Preparing for an unknown future
Recent stories on cable bandwidth have outlined how cable system operators today are looking at efficiency gains that will enable them to offer additional high-definition channels, increase data speeds via DOCSIS 3.0 channel bonding and expand their business services offerings. While they are preparing for these near-term service launches, it would be prudent for them to keep one eye on the horizon.
Figure 1: Today’s typical 500-home node capacity.
The high-powered fiber network being deployed today by Verizon, the potential for a similar architecture to be deployed by AT&T and the evolution of how consumers will use that capacity are signals for the cable industry. Assuming that the campus dorm room of today–where the Internet is the central resource for entertainment, communication and expression–becomes the home of tomorrow, we can be certain that consumers will be seeking "what-they-want-when-they-want-it" access to fast, full-screen, high-quality high definition video, and a host of other communications applications.
Telcos are likely to leverage their capacity advantages to promote video quality and time-shifted video, playing right into the heart of the changing viewership patterns we are seeing today. Picture a landscape in which cable's competitors have numerous qualitative and quantitative differentiators, including:
- Superior picture quality on non-IP video feeds that will become increasingly important as consumers opt for larger televisions and HD services expand. Today, to provide capacity for new, emerging services, cable operators are compressing, muxing and tweaking their video services, which could give telcos a quality advantage on a 50-inch HD screen.
- A greater focus on IP video feeds. The days of clicking on a video link and expecting a two-inch by two-inch choppy feed could be replaced with a fast, full-screen, high-quality HD feed.
- The upstream capability to enhance the consumer experience on interactive video gaming and Slingbox-type "placeshifting" services (i.e. services that allow viewers to access local television or even personal content from anywhere in the world).
In addition, the telco fiber builds will enable them to push new service areas without thinking twice about the capacity to deliver them. For example:
- Focus on unicast–offering a separate HD stream (or two) to every household on the network. This would facilitate tailored advertisements based on individual users' demographic data (i.e. making bandwidth the key driver behind advertising revenue growth).
- Video surveillance, or other constant upstream bit rate services from the home.
It would be economically dangerous for MSOs to ignore the capabilities of the new fiber deployments. While the economics are often challenged, the telcos are coming and residential services are a cornerstone of their strategy. In our view, cable system operators would be wise to prepare to rapidly deploy competitive services that will retain current subscribers and attract new customers in the face of this telco threat.Cable's response
An advantageous approach would be for MSOs to focus their response on developing a high degree of strategic bandwidth flexibility. Rather than deploying solutions that focus completely on bandwidth efficiency, cable operators would be best served to be ready with a plan to match telecom service providers for the foreseeable future, and ideally to be able to roll the plan out rapidly, if needed.
Figure 2: BPON/GPON capacity.
Many of today's solutions offer temporary relief, but do not give MSOs a truly competitive weapon. Switched digital video (SDV) provides only temporary relief from downstream capacity issues and does nothing to address the upstream issues that MSOs will face.
For example, switching 200 digital broadcast programs with 50 percent peak usage among each service group will result in approximately a 50 MHz or 300 Mbps savings. As large as that appears on the surface, it equals only 300 kbps/home in a 1,000-home service group. When operators choose to use targeted, narrowcast advertising opportunities with SDV, the savings will be even less.1
The other issue is that, while it is true that SDV will "save" more relative spectrum as digital programming, especially HD, takes up even more spectrum, the aggregate bandwidth demand will move higher. For example, if 100 of the channels being switched convert from SD to HD (a highly aggressive assumption), the 50 MHz of gain from switching the SD becomes a 25 MHz net loss in spectrum because of the increased data rate for HD.
This is not to say that operators should not implement SDV–it is an essential tool to enable virtually unlimited content availability. However, cable operators must realize as content evolves from SD to HD as the default format, switching is not likely by itself to compensate for the greater data rate required per stream in the overall picture for spectrum and bandwidth.A look at competitive networks: Delivering capacity
At the core of the matter is how much bandwidth (both upstream and downstream) a network service provider must provide to a consumer in order to remain competitive. Rather than trying to guess the impact of emerging services or applications, this article compares the relative capabilities of the networks being envisioned by cable companies and telcos. Most of the incremental technologies being deployed by cable companies today (SDV, advanced video compression, etc.) are techniques available to both cable and telcos, not differentiators. Thus, this article looks only at the weapons being deployed and normalizes those weapons on the basis of network capacity per home passed.Current MSO networks
Admittedly, no two cable networks are identical and it is difficult to generalize when speaking of cable technology. However, for the purposes of this analysis, we will assume a 750 MHz cable plant that utilizes 256 QAM downstream and 16 QAM upstream. In this scenario, the maximum potential network throughput (again irrespective of services) is about 4,400 Mbps downstream and 90 Mbps upstream. This represents the total network capacity that can be shared among the subscribers in each service group.
Assuming a 500-home node, the total network capacity on a per-home-passed basis, is roughly 8.8 Mbps/HP downstream and 0.18 Mbps/HP upstream. Remember, this figure represents raw network capacity, considered independently of how it is used to offer services.
An 860 MHz network, by comparison, has a capacity of 10.2 Mbps on the downstream, while the upstream is unchanged at 0.18 Mbps/HP.
Finally, a 1 GHz network has a capacity of 11.9 Mbps on the downstream, while again, the upstream remains unchanged at 0.18 Mbps/HP.Verizon FiOS
When analyzing Verizon's FiOS deep fiber deployment, we must break the network capacity into two separate buckets: a Passive Optical Network (PON), co-
deployed with a separate RF overlay network with a 50-870 MHz bandwidth. Essentially, Verizon is using the PON-based architecture for the data, voice, and perhaps transactional video traffic, while the RF overlay is used for downstream video, using the same signal types as today's cable HFC networks. For the voice and data portion of the network, early implementations of FiOS have used the Broadband PON industry standard (BPON) that provides 622 Mbps downstream and 155 Mbps upstream capacity. The BPON standard creates a service group size of around 32 homes which translates into approximately 19.5 Mbps downstream and 4.5 Mbps upstream per home passed.
As FiOS adopts the newer Gigabit PON standard (GPON) architecture, capacity will increase to 2.4 Gbps downstream and 1.2 Gbps upstream. This standard utilizes a 64 home service group and gives the operator more flexibility in the deployment, while providing 37.5 Mbps/HP of total downstream network capacity and roughly 18.75 Mbps/HP on the upstream.
This entire capacity is available to support voice, data and new emerging video, such as interactive video or IP Video, while the RF overlay network (whose transmission characteristics are similar to today's cable networks) can be devoted to video. This gives Verizon an additional 5.2 Gbps of downstream network capacity–80 Mbps/HP with 64-home GPON groups or 160 Mbps/HP with 32-home BPON groups.
Figure 3: BPON/GPON with RF overlay capacity.
Thus, the total per-home-passed FiOS network capacity, including both PON and RF networks, is 100 to 200 Mbps on the downstream and 5 to 18.75 Mbps on the upstream as illustrated in Figure 3.Cable strategies to match
This article will discuss the four viable strategies that will give MSOs capabilities comparable to Verizon's FiOS network.
1 - Fiber Deep –This strategy consists of splitting nodes to 125 homes (thereby reducing the number of homes sharing a given amount of bandwidth), increasing the downstream bandwidth to 1 GHz, and changing the ratio of downstream to upstream bandwidth (mid-split conversion). This strategy delivers approximately 44 Mbps/HP downstream and 2.4 Mbps/HP upstream–less than FiOS but a major improvement.
2 - Fiber Very Deep –This differs from the previous strategy only in that nodes are split to 64 homes. With that change, the downstream capacity is increased to 87 Mbps/HP and the upstream capacity is increased to 4.7 Mbps/HP.
3 - Fiber Shallow –This strategy includes reducing node size to 250 homes and adding spectrum overlay of an additional 700 MHz of downstream and 200 MHz of upstream capacity. This strategy enables approximately 35 Mbps downstream and 2.8 Mbps of upstream capacity per home passed. This gets close to the capacity offered in the first strategy with less capital expenditure in fiber builds.
4 - Fiber Deep with Spectrum Overlay –This differs from the third strategy only in that nodes are split to 125 homes. With that change, the downstream capacity is increased to 71 Mbps/HP and the upstream capacity is increased to 5.7 Mbps/HP.
Thus, an overlay of spectrum on a 750 MHz plant at 250 homes provides almost as much total bandwidth as a 1 GHz enhancement plus mid-split to 125 homes at less than half the cost. Like-wise, overlaying spectrum on a 750 MHz plant at 125 homes is much less expensive and provides a similar throughput to a 1 GHz enhancement plus mid-split to 62 homes.
Figure 4a: Cable’s alternatives.
It would appear then that the most capital efficient way to increase overall plant throughput is to first provide additional spectrum via overlay technologies and then move that architecture deeper as more throughput is required. This gives cable operators the capacity they need today without the expensive fiber builds required to support the smaller node sizes. Remember, we started at a 500-home node for this analysis. If the actual plant is closer to 1,000-home nodes, the relative economics for the solutions that include spectrum overlay look even better.
The baseline system is 750 MHz, 500-home node.
There are significant differences between the PON architecture and HFC. The most significant difference may be the shared access model of the cable networks versus the dedicated access of the PON scheme. There are pros and cons to both, but conceptually, cable operators may need less total throughput over time than their telco counterparts. Overlaying spectrum, even with relatively large nodes, provides more effective bandwidth when considering this element of service delivery.
Figure 4c: Adding this spectrum comes at a cost.
For some services, greater network efficiency is gained when service groups are larger than the physical sub-nodes. The reason is two-fold: the statistics of shared resources and the cost of serving each group.
First, the statistics of shared resources are less efficient as the group of users shrinks. This is well documented in traditional telephone literature. To understand the phenomenon, consider a VOD service in which the probability of a given user accessing the service in the busiest hour is 10 percent. If the service group is very large, it will be sufficient to provide one-tenth as many streams as there are customers. At the other extreme, however, with fewer than five customers, an operator will have to provide nearly as many streams as customers because of the higher possibility of multiple VOD users overlapping. For intermediate service group sizes, the ratio of streams to customers decreases smoothly as the service group gets larger and their behavior more closely approximates the statistical average.
Secondly, each service group will require a separate port (physical or logical) in the headend service-provisioning equipment and a separate group of QAM modulators and demodulators. This adds to the cost of creating small service groups.
To the extent that service groups are larger than physical nodes, nothing is gained by further node splitting, since the same signals will be present in each sub-node. This would generally be true for such services as telephone, VOD and SDV services–all are inefficiently provided to service groups smaller than 500 homes.
In addition, most operators likely will continue to broadcast the most popular video channels, whether analog or digital, because there is no advantage to be gained by switching any program if the probability that at least one customer in each service group will be watching it is high.
Given all these factors, a better way of comparing the impact of spectrum overlay technologies relative to the Fiber Very Deep bandwidth enhancement method is to look at what happens when only a portion of the available bandwidth is dedicated to those switched services that benefit from node sizes as small as 64 homes. Figure 4 shows, for each upgrade technology, how much bandwidth is available for broadcast and larger-service-group services as a function of the bandwidth set aside for small-group (fewer than 500 homes) services.
As the figure shows, the available "common" or bandwidth is much greater for the spectrum overlay approach regardless of how much bandwidth is assigned to small-group switched services. Thus, overlaying spectrum provides more effectivebandwidth than node splitting combined with only a moderate bandwidth increase.
Figure 5 demonstrates an especially important finding. What it shows is that cable operators can get more from their networks by taking fiber less deep and overlaying spectrum than they would from taking fiber more deeply into the network with the existing architecture. We believe this has the potential to save operators a significant amount of capital expenditures over the next 5 to 10 years. They can potentially reduce their overall fiber construction spending (i.e. staying at 250-home nodes) while competing head-on with the telcos bit-for-bit.Summary
Some might say this analysis is not even-handed–if the telcos are building a bazooka when all that is needed to compete is a
rifle, then their financial performance will be mired by this overinvestment and MSOs will come out the winner. Similar questions were asked of Intel in the mid-80s: Why were faster and faster chips developed when consumers did not have applications that required them? Somehow, our PCs are still slower than we would like.
Cable system operators need to think broadly about network evolution options. Just as in personal computing, where there is an endless race between faster chips and the applications that run on them, most incremental technologies in cable address only near-term service rollout plans. This gives the telcos an opportunity to exploit and differentiate using the greater capacity that their new fiber network delivers.
It would be wise for cable system operators to think proactively about ways to ensure that their networks can match the competition's capacity, so that they can maintain maximum flexibility to deliver what customers want, when they want it. The analysis in this article shows that with spectrum overlay technology, MSOs' coaxial plants will have plenty of horsepower to stay in–and win–the game.