety in the near future in most major cities.
We started discussing a project last month where I needed to move 1Gbps or more of bandwidth 20 miles for a county wide system. There is new wideband equipment being released as we speak and I haven’t finished my homework so I’m going to table that phase of the project for a while. Since the project has several components, I thought we might jump over to the third tier, the last mile, since it directly ties into public safety.
As many of you live in areas where your vegetation grows over 8’ and doesn’t make you look like a porcupine when you bump into it, I thought tackling a project like that would be a good idea. Living in the desert has made me oblivious to the fact that there are cities across the country where houses are surrounded by trees that could double as space elevators and should have lights on top to warn airplanes away. Throw in vegetation that comes with its own zip codes, and there are places where the term Non-Line-of-Sight (NLOS) takes on a whole new meaning. Every cell tower that I saw in the area was so tall I figured I was coming down with magic beans if I ever had to climb one. I also learned that approaching someone getting out of a truck that came with a factory gun rack was probably not the best idea.
However, as high as all the cellular towers were around the county, there is no chance that 80% of the houses are ever going to see one from their front window. That means we are left to figure out how to get through the trees. In Chapter 17, we talked about how 900MHz can punch through vegetation, houses, and really mean neighbors. Since that article was written, some of the data I’m reading on current 900MHz deployments is pretty promising. However, deploying 900MHz can be kind of like trying to give a cat a bath, you get a trip to the emergency room and the cat just gets really ticked.
A quick review of 902-928MHz band means that we only have 26MHz of bandwidth to work with. By comparison, 5.8GHz has about 100MHz of spectrum to work with and 2.4GHz has approximately 60MHz. In practical application, a 5.8GHz system can deploy 4-5 APs with non-overlapping 20Mhz channels and a 2.4GHz tower can have 3 non-overlapping 20MHz channels. The 902-928MHz band is a little tighter so we only have room for one channel if we are using a standard 20MHz wide channel.
That means we need a different strategy. Assume that our coverage area is 360 degrees. Most wireless products in 902-928MHz band use down-converted WiFi chipsets. That means the throughput will be the same as 802.11a and 802.11g radios with the same channel widths. Motorola uses an FSK scheme instead of an OFDM modulation scheme that trades off a lower throughput for a better s/n ratio.
Both systems have a bandwidth rate in Mbps that is numerically about ½ the spectrum size. For example, Motorola uses an 8MHz channel and has a maximum capacity of 3.3Mbps total aggregate. WiFi down-converted systems with a 10MHz wide channel will have an aggregate throughput of about 5Mbps. Some Wifi systems have a little higher throughput but all of this assumes a connection at the highest level modulation rate. The other side of this equation is the whereas Motorola will work with a s/n ratio of 3dB, a WiFi system needs a minimum of 10dB.
802.11N 2x2 MIMO has a different formula. Not only is the efficiency of the protocol better than 802.11a/g, the processors are typically faster, more efficient, and have better sensitivity ratings than older a/g chipsets. From there, throw in the fact you are transmitting two signals in the frequency spectrum. The end result is such that the rule of thumb of throughput is approximately quadruple the channel frequency width. A 10MHz channel width should have a theoretical throughput of about 40Mbps.
The lack of channel bandwidth in the 900Mhz band is one reason that it never caught on very well. The other reason is that the noise floor is significantly higher in the 900MHz band than it is for other unlicensed bands. Because of the NLOS nature of the band, it’s heavily used for SCADA, baby monitors, cordless phones, ham operators, and many other things. In any major city or suburb, you probably have noise levels in the 50-60dBi range which makes it very difficult for most WiFi based chipset radios to operate.
The third problem with the 900MHz band is that laying out a network design for a large area is significantly more difficult. In 5.8GHz, the antenna beam patterns are pretty well defined, the signal can’t penetrate lace curtains, and it has limited reflection characteristics compared to the 900MHz. For those reasons, it’s relatively easy to define optimal tower locations. 2.4GHz sort of falls in between but the noise figures for 2.4GHz will still be lower than 900MHz at ground level. At the tower locations, I can imagine it’s a toss-up as to which frequency band will have a higher noise level between 900MHz and 2.4GHz depending on the population density around the tower. 900MHz on the other hand, bounces around like a Superball in bathroom, needs a Fresnel zone the size of Delaware, and just feels a little tickle when busting through a brick wall.
Verizon is probably about halfway to deploying their 700MHz LTE system across the country. The difference between them and us is years of experience in 800MHz and software that costs more than my last car. In addition, the antennas they use can be remotely adjusted on the fly in micro increments to fine-tune coverage zones. Since I haven’t heard anything about WISP operators learning to levitate, that means lifts or tower climbs, both of which are significantly more difficult and time consuming than moving a mouse.
With these issues, why would anyone want to use 900MHz band? Well, that is because we really don’t have much of an option. It’s our only option through Sherwood Forest and we need to figure out how to make that work. Let’s first define the environment as being rural which reduces the variables and takes the noise figure off the table. Then let’s assume we are covering about 2 miles in every direction and we have trees all around. Throw in the decision to limit users to a maximum download speed of 4Mbps and upload of 1Mbps for Round 1 and the system design gets a little easier.
The first issue is the AP configuration. Since 802.11n 2x2 MIMO is pretty much standard in the WiFi industry today, the use of it in 900MHz is unique. The Ubiquiti Rocket M900 with a dual-polarity 13dBi 120 degree sector antenna makes an attractive option. Assuming a 20MHz wide channel, the AP should be able to support a theoretical maximum of 60-100+Mbps of aggregate bandwidth. However, we have to put 3 of them on the tower to cover 360 degrees and I guarantee we aren’t the only squatter on that land. That means that we are going to have to limit the channel size to 5Mbps to avoid overlapping channels and minimize interference. It also means that the maximum theoretical throughput is about 20Mbps and from my testing, I would assume 10Mbps per AP.
If we use the 10-1 client/bandwidth formula, then we should be able to support 20 users per AP with a 4Mbps down and 1Mbps up scheme. That’s also 60Mbps of total capacity per tower. Those aren’t bad numbers if you are charging $40 per month per client. If you are using Water Tanks for example, where you can isolate the antennas from each other, you may have the option of using 10MHz channels. There are also antenna shields from sales@rfarmor.net that promise that you can use 10MHz channels and radios on the same tower and they won’t interfere with each other. However, this could also be accomplished with GPS synchronization with close to similar results if the Rockets every support this feature. I’m a fan of as much isolation as possible with multiple APs on the same tower regardless of synchronization so they are in my budget list.
This is where we tie all this together. 900MHz in rural areas is also a great poor man’s Public Safety System. With the right RF engineering and other design features, the same system could deploy 2-10Mbps or more to a police vehicle. Fast handoff hasn’t been resolved yet but I’m working on a couple solutions as I’ve mentioned in other articles. A system like this should cost about 1/20th to 1/100th the cost of an equivalent LTE system. I’m not suggesting that this type of system be deployed in any cities or suburban areas that have high noise issues in that band or that plan on deploying a 900MHz smart grid system, but since over 90% of this country is still rural, this capability can be added for at a cost of under $3000 per AP with each AP covering 4-36 sq miles. The cost for the cars is less than $400 without fast handoff or redundancy. Compare that with a typical LTE or WiMax deployment that the police clearly have no money for, and it’s a pretty good alternative for small towns and rural counties.
900MHz is both magical and a pain to work with. It opens up new opportunities in rural areas where wired fears to tread for financial reasons. At the same time, it’s also a little more expensive than traditional 5.8GHz PTP systems but a lot less expensive than most street level 2.4GHz muni systems. However, when compared to 3G speeds, 5GB caps, spotty coverage, or even worse, no coverage, it’s still a far better option. The application of 802.11n 2x2 MIMO technology now makes it a more viable technology for rural areas.

Saturday, February 5, 2011

Chapter 21: Let’s do it for the Children


I get a lot of calls from people interested in starting a WISP. Since I look at each deployment as a challenge both technically and financially, it’s very interesting to me to hear the scenarios and the expectations. On the other hand, I’m also involved in several industry blogs centered on different products, and I have had the opportunity to listen to many experts with years of experience describe their methodology and philosophy behind their designs. What’s most fascinating is that the modern WISP operator has developed into the medical equivalent of the general practitioner or the proverbial jack-of-all-trades which has resulted in many successful deployments. Keep in mind that most profitable WISPS are built around PTMP design instead of a municipal mesh model. I’m hoping Guerilla WiFi can change the model.
One thing I learned a long time ago is that being vendor agnostic is a lot cheaper than blind loyalty. Before you start a WISP, first look at all the products that are on the market that are appropriate. Listen to what others say, especially those that have proven financially successful models. However, keep in mind that many people will blindly follow manufacturers like lemmings over a cliff, or worse, suggest that you go with them if they are already invested in a particular vendor. Make sure you talk to operators that are using different types of products.
During the market transition of Microsoft Word for Windows from WordPerfect for DOS, I saw staff ready to quit their jobs if they were forced to change to Microsoft Word because of their comfort level with Word Perfect. During one particularly testy Word Perfect to Word upgrade at a law office, a para-legal threatened me with an optional surgery procedure involving her keyboard that not only was I sure she wasn’t medically licensed to perform, I wasn’t even convinced it was anatomically possible. Unfortunately, the world doesn’t stand still in the high-tech industry and complacency means obsolescence as demonstrated by IBM, Sun, Lotus, and even Microsoft (Balmer needs to get a better cologne because whatever he is wearing is making his senior staff disappear faster than Microsoft’s consumer base). Try highlighting Lotus administration in your resume and see how many job offers you get. In the wireless industry, there are established players such as Motorola and Dragon that are getting hurt by new players such as Ubiquiti, Tranzeo, Microtik, and SAF who are providing products with significantly lower prices. That should never be the only criteria for selecting a product since every product has some feature that may be critical for a design as we shall see when I get to the backhaul section.
The model for Guerilla WiFi wasn’t based on the idea that WISPs can only compete in the remote areas. It was designed with the idea that profits should be generated in 12 months or less, regardless of location, and that the complete ROI of a new design shouldn’t be much more than 2 years. This should make it a no-brainer business model for investors. But wait, there’s more. Buy it today and you get the following bonus, it’s cheap to maintain so investors or governments shouldn’t go broke keeping it running, and better yet, it’s good for the children. This model also comes from the idea that it will take products from multiple vendors to make work.
Recently, I was approached by an economic group looking to improve coverage in their county. The basic statistics were that the county had 25,000 households with 10% of the households having no high-speed internet or only cellular and satellite options. 47% of the 25,000 households had high-speed internet installed between cable and DSL, and were paying about $36 per month although I don’t believe that included taxes. Average annual income level in the county is about $43,000. The total county area is a little less than 500 square miles.
I ran some calculations and using the PTMP Guerilla WiFi model and came up with the following numbers using a bandwidth to user ratio of 12-1 and a target of 5Mbps per customer average. Profitability should be achievable in less than 12 months and ROI in 30 months with a continuous linear growth factor through that period. In 36 months, the model predicts a gross profit percentage of 50% or more by month 36. Those numbers alone made it a no-brainer in terms of a simple investment opportunity.
However, the next part ticked me off and should embarrass the telcom industry, the political establishment, and be a wake-up call for our country. Data provided by the economic council showed this disturbing fact, this county has a middle school where only 12% of the students have the option of high-speed internet at home. Now it just became personal.
This is truly an embarrassment for a nation that used to be a leader in innovation, manufacturing, and education. Other schools had higher percentages but the reality is that until all students in this country can access affordable, high-speed internet at least for educational purposes, then we need to do better as a nation. If I was King for a day, I would host a conference, not with a bunch of politicians or leaders of nationwide cellular companies, but with the guys who have built and operate profitable WISP infrastructures all over North America. The knowledge and talent built up by guys who can climb a tower, design a network, map an RF plan, and run multi-million dollar businesses is what this problem needs.
Here’s an idea, take a cue from the Sopranos and carve up the country into areas of responsibility for the WISPS. Find the areas that wired services and monopolistic incumbents aren’t providing competitive options, define a fixed price for each client installation with guaranteed bandwidth requirements, get rid of all the union rules and bureaucratic paperwork that government jobs require (which is a complete waste of time and lowers wages due to overhead) and turn these guys loose. I guarantee that installations would be starting in weeks and the job would be completed within a year. Want a stimulus program idea, this is it. When it’s all over and the politicians want to know how it was done, just say “fuggetaboutit, you don’t want to know. Enjoy your $10 Internet and don’t ask any questions”.
Of course I’m joking, but there is simply no reason for 88% of students in a school not to have access to high-speed Internet anywhere in the United States, politically or economically. Being the capitalist that I am however, I realized that this situation wasn’t going to be solved with politicians directing government funding as evidenced by the last round of government grants. We wasted 7 billion dollars in the last couple years and it didn’t make a dent in the bandwidth needs in our country. We spent millions of dollars of that money just trying to figure out where the holes were.
There needed to be a business case which supported providing internet to these students at a cost which makes it attractive to investors ,(not that the federal government would have any interest in that unless it involved federal employees or union workers to run it, no self-interest there), but yet provide 100% coverage for students. After a little more thought on the problem, three more Diet Cokes, and several of my favorite snacks, the OREO cookie (a shameless plug but maybe Coke and Nabisco will send me a 24-pack of the bubbly and a bag of double-stuffed cookies to power my idea processor), I came up with a methodology I have named, “Education Everywhere”. With slight modifications, the Guerilla WiFi model can easily accomodate “Education Everywhere” capabilities to these students for less than $10 per month and still produce a profit for investors. At that price, there is no longer an excuse not to have high-speed bandwidth for almost every student in the country.
With a little cooperation from the schools, Education Everywhere could also tighten up the bond between parents and education as well as expand distance learning options. The $7 billion dollars worth of grants that the government gave out could have installed 8,500,000 residential locations and subsidized them for 3 years. On the other hand, it could have provided “Education Everywhere” bandwidth to students for the next 6 years or more. We could have covered 40-60% of the 100 million users the federal government claims don’t have high-speed internet. If a President ever comes out in a State of the Union speech leading with those types of numbers, I will sit and listen to the rest of the speech without saying a word. Okay, maybe not since political speeches are boring and have more fantasy in them than the Sci-Fi channel, but you don’t win the future if you can’t produce a quality education with the right high-tech tools. I just want to do it without the country going into debt further.
As much as I’m excited about the “Education Everywhere” program, I’m going to hold back on writing more about it until we start deploying it in a few months. Instead, let’s get back to the details of how to deploy a system to cover 500 square miles and makes a profit. There are three basic parts to the infrastructure:
  1. Data Center Infrastructure
  2. Backhaul
  3. Last-Mile
Starting with step 1, I have discovered so far that data centers in other parts of the country are apparently more proud of their bandwidth than we are in the Southwest. Prices in Phoenix range from $1-$2 per 1Mbps depending on the pipe size. My first quote for one particular data center started at $11-$14 per 1Mbps. It took me 14 pages of reading to find those numbers behind a very professional looking proposal format. I would rather the company save the money used to develop this great looking proposal and get the pricing down to $3 per Mbps or less. I’m buying bandwidth, not proposal writing templates. I know of WISPs that are paying $80 per Mbps in Canada due to their monopolistic structure and who are probably wishing they could get these prices but hey, that’s what capitalism is all about. $11-$14 isn’t going to fly with this model so I will be back on the phone tomorrow wondering why it’s taking data centers a week to respond to my contact us forms. Fortunately there are several more options and some of the verbal quotes I am receiving now (backed up by what I’m sure are more very lengthy professional looking proposals) are getting down into the $6 per Mbps region.
Based on our 12-1 ratio, 5Mbps minimum, and the expectation that we are going to have 2400 clients deployed in the first year, our required bandwidth at the data center is about 1Gbps. What a coincidence that number happens to coincide with the bandwidth numbers of 80GHz and licensed radio systems. Here is our first decision point based on terrain. If our backhaul distance is 20 miles, we have to decide if we want three to four 80Ghz hops (I’m looking at Bridgewave for this model). However, redundancy for this model gets expensive with this many hops. The second option if the terrain permits it, would be to shoot the entire distance with a licensed frequency radio from Dragonwave or Bridgewave. I haven’t looked at the path calculations or the terrain maps yet but I’m guessing Dragonwave can do a 20 mile shot at 1Gbps at lower frequencies. Not sure what I can do at 18GHz yet but I’m running the numbers now. If not, I know they can do 2 hops for certain. Dragonwave can also go out to 2Gbps or more with compression which makes it a frontrunner in this project if I choose this model. I’ll have the backhaul decision made before the next article, along with a redundancy plan.
Although the primary goal of this project is based on good old capitalistic motives, if there is a chance of doing the right thing simultaneously, so much the better. Internet isn’t a taxpayer right any more than happiness or health care, but as Americans, we need to figure out how to make this happen for the kids. What’s funny is that budget cuts across the country aren’t the worst thing that can happen since that means inefficient government programs will simply not have the funding. Simultaneously, investors sitting on the sidelines might start looking hard at investing into a market that not only has a huge profit potential, but the opportunity to move the country forward again in the area of technology. I’m very grateful to the economic council that opened my eyes to the situation in their area of the country and applaud their efforts in helping me make this happen. Hopefully many of you in the Wireless Industry that have insight into the same issues facing students in your areas will now look at ways to break down the Digital Divide wall also with other innovative ideas that we can share with the industry.

Thursday, December 23, 2010

Chapter 20 – Can’t we just all get along!


I think it’s time to set the municipal alarm and wake the industry back up. Between Verizon and Sprint, I coiuldn’t figure out which one has the worst marketing department. I do know these companies keep supporting more reasons to push WiFi. The technical divisions of those companies finally deploy high-speed gun and the marketing departments promptly shoot themselves in the foot with it by either limiting users to 5GB per month or throttling to 256Kbps which is slower than EV-DO. I can’t even in good conscious beat up AT&T since everyone else, including Consumer Reports, is already doing that. I think it’s more of an admission that 4G is being overhyped and oversold. Apparently getting bandwidth to the tower is a lot harder than getting bandwidth from the tower to the user. It doesn’t matter how much horsepower the car has if you can’t get gas in the engine. I think I will keep my unlimited bandwidth EV-DO card a few more months until Clearwire shows up in our city with an investor in tow. Right now I’m not even sure that a 12-month contract has any value with them.
The original idea behind mesh was municipal deployments that were to return profits back to investors. Earthlink and MetroFi weren’t pouring money into city after city for charitable reasons although they used different revenue sources. Earthlink wanted the users to pay and MetroFi wanted, well, actually I don’t know who MetroFi was looking to pay the bills for them. Apparently neither did they although I’m sure the spreadsheets looked good to the investors. The only decent revenue they got was when Microsoft kicked in a bunch of money to be their partner. Unfortunately the RF engineers were left out of the meeting between investors and the IT department. Usually it’s because everyone assumes that the engineer will tell the truth, but we will get to xG technologies later.
Sadly, the municipal wireless industry as a whole has resigned itself to the government or the utility industry as being the only clients. I applaud the idea that government is trying to become more efficient as those two words in the same sentence are typically an oxymoron. Unfortunately, the idea of municipal wireless as another last-mile pipe to the internet is now the exception instead of the rule. It seems that no investor is willing to take on a municipal deployment as a straight investment. Even the WISPS that are successful hide in the nooks and crannies of remote areas where wired service can’t compete. These areas are slowing drying up in the U.S. as an area population increases, which means wired services eventually come into the area. I have seen reports where this is the only areas that can be made profitable for wireless. In other countries, wireless is highly preferable where not only is there little wired infrastructure in place, the theft of wire is an even bigger problem.
Most of the hardware manufacturers aren’t helping things by promoting the profitability of municipal wireless. The problem, and I completely understand it, is that municipal deployments don’t use enough hardware to justify a low cost. For example, if our 20 mile area had 32 APs per square mile (I’m jacking up the number simply for effect), then the total number of radios is 640. If those radios cost $100 per unit as I have proposed, the total sale is a pathetic $64,000 by mesh radio standards. Consider a 50% gross profit and the manufacture only made $32,000. To make this idea even close to palatable, the manufacturer needs to sell about 5 cities per month which isn’t going to happen, especially with the cost of sales. Jump the price of the APs to $1000, and the project cost just went to $640,000 with the manufacturer making about $320,000. Do six of these a year, and you have the start of a profitable operation. Take it to another level from $2,000-$12,000 and the numbers are staggering. This is why mesh radios cost so much and why there is no motivation for the mesh vendors to reduce the price of their products. As long as government is around to buy their equipment, these products are not going to get cheaper. In reality, I can’t blame these companies as I’m all about profit and if this provides them the highest profit potential, so be it. This just means there are market holes. This is also why my focus on profitability doesn’t include most of these products any longer.
Several companies have the capability to embrace this potential market quickly while the window is still open and can turn a very good profit if approached correctly. For example, SkyPilot ported their firmware to use lower cost CPE devices from Ubiquiti. The goal was the firmware licensing generated more revenue. That was a phenomenal idea and I was ready to jump on the bandwagon since their firmware is extremely solid. Unfortunately, they tied it to their legacy 12Mbps AP product that greatly limited total throughput and increased the cost 50-100% per square mile. 12Mbps just doesn’t work for me any longer with the advent of HD video and video streaming now the norm.
Mesh Dynamics went even further. Mesh could use the Bullet radios as APs for example. This prompted me to use them in a local bid situation. It was a great move except they haven’t moved further into the 802.11N products yet. At this point, it’s almost necessary to have an 802.11N product in this market for several reasons. Hopefully if Mesh Dynamics or SkyPilot don’t get there, maybe somebody else will step up.
Smart Grids are also coming along at just the right time for the market to see resurgence. A smart grid is a small piece of a large pie. If a city is spending $100-$300 per house for a wireless meter and the city has 20,000 homes, then $100K per square mile isn’t a big issue. Especially since the revenue per house is somewhere around $30-$100 per month. Utilities are looking at a 10 year ROI and $2,000,000 spread out over 20,000 homes for 10 years works out to less than $1 per month per household. Don’t think that the Smart Grid operators are going to let you watch the NFL package on your phone over WiFi. Security issues and absolutely no financial motivation to open that network up to public WiFi kills that idea before it starts. I think Burbank may have that option, but it’s going to take a lot of work to get many cities to open that up. Stuxnet also has a lot of people running scared right now in the SCADA industry unless you are a centrifuge in Iran. In that case, you are running like a cheap Rolex knockoff, way faster than you should.
So where does that leave those of us preaching that wireless technologies can compete with wireline services? Well, we had to first figure out how to get the Capex and operating expense down. Then we had to figure out how to get the performance up. As an engineer and business owner, I used “Tales from the Towers” to demonstrate both those solutions. However, those were technical and financial. The reality is still that either the city or the power companies own the vertical assets and you either work with them or you use a relay methodology off of individual clients.
So is it possible to get two systems, public and private, to exist in the same town? That’s going to be difficult at best, although not impossible. It’s better if the private and public industry work together but because of what happened in the early days of municipal mesh, that’s going to be like convincing the Democrats not to raise taxes. This is especially true if private industry tries to get the municipality to pay for some of the infrastructure. So let’s try and figure out how to make public and private work together.
Wireless frequencies are a fixed asset. With a 20MHz wide channel we only have 7-8 useable 20Mhz frequencies in 2.4GHz and 5.8GHz. That means for 360 degree coverage in 2.4GHz which usually only uses 3 channels, we need three 120 degree sector antennas. However, let’s say we want six 60 degree antennas instead to reduce the noise and increase gain and we don’t have a beam-forming option yet. This is the methodology that Motorola uses. That means 2 radios have to be on the same channel. If these antennas are on a pole, two APs on the same frequency are going to interfere with each other. This is where GPS synchronization pulls a Superman act and saves the day.
GPS synchronization gets the AP to talk to satellites, defines a timed reference point, and then unlike Congress, gets all the APs to work together for the common good. The common good in this case is to make sure that all radios transmit and receive at exactly the same time. The end result is that two radios that are within a few feet of each other on the same pole won’t be transmitting and receiving at different times. This allows frequency reuse without cutting the throughput. Theoretically, a pole could have 10 APs on the same channel, each with 36 degree sectors. Okay, I’m getting into beam-forming at that level but there is a difference.
Beam-forming is one processor/antenna defining several directional signals to create PTP connections with clients. GPS synchronization allows several APs, thus multiplying total throughput, on a pole and then use whatever antennas are needed for the application. If mixing sector antennas and directional antennas is appropriate, then GPS synchronization allows that without causing interference between the APs.
Where I’m going with this is it’s now possible to assign one channel to one party and a second channel to a different party and neither party really gets hurts by the redundancy. That would be as opposed to a MAD (Mutually Assured Destruction) methodology which is where Motorola and Ubiquiti WISPs are heading now. GPS synchronization is a proprietary technology which means Ubi and Moto aren’t going to be working together. I don’t see either company going out of their way to make that happen, so plan accordingly.
Two more quick things before I have to get some real work done today. Let’s start with xG technology which is a fascinating study in how investors put money into technology that clearly violated accepted rules of wireless. It’s the wireless version of snake oil. The first clue is that they wouldn’t release technical data. In this case however, it was the engineers who now have violated the old wives tale that engineers never lie. The fantasy that was xMax has now gone the way of Santa Claus (you knew I had to get a holiday reference in there somewhere) and we now know it was never real, regardless of how many baloney patents were filed. The reality is that xMax is simply a narrow-band 900Mhz frequency hopping radio. Oh wait, that’s what it is today since the original concept never worked to begin with, or did it? It got investor money and market cap into the 1.5 billion dollar area. xG finally released a product which is a generation behind 802.11n protocol and several times slower. I’m just ticked at the engineers who have now damaged the profession. The stockholders will handle the legal stuff with the company. What’s really sad is the U.S. military bought this bogus product which provides more legitimacy and rewards the concept of bait and switch.
How about simply modifying a 900MHz 802.11N AP with a proprietary frequency hopping firmware for oh, I don’t know, for $200 per AP. Wait, that product exists. Yes, the hopping rate is different but that’s a simple firmware modification. Hmmm, $200 versus the cost of an xG base station. That would be as opposed to the original price point of the xMax base station of $350,000. Wow, I have heard of techonology getting cheaper but…. BTW, the cognitive feature could be handled simply with a second or even a third AP sniffing which drives the price to $400.
The last is the mobility component. Several vendors have mobility as part of their product line. BelAir, Tropos, etc… have fast handoff capability for moving vehicles. To accomplish this, they developed proprietary extensions to the 802.11 communications protocol. I’ve been working on Guerilla WiFi accomplishing the same thing. I’ve proposed using a Peplink Max Mobile router which has built-in WiFi. It allows up to 3 paths to create a VPN tunnel back to a central location. You have to use a Peplink router on the other end but it works pretty well. I’ve tested it with EV-DO and GE-MDS Mercury radios for my dual-paths. You can even add the internal 2.4GHz b/g WiFi card for 3 paths if it can find an AP to connect to.
However, the cost of this setup starts heading north of $10K per car. I wanted Guerilla WiFi to have the same capability at a much lower cost. I’ve come close with some caveats. Peplink makes the CarFi which works pretty well at slow speeds (sub 25) or when parked. It takes 3-5 pings to reconnect or about 5 seconds on average. Not bad but still slow. Using the same radio on both sides such as the Ubiquiti Bullet takes slightly longer. However, the solution is forthcoming which means Guerilla WiFi should have a 50ms handoff pretty soon. I’ll keep you updated after I get to test further. As it stands today, if police can forgo connectivity above 25mph, then a $10K per square mile system and a $600 cost in the car gives them 100% outdoor coverage with 1-60Mbps of bandwidth.
Check out the new sections on Muniwireless that focus on smart phones and tablets. We will cover some recent testing there.

Tuesday, November 30, 2010

Chapter 19 – Catch a Wave-Guide and You are Sitting on Top of the World


The Beach Boys are going to hate me for this but I’ve been waiting for years to use that line. I also wanted to title it, “Look Ma, no mesh” but I should have used that one several articles ago. It’s not that I have anything against mesh as there is an application for almost every technology. At this point in the industry and the economy, however, it’s time to get past a word very few non-technical people understand and the excessive associated cost of it. Anyway, this article is about wave-guide antennas so let’s get back to that.
Vivato was a wave-guide based antenna. I have had Securawave wave-guide antennas installed for about 7 years. I became a believer when I connected a car at 2 miles and my laptop inside a Jack-in-the-Box at 1 mile. The horizontal polarity was a huge advantage since most wireless APs were vertical polarity. Securawave is no longer in business so I’m hoarding the last few units I have to support units I have in the field. Not that solid aluminum blocks have a tendency to fail as I suspect they will last longer than the buildings they are mounted to (they just don’t make them like this anymore), but in case of physical damage. I don’t need dual-polarity and 2x2 MIMO yet in these areas because Qwest hasn’t figured out how to get more then 3Mbps over the so-called HD Internet services and I’m not ready to mortgage my house for their other service offerings. Of course it’s hard to dial the phone when your eyes are tearing up from laughter when they advertise their new 40Mbps service. 18 months ago they couldn’t even keep a 640Kbps DSL line running properly in the middle of Phoenix less than 1 mile from Sky Harbor Airport.
Historically, wave guide antennas were expensive to make which is probably why it didn’t have more popularity. Ubiquiti seems to be bringing it back with its new omni-directional dual-polarity antennas. We know that dual-polarity has better penetrating and range capability than single polarity. So the idea of extending that into an omni-directional antenna seems like a great idea. Since there isn’t any other 2x2 MIMO omni-directional dual-polarity antennas that I know about, this is really cool from a technical, design, and financial standpoint. It’s also how we are going to keep Guerilla WiFi cost-effective and make it better.
The new Ubiquiti omni-directional antenna isn’t a true wave-guide antenna. Since it needs both polarities, half the antenna is basically two 180 degree vertical polarity sector antennas back to back with dual 180 degree wave guide antennas. Since it’s not out yet, I haven’t tested the unit but the pre-spec guess is it’s around 12-13dbi. That means the effective LOS range with dual polarity is going to be about 30% farther than an omni with 15dBi of gain for a couple of reasons. However, real world performance is going to be significantly better since dual-polarity will definitely penetrate vegetation better, reduce noise off-polarity, and reduce fading. If the client is using a dual-polarity indoor radio, then not only will range be better, noise will be significantly reduced even further. I smell a huge performance improvement in the air for Guerilla WiFi.
Let’s go back to Chapter 1 of Tales where we designed a $10K per square mile system. In that design, we used a 15dBi omni-directional antenna with a single polarity omni-directional. This design was using a single stream 802.11 b/g/n design. With the new dual-polarity omni-directional antenna, we still use a single radio for our AP but we have doubled the throughput with a 2x2 MIMO stream. In addition, we have doubled the throughput of every hop and added an additional hop. Even at the 4rd hop we are still delivering up to 10Mbps. Keep in mind that at a 10-1 oversell rate, that means that you can sell twenty clients 5Mbps at the end of the chain. Although pricing hasn’t been released, I’m pretty sure this antenna with a Rocket M2 radio will still cost less than $300. Double the performance of the original Guerilla WiFi at no additional cost and it’s better than triple coupon day at my local grocery.
Think about that for a moment. If we only had one egress point for our network, our base network was limited to 3 hops with the same capacity at the end. This single antenna, which allows us to change from a single stream 802.11b/g/n radio to a 2x2 AP, doubles that throughput, thus extending our single AP model out even further and doubling bandwidth down the chain. Our egress point can also use multiple radios which can triple the throughput to 3 times that for $300-$600 more without even getting into out next topic, GPS Synchronization. So for less than $11K per square mile, the system now supports 180-300Mbps, depending on the clients. So if your town is 20 square mile and this whole system cost $250K to put in, then it’s a no-brainer simply for cameras, security, mobile access, and department efficiency. Of course, if it goes through federal funding, has government engineers add in the fact that it has to support mesh (pointless in this and most designs but adds significant costs), throws in ridiculous temperature requirements like -30C for Phoenix or 80 degrees centigrade when -75 degrees centigrade would work fine, and requires copious amounts of paperwork to tell 14 different government agencies that you are paying wages equal to union scale even though you pay your guys more than that, then it’s going to cost $1,000,000 (my English teachers just had heart attacks over that sentence). If you can sense my frustration with federal rules that require union contractor shops to do work that is clearly better suited to IT companies, then you are very astute. This is why projects involving the government cost so much and take so long.
If this is a for profit network, $250,000 to cover 20 square miles with this level of bandwidth is pretty impressive. It’s fairly easy and cheap to add 100Mbps backhaul to each square mile for a total of 2GBps for the entire system. Realistically, I would guess that you would get about 500Mbps before the price starts going up for full-duplex links. Keep in mind that we are back to the core idea of an inexpensive municipal deployment.
Using numbers from previous articles, let’s assume 800 potential clients per square mile. If we get 10% of that base at $30 per month, that’s $48,000 per month. With this type of system, only a very small percentage of clients are going to need truck rolls. Total revenue on this network is almost $600K per year. If users have to get client radios, then they start at $30 for 802.11n 1x1 Vertical Polarity CPE’s. For about $80, you can include a window mount and get 2x2 MIMO. In most environments, I would be surprised if truck rolls needed to be done on more than 10% of the clients. Even if you add in the cost of the client radios, this system should be cash flow positive at 800 clients and should pay for itself at 2400 clients within 12 months. This just covers residential and doesn’t even get into business revenue. The numbers are now speaking to me so it’s time to kick the venture capital market back into high gear.
Also consider this, the 16 AP per square mile strategy now covers a higher percentage of deployments, especially profit oriented ones. If you live in the middle of high-density city, then you would want more APs per square mile for density or you fall back on the super AP concept described in previous articles. Since WiFi is far more unpredictable that PTP RF modeling, there is nothing wrong with installing 16 AP’s and then site surveying to see if there are coverage gaps than hinder more revenue.
Cell phone companies don’t cover every square inch of every house on the planet. It’s not cost effective. They deploy with the best models they have and then decide if it’s profitable after field testing to fix poor signal areas. WiFi should be deployed the same way. Put up 16 AP’s, field test, check the areas with poor coverage, and then decide if it’s worth another $300 in equipment to cover that area. If an area has higher usage, add in a triple radio AP upgrade for a few hundred dollars more. If an area really just needs more signal gain, look at adding a beam-forming AP just for a specific direction. There are many options but all of them would be based on sound profitability principals. There is nothing wrong with walking away from 2 customers that might cost $3000 to add additional infrastructure to cover.
I have 2 “Peeves of week” I have to get off my chest. The second is 99.999% uptime. I was asked to design a system where I have to guarantee the CPE’s have 99.999% uptime. Since the wireless industry has new equipment out every few months, very little of what I would deploy today has enough history for me to put my reputation behind that request. I’m not talking about PTP full-duplex $10K and put links but $50-$400 CPE’s. First off, anything less than 802.11N is too old and too slow for this type of video application. Second, anything that’s 802.11N hasn’t been around long enough to know how it’s going to run 3 years from now. It’s kind of a catch-22 situation. That means using cameras with built-in recorders that are going to cost three times as much or more than using lower priced cameras with CPE’s. Unfortunately there is no product history out there than guarantees that. The reality is that all the radios I work with rarely, if ever, just simply go offline if installed correctly. That doesn’t mean that they aren’t going to down for planned firmware upgrades or other system changes. However, the 99.999% uptime request didn’t stipulate whether planned maintenance was included.
I’ve already mentioned my first pet peeve, municipal bids that ask for mesh when every AP connected has a directional antenna. To everyone who keeps adding this expensive request into these bids, it’s a waste of money taxpayers’ money and costs the city a lot more to support in the long run. When you add a directional antenna to an AP or CPE, it’s a PTP or PTMP design, regardless of what firmware is on there. There is no mesh because the radio can’t connect to anything it’s not pointed at. By adding the mesh requirement, you are either getting the most expensive product out there or White Box APs with custom open-source mesh firmware that is cheaper. I’m not saying there isn’t a place for mesh, but don’t eliminate other options like WDS. It’s not your money you are spending; it’s ours, the taxpayers. Let the industry and a wireless engineer decide what the best product for the design is. It shouldn’t be the salesperson that took you out to lunch last week and has shiny brochures. If mesh is appropriate and the best fit, let the companies bidding on the project put that down. If WDS will work just as well or a PTMP design is better, they will bid that. So can anyone guessed what crossed my desk again this week? I know I’m beating my head against the wall on this since arguing with government is like trying to tell a 3 year old that candy isn’t good for them.
So we have come full circle on Guerilla WiFi from $10K per square mile to $50K per square mile and now back to $10K per square mile with double the performance. If this doesn’t kick the industry back into high gear, I’m not sure what will. Next we will cover GPS sync and how that affects deployments strategies.

Wednesday, November 24, 2010

Chapter 18 - Demystifying Beam-Forming


Based on some of the emails I’m getting, the biggest complaint is not enough details and without them, it’s difficult to implement the idea. Fair enough, I’m definitely guilty of that. In my defense, please understand that I want to have a personal life. Not that I have one now, as most WISP’s can attest, but if I had to put down every single detail on every single project or idea, I would be writing from here to Kingdom come. Except for the 6 guys on the planet like me that think they ought to make a movie out of every article in EETimes (I think a lower noise figure on a 741 OP-Amp makes a compelling plot line), most of you would be cancelling your Ambien prescriptions. The reality is that the details only matter to technical people and those are called White Papers. I’ve written a couple but I can’t sit still long enough to finish dinner, let alone do a 20 page technical document. My wife says that if I had put as much time into my homework as I have these articles, I would have my Master Degree by now. Of course, if University of Phoenix gave me real life credit for standing in a man-lift hanging antennas 80 feet in the air, I could get my Doctorate without ever attending class.
Let’s get back to the really cool stuff like beam-forming. We covered 900MHz and both the positive and negative aspects. Let’s move on to the category of what’s old is now new again, beam-forming. Vivato pioneered the idea several years ago but the financial model, coupled with irresponsible pre-marketing, didn’t really work in the real world. However, they did get the FCC to change the rules which is going to be a very good thing moving forward. The difference is how the different implementations of beam-forming are implemented and the effect in the real world, technically and financially.
Long before Vivato, the Super Scanner antenna from Antenna specialist tried to get an antenna to function in both a directional and omni-directional mode with the turn of a knob. Some of the manufacturers such as Wavion and Netronics sort of copied this model by using a ring of vertical omni-directional antennas. They can get up to 12dBi of antenna gain off of 7-9dBi antennas. Currently these units are limited to 802.11b/g bands and vertical polarity.
The next kind of design was done by SkyPilot where they simply rotated connection to one of 8 integrated sector antennas at any one time. This gave them up to 44dBm of EIRP in the 4.9-6.0 GHz bands. Ruckus, Netgear (with Ruckus helping), Cisco, and others are now integrating the antennas on circuit boards but at a lower gain. There will be several antennas cut into the boards or some other internal antenna hardware to add vertical and/or dual-polarity to the board that provide directional coverage. These designs are achieving up to 10dBi in gain with the bigger advantage being the noise reduction from other directions. They use the model of picking the 2 best signal antennas to receive the signal simultaneously since doubling antenna capture area provides up to a 3dB increase.
Vivato is still the technical King here in this realm though, although they weren’t covering 360 degrees. Vivato achieved 24dBi of gain in a 100 degree pattern in 2.4GHz. The catch is they did it for $8,000-$15,000, which meant I could either buy a new van for my company or 2 APs. That sort of limited the market for the product to the point that they are no longer in business, which is also what happens when you don’t listen to your vendors trying to sell the equipment. However, that 24dBi of gain is a huge advantage for reaching underpowered smartphones. Since Ubiquiti has now announced that they are releasing a beam-forming product over the next 2-4 months, the question is, does that have value for a municipal design?
Getting back to today, Wavion’s newest product, the WBS-2400 is probably the closest to the Ubiquiti beam-forming radio. It’s also very similar to Altai A8-Ei product with multiple radios, sectored antennas, and one enclosure. Both products tighten up the beam and assign a single radio per antenna. Unfortunately, both these products are still b/g.
Going back to the formula that distance doubles for every 6dBi of gain, if a 2.4GHz beam-forming AP has 21dBi of gain, or what I’m guessing the new Ubiquiti units are going to be, that almost quadruples the range of most omni-directional AP’s. Throw in the fact that noise will be reduced by at least -20dB, and the s/n ratio goes berserk. The effective range increase, and this is only my best estimate in the worst scenarios, is probably a factor of 8-16 in open air environments. This means hitting a smartphone at 3000-4000 feet or more is possible.
Here is where it gets interesting though. Based on past experience, I’m guessing the Ubiquiti beam-forming AP will be less than $300. That is 10 times less than other beam-forming products, none of which support 802.11N 2x2 MIMO. Heck, it’s less than I paid for a baseball bat last season.
We were all trained to think in terms of APs needing to cover 360 degrees for municipal deployments. Every product out there is designed for that. The end result of this is when the density reaches a certain level, the self-interference is basically self-defeating. I’ve seen proposals that suggest 60 APs per square mile for example. Central management systems can dynamically adjust power output between APs to reduce the interference and can simultaneously adjust power output on a per packet basis to reduce the interference even further. However, all this costs which effectively drives the price up to $150,000 per square mile or more for this level of performance. However, what in the playbook says that every AP has to have omni-directional coverage?
Throughout this series, we’ve been driving towards a low-cost, high-capacity design that can be commercially deployed. The original idea was trying to develop an inexpensive Super AP. I’ve gone from $200 to $2000 and was working on an idea that would have driven the cost to a retail level of $4000. It supported 1Gbps per AP and 700+ simultaneous clients, but still it got away from my original idea which was to lower the bar on the per mile cost, which it didn’t accomplish. It basically comes down to the following, are you trying to compete with cable which can deliver up to 50Mbps or delivering Internet to people who have to suffer with dial-up, or worse, DSL delivered by Qwest in Arizona (okay, bad joke).
Beam-forming lets us start with a new playbook. The first thing you do in a playbook is define your assets (okay, I never played organized football so I’m making this part up.). Using the Ubiquiti beam-forming product (2.4GHz version), we know roughly that we will have about 21dBi of gain, 90 degree coverage zone, 16 degree beam pattern, GPS synchronization, and 2x2 dual-polarity 802.11N MIMO. Since we can now hit a phone at 2500 feet without much of a problem in open air, the design mentality changes. Even with 21dBm of gain, we still aren’t going to penetrate several buildings. However, we probably bought ourselves another house or two. By changing the angle of transmission, ala cell towers, and to be fair to Vivato’s deployment design in Spokane, tops of buildings aiming down, and all of a sudden we are hitting several houses down the street. Throw in 2x2 MIMO dual-polarity, and the connection distance goes way up.
Because of the price point, I’m guessing Ubiquiti’s product uses the same type of design as SkyPilot in terms of switching between multiple feed points on the antenna. The difference is that each feed generates a different angle of radiation versus being a completely separate physical antenna. Ruckus hits multiple feeds also with circuit board based antennas.
Although the tops of buildings aren’t ideal for sector antennas due to amount of AP’s that are going to get picked up, if you aim it towards the horizon, angle it down 60 degrees or more, you still get coverage down the street and pretty impressive building coverage. At that angle, you are reducing attenuation from outer wall penetration and even taking the route of shooting through the roof. This doesn’t work all that well in areas with trees, but if the buildings are made out of concrete, brick, or stucco, then this might be the best option. There are a lot of cities where the street lights are decorative and there is no way to install an AP down at ground level, so secondary options are a must. Beam-forming makes this design more useable by effectively reducing the beam pattern down way down.
This model might work in a different way also. Think in terms of 3 dimensions instead of 2. Larger buildings or apartments in downtown areas might be 10 floors or more. Placing these antennas on the side of one building several stories up and aiming into another building let’s you build a vertical WiFi Although attenuation has been a killer, 21dB of gain makes up for a lot of wall and window attenuation. In addition, dual polarity will improve penetration even further. If you pick up 30 customers in a building at $20 per month, 3-4 antennas pointing into the building get paid for in a couple of months. I would have also said it’s significantly cheaper than AP’s in the building until I saw another product that I’m testing (to be discussed later). However, when you add in the savings on pulling cable, it’s a no-brainer.
Irregular areas with streets curving all over the place are another option for this type of design. Intermixed homes and small business areas with mostly 1-2 story homes and 2-5 story buildings work well for this design. Outdoor coverage is universal with indoor coverage hitting around 80% or more with careful planning. Keep in mind I’m assuming everyone wants to give you free vertical assets to enhance the system.
Bringing us back to earth gives us most of the real-world deployments, street lights. This is where the GPS synchronization really kicks in and is now the direct comparison to Wavion or Ruckus. Placement of four of these radios on a pole to cover 360 degrees provides the absolute farthest range of any AP on the market. This is only possible with GPS synchronization to keep them from interfering with each other. It effectively provides twenty four 16 degree beams shooting out in 360 degree pattern. The cost of this super AP will be about $1500 which isn’t a lot less than Wavion or other beam-forming 360 degree APs. It does have significantly longer range through, by a factor of 9dBi and dual-polarity which is worth another 3dB in my field testing. With the IPad supporting 802.11n and the Beam-Forming antenna supporting dual-polarity, this would be a potent combination for off-loading some bandwidth.
If we duplicate this design with Altai or Ruckus, the cost is now over $10K per AP. There are other manufacturers such as Motorola that use sector antennas but then the antenna gain drops back to 12dBi or less and their beam patterns are 120 degrees. The best way to reduce noise is to decrease the beam pattern which it also doesn’t do. Effectively we get a 9dBi increase and 20+dB of noise reduction for a cumulative 29dBm improvement in s/n ratio. Regardless of what wireless philosophy you follow, that’s huge.
Here is one of my philosophies though, “If you spend too much, all you have lost is a little money. If you spend too little, you could lose everything.” Guerilla WiFi is based on the concept of being the most cost-effective way to deploy and what I feel is the only way to make municipal wireless a self-sustaining profitable model. That doesn’t mean it’s the best way in every situation. I’m describing equipment that isn’t even out on the market yet and isn’t field tested. Wavion, Ruckus, Altai, Bel-Air, Motorola, FireTide, Cisco, Enterasys, and a host of other vendors are proven products that have been deployed and tested. I do think these vendors are making it difficult to meet the financial needs of a profitable municipality model which is where Guerilla WiFi comes in, but the products are solid. This models works for me and others who are willing to spend the time to work through early firmware issues and fine-tune the models during deployment. It’s kind of like the difference between Linux and Windows. Since Ubiquiti uses a common firmware platform across all of their outdoor products and has made great strides towards stability and features, I expect the new products to use the same firmware base which means , the firmware. However, there is still a long way to go towards a turn-key municipal deployment without a lot of engineering intervention. Based on the cost/performance ratio, I believe it’s worth it. But realistically I spend a lot of time testing and adjusting. Most customers are going to expect a more turn-key solution. However, if it’s your money, then this solution provides the best chance of financial success.
With this AP design, $1500 per AP is getting pretty pricey. However, it will also combine with our next article, the 360 degree dual-polarity 2x2 MIMO AP for under what I’m guessing is going to be $300. I’ll try not to wait 3 weeks to make this happen.
One other note that I want to mention is that Mike Ford is no longer at Ubiquiti. Mike was a central figure that seemed to have his hands in everything from testing to customer support. Over the last 3 years, he has become a friend that always went the extra mile to make sure I had the resources I needed for my clients. He was also instrumental in some of my decisions on my deployments. Although I don’t know his future plans, I am very sorry to see him go.

Wednesday, November 3, 2010

Chapter 17 - Who needs White Space?


It’s time to step up our game. There is no problem generating massive bandwidth from an AP location. We have proven that fact. What we haven’t figured out yet is how to leap tall trees in a single bound or walk through brick walls. If you are willing to add in another 20Mhz of super-powerful, wall-penetrating, obstruction busting, tree smashing signal, then we have solved the problem.
You are thinking I’m going to jump on the White Space bandwagon. I’m sure I will someday but there’s a lot of work that has to happen before that option is available. However, the infrastructure we have designed is pretty flexible and inexpensive. When White Space becomes feasible and cheap (my favorite word), we can add it. In the meantime, we have two other options. I think I will save the best for last though.
White Space is touted for 2 reasons, extended range and building penetration. It also has one big disadvantage. In major cities, there may not be a lot of channels available according to Spectrum Bridge, for White Space to operate. In addition, limited power output from clients is still going to limit high-bandwidth range back to the AP. I also see the 6MHz channels being an issue with bonding being the same problem as trying to run 40MHz outdoors.
Before we jump into this though, let me make note that Ubiquiti just released a stack of new products that are game changers in WiFi, indoor and outdoor, Video Surveillance, and cellular service. These technologies cover everything from Beam-Forming to GPS sync to dual-polarity omni’s. I’ve had to hold back due to NDA’s but we will now start covering how these technologies can be integrated into our Guerilla WiFi design to take the systems to a whole new level. Most of these products are several weeks away from shipping so we have time to develop our system. I do a have a pair of the 900MHz M series 802.11N 2x2 MIMO AP’s in my hands and that is the topic of this article.
Option one is 900MHz. Yes, it’s crowded, noisy, and seriously overused. So is a Japanese subway but people still use it because it’s the best option. Until now though, the best WISPs systems limited users to about 3Mbps under ideal conditions with APs limited to about 7Mbps. There really wasn’t a lot of development in that frequency band due to the interference and reduced band size as compared to all the other unlicensed options.
900MHz has close to the broadcast properties as White Space except for the vastly higher interference. In a municipal system with 16 AP’s per square mile, AP’s are within 600’ so range of everybody. Unfortunately, 2.4GHz can’t penetrate obstructions very well. Brick or Stucco buildings that will suck the life out of 2.4GHz are merely a few dB of loss to a 900MHz signal. At that range, trees effectively disappear. Junior’s frequency hopping baby monitor and the SCADA transmitter hanging on your water meter is now more of a problem than obstructions. In 900Mhz, with 802.11N 2x2 MIMO now being applied, bandwidth isn’t the issue any longer. The biggest problem is interference.
Everything has a threshold though and you just have to find it. With parents, it’s how low your grades can go before you become the prisoner of the bedroom Alcatraz and your friends start filing missing person’s reports. With RF, it’s the difference between the signal and the noise and how efficiently we use the bandwidth. Now toss in a very narrow band, 26MHz, and the total bandwidth throughput from any AP is going to be limited. Oh, yea, did I mention that we only have to go 600 feet?
802.11N isn’t just for 2.4GHz and 5.8GHz. It just hadn’t been applied before in the 900MHz band. Ubiquiti just released several new 900MHz products with 802.11N 2x2 MIMO protocol. Using a 5MHz channel will allow up to 4 APs to operate on one pole, which each one providing up to 20Mbps. However, using buildings and some shielding, might allow up to 4 channels of 10-20MHz in more remote areas to allow up to 300Mbps. I’m sort of guessing here but when I get one of the base station sector antennas in, I will do some testing to determine what would be needed for isolation.
Keep in mind through all of this that we are still dealing with an AP/sector antenna combination that cost less than $500. If budget is an issue and the city thinks that a several 4’ antennas on a pole aren’t their idea of aesthetically pleasing, then look at using 4 of the Nanostation M900 Loco’s. They are very small, and although rated at a 60 degree beam pattern, they can easily cover 90 degrees with a small drop off in antenna gain. In fact, the beam pattern for these radios is way over 90 degrees at 7.5dBi of gain.
The only problem here is that you need one of these radios in or on the house since there is no portable device that can support 900MHz. That starts getting expensive at $200 for the radio and another $50-$150 for an indoor WiFi device for wireless coverage. However, the problem of building penetration is completely solved.
If I haven’t mentioned it before, we only have to go 600’ from an AP location if we have 16 poles per square mile. Realistically though, if we have 16 APs per square mile, I would probably only use a maximum of 4 AP locations with the 900MHz radios for budget reasons. That means we might have to go 1300’. Of course, we are using a proprietary polling scheme and a dual-polarity signal to go that far. If the noise floor starts at -65, the signal level needs to be at -55 or better. At 1300’, even with obstructions, our signal level should easily exceed that.
Trilliant and other Smart Grid companies are releasing MOAB (Mother of All Bombs) 900MHz radios that are up to 1W for residential installations. If you think that a few towers can cause interference, try fighting tens of thousands of radios dropped into the middle of your coverage zones. Motorola 900MHz WISPS from here to Canada are getting hammered and there isn’t a lot they could do about it. There are going to be cities where running 900MHz WiFi may not be feasible. Don’t panic yet, we will take our 2.4GHz game up a notch also.
There are two advantages to the Ubiquiti 900MHz radios to fight interference. One is the dual-polarity MIMO design. In the city though, most radios are using antennas with such low gain, polarity isn’t going to make a lot of difference. However, the M900 product line came out simultaneously with the ability to frequency hop. That means you have 4, 5MHz channels to jump around with at 300ms rates to avoid noise try and punch a signal through. If the Smart Grid density is too high, then even that isn’t going to matter but right now it’s the best option available.
The second advantage is AirMax. AirMax will simply ignore other packets in the band and also eliminate the hidden node problem. Although interference is interference, AirMax AP’s won’t slow down acknowledging other APs in the band.
Ahh, but what works for city folk works even better for country folk. The dreaded trees of death for 2.4GHz and 5.8GHz are merely pin pricks to 900MHz. Toss in the dual-polarity 2x2 MIMO design and now signal will punch through the forests like Ray Lewis through an NFL helmet. Expand out the channel to 10 or even 20MHz, and throughput for a single AP could go as high as 80Mbps.
On the muni-wireless issue, we can assume that designating 4 AP sites per square mile will add about $1000 or about $4000 per square mile. It also adds 320Mbps of total capacity per square mile. Add in the CPE Capex of $200 per client, assume 25 clients need this radio to avoid a truck roll, and you have an additional $5000. If you truck roll, add another $3750 in the Capex column for those of you keeping track.
Based on those numbers with a $30 per month fee and a free install, it will take 17 months to recoup the Capex. Of course, we want to charge $100-$200 for an install to offset some of those costs. In areas where municipal staff thinks antennas are cool and interference is minimal, we could even use four 900MHz dual-polarity sector antennas with 13dBi of gain. That will provide over 6 times the coverage area which might reduce the AP locations from 4 to 2 but it will more than double the cost per AP, which is a wash. There will be scenarios where either option will be more appropriate.
Our second option is the new beam-forming radios. I’ll cover that in more detail in the future. Since the 2.4GHz versions of these units won’t be out for another 4 months or so, there is no hurry. However, they add another 4-6dBi of gain over a sector antenna and 8-9dB of antenna gain over any other beam-forming AP other than Vivato. Add in dual-polarity, which my field testing shows to be worth up to 3dBi more useable gain, and a 16 degree beam pattern to reduce noise (I’m extrapolating from the 5.8GHz beam-forming unit that was announced. The final specs may vary.) and that that’s enough gain to penetrate an extra wall or almost quadruple the coverage distance to a client. I’ll go into the difference between between all the beam-forming units on the market in our next article.
Now we have even more tools to play with for Guerilla WiFi. Theoretically, it wouldn’t be hard to build a 1Gbps AP to work across multiple frequencies with beam-forming and GPs for less than $4000. Taking this concept even further, it also wouldn’t be hard to create a load-balanced, business quality, multi-frequency design that could bond these frequencies together for very high-capacity throughput. There are other variations of this for backhaul, redundancy, and uptime. It’s possible, with a little networking work, to create a mission critical design capable of delivering tens or hundreds of MB’s to a CPE for less than $400 on the CPE side. This type of system could easily deliver 99.999% uptime using unlicensed frequencies, even with scheduled maintenance. And don’t get me started on 900MHz mobile options. The hits just keep on coming.