Chapter 14 - Money Talks

I have gotten calls from many people that want to start a WISP company, whether using a mesh/muni model or a PTMP system. Although I believe there is no better time than now, that doesn’t mean it’s easy but I hope to prove that it’s financially feasible everywhere. Competing against satellite or cellular services like EVDO, WiMax, or even LTE is a no-brainer in areas that have no wireline services. I’ll cover the new Sprint/Clearwire LTE service just announced for Phoenix in another article. I can build small, profitable financial models all day that can provide superior services even at subscriber/bandwidth models of 10-1. There are also other services like VoIP that can be provided but for the beginning of this analysis, we are going to focus on Internet services only. However, let’s take this into the professional corporate/investor level and see what happens.

Breaking this down into 3 areas and 2 types of design models to cover areas. These areas are a gross oversimplification because each area has different RF models that also have to be considered based on how RF friendly they are. However, for the sake of argument, let’s use these numbers:

1. 
   
   Rural – 20 or less potential subscribers per square mile
2. 
   
   Suburb – 600 potential subscribers per square mile (2700 people per square mile in Phoenix)
3. 
   
   City – 3,000 plus potential subscribers per square mile (Boston 12000 people per square mile)

The 2 basic types of models are PTMP and WiFi. So, let’s look at how each of these models can be deployed profitably. We covered inexpensive WiFi systems early on and I still stand by that model, but let’s analyze it a little more critically and see where it fits in the big picture. More important than anything, is it possible to build a system that has the financial strength to be a growing and profitable company versus Joe Technical’s weekend hobby?

As many WISPs have successfully demonstrated, clearly it’s possible to be profitable if you market in some rural areas. Assuming a starting company of 4 people, the revenue generated has to be around $40,000 per month to be profitable at the low end. I’m just summarizing some of the spreadsheets that I have used so you will have to take my word on the cost structure. With an average monthly rate of $30 per month per client, we calculate that it will take 1333 clients to make that type of revenue each month. By all accounts, that’s a pretty impressive size to jump right into. Keep in mind this doesn’t include your original Capex and how much investment you need to get to 1333 clients. It’s not going to be cheap.

If you are a hot-spot provider, you get revenue from hourly, daily, and weekly rates. Those markets are also shrinking in the U.S. for the most part but some new ideas for phone users are coming forth. In this discussion, we are going to focus on the monthly subscribers. There are also other revenue sources such as installation revenue, business rates that are higher than personal subscriber rates, and other normal ISP types of services. Additional services might require additional staff with more expertise, thus raising the associated monthly costs.

Rural markets are easy. Set up a PTMP system on whatever inexpensive vertical assets you can get access to. If there is not a lot of vegetation, these systems can handle36 square miles easily per tower and up to 300+ users per tower. This assumes you don’t already have competition in the area and there is no interference. You just need 7 towers with 200 users on each tower cover your monthly expenses and be profitable. If there are any of these areas left in the U.S., let me know but I’m not going to hold my breath.

However, there are some rural markets that are underserved simply because of distance, vegetation, and costs of deployment. They may not have 1300 clients, but find a few of them with a couple hundred clients per area and the plan still works. Some of these areas were just not reasonable to do with 2.4GHz or 5.8GHz. They may have been a good 900MHz radio option but limits in equipment, the band, and interference mean that most of the 900MHz products either only delivered 1.5Mbps or the cost of deployment was too high. Newer 802.11N 2x2 MIMO equipment that is hitting the market now should allow for an improvement in throughput in these areas with client bandwidth similar to anything available in the 2.4 and 5.8GHz bands and cost far less. Depending on the design, it should be reasonably easy to promise 10Mbps to a client on the wireless link depending on back end bandwidth. 900Mhz isn’t an ideal band due to limited spectrum width and interference in city or even suburban deployments. However, it extends out the range of clients in dense vegetation areas. In most rural deployments without too much interference, I would expect bandwidth off a centralized tower to max out around 160Mbps with the right setup in 900MHz.

Let’s move into the suburbs now. Most suburbs are served by cable, DSL, or a combination of both. To be honest, DSL is simply not living up to the hype of the marketing division. In Phoenix, Qwest advertises speeds up to 20Mbps when they are lucky to hit 3Mbps, even within 1 mile of Sky Harbor Airport in the middle of the city. Move 300 feet and not only can they barely deliver 640Kbps, they can’t keep it running more than a month or two before it to crashes again. In my case, after I complained for the umpteenth time, they told me with one more complaint they would pull out of our business complex leaving me with no wireline high-speed bandwidth. I had to move my office just to get 3Mbps after several years of subpar service. Even if you are satisfied with your DSL service, it only takes one bad technician adding one more client in your area to cause it slow down or crash again. If this is the main provider of service in your area, I say let the best technology win. I will take WiFi based wireless over any area where DSL is the only technology available. That alone means that major cities still have opportunities.

Cable is another matter. Cable companies are promising huge amounts of bandwidth today and for the most part are very technically stable. In Phoenix, I have a 20Mbps circuit. However, I get around 11Mbps on Speedtest or Speakeasy most of the time. Other times I have seen it down to 1Mbps. Even though you may pay for a specific level of bandwidth, there is no SLA that you will be delivered that level as opposed to a business level SLA agreement. The reality is that the bandwidth advertised is the burst speed or web browsing speed. If you download or try to move large files such as video or a file transfer, that speed will be reduced significantly. However, other than fiber to my house which I doubt I will see in my lifetime, it’s the usually the fastest option.

In either case, this isn’t a market where 1Mbps is going to fly. Even Grandma and Grandpa are watching NetFlix. Google TV is no longer an urban legend either. Throw in every game machine out there downloading movies and unless you are willing to run with the big dogs, this is not a battle ground you want to enter. I’m talking about a wireless service where you plan in advance on delivering 5Mbps average to everyone with peaks up to 50Mbps or more and no more than an 8-1 bandwidth to user ration. 802.11 b/g/a won’t fly here. You have to be ready to invest in bringing a large bandwidth pipe in and the subsequent costs of deploying 1333 users. I have numbers on both PTMP and Muni-WiFi models but to simplify this model and since the area is generic, let’s assume a 50-50 model.

Since we are estimating 600 households per square mile (houses and multi-dwelling unit) and assuming we get 20% of that market, we need about 10 square miles to meet our $40,000 revenue total. That also means we need about 4 towers or building assets to ensure we have LOS to all the locations. Estimating a cost of $25K for the backhaul (assuming the entire system is wireless) gives you 2Gbps. However, it might make more sense for a local fiber, MPLS, or even cable backhaul although using a company for your backhaul that you plan on competing against may not be the best idea. Unless you consider bringing an anti-trust lawsuit against multiple telcom providers while they drive you out of business a fun time, you probably want to find other local carriers. In the end, we have determined that the tower installations will cost about $125,000.

Since half the clients are WiFi, we are still going to have to install 16 APs or more per square mile. I’m going to use that number along with some numbers on designs I’m working on now to come up with $30,000K and 360Mbps per square mile. Each AP will deliver up to 100Mbps to start with and if you go back to previous articles, can deliver several hundred Mbps if needed. Since we have the vertical assets, backhaul to the street lights is taken care of. The problem here is that it still costs us $300,000K and is the largest part of the Capex. Additional revenue sources have to be found here to justify this but we will discuss this later.

Our numbers assume that half of the clients will require truck rolls. That will cost $195,000 to deploy. Considering that each technician can do about 4 installations a day, it will take 160 man days to install enough clients. Since there is only about 20 workdays per month, it will take 8 crews to get up to this number within a month. Most of us would have to outsource this unless you have several friends with days off who also happen to work as installers for Direct TV. The revenue on installations will range from $0 to $200 depending on the market. Assume $100 and the net cost of deployment is going to run about $60K.

The other 650 clients can be installed any time since they are WiFi based. Either they can connect directly or some type of CPE device can be made available. Let’s call that a no Capex cost since $50 will come pretty close to the actual cost.

To summarize, the Capex for this model costs $530,000 to deploy. Even if we net $10K per month which is reasonable, this isn’t going to fly financially with a 4.5-year ROI. This is also a maximum bandwidth system than can compete directly with Cable/DSL/LTE or anything else out there for the near future. Scaling it out further doesn’t improve the ROI but it sure increases the revenue stream. So, how can we make it more cost effective? There are two ways, reduce the cost or find more revenue.

Start with the idea that we only need 100Mbps per square mile in the beginning. This reduces tower costs down to $15K per square mile for a backhaul system that will support 720Mbps instead of 2Gbps. We just saved $75K. It also drops the per mile costs of the WiFi system to about $20K per square mile. That’s another $100K on the savings side.

There isn’t much that can be done on the truck rolls other than to consider the option of doing the installations in house. That saves about $50 per installation but it may take 3-6 months to get up to 650 installations. That’s also more reasonable in most models that I have developed. What I didn’t take into account is the advertising necessary for this speed of penetration but there are several cost-effective ways to do this. I’m leaving this number alone for that reason.

The cost is now down to $360K. The ROI is down to 3 years but is still not that reasonable except for this, cable companies are not reducing their rates. In fact, they keep going up along with the bandwidth. One of the reasons is that they are being squeezed on the TV side by content providers demanding more revenue per users. At the same time, it’s much more difficult to raise TV cable rates due to competition from satellite providers and local ordinances. People are also dropping land lines faster than a Bugatti Veyron because of cellular phones. As much as I complain about cruddy DSL service, the reality is that it has also taken many of the Internet clients from cable due to low cost, which helps to drive the prices down. You do get what you pay for however. Hey DSL companies, here is an idea. Instead of trying to bring fiber to the home which you clearly can’t cost justify, how about just bringing it down the street so my little modem doesn’t have to connect 3 miles away across some 20 year old wires. If anybody needed a wireless option, the DSL companies do and they have boxes on almost every street. Having those assets for a wireless design would be my wildest dreams but for some reason, they can’t get off the wired mentality.

There are other revenue sources for this that I have covered in previous articles. Think through some ideas with cell phones, hot spots, multi-dwelling buildings, backhaul, VoIP, other ISP and business services, and the ROI actually starts dropping to about 2 years or less. If you have an existing company that already has sales people, project managers, and office staff in place, then the costs of getting to this point is pretty reasonable. Peg that at about $50K. Starting this project from scratch probably means about $150K. That adds 6-18 months on the ROI. Scaling the system and deploying more slowly puts out the ROI but reduces the Capex as the system starts paying for itself in about a year. There are many ways to play with the numbers but the bottom line is this, it’s now possible to compete with wireline services in any market. I’m also basing my numbers on what equipment can be bought today. I am pretty sure tomorrow may bring many more surprises that will change the financial and technology foundation of WiFi and that tomorrow is far closer than we think.

Chapter 13 - Interference isnt a Hockey Penalty

Unlicensed frequencies mean that interference is a way of life in most major cities. The question is what to do when all the frequencies you plan on using or even radios that are in operation start having errors. I just came from an installation like that. In addition, then I was asked to design an expansion to the system. One of the clues was that interference is a problem is when things work for a couple weeks and packet errors in the log jump from nothing to tens of thousands.

By default, most of us set up our networks with 20MHz wide channels. That’s the default for 2.4GHz WiFi and usually the default for most 5.8GHz deployments. However, what happens when we do a site survey and 500 APs show up on the list? If the design is already deployed, your options are limited. You basically have three:

1. 
   
   Find the least interfered channel. I usually don’t hold much hope out for that.
2. 
   
   Increase power at the possible expense of reduced modulation rates – which should also be titled “How to make new friends”
3. 
   
   Reduce channel width down to 10 or even 5 MHz – not an option with WiFi hotspots
4. Play FCC Russian Roulette and and use channels you aren’t supposed to be on with way more power than are legally allowed in those bands anyway. It seems that for those of people violating FCC rules, they must figure if you are going to break one rule, might as well break two. I have even seen systems set up by consultants and utilized by police departments on these frequencies. If the FCC ever drove through towns with sniffers they way Google did with cameras, the fines alone would solve the National Debt.

If you control both sides of the radio equipment equation, a combination of these together might squeeze out better performance and minimize the interference with a little patience. Sometimes you just have to take the best option out of bad choices. For example, is it better to have a 20MHz wide channel with a few errors or a 10MHz wide channel with no errors? That same argument applies with a 20MHz wide channel with more power which might reduce modulation versus a 10MHz wide channel with lower power and higher modulation. Keep in mind that you are now running your design on the edge so that any new interference will probably degrade your system even further. On the other hand, you might be disrupting someone else enough that they move to another band.

Designing the system from the beginning provides more options than having to fix an existing system. Let’s get something out right now. If tomorrow you decided to deploy a PTMP system with wide angle sector antennas in an unlicensed frequency band in a major city, you are not going to be the first one in those bands. This means expect interference and plan for it. Use the first three ideas and let someone else who should be working at McDonalds instead of consulting in the wireless industry, use option 4. If you get bored, site survey the band and give the FCC a little heads up on what you found in the 5.0-5.6GHz bands.

The strategy you develop from here is going to depend on the type of application. If the application is WISP services where you are trying to cover a general area, good luck. I’m not saying it can’t be done, it’s just going to be about as easy as washing a cat in the tub. Throwing up 90 degree or more sector antennas or even worse, omni-directional antennas is simply inviting problems. Not only are you going to tick off everyone around you, they are going to return the favor out of pure self-preservation.

Let’s start with the idea that deploying anything with 802.11a/b/g today is also simply a waste of bandwidth. Go right for 802.11n and try to go for 2x2 MIMO. However, keep in mind that as hard as the standards body tried, 802.11n simply doesn’t work well with legacy devices. It’s not that they won’t function, it’s just the legacy devices slow down 802.11n radios and simultaneously 802.11n will interfere with 802.11a/b/g radios. Throw in some manufacturer proprietary settings and things get even more interesting. Other than the basic ideas above, is there a way to work in a high-interference environment?

Start by thinking of your English 101 class in college. Charlemagne said, “let my armies be the rocks and the trees and the birds in the sky”. Let’s change that to “let my shields be the trees and the buildings”. Even though RF engineers in 2.4GHz frequencies and above look at vegetation and buildings as the enemy, they can also be allies in a design. Instead of fighting RF interference in the open air 10 floors or more above the city, take the battle to the ground where the buildings and trees block interference. Taking an RF survey 300’ in the air is a whole lot different than taking one standing on a corner block surrounded by trees and buildings.

So now you must be thinking that we are back to metropolitan WiFi. Actually no. WiFi assumes 2.4GHz and even at ground level is probably pretty congested. However, there is an FCC rule called the 3-1 rule pushed through by Vivato with the FCC. We pretty much know the current omnidirectional rule for WiFi, 30dBm radios with 6dBi omni-antennas. However, the 3-1 rule means that in a PTP link we can increase the antenna gain by 3dB if we reduce the power output by 1dB. If we are going PTP at ground level in 2.4GHz where we know interference is going to be an issue, some really directional equipment might work pretty well. It shouldn’t be too hard to get a 52dBm signal to go 2 blocks with directional antennas 15’ off the ground. This can be achieved with a 30dBm antenna and a power output of 22dBm. Obviously I’m joking but the point I’m trying to make is that highly directional antennas at ground level in any band will be very effective, even in high-noise environments. Drop the channel width down to 5MHz with a 2x2 MIMO 802.11N radio and your interference issues will probably become a memory. Some poor schmuck in the middle of this link might get hammered however, especially if he is using a legacy device. Let your conscience be your guide here.

5.8GHz works even better in this environment if the trees aren’t in the way. Testing has shown that 2.4GHz 2x2 MIMO dual-polarity will punch through trees but 5.8GHz works about as well as a political candidate does once he gets elected. It might work or it might not. Either way it’s not reliable. The buildings will also kill 5.8GHz signals from above, leaving the ground wide open. There aren’t a lot of indoor APs using 5GHz bands yet but with up to 53dBm of signal output, interference will be a memory. In reality, you really don’t need or actually want 50+dBm of signal to go 2 blocks unless you are shooting through dense rainforest but the concept of using highly directional antennas at ground levels might solve some problems.

Taking that further, if you are using a proprietary polling scheme like Motorola or Ubiquiti or frequency hopping, the noise those APs will generate will be massive. Using this technique in a financially feasible manner, assuming you don’t have the government’s open pocketbook backing you, requires either a mesh type radio with directional antennas or a generic radio that can be either an AP or a CPE device, depending on settings. There are also many variations of this that can work quite well also.

I am finding that almost every environment that I’m going into is challenging me to come up with new variations of designs to be successful either technically or financially. The old hub and spoke model is much more difficult to deploy in a city today when there are thousands of radios already deployed with many of them not following the FCC rules in either frequency or power output. Rising or high internet costs coupled with lower equipment costs are now opening up new competitive markets for wireless that require new financial models that actually work.

This design is specifically oriented towards a fixed location model in a high interference environment with talk buildings. Dynamic location designs have to be done differently because you don’t have fixed CPE locations. We covered this in some of the previous articles. Omnidirectional antenna use in a municipal environment just doesn’t fly any more. In lower population density environments, things are different. In the middle of a major city however, use the environment to your advantage.

On a side note, it’s been pointed out to me that I should probably proofread my posts a little further. I notice a few “minor” typos that I miss in my haste to get these done. Unfortunately this isn’t my day job and my wife who clearly proofed the first few, just doesn’t have time to keep me from looking like I never took a technical writing class before. She also hates the first person style in a technical document and gave me the Rollie Eye every time she took out a “you” and “we”. So, I’m down to proofing myself and I’m hoping that content is more important than grammar. Proofing these takes twice as long as writing them. So, if you the reader are willing to let the grammatical errors slide by, I’ll try and get them out a little quicker.

"Rory Conaway" - Triad Wireless