NEXT GENERATION - CABLE ACCESS NETWORK
AN EXAMINATION OF THE DRIVERS, NETWORK OPTIONS, AND MIGRATION STRATEGIES FOR THE ALL-IP NEXT GENERATION – CABLE ACCESS NETWORK
Authors:M. Emmendorfer, S. Shupe, D. Cummings, T. CloonanContributors:Z. Maricevic, M. Schemmann, B. Dawson, V. Mutalik, J.Howe, A. Al-Banna,and F. O'Keeffe
ARRIS
ABSTRACT to rise at an alarming rate. Cable Operators like the United Kingdom's Virgin The Cable Industry is facing a Mediaannounced in April 2011 an Internet decade of unprecedented change in the areas speed trial of up to 1.5 Gbps downstream of video and high-speed Internet services. and 150 Mbps upstream [1].The cable This change,driven by competition and competitor Verizonis reportedly exploring consumer demand, will transform the cable plans to upgrade its FiOS system to XG- network end-to-end. This paper will focus PON, the 10 Gbps downstream and 2.5 entirely on what we are calling the Next Gbps upstream technology [2]. New Generation Cable Access entrants in the video distribution space are Network,examining the business drivers, capitalizing on the network investments network options, and migrations strategies in made by the telecom industry, forcing the access layer of the data and HFC changes in their video delivery network as network to provide more IP-based capacity well as the high-speed data network. A key to and from the home. The document covers challenge the cable industry will face in the in-depth the core business drivers and the future will be offering PON-like IP-based technical options spanning animmense area capacity in the downstream and the of network disciplines and technologies, upstream to consumers, while leveraging thus we have included a comprehensive their existing coaxial network. executive summary at the conclusion of the report. Some of the most often asked questions by cable industry forward-looking The analysis includes the allocation planners reflect the key challenges the of existing spectrum and possible future industry is facing for this decade and spectrum expansion to accommodate beyond. Some of these challenges and consumer demand. Cable Operators and questions include: 1) How long willthe their competitors are enabling consumers to current spectrum split and 500 HHP change their viewingoptions for video nodelast? 2) What are the network services and the usage of the high-speed technology and architecture options and Internet network.In the area of high-speed what are the pros and cons? 3) How long Internet service, competition and consumer will each of these new network architecture demand is increasing the service speed tiers options last? 4) What are the financial offered,and network traffic usage continues
impacts of the options?5) What are the best each network architecture and technology, ways to leverage previous, current and so industry leaders may make an informed future investment? decision.These topics under consideration comprise several network technology This paper will seek to provide some disciplines, which are often separate areas of visibility and answers to these questions and concentration. This paper is by no means key challenges. The paper will focus conclusive; some of the areas under entirely on the network aggregation and examination have not had significant study access layer including the CMTS, HFC and or have the absence of products to home network. This paper will provide sufficiently examine and forecast the best some predictions for service tier and traffic path.There are also timing considerations growth, which serve as the drivers for and business trade-offs that will need to be network capacity and network utilization considered. forecasts that are used to predict the timing of the network changes and investment. We INTRODUCTION – PLANNING FOR THE NEXT will examine the network technology and GENERATION – CABLE ACCESS NETWORK network architecture options from spectrum A major challenge the cable industry splits, data MAC and PHY technologies as will face in the future will be meeting the well as network architecture options. This needs of the consumer and addressing the paper will consider the capabilities of a drop competitive threats of PON/FTTH systems in upgrade with an effort to maintain a 500 all while leveraging the existing coax to the HHP service group and typical number of home. This will mean significant changes in actives and passivesto determine the the use of network technologies, spectrum viability and impact for upstream spectrum allocation, and overall network architectures. Planning for the Next expansion. The funneling effect must be Generation – Cable Access Network is considered in the analysis for the NG Cable extremely difficult as this spans across Access Network. The paper provides several network disciplines within the cable analysis and comparison of some of the industry and even technologies outside or network elements under consideration. The not widely deployed in cable, such as PON, paper introduces a term called Digital Fiber Wireless, and EoC. The span of network Coax (DFC) as a next generation technology disciplines also reaches into the network elements and underlying sub- architecture, which may augment the HFC systems such as MAC layer, PHY layer, media conversion style architectures that HFC optical transport components, as well utilize centralized data access/aggregation as several access radio frequency layer equipment. technologies end-to-end such as amplifiers, passives, and coaxial cable. What is proving We considered a couple of migration to be a significant challenge is the increase strategies as more viable than others and dependency between all of these while not picking a particular end-state traditionally separate network disciplines as approach, our position is to examine the part of the new cable access network architecture. In years past these options and document the pros and cons of technologies functioned in many ways
independent of one another. This next Flexibility to accommodate generation cable access architecture will incremental allocation of likely migrate to more IP based spectrum in IP/Bandwidth for smooth transition the downstream cannibalizing existing strategy and pay as you grow or just technologies which are non-IP based in time network planning creating a more efficient and competitive DOCSIS Backwards Compatibility network transport platform to compete with leverages MAC/PHY channel PON on the downstream and simply have bonding groups previously deployed the versatility of IP based technology. The and occupying spectrum yielding upstream will need more spectrum and it is investment protection delaying or the overall spectrum allocation and avoiding significantly costly placement of this new spectrum, whichwill approaches to find new spectrum have the greatest impact on the cable Investment protection by re-using industry for decades to come. spectrum already in service, DOCSIS, HE lasers/receivers, and The challenge we have is predicting CPE (STB/Data) as much as possible the timing of the change in the network and Leverage network passives the most how long each change will last. numerous OSP element Additionally and most importantly what are Avoid costly and unnecessary fiber the impacts of each of the upstream builds spectrum options that may be considered for Keep the OSP as Simple as Possible the future. This paper will provide for as Long as Possible predictions, such as the drivers for the use of Leverage High densities and the spectrum in the downstream and economies of scale upstream. The paper considers the spectrum allocation options and predicts how long Importance of Backward Compatibility each will last beginning with the current with DOCSIS 3.0 and Any Successor sub-split options and several spectrum splits which add new upstream capacity and how The authors of this paper believe that long these will last. The report provides a DOCSIS and any successor should consider technical comparison of the upstream the value of backwards compatibility spectrum options and the impacts that each has from services to overall network especially across channel bonding groups. architecture and cost. This assures previous and future investment may be applied to create a large IP based Our Goals for Next Generation – Cable bandwidthnetwork while not stranding Access Network include: previous capital investment and spectrum. Achieve upstream bandwidth The use of channel bonding leverages every requirements through this decade MHz, which are finite and not free, this is all Achieve downstream bandwidth towardsan effort to create one large IP pipe requirements through this decade to and from the home. The use of Continued versatility to backwards compatibility has benefitted the accommodate advances in cable industry as well as other industries networking technology without which use technologies like IEEE Ethernet, massive changes to the outside plant WiFi, and EPON creating consumer network.
investment protection, savings,and a smooth and most importantly the spectrum migration strategy. The adoption of allocation options to forecast network backward compatibility simply allows the capacity. Then after considering the Service MSOs to delay and perhaps avoid major and Traffic growth we measured this against investment to the network such as adding the network capacity options, in the section more data equipment, spectrum, node splits, referred to as Network Utilization and or running fiber deeper. Capacity Planning. In this section we forecast the timing and duration of each Overview of Our Methodology For network step. In step four we examine Network and Capacity Planning several of the network technology and In our analysis and in the structure of architecture options under consideration. this paper we have examined the Next The Network Migration Analysis and Generation-Cable Access Network in Strategies consider all of the factors of the several steps as captured in the illustration aforementioned steps and provides some below (figure 1). As shown in the analysis of possible migrations strategies to illustration our process was to first address the competitive threats and determine the future requirements. The first consumer drivers. The migration strategies step examined the service tier and traffic selected by the cable operators are growth estimates based on a model that dependent on many factors, and there may captured a thirty-year history to make an not be a consistent approach selected across attempt to predict the future network needs all MSOs. In fact, within a given MSO the for perhaps the next two decades. In the analysis may vary by market. Our analysis second phase we considered the technology measures the costs of several network options.
FIGURE1: METHODOLOGY FOR NETWORK AND CAPACITY PLANNING
may employ a 750 MHz to 1002 MHz SERVICE TIER AND TRAFFIC GROWTH systemequivalent over 4.3 - 6 Gbps of ESTIMATES downstream capacity but it is unknown if all Consumers and Competition Are Driving of this capacity is used. The PON network Change is used for IP based services like Internet, telephone and perhaps on-demand unicast The MSO’s competitive landscape video transmission. If we consider both the has changed rapidly in just the last 12 PON system as well as the video overlay months especially from Over The Top system, the FiOS network capabilities may (OTT) video providers such as Apple TV, reach ~14 Gbps+ of downstream throughput Amazon, Hulu, Netflix and others entering (XG-PON 10 Gbps + 750 MHz at the On-demand video market. In many approximately 4 Gbps+) and upstream ways the consumer electronic companies reaching 2.5 Gbps). This capacity may be like Appleare becoming service providers more throughput than is needed for many enabling the video experience across all years or even decades to come based on the platformsand across any carriers’ network. modeling in the following sections. This The OTT competition affects the MSOs in level of capacity may not be needed until the lost revenues for On-Demand services and year 2025-2030. perhaps a reduction in the subscription service. Adding to the lost revenue is The cable network has a massive increased costs to the high-speed data amount of capacity perhaps up to 6 Gbps to network due to increased consumer usage. the home and perhaps 100 Mbps from the home. The cable industry is making The recent completion of Verizon’s investments in IP based video delivery FiOS roll out will undoubtedly remain a technology and expanding the high-speed threat to the MSO’s triple play offering. Internet IP capacity as well. The coaxial Additionally it was reported that Verizon network is very nimble and may increase the will consider an upgrade to their FiOS spectrum allocation beyond the current network to the next generation Passive levels in either direction. This important Optical Network (PON) technology known fact is covered in detail in this paper. The as XG-PON, the 10 Gbps downstream and amount of capacity needed in each direction 2.5 Gbps upstream system [2]. This could is projected over a period of nearly two replace the earlier generation B-PON (622 decades as well as several technical options Mbps down and 155 Mbps up) and the G- are explored. PON (2.5 Gbps down and 1.25 Gbps up) systems. The Verizon FiOS network also Upfront Disclaimer on Service Tier and uses what is known as the video overlay Traffic Growth Estimates network along with the PON technology. In this report we will be making The video overlay network provides network traffic predictions for the next two broadcast video services using technology decades and we acknowledge that these similar to cable systems. The video overlay numbers are highly debatable.Theseforecasts
may not match any particular cable or Video Service and Delivery telecom provider.The modeling for the Assumptions(Downstream Only) Internet portion of the traffic is based on modeling, which goes back nearly thirty We could have considered many years. This model illustrates Data Service factors for the video service network Tiers offered to consumers increase at about requirements. We could have done a year- a 50% compound annual growth rate by-year prediction of the allocation of linear (CAGR) and this model also is used to programming, VoD, SDV, SDTV, HDTV, forecast actual consumer traffic usage which 3DTV, amount of in-home pre-caching, and also grows at roughly a 50% CAGR. The service group size and number of tuners, etc, data service portion of the model is but we did not consider all of these areas predictable but at some point as with individually as these may vary widely Moore’s Law, the growth rate for Service among MSOs and over time. Tiers Offered to Consumers as well as traffic usage may not continue on this We simply will assume that Video trajectory for another 20 years.We are only Services will use all available capacity not using these Service Tier and Traffic Growth being used by the High-Speed Internet Estimates as ―rough ballpark numbers‖ to Services. We will however make some allow discussion and forward planning.The forecast for what could be considered a Network forecast will include Video minimum allocation of capacity for an MSO Services offered by the cable provider as deliveredvideoservice, below are our Video well asHigh-Speed Internet Services. assumptions and traffic forecast.
Video Assumptions Take-rate of the service 60% Viewers are actively watching a program during the busy-hour/busy-day 60% Average video viewers per active home 2 Linear Service (Broadcast) 0% On-Demand Service (Unicast) (this worst-case assumption creates the 100% biggest BW challenge) Average program bandwidth (assumes mix of SD, HD, and 3D in 10 Mbps MPEG4) HHP Fiber Node or Service Group (SG) 250 FIGURE 2: VIDEO ASSUMPTIONS FOR FUTURE CAPACITY PLANNING
Video Calculation FIGURE 3: VIDEO TRAFFIC ALLOCATIONS FOR 250 HHP/Node * (0.6 take-rate) * (0.6 active) 1.8 Gbps per 250 HHP SG FUTURE CAPACITY PLANNING * (2 viewers/active home) * (10 Mbps/viewer) or 3.6 Gbps per 500 HHP SG
The video service is projected be a figure also attempts to predict the max unicast offering and the model essentially service tier we may see in the future, if the will always reserve or allocate 12 Mbps per growth trend aligns with the preceding video subscriber (1800 Mbps / 150 video years. Perhaps we will allocate the entire subscribers) as illustrated in the figure 2 and 750 MHz downstream spectrum or 3. However, like today video services will equivalent to Internet services by 2023. As dominate the spectrum allocation compared illustrated in the figures below, the to High-Speed Internet for nearly the entire downstream and upstream modeling began decade. The modeling in the remainder of with the dial-up era, moving into the this paper assumes that video services will broadband era and now the DOCSIS channel consume all of the bandwidth that High- bonding and PON eras. This model assumes Speed Internet does not require, however the a 50% CAGR for the Internet service tier. model reserves the 12 Mbps per video sub as a minimum allocation for a video service Max. Data Service Tier Offering Downstream offered by the MSO, unless otherwise stated. 100G ~85 Gbps?
10G Certainly the MSO’s high-speed data 4.3 Gbps DS Limit
1G The past 25-years shows a subscribers may use the data network to constant bandwidth increase 200 Mbps of about a 50% CAGR 100M view video content on devices like tablets, 50 Mbps 12 Mbps handhelds, TVs, PCs, and other devices. 10M 1 Mbps 1M 5 Mbps 256 kbps 512 kbps 100k 56 kbps High-Speed Internet Service Tier Offered 28 kbps 128 kbps 33 kbps 10k 9.6 kbps (Downstream and Upstream) 14.4 kbps 1.2 kbps 2.4 kbps 1k
MaxDSPermitted Bandwidth Modems(bps) for 300 bps The network traffic estimates need to 100 The Era of The Era of The Era of The Era of DOCSIS 3.0 DOCSIS 4.0, 10 Dial-Up DOCSIS Channel Fiber Deeper, consider the downstream and upstreamhigh- Modems 1.0-2.0 Bonding and FTTx 1 speed Internet service tier, in other words 1982 1986 1990 1994 1998 2002 2006 2010 2014 2018 2022 2026 2030 the data speed package the MSO offers to Year FIGURE 4: MAX INTERNET DATA SERVICE consumers. The highest data speed offered TIER OFFERING DOWNSTREAM in either direction is a determining factor for sizing the network. TheHigh-Speed Internet service tier and traffic will grow considerably during this decade moving from perhaps four6 MHz channels downstream, which is less than 4% of the MSO’stotal spectrum allocation and may grow to perhaps 40-50% in the next 10 years. The high-speed Internet service tier offering will be a key contributor to overall bandwidth drivers. The figure below shows a thirty-year history of the max bandwidth offered or available to consumers. This
Max. Data Service Tier Offering Upstream
100G
10G ~17 Gbps?
1G ~160 Mbps US Limit (Euro) 100M ~90 Mbps US Limit (North America) Assume the US Tmax is always 10M 1/5 of the DS Tmax (ex: 50 Mbps DS tier has 10 Mbps US today) 1M
100k
10k
1k
100 Max Max BandwidthUS Permitted Modemsfor (bps) The Era of The Era of The Era of The Era of DOCSIS 3.0 DOCSIS 4.0, 10 Dial-Up DOCSIS Channel Fiber Deeper, Modems 1.0-2.0 Bonding and FTTx 1 2006 2010 2014 2018 2022 2026 2030 1982 1986 1990 1994 1998 2002 FIGURE6: COMBINED INTERNET MAX Year 2017 2018-19 SPEED PREDICTIONS FIGURE 5: MAX INTERNET DATA SERVICE
TIER OFFERING UPSTREAM There has been a significant increase The table below captures the year- in the services offered resulting in an up tick by-year predictions of the downstream and off the linear progression. Additionally, upstream service projections from the announcements from cable operator Virgin figures above. This table will be used for Media of an Internet speed trial of up to 1.5 the capacity requirements found in the Gbps downstream and 150 Mbps upstream Network Utilization and Capacity Planning [1] and Verizon reportedly exploring plans section later in this document. It is to upgrade its FiOS system to XG-PON, the uncertain if the Max Service Tier trends will 10 Gbps downstream and 2.5 Gbps upstream continue for the next 15 years at a 50% technology may further move the model CAGR. The service offerings will, from higher [2]. The rollout of downstream time-to-time, not maintain alignment with channel bonding was a key contributor to the projections. Typically leaps above the the expansion of the service offering as well line happen when there are major as PON. As upstream channel bonding is technology advances, such as dial-up to deployed in the near term we expect an cable modem/DSL, then to channel bonding expansion of the upstream max service tier and PON.So, if we analyze where the to increase as well, perhaps initially at a telecom industry is today with their max higher rate that the 50% CAGR as the model downstream and upstream service offerings has captured over the last thirty years. This this may not be in alignment with the is critical information for the network predictions. planners; any acceleration in the service tier offered would change the predictions we have captured in this paper affecting the estimated migration timeline. The expansion of service tier often leads to
higher per customer bandwidth usage or network traffic.
High-Speed Internet Bandwidth Per Subscriber (Downstream and Upstream)
In addition to the service tier offered to consumers, the actual usage of the network by the consumers is a critical factor for network planners. This is known as the bandwidth per subscriber (BW per Sub). The determination of bandwidth per sub, is a measurement of the total amount of bandwidth or traffic in a serving area divided by the number of consumers in the FIGURE 7: DOWNSTREAM BANDWIDTH PER serving area, this may be measured during SUBSCRIBER busy hour(s).The bandwidth per subscriber
is measured in the downstream and upstream direction. The downstream is currently measured at a 220 kbps per subscriber and the upstream at 36 kbps per subscriber, as illustrated in figure 7 and 8. The rate of growth is projected at a 50% CAGR. The bandwidth per subscriber and the CAGR may vary, however these numbers seem reasonable for the North American market. These numbers are used for planning purposes in this paper.
FIGURE 8: UPSTREAM BANDWIDTH PER SUBSCRIBER
Summaries for Service Tier and Traffic Growth Estimates
The video service offering will evolve over time from broadcast to unicast. The model plans for 3.6 Gbps of video traffic in a 500 HHP service group and 1.8 Gbps in a 250 HHP serving group. The
model will use both High-Speed Internet PHY layer efficiency estimates additional projections, like the Service Tier Offering MAC layer overhead has not been and bandwidth per subscriber to predict calculated. The authors wish to express that Network Utilization and Capacity Planning. ARRIS is not aware of industry plans to adopt OFDM in the DOCSIS standard. Additionally, ARRIS has conducted internal studies for years examining the possibilities NETWORK CAPACITY of an OFDM based systembut we have no The network capacity of the cable plans to incorporate OFDM based systems access network is determined by the amount into our products. of spectrum available and the data rate possible within the spectrum. There are PHY Layer Throughput Assumptions many factors that determine the amount of DOCSIS QAM Based spectrum and data rate possible such as the There are three figures which location of the spectrum, noise, PHY/MAC captures the assumptions of DOCSIS QAM layer efficiencies possible, and several other based system. The first calculates the factors. We have modeled several spectrum DOCSIS 256QAM downstream, figure 9. allocation options as well as data rate The remaining two tablesmodels the possibilities. The analysis below captures upstream using DOCSIS 64QAM and the PHY layer throughput assumptionsof DOCSIS 256QAM each assumes ATDMA, DOCSIS QAM (Quadrature Amplitude figure 10-11. These tables measure the PHY Modulation) forthe downstream and layer spectral efficiency of DOCSIS QAM upstream. The analysis also considers a based solutions. These are used to calculate DOCSIS OFDM (orthogonal frequency- the network capacity of the cable network division multiplexing) based system that considering several spectrum options. could emerge in the future. These are again
FIGURE9: DOWNSTREAM DOCSIS 256QAM
FIGURE 10: UPSTREAM DOCSIS 64QAM
FIGURE 11: UPSTREAM DOCSIS 256QAM
hardware. ARRIS believes that the DOCSIS A key take away is performance gap specifications should be modified to include between 256QAM PHY and 64QAM layer 256QAM upstream as well as 1024QAM in efficiencies. The assumptions for 64QAM the upstream and downstream. at 4.10 bps/Hz would require 28% more spectrum and DOCSIS channels to maintain DOCSIS OFDM Based the equivalent PHY layer throughput. The For analysis purposes the paper use of DOCSIS 256QAM for the upstream provides measurements using is not part of the DOCSIS standards, OFDM/OFDMA, again OFDM is not part of however some CMTS and CM products the DOCSIS standards. support this modulation profile in
FIGURE 12: OFDM 1024QAM ANALYSIS
FIGURE 13: OFDM 256QAM ANALYSIS
carrier QAM and OFDM. These are subject In the figures above 256QAM was to debate. analyzedusing estimates for PHY layer efficiency comparing DOCSIS single carrier Downstream Capacity 256QAM and DOCSIS OFDM 256QAM. The most critical determination for The analysis for the OFDM based approach the capacity of the network is the amount of shows a slightly higher PHY layer spectrum available. The determination of efficiency. The actual performance of either the downstream capacity will assume the in real-world deployments is unknown. eventual migrations to an all IP based There are many attributes and assumptions technology. The migration to all IP on the than can be modified; we used an estimate downstream which will optimize the that we considered to be fair for single capacity of the spectrum providing the versatility to use the network for any service
type and provide the means to compete with MHz. Certainly there are other spectrum PON and the flexibility to meet the needs of options that could be considered such as the future. This table provides capacity moving the downstream above 1 GHz and projections considering: 1) the upstream other spectrum options for the upstream. spectrum split, 2) the use of DOCSIS QAM This table will calculate the estimated or DOCSIS OFDM, 3) several downstream downstream PHY layer capacity using spectrum allocations from 750 MHz to 1002 several spectrum options.
FIGURE 14: DOWNSTREAM NETWORK CAPACITY ESTIMATES
The model used a lower order modulation assumption for QAM but high order modulations are certainly possible. The spectrum capacity of single carrier QAM and OFDM may actually be similar, however more real-world analysis is needed to accurately measure the performance of both technologies.
Upstream Capacity current, and future investments made by the MSO to be applied to a larger and larger The upstream capacity bandwidth pipe or overall upstream measurements are more complicated and not capacity. If backward compatibility were as straightforwardas the downstream not assumed the spectrum options would capacity projections. In the table below, have to allocate spectrum for DOCSIS QAM many of the spectrum split options were and separate capacity for any successor, evaluated considering several PHY layer resulting in a lower capacity throughput for options and modulation schemeswithin each the same spectrum allocation and would spectrum split. compress the duration of time the same These are some key assumptions spectrum may be viable to meet the needs of about the upstream capacity estimates: the MSO.
Sub-split spectrum region considered The upstream capacity 22.4 MHz eligible for channel measurements found in figure 15compares bonding various spectrum splits, modulation types as well as single carrier QAM and OFDM. Sub-split spectrum was calculated with only DOCSIS 3.0 64QAM The spectrum splits found in the table include Sub-split, Mid-split, High-split Sub-split channel bonding spectrum (200), Top-split (900-1050), and Mid-split counted in capacity summaries with with Top-split (900-1050). The Top-split any new spectrum split options above 1.2 GHz were not calculated All estimates use PHY layer in this table. efficiency estimates additional MAC layer overhead has not been The spectrum split, PHY, and calculated. modulation type are examined in figure 15 to determine the ―Total PHY Channel Bond An important assumption is that the Capacity Usable‖, found on the last column. upstream capacity measurements assume This was intended to delineate between that spectrum blocks from the sub-split single carrier QAM and OFDM omitting the region and any new spectrum split will all MAC layer throughput calculations. Traffic share a common channeling bonding engineering and capacity planning should domain. This is essentially assuming that consider the MAC overhead and headroom backwards compatibility is part of the for peak periods. Similar to the examination upstream capacity projections. The of the downstream capacity projections upstream capacity projections for each split above, the upstream projections illustrate will assume DOCSIS QAM and if adopted that OFDM has more capacity compared to in the future DOCSIS OFDM based systems QAM; this may not be the case in real-world will all share the same channel- deployments. bondinggroup. This will allow for previous,
Technology Data Spectrum Summaries Channel Bond Data Rate Capacity Rates Per MHz Total PHY DOCSIS DOCSIS DOCSIS DOCSIS Channel MSO Upstream Channel Total QAM 64 Sub- QAM 64 QAM 256 OFDM Split Type Bond Bonding Bandwidth Summaries Upstream Sub-split NEW NEW DOCSIS DOCSIS DOCSIS split Total Total Total Spectrum Spectrum Spectrum Spectrum QAM 64 QAM 256 OFDM Spectrum Capacity Capacity Capacity Capacity Usable for Likely Only Usable for Usable for Usable Usable Usable Likely Only Data Rate Data Rate Data Rate Total US Channeling used for DOCSIS DOCSIS Data Rate Data Rate Data Rate used for Usable Usable Usable (Usable) Spectrum Bonding DOCSIS 3.0 QAM OFDM Per MHz Per MHz Per MHz DOCSIS 3.0 (Mbps) (Mbps) (Mbps) Sub-split DOCSIS QAM 64 37 22.8 22.4 4.1 5.2 6.28 92 - - - 92 DOCSIS QAM 64 80 65.8 22.4 43.2 4.1 5.2 6.92 92 177 - - 269 Mid-Split DOCSIS QAM 256 80 65.8 22.4 43.2 4.1 5.2 6.92 92 225 316 DOCSIS OFDM 80 65.8 22.4 43.4 4.1 5.2 6.92 92 300 392 DOCSIS QAM 64 195 180.8 22.4 158.4 4.1 5.2 6.92 92 649 - - 741 High-split (200) DOCSIS QAM 256 195 180.8 22.4 158.4 4.1 5.2 6.92 92 824 916 DOCSIS OFDM 195 180.8 22.4 158.4 4.1 5.2 6.92 92 1,096 1,188 DOCSIS QAM 64 187 172.8 22.4 150.4 4.1 5.2 5.54 92 617 - - 708 Top-split (900-1050) DOCSIS QAM 256 187 172.8 22.4 150.4 4.1 5.2 5.54 92 782 874 DOCSIS OFDM 187 172.8 22.4 150.4 4.1 5.2 5.54 92 833 925 Mid-Split + Top Split Mid-split + Top-split (DOCSIS QAM) 230.4 216.2 22.4 193.6 - - - 92 617 225 933 (900-1050) Mid-split + Top-split (DOCSIS QAM and OFDM) 230.4 216.2 22.4 193.6 - - - 92 225 833 1,150 FIGURE 15: UPSTREAM NETWORK CAPACITY ESTIMATES
NETWORK UTILIZATION AND CAPACITY A very important point is that the PLANNING network architecture and performance If you are wondering how long a characteristics of the plant in the real world spectrum split may last or the sizing of the will determine the spectrum capacity to be service group in the downstream or used. The determination of the network upstream this sections will provide some architectures that may work at various spectrum splits, modulations, and number of estimates for consideration. In this section carriers was a critical finding of this report. of the report the network utilization estimates and capacity planning forecasts We have modeled the network architecture and performance assumptions to estimate are examined. This section will predict the year and potential driver for network the modulation and capacity possible for each spectrum split. This allowed us to change. The information found in this section will be based on the findings of the determine the overall requirements and impacts to cost of the various split options preceding sections, which forecasted the and the ability for the spectrum split to meet service usage for video and High-Speed the business needs of the MSO. The Internet as well as network usage on a per- network architecture requirements and subscriber basis. Additionally this section impacts for each spectrum split will be will use the network capacity estimates for found in the sections called ―Network the downstream and upstream. Technology and Architecture An important attribute of cable Assessment‖and the cost assessment section systems is that the HFC optical and RF called ―Cost Analysis‖. network as well as the data access layer network like the DOCSIS CMTS allows for
upstream and downstream capacity upgrades
may be made separately, where and when by about the year 2023 the existing needed per service group. The report downstream spectrum would be entirely separates the utilizations and capacity needed for High-Speed Internet Services. It planning results for the downstream and again should be stated that these are just upstream to take advantage of this key prediction for the next decade or more, it is feature.A key factor for the calculations will uncertain if speeds would be desired or be the service tier growth forecast and the offered at the levels shown. per subscriber usage, which have been separated as well. As stated previously these are just predictions and there are many factors that may influence change and the rate of change, so these findings should just be used for discussion purposes only.
The Downstream
The downstream network capacity drivers will be separated into High-Speed Internet Max Service Tier plus Video Traffic Predictions and another measurement will FIGURE 16: DOWNSTREAM SERVICE TIER be forEstimated Bandwidth per Service AND NETWORK CAPACITY ESTIMATES Group.
Capacity Planning for High-Speed Internet Estimated Bandwidth Per Service Group Max Service Tier plus Video Traffic (Downstream) The upstream and downstream High- There are several contributing factors SpeedInternet service tier growth by year used to forecast the capacity for a service and direction is used to forecast the date group. They include, the size of the service when the downstream may be at capacity, group, take rate of the services, estimated see figure 16. The HFC downstream per subscriber data usage, and the allocation capacity assumptions will usethe equivalent of capacity for an MSO managed video to a 750 MHz system, approximately 700 service offering. The model defines a MHz of usable downstream spectrum to service group as a collection of HHP measures the date the capacity threshold is beginning at 1,000 HHP to 63 HHP. We use reached. The table shows that the MSO may the modeling projections from the previous offer a 2.2 Gbps Downstream High-Speed section and apply the capacity capabilities of Data Internet service tier and support the 750 MHz system or equivalent. The capacity for a managed video service analysis predicts that a 500 HHP service package of about 1.8 Gbps, the year 2021. group will meet the capacity needs for the Additionally, if the High-Speed Internet future and a migration to each 250 HHP growth rates remain at a 50% CAGR, that service group will last a full decade. The
estimated bandwidth per service group is a 500 HHP and 250 HHP service group to measures based on the high-speed Internet estimate the date of the migration to a user traffic, the call outs in figure 17 capture smaller service group. the video allocation estimates used for the
FIGURE 17: DOWNSTREAM SERVICE GROUP CAPACITY PLANNING ESTIMATES
The Upstream under examination will last. In figure 18, the upstream max service tier used in the The network utilization and capacity Service Tier section earlier in the paper is planning forecast of the upstream may meet assessed with the network capacity data the capacity limits of the sub-split 5-42 in estimates for the split options found in the North America and Europe’s 5-65 within immediate preceding section. The service this decade. Surprisingly, it could be the tier estimates along with the capacity network utilization or traffic at a 500 HHP estimates for each split option is used to service group which meets the throughput predict the year each upstream split option capacity of the sub-split spectrum. The will be at or near capacity.We again wish to section below examines the spectrum split point out that these are just estimates used options and the timing impacts. for planning purposes. Capacity Planning for High-Speed Internet Max Service Tier This section captures the duration of time each of the upstream split options
Year 2022: High-Split 200 or Year 2017: North America Top-Split (+Mid-Split) along 5-42 HSD Service Tier with DOCSIS QAM type Exhausted Touching the technology carries the industry Actives to add more upstream into the next decade & beyond Spectrum Required
Year 2018-19: Euro 5-65 HSD Year 2023: Upstream HSD Service Tier Exhausted Max Service Tier Prediction of Touching the Actives to 995 Mbps Upstream Service adding more upstream consumes High-Split 200 or Spectrum Required Top-Split (+Mid-Split) Using DOCSIS QAM (~900 Mbps + Year 2020: Mid-Split 5-85 with or OFDM (~1.1 Gbps +) either DOCSIS QAM or OFDM is approaching capacity and in 2021 would be out of capacity Year 2024: MSOs touch the driven by Service Tier Growth passives to increase spectrum thus more upstream Spectrum above 1 GHz to achieve higher is Required (High-Split or Top- Upstream Speeds Split)
FIGURE 18: UPSTREAM SERVICE TIER AND NETWORK CAPACITY ESTIMATES
and service group size that may sustain the Estimated Bandwidth Per Service Group traffic load and this table will also mention (Upstream) when service tier projection will force a split It is estimated that the major factor or spectrum increase. that may cause pressure on the upstream split options will be the capacity caused by It should be observed that a 500 traffic usage or bandwidth per service group. HHP service group with 250 subscribers will The table below estimates the traffic last for a decade or more, however this generated by the users in a service group. assumes that spectrum increase like that to The network capacity estimates of each split mid-split may happen in the year 2015 option is then used to determine the year and driven by traffic growth, this will be 2 years service group size that may sustain a given before sub-split is projected to run out split option. In the table below and because of service tier growth. Thus discussed later in this paper are some moving to mid-split in 2015 would allow the assumptions to the usage and indeed relation 500 HHP service group to be leverage until of the spectrum options to the service group about the year 2019. If high-split (200) is size at the upstream optical domain level. added in 2019 this may allow the 500 HHP This table highlights the year, split option, service group to remain until 2021-2022.
Year 2021-22: Year 2015: 500 High-split (200) or HHP Node Split Top-Split ―or‖ Add Spectrum (900-1050 +Mid- Split) with Year 2017: Sub- DOCSIS QAM split Exhausted last through 2021 Because HSD with OFDM 2022 Service Tier then a 250 HHP projections SG in 2023
Year 2020-21: Year 2019: 500 HHP Node with Top-Split (1250-1550) + Mid-Split DOCSIS Mid-Split at QAM Exhausted capacity with 500 HHP Node will Year 2020-21: need to move to Mid-Split 125 HHP meet Exhausted service tier and Because HSD traffic projections Service Tier projections Year 2024: Service Tier Drives need to exceed 1 GHz ―not‖ Traffic FIGURE 19: UPSTREAM SERVICE GROUP CAPACITY PLANNING ESTIMATES
900-1050 is consider yielding ~930 Mbps The figure 19, does not capture all of this has slightly more capacity than High- the options but may be used for planning split (200) and the passives are not touched based on traffic forecasts. The selection of with this Top-split option and avoids the Top-split 900-1050 with a 500 HHP service STB out of band communications challenge. group has an estimated capacity of ~700 Mbps coupled with the sub-split. The Top- Summary of Capacity Planning split 900-1050 option with Mid-split may It is very important that the reader yield a capacity of ~930 Mbps in a 500 HHP understands that our assumptions use HHP service group given the assumption per service group, this may be a physical documented in the following section, see node or a logical node which uses figure 23. The use of the Top-Split options segmentation to meet the sizing level. at 1250-1550 will not be able to use the high Another very important consideration is that order modulation if we assume a 500 HHP the model assumes that over time that full service group, however there is more spectrum would be allocated to the service spectrum available. The Top-split 1250- group to meet the capacity projections for 1550 option with Sub-split is estimated to user traffic in the downstream. The have a capacity of ~500 Mbps and with upstream split options will have a direct Mid-split ~725 Mbps. All of the Top-split relationship to the network architecture to options are viable for a 500 HHP service include the size of the service group, number group but at lower order modulation when of actives, passives, cable portion of the compared to the low frequency return network. options of Mid-split or High-split. If we assume Mid-split is a first step and Top-split
NETWORK TECHNOLOGY AND 500 HHP Node Long-Term Viability ARCHITECTURE ASSESSMENT Our analysis finds that upstream and downstream bandwidth needs may be met The goal of any cable operator is a while leveraging a 500 HHP node service drop in upgrade to add spectrum capacity group for a majority of this decade and even when needed. This saves time and money in beyond. The maintaining of a 500 HHP resizing the network such as node and service group is of immense value to the amplifier location and spacing. Adding MSOs. The ability to solve capacity network elements or changing network changes while maintaining the node size and element locations will impact cost for spacing enables an option for a drop-in electrical powering requirements. Ideally, capacity upgrade. the upgrade would touch the minimum number of network elements to reduce cost If the goal is to achieve 1 Gbps and time to market. In the section, the capacity upstream this may be achieved technologies, systems and architecture using a typical 500 HHP node service group options are explored. The paper will with 30 actives and 200 passives, and over 6 examine some of the pros and cons of miles of coax plant in the service area as several technologies and architectures, fully described later in this paper, see figure which could be used to provide additional 23. capacity. The existing 500 HHP node has Overview of Important Considerations long-term viability in 750 MHz or higher and Assumptions systems providing enough downstream capacity to last nearly the entire decade. In This report has highlighted some the upstream a 500 HHP node is predicted to important areas for network planners to last until mid-decade when the sub-split consider while making the decisions for the spectrum may reach capacity and then a next generation cable access network. choice of node split, node segment or add Avoidance of Small Node Service Groups or spectrum like mid-split to maintain the 500 FTTLA HHP service group are options. The The analysis and conclusions found physical 500 HHP node service group may in this report indicates that the need for remain in place with High-split (200)or Top- smaller node groups with few actives and Split with Mid-split providing 900 to 1 Gbps passives such as Node +3 or even Fiber to capacity. the Last Active (FTTLA) is not required to 1 GHz (plus) Passives - A Critical meet capacity, service tier predictions or Consideration for the Future network architecture requirements for this The industry will be considering decade and beyond. several spectrum splits and special consideration should be made to the most numerous network elements in the outside plant, the passives. Avoiding or delaying
modification to the existing passives will be avoid upgrades that may not be needed until a significant cost savings to the MSO. the 2020 era when the MSOs may pursue Below are key factors about the 1 GHz spectrum above 1 GHz. If the 1 GHz passives: passives are considered and the desired use is over 1 GHz we believe that 1050 MHz is Introduced in 1990 and were rapidly obtainable. There will be challenges with adopted as the standard AC power choke resonances, which may This was prior to many major impact the use of these passive greater than rebuilds of the mid-late 90s and early 1050 MHz with predictably. 2000s Prior even to the entry of 750 MHz The Value of Time optical transportandRF amplifiers/ The legacy STB out of band (OOB) products in the market place communications which uses spectrum in the Deploymentof 1 GHz passives that High-split area will be a problem for this would have more capacity than the split options; however a mid-split as the first electronics would have for nearly 15 step will provide sufficient capacity for years nearly the entire decade according to our Passives are the most numerous service and capacity predictions. The network element in the Outside Plant thinking is that another decade goes by and (OSP) the legacy STBs may be few or out of the Volumes are astounding perhaps as network all-together. If the STBs still many as 180-220 behind every remain in service another consideration is 500HHP Node or about 30 per every that these legacy STB may be retrieved and plant mile (perhaps 40-50 Million in relocated to markets than may not need the the U.S. alone) advanced upstream spectrum options. Yet, another consideration is a down conversion 1 GHz Passives may account for of the OOB communications channel at the 85% of all passives in service today homes that have legacy two-way non- Vendor performance of the 1GHz DOCSIS set-tops. Passives will vary and some support less than 1 GHz Overview Of Spectrum Splits Our internal measurements indicate that most will support up to 1050 The spectrum allocation options MHz should consider the impact to the overall Taps in cascade may affect capacity, end-to-end system architecture and cost. thus additional testing is required The solutions should also consider the timing of these changes as this may impact cost. The end-state architecture should be Assessment of the Passives considered for this next touch to the HFC. The authors believe that special We do not need to solve next decades consideration should be given to solutions problems now, however we should consider that leverage the existing passive. This will
them as part of the analysis. The MSO has split. In figure 20, the Top-split (900-1050) several spectrum split options available and optionshas a 150 MHz block of spectrum some are examined in this paper. The figure allocated for guard band between 750-900 below is an illustration of some of the MHz and 150 MHz block of spectrum spectrum split options; it also depicts a few between 900-1050 MHz for upstream. other options, such as Top-split with Mid-
FIGURE 20: SPECTRUM ALLOCATION OPTIONS
Mid-split DOCSIS QAM capacity approaching ~316 Mbps Overview Avoids conflict with OOB STB The Mid-split Architecture is defined Communications as 5-85 MHz upstream with the downstream Lowest cost option starting at approximately 105 MHz; this may High order modulation possible also be referred to as the 85/105 split. The 256QAM perhaps higher mid-split architecture essentially doubles the current upstream spectrum allocation The use of 256QAM translates to however this may triple or even quadruple fewer CMTS ports and spectrum the IP based capacity. The capacity increase (using 64QAM would require in data throughput is a result of the high- approximately 28% more CMTS order modulation and all of the new ports and spectrum) spectrum may be used for DOCSIS services, DOCISIS systems already support which is not the case with the sub-split this spectrum (5-85) spectrum that has generally accepted Some amplifiers support pluggable unusable spectrum and legacy devices diplexer filter swap consuming spectrum as well. Some existing node transmitters and headend receives may be leveraged Pros Does not touch the passives Sufficient bandwidth to last nearly Upstream path level control is the entire decade similar to the Sub-split (~1.4 times the loss change w/temp);
ThermalEqualizers EQT-85 enables Pros +/-0.5 dB/amp delta Operates effectively at a typical 500 HHP node group using 256QAM Cons (see details in the sections later in Impacts Video Service (in low this paper) channels) The use of 256QAM translates to Reduces low VHF video spectrum fewer CMTS ports and spectrum Throughput over 300 Mbps is less (using 64QAM would require than the newer PON technologies approximately 28% more CMTS Assessment ports and spectrum) The selection of Mid-split seems like DOCSIS QAM capacity approaching an excellent first step for the MSOs. This ~916 Mbps split option has little impact to the video DOCSIS OFDM capacity exceeds services and does not impact the OOB 1.1 Gbps STBcommutations. This spectrum split may Very low cost spectrum expansion last nearly the entire decade, allowing time option, especially considering similar for the MSOs to assess future splits, if capacity Top-split options (STB required, and the impacts to other split OOB cost was not considered in the optionat that time. analysis) Lowest cost per Mbps of throughput High-split (200) Some existing HFC Equipment Overview supports High-split like node The High-split (200) Architecture is transmitters and headend receivers generally defined as 5-200 MHz with the DOCISIS systems already support downstream starting at approximately 250- some of this spectrum (5-85) 258 MHz. Though other High-split options Passives are untouched may be considered above 200 MHz these High-split provides sufficient were not part of the examination. High-split upstream capacity and the ability to is being considered because full or partial maximize the spectrum with very analog reclamation is underway or planned high order modulation by cable operators. This will allow a High-split (200) does not waste a lot smoother transition when considering of capacity on guard band consumption of existing analog spectrum. Level control using Thermal As with mid-split DOCSIS 3.0 Equalizers EQT-200 (~2.2 times specifications systems may be used; Sub-split cable loss) however, to take advantage of thespectrum, additional development is required. Cons Conflicts with OOB STB Communications if DOCSIS Set-top box Gateway (DSG) is not possible
Takes away spectrum from Video upstream band. The total upstream capacity Services (54-258 MHz) may be either 187 MHz or 230 MHz Takes away spectrum from Video depending on the lower band frequency devices (TVs and STBs) return selected.The downstream would Potentially revenue impacting begin at either 54MHz or 105 MHz and because of spectrum loss supporting terminate at 750 MHz in the current analog video service tier specification.All of these architectures will Downstream capacity upgrade from share a 150 MHz guard band between 750- 750 MHz to 1 GHz to gain back 900 MHz, this may vary in the end-state capacity lost to upstream proposal howeverthese defined spectrum splits will be used for our analysis. The Top- Assessment split (900-1050) with Sub-split and with the The use of high-split would impact Mid-spilt option are compared in a table OOB Set-top Box communications for non- called Spectrum Allocation Comparison, DOCSIS Set-top Gateways were not figure 21. The placement of additional possiblein the upgraded service area. If the upstream atop the downstream has been deployment of High-split (200) is planned considered for many years. The Top-split later in time,this may allow these older (900-1050) approach may be similar to a STBs to be phased out. This split provides Time Warner Cable trial called the Full lots of bandwidth with a minimal amount of Service Network in the mid 1990’s, which is prime spectrum wasted for guard band. believed to have placed the upstream above the 750 MHz downstream. These are some If the main challenges with the use of the pros and cons of Top-split (900- of High-split are overcome, this seems like 1050): the ideal location for the new upstream(technically). The economics are Pros also compelling for High-split (200) against Operates effectively at a typical 500 the other split options considering just the HHP node group but with no more network access layer. If the STB Out of than 64QAM (see details in the Band (OOB) and analog recoveryneed to be sections later in this paper) factored into to the High-split, the cost Top-split with Sub-split DOCSIS analysis will change. QAM capacity ~700 Mbps given a Top-split (900-1050)with Sub-split or Mid- 500 HHP Node/Service Group split Top-split with Mid-split DOCSIS QAM capacity ~933 Mbps given a Overview 500 HHP Node/Service Group (equal A new spectrum split called Top- to High-split) split (900-1050)defines two separate Top-split with Mid-split DOCSIS spectrum bands, which may either use sub- OFDM capacity exceeds 1.1 split or mid-split plus the new spectrum Gbpsgiven a 500 HHP Node/Service region of900-1050 MHzfor a combined Group (equal to High-split)
With Sub-split ―no‖ video services, High-split has 12% or more capacity devices, and capacity is touched for revenue generation when With Mid-split has a ―low impact‖ to compared to Top-split (900-1050) video plus Mid-split, this is because the STB OOB Communications are not guard band requirements waste affected bandwidth Passives are untouched (only Top- Will require more CMTS ports and splitthat avoids touching passives) spectrum when lower order Existing 750 MHz forward modulation is used, perhaps 28% transmitters are leveraged more CMTS base on the efficiency comparison estimates found on Cons figures 10 and 11. No products in the market place to Upstream is more of a challenge determine performance or accurate compared to using that same cost impacts spectrum on the forward path The analysis estimates that Top-split Upstream is more of a challenge (900-1050) is about 1.39 times the compared to using that same costof High-split (200) with a 500 spectrum on the forward path (cable HHP node architecture loss ~5x Sub-split, 2.3x High-split; The analysis estimates that Top-split ~+/-1 dB/amp level delta w/EQTs is (900-1050) with Sub-split is about unknown) 2.4 times the costof High-split (200) Interference concerns with MoCA with a 125 HHP node architecture (simply unknown scale of impact but Achieving similar capacity of High- may affect downstream in same split (200) and with Top-split (900- spectrum range) 1050) with ―Sub-split‖ will require a 125 HHP service group (node) which Assessment is a major cost driver.(Note the use The Top-split (900-1050) options are of Mid-split with Top-split will being considered because option keeps the provide 900+ Mbps more than High- video network ―as is‖ when considering sub- split) split and has marginal impact if mid-split is Spectrum Efficiency is a concern used. The Top-split 900-1050 option has because of guard band (wasted additional benefits in that the Set-top box spectrum) and lower order out of band (OOB) challenge is avoided and modulation (less bits per Hz) this optiondoes not touch the passives. This resulting in lower throughput when Top-split is estimated to cost more than the measured by summing the upstream High-split; estimated at 1.3 to 2.4 times and downstream of Top-split (900- depending on the architecture selected. The 1050) and High-split using similar MSOs will just begin to evaluate this option spectral range(see figure 21). against the others.
Top-split (1250-1550) with Sub-split or Mid- Mbps given a 125 HHP split Node/Service Group With Sub-split ―no‖ video services, Overview devices, and capacity is touched The Top-split (1250-1550) With Mid-split has a ―low impact‖ to Architecture will be defined as part of the video 1250 – 1750 MHz spectrum band. In our STB OOB Communication is not analysis we limited the amount of spectrum affected allocated for data usage and transport to 300 Placing the upstream spectrum MHz and defined the placementin the 1250– beginning at 1250 MHz and up 1550 MHz spectrum band. The allocationof allows for the expansion of capacity 300 MHz provides similar capacity when without impacting the downstream compared to the other split option. The main consideration for this Top-split option Cons is that it avoids consuming existing Passives must be touched downstream spectrum for upstream and Smaller nodes or upstream Service avoids the OOB STB communication Groups perhaps a 125 HHP will be channel. required to approach or exceed the 1 Pros Gbps speeds comparable to High- May operate at a typical 500 HHP split (200) and Top-split (900-1050) node group but estimated to use Highest cost solution compared with QPSK, unless HHP is reduce to 125 High-split and Top-Split. then it is estimated the 16QAM may The Top-split (1250-1550) with Sub- be used (described in more detail in split is about 3 times the cost of the network architecture and cost High-split (200) for similar capacity sections of this report) without consideration to the DOCSIS Top-split 1250-1550 with Sub-split layer. DOCSIS QAM capacity ~500 Mbps Will require more CMTS ports and given a 500 HHP Node/Service spectrum when lower order Group modulation is used, perhaps 90- Top-split 1250-1550 with Mid-split 100% more CMTS base on the DOCSIS QAM capacity ~725 Mbps efficiency comparison estimates given a 500 HHP Node/Service found on Figure 11 compared to the Group use of 16QAM estimated at 2.7 Top-split 1250-1550 with Sub-split bps/Hz. DOCSIS QAM capacity ~912 Mbps No products in the market place to given a 125 HHP Node/Service determine performance or accurate Group cost impacts. Top-split 1250-1550 with Mid-split Return Path Gain Level Control: DOCSIS QAM capacity ~1136 (cable loss >6x Sub-split, 2.8x High-
split; +/-2 dB/amp w/EQTs is addto the cost of the solution. The use of unknown) lower order modulations will require more Interference concerns with MoCA CMTS upstream ports and more spectrum, (simply unknown scale of impact but which will impact the costs of the solution may affect downstream in same as well. Additionally, the conditioning of spectrum range) the RF components to support above 1 GHz may add to the costs of the solution. Assessment However determining the financial impacts The Top-split (1250-1550) with Sub- of performing ―Above 1 GHz plant split is about 3 times the cost of High-split conditioning‖ is unknown and was not (200).The placement of the return above considered in the financial assessment found 1 GHz requires the passives to be replaced later in this report. If we consider the or upgraded with a faceplate change. There service and network capacity requirements are approximately 180-220 passives per 500 for the upstream and downstream for the HHP node service group. It is estimated that next decade and beyond, the cable industry the node service group of 500 HHP may be should have sufficient capacity under 1 leveraged initially, however the GHz, which is the capacity of their existing requirements for higher capacity will force network. smaller node service group, which will
High-Split Top-Split Top-split (900- Top-split Top-split (1250- Name of Spectrum Split Mid-Split (200) (900-1050) 1050) & Mid-split (1250-1550) 1550) & Mid-split) 5-42 & 900- 5-54 & 1250- Upstream Spectrum Range 5-85 5-200 5-85 & 900-1050 5-85 & 1250-1550 1050 1550 Downstream Spectrum Range > 105 MHz > 258 MHz 54-750 105-750 54-1002 105-1002 Upstream Spectrum Bandwidth in MHz 80 195 187 230 337 380 Downstream Spectrum Bandwidth in 897+ 744+ 696 645 948 897 MHz Guard band Spectrum Allocation 20 58 162 170 260 268 (wasted spectrum between US/DS) Upstream PHY Layer Spectral Efficiency Capacity (assume QAM & 316 916 708 933 500 725 500 HHP SG) in Mbps Downstream PHY Layer Spectral Efficiency Capacity (assume QAM) in 5651 4687 4385 4064 5972 5651 Mbps Total PHY Layer Spectral Efficiency 5967 5603 5093 4997 6472 6376 Capacity Usable (Up+Down) None, Yes because Mid- Yes because Mid- Assumes Video Service Impact Yes Yes split used in None split used in existing 750 Lowband Lowband System Video Spectrum Loss Location 54-105 54-258 750-860 54-105 None 54-105 (assuming 54 MHz-860MHz usable) Assumes Video Spectrum Loss in MHz 51 Mhz 204 MHz existing 750 51 MHz None 51 MHz (assuming 54 MHz-860MHz usable) System OOB STB Communications ANSI/SCTE 55-2 2008 [4] (70 - 130 MHz) Not likely, however some STBs ANSI/SCTE 55-1 2009 [5] (70 - 130 may be hard coded within the MHz) Impacted No No No No mid-split range (75.5 and Some STBs may be hard coded within 104.25 MHz) the mid-split range (75.5 and 104.25 MHz)
Estimated Node Service Group Size per Return Laser in HHP (estimate & 500 500 500 500 500 500 may vary) Maximum number of Actives 30 30 30 30 30 30 Supported (estimate & may vary) Maximum number of Passives 200 200 200 200 200 200 Supported (estimate & may vary) Headend Optical Transmitter (Requirements, Replacement, Up to MSO Up to MSO Up to MSO Up to MSO Up to MSO Up to MSO Leverage) Headend Optical Receivers May Be (Requirements, Replacement, May Be Leveraged Replace Replace Replace Replace Leveraged Leverage) Nodes Optical Side (Requirements, May Be May Be Leveraged Replace Replace Replace Replace Replacement, Leverage) Leveraged Node RF Side (Requirements, Replace Replace Replace Replace Replace Replace Replacement, Leverage) Best Case: Amp is removed from service if pluggable Amplifiers (Requirements, diplexer filter swap is supported Replace Replace Replace Replace Replace Replacement, Leverage) (this is not a field upgrade). Worst case replace the Amp Passives (Requirements, Leverage Leverage Leverage Leverage Replace Replace Replacement, Leverage) House Amplifiers (Requirements, Replace Replace Replace Replace Replace Replace Replacement, Leverage) Achieve Cable Modem Value not to Yes Yes Yes Yes Yes Yes exceed 65 dBmV CPE Cost Impacts $ $$ $$$ $$$ $$$ $$$ Interference Concerns with MoCA [6]: MoCA 1.0 and 1.1 Operating frequency 850 – 1500 MHz No Yes Yes MoCA 1.1 Annex Interference Interference Yes Interference Yes Interference No Interference Concerns with Interference Expanded operating frequency of 500 Concerns Concerns Concerns with New Concerns with New New Upstream Concerns with MHz—1500MHz with New with New Upstream Upstream New Upstream MoCA 2.0 Upstream Upstream Expanded operating frequency from 500 MHz – 1650 MHz FM Radio Band, DTV and Aeronautical frequencies - avoidances of these Yes No No bands reduces the overall spectrum Interference Interference No Interference No Interference No Interference Concerns with Interference bandwidth available for data services. Concerns Concerns Concerns with New Concerns with New New Upstream Concerns with Areas affected may have lower order with New with New Upstream Upstream New Upstream modulation and smaller service group Upstream Upstream to attain desired capacity level. [4], [5], [6] FIGURE 21: SPECTRUM ALLOCATION COMPARISON
return path amplifiers that combine signals Characterization of RF Components onto a single return path (for a non- segmented node). For now we will ignore The network components that most the gateway noise, since in principle it could affectsignals carried above 1 GHz are the be made zero, or at least negligible, by only coaxial cable, connectors, and taps. The having the modem return RF amplifier characteristics of these components are turned on when the modem is allowed to critical, since the major goal in a next ―talk‖. generation cable access network is to leverage as much of the existing network as The RF return path amplifier noise possible. funneling effect is the main noise source that must be confronted; and it cannot be turned Before getting into the specifics off! This analysis is independent of the about the RF characterization and frequency band chosen for the ―New Return performance requirements, it is worthwhile Band‖ (e.g., Mid-split 5-85 MHz; High-split to establish the quality of signals carried 5-200 MHz; or Top-split with UHF return), above 1 GHz and below 200 MHz. The although the return path loss that must be bottom line is that while return path signals overcome is dependent on the highest can be carried above 1 GHz, they cannot be frequency of signals carried. For a first cut carried with as high order modulation as is at the analysis, it suffices to calculate the possible at lower frequencies. For example, transmitted level from the gateway required if the goal is to meet similar return path data to see if the levels are even possible with capacity the signal carriage above 1 GHz is readily available active devices. The obvious possible using 16QAM with about 300 MHz way to dramatically reduce the funneling of RF spectrum(47 channels of 6.4 MHz noise and increase return path capacity is to each). Whereas below 200 MHz 256QAM segment the Node. That is not considered is possible (due to lower coaxial cable loss) here to assess how long the network remains and only 24 channels occupyingabout 180 viable with a 4x1 configuration, a 500 HHP MHz spectrum is required, using rough node service group. estimates. Additionally, the over 1.2 GHz solutions may require a 125 HHP service The thermal mean-square noise group to support 16 QAM, where as the 200 voltage in 1 Hz bandwidth is kT, where k is MHz solutions may use a 500 HHP service the Stefan-Boltzmann constant, 1.38x10^-23 group, this is a key contributing factor to the J/deg-K, and T is absolute temperature in cost deltas of the split options. degrees Kelvin. From this we have a thermal noise floor limit of -173.83 dBm/Hz. For a Path Loss and SNR bandwidth of 6.4 MHz and 75-ohm system, In a typical HFC Node + N this gives -57.0 dBmV per 6.4 MHz channel architecture, the return path has many more as the thermal noise floor. With one 7 dB sources for extraneous inputs, ―noise‖ than noise figure amplifier in the chain, we the forward path. This includes noise from all the home gateways, in addition to all the
wouldhave a thermal noise floor of -50 dBmV/6.4 MHz channel.
Two amplifiers cascaded would give 3 dB worse, four amplifiers cascaded give 6 dB worse than one. And since the system is FIGURE 22: MODULATION AND C/N balanced to operate with unity gain, any PERFORMANCE TARGETS amplifiers that collect to the same point also The table below, figure 23, increase the noise floor by 10*log(N) dB, documents many important assumptions and where N is the total number of amplifiers in assumed node configuration conditions. An the return path segment. For a typical important assumption is the CM maximum number of 32 distribution amplifiers power output level of +65 dBmV into 75 serviced by one node, this is five doubles, or ohms. What this means is that if many 15 dB above the noise from one RF channels are bonded (to increase the amount Amplifier, or -35 dBmV/6.4 MHz of data transmitted), the level of each carrier bandwidth. The funneling effect must be must be decreased to conform to the CM considered in the analysis for the NG Cable maximum power output constraint. Two Access Network. channels bonded must be 3 dB lower each; If the return path signal level at the four channels must be 6 dB lower than the node from the Cable Modem (CM) is +15 Pout(max). Since the channel power levels dBmV, it is clear that the Signal-to-Noise follow a 10*log(M) rule, where M is the Ratio (SNR) in a 6.4 MHz bandwidth is 50 number of channels bonded to form a wider dB; very adequate for 256QAM or even bandwidth group. For 16 channels bonded, higher complexity modulation. But if the each carrier must be 12 dB lower than the Return path level at the node port is 0 Pout(max). dBmV, the SNR is 35 dB; this makes For 48 channels bonded, each must 256QAM theoretically possible, but usually be 16.8 dB lower than the Pout(max). So for at least 6 dB of operating margin is desired. 48-bonded channels, the level per channel is If only -10 dBmV is available at the node at most 65 dBmV -17 dB = +48 dBmV. If return input, the SNR is 25 dB; and so even there is more than 48 dB of loss in the return the use of 16QAM is uncertain.This path to the node return input, the level is <0 illustrates (figure 22) the very high dynamic dBmV and 64-QAM or lower modulation is range of ―Pure RF‖ (about 15 dB higher than required. The node and system configuration when an electrical-to-optical conversion is assumptions are as follows. involved).
FIGURE 23: TYPICAL NODE ASSUMPTIONS
Cable Loss Assessment seen below, the loss at 2 GHz is about 5% higher than expected by the simple sq-rt(f) rule. The graph below illustrates a slightly lengths of 1/2‖ Two different more than twice the loss at 2 GHz compared diameter hardline coax were tested for to 500 MHz, see figure 24. Insertion Loss and Return Loss (RL). The loss versus frequency in dB varied about as the square root of frequency. But as can be
Slightly more than twice the loss at 2 GHz compared to 500 MHz.
FIGURE 24: DISTRIBUTION COAXIAL CABLE – INSERTION LOSS VS. FREQUENCY
Supplier A
FIGURE 25: DISTRIBUTION COAXIAL CABLE – RETURN LOSS VS. FREQUENCY
In the plot (figure 25),the coax manufacturers, values, and number of Return Loss (RL) did not vary as expected outputs. Most were designed more than ten above 1200 MHz. This appears due to an years ago, well before >1 GHz bandwidth internal lowpass matching structure in the systems were considered. One of the serious hardline-to-75N connectors (apparently for limitations of power passing taps is the AC optimizing the 1-1.2 GHz response). The power choke resonance. This typically is connectorsare an important element to return around 1100 MHz, although the ―notch‖ loss with signals above 1 GHz. frequency changes with temperature. Tap response resonances are typical from ~1050 to1400 MHz.A limitation of power Tap Component Analysis passing taps is the AC power choke resonance. This is an important finding when leveraging the existing passives; Taps are the components with the therefore the use above 1050 MHz may not most variability in passband characteristics, be predictable or even possible. because there are so many different
FIGURE 26: 27 DB X8 TAP - INSERTION LOSS VS. FREQUENCY: ALL PORTS
FIGURE 27: 27 DB X 8 TAP - RETURN LOSS VS. FREQUENCY: ALL PORTS
Even the newer, extended bandwidth in the 1050 MHz to1300 MHz range. taps, with passband specified 1.8 GHz or 3 Especially on the tap coupled port. GHz, the taps usually have power choke However, most Taps work well to ~1050 resonances (or other resonances, e.g., MHz. inadequate RF cover grounding) resonances
FIGURE 28: 11 DB X 2 TAP- INSERTION LOSS VS. FREQUENCY
FIGURE 29: 11 DB X 2 TAP - RETURN LOSS VS. FREQUENCY
Nearly all taps exhibit poor RL characteristics on all ports above 1400 MHz. Some are marginal for RL (~12 dB), even at 1 GHz. Therefore tap cascades must be tested and over temperature to verify theactual pass band response due to close by tap reflections.
HFC Optical Return Path Transport up carefully to produce the desired RF Architecture output at the receiver when the expected RF level is present at the input of the As we have analyzed several areas transmitter. Any change in the optical link of the network to assess the impact and budget will have a dramatic impact on the requirements to support additional upstream output level at the receiver unless receivers capacity we will now turn to the optical with link gain control are used. Also, as portion of the HFC network. The optical with any analog technology, the layer will be examined to support the performance of the link is distance additional upstream capacity. We will look dependent. The longer the link, the lower at two classes of HFC optical transport, the input to the receiver, which delivers a analog return path and the second type is lower C/N performance. The practical digital return, which is commonly referred distance over which an operator can expect to as Broadband Digital Return (BDR). to deliver 256QAM payload on analog Overview - Analog Optical Return Path return optics is limited. Analog return path transport is Assessment accomplished with a Distributed Feedback The analog return transmitter will (DFB) laser located in the node housing and work well for the low and high frequency an analog receiver located in the headend or return. Analog return path options should be hub. Analog return path transport is available for the higher frequency return considered as a viable option for sub-split, options at 900-1050 MHz and 1200-1500 mid-split, and high-split returns. MHz. However the cost vs. performance at Pros these frequencies when compared to digital The chief advantage of this method alternatives may make them less attractive. is its cost effectiveness and flexibility. If the There will be distance limitations and analog return optics are in use in the field EDFAs will impact the overall system today, there is a good chance that they will perform noise budgets. The distances of 25- perform adequately at 85 MHz and even 200 30 km are reasonable and longer distance MHz loading, if required in the future. This would be supported. would allow an operator to fully amortize Overview - Digital Optical Return Path the investment made in this technology over Digital return path technology is the decade. commonly referred to as broadband digital Cons return (BDR). The BDR approach is There are drawbacks to using analog ―unaware‖ of the traffic that may be flowing optics. Analog DFB’s have demanding over the band of interest. It simply samples setup procedures. RF levels at the optical the entire band and performs an analog to receiver are dependent on optical digital conversion continuously, even if no modulation index and the received optical traffic is present. The sampled bits are level. This means that each link must be set delivered over a serial digital link to a
receiver in the headend or hub, where a Return Receivers aren’t easily converted to digital to analog conversion is performed new passbands. It requires ―forklift and the sampled analog spectrum is upgrades‖ (remove and replace) of these recreated. optics when moving to 85 MHz and 200 MHz return frequency. There is currently no Pros standardization on the Digital Return There are a number of advantages to modulation and demodulation schemes, or the BDR approach. The output of the even transport clock rates. receiver is no longer dependent on optical input power, which allows the operator to Assessment make modifications to the optical It is more difficult and therefore multiplexing and de-multiplexing without more costly to manufacture BDR products. fear of altering RF levels. The link This may be a driver to use DFB products performance is distance independent – same for the new returns. The selection of BDR MER (magnitude error ration) for 0 km as products may be driven by distance and for 100 km. The number of wavelengths performance requirements. Another driver used is not a factor since on/off keyed digital to move to BDR will be when there is near modulation only requires ~20dB of SNR; cost parity with DFB, today this is the case thus fiber cross-talk effects do not play a with the 5-42 MHz optical transport systems role in limiting performance in access-length and this may be the case in the future with links (<100 km) the new spectrum returns.
The RF performance of a Digital Review of HFC / Centralized Access Return link is determined by the quality of Layer Architecture the digital sampling rather than the optical The HFC has been around for nearly input to the receiver, so consistent link performance is obtained regardless of two decades and has evolved to include optical budget. The total optical budget many technologies and architectures. Some capability is dramatically improved since the of these include analog and digital optical optical transport is digital. This type of transmission technologies and HFC transport is totally agnostic to the type of architectures such as Node+N, Node+0, traffic that flows over it. Multiple traffic RFoG, QAM overlay, full spectrum, etc. But classes (status monitoring, set top return, regardless of HFC technology and DOCSIS, etc) can be carried simultaneously. architecture selectedthe function of an HFC class of network remains constant, it has Cons always been a ―media conversion The chief drawback to BDR is the technology" using analog, and today digital fact that nearly all equipment produced to methods,to joindissimilar media types like date is designed to work up to 42 MHz. fiber and coaxial. The HFC architecture Analog receivers are not useable with being a ―conversion technology‖ allows the Digital Return transmissions. Further, the outside plant to remain relatively simple and analog-to-digital converters and Digital very flexible to changes at the MAC and
PHY layers. The HFC architecture is also a change the HFC architecture. HFC ―centralized access layer architecture‖ where remained simple and carried the next all of the MAC/PHYprocessing takes place generation data technology through it at the headend. transparently; these are clear examples of its flexibility. Additionally, HFC may have The HFC architecture has proven to carried all of these technologies be a valuable asset for the MSO, enabling simultaneously to support a seamless the evolution to next generation access layer migration to the next generation; a clear technologies while avoiding changes to the example of versatility! The evolution of the HFC layer of the network, with the cable network primarily is achieved by exception of adding spectrum and capacity. changing the bookends and not the plumbing Examples of this transition include analog in-between. It is hard to imagine the impact video systems, digital video systems, if the outside plant, such as nodes, needed to EQAM, UEQ, SDV, CBR voice systems, be changed to support each next generation pre-DOCSIS data systems, DOCSIS 1.0, MAC/PHY technology that came along over 1.1, 2.0, 3.0 and so on. This entire multi the last 20 years. decade transition did not fundamentally Broadband HFC HFC Node Actives Passives Systems Optical Systems
Optical Xmtr Optical Rcvr
Broadcast C o D Services m
Filter i Optical Xmtr Optical Rcvr b
Highpass i p n Network e l r Access / e Aggregation x Optical Rcvr Optical Xmtr e (CMTS/CER) r
Coax Fiber Coax
Layer 2+ Media Layer 1 Optical Transport Media or Digital MAC/PHY Conversion Carries RF Signals Conversion Coax Carries RF Signals
MAC/PHY Optical Systems Fiber Optical Node Performs Layer Separate Performs Media or Carries Media or Digital than Optical Digital Conversion Analog or Conversion Transport Digital Layer Optical Layer is Modulated Optical Layer is Simple Simple and with Signals and with no MAC/PHY no MAC/PHY Layer Layer and limited or no optical OAM FIGURE 30:HFC A “MEDIA OR DIGITAL CONVERSION ARCHITECTURE”
Introduction to DFC (Digital Fiber Coax) The underpinning of this style of / Distributed Access Layer Architecture distributed access layer architecture is not new and goes back to the CMTS in the node As we examine the future to support discussions in the late 1990’s and early higher IP upstream data capacity and a 2000’s for the cable industry. These transition of the downstream to more IP concepts arose when the DOCSIS capacity for data and IPTV, the underlining technology was emerging to replace architecture of HFC and centralized access proprietary data and CBR voice systems and layer may be placed into question. This is was also considered during a period when certainly nothing new. With each major HFC upgrades were still underway or shift in technology or major investment planned. These distributed access layer planned, the question of centralized vs. architectures were discussed again in the late distributed appears. We will examine this 2000’s when DOCSIS 3.0 was emerging, class of architecture we are calling Digital this time referred to as M-CMTS (modular- Fiber Coax (DFC). CMTS) or P-CMTS (partitioned CMTS), The Digital Fiber Coax Architecture which could place some CMTS functions in label is for a network class which differs the node, perhaps just the PHY layer. from HFC in that MAC/PHY or just PHY Again, in late 2010 with the announcement processing is distributed in the outside plant of DOCSIS-EoC (Ethernet over Coax) from (node) or MDU and also uses ―purely Broadcom, placing the CMTS in the node or digital‖ optical transport technologies such MDU revived the industry debate. Though as Ethernet, PON, or others to/from the DOCSIS-EoC is mainly focused on the node. The industry may determine to call China and worldwide MDU market. In this class of architecture something else, but addition to the CMTS in the node, the cable the functions, technology choices and industry has considered QAM in the Node architectures are different than HFC. as well.
As described in the following The diagram below illustrates an sections there are many technologies and example of Digital Fiber Coax Class of architectures that could all be categorized Network Architecture using CMTS as the asthe class of architecture we defineas coax technology and PON or Active Digital Fiber Coax (DFC). This term may Ethernet as the optical transport to the node. certainly change and is just used for The illustration is a node but could be an discussion purposes within this document, MDU. Additionally any optical technology however it is clear that the functions of DFC or coaxial data technology could be are not similar to HFC, the industry should employed, as discussed in detail below. consider naming this class of architecture.
HE Broadband HE Optical Node Actives Passives Systems Systems
X P-CMTS - Network Ethernet DFC C Access / Optical (DS) o pD Narrowcast m li Aggregation (PON, And/or b Fiber i ep (CMTS/PON/ Ethernet P-CMTS n l (US) e x CER) etc) r e rx e r New DFC Node Functions
Optical Layer 2+ MAC/PHY (PON, AE, other) DFC Node Function Coax Carries RF Signals
(has more than Media or Digital Conversion) Optical MAC/ The is “No Media Fiber Carries Optical Node PHY is the Conversion Digital Performs MAC/PHY Optical Transmission or just PHY Distribution Optical Layer is Functions Transport not Analog Layer for DFC Modulated or Coax portion Layer Digital HFC Optics is a P-CMTS some portion of CMTS Functions at node FIGURE 31: DIGITAL FIBER COAX A “PHY OR MAC/PHY PROCESSING ARCHITECTURE”
The actual development of a CMTS CMTS-like functions in the node or MDU in the node, referred to as DOCSIS EoC by gateway. Broadcom, is a direct response to several other technologies morphing to be used as a The functions of EoC technologies coaxial access layer technology. While are similar in many ways to DOCSIS, as DOCSIS was designed from the ground up most have a device, which functions like the to be a cable access layer technology, the CMTS, a central controller for scheduling only architecture that a DOCSIS system was network resources in a multiple or shared designed for was a centralized access layer access network with end points like approach, to be carried over an HFC modems. The EoC architecture uses a fiber network. The centralized access layer connection, likely Ethernet or PON to the approach is a valuable approach to MSOs node or MDU, where this transport is that have full two-way HFC and customers terminated and the data is carried to/from spread over vast outside plant areas. the CMTS-like function in the node/MDU However, not all MSOs worldwide have full and to/from customers over the coax. two-way, high capacity, and suitable HFC Ethernet over Coax may be networks for data service; thus making a considered as an access layer technology centralized CMTS architecture a challenge. where many consumers gain access to the A network architecture or suite of service provider’s network. However, some technologies used over coax referred to as of the technologies in the EoC space may ―Ethernet over Coax‖ (EoC) emerged to have started as home networking compete against DOCSIS; and the technologies such as MoCA, BPL, architecture was distributed, placing the HomePlug, HPNA, G.hn, HiNOC, WiFi over Coax, and more. The placement of any
of these technologies in a node to interface Architecture Options for DFC with coax in our view is not HFC style The DFC architecture could consist architecture, but rather DFC style of MAC/PHY or simply PHY functions in architecture, as the MAC and the PHY the Node. The architecture could support processing takes place at the node. downstream and upstream functions or just a single direction. Overview - DFC is a New Architecture Class for Cable Examination of DFC with EPON and P- There are two different Fiber to the CMTS Node (FTTN) architectures, which utilize We have examined many layers of coax as the last mile media. If we consider the network architecture and considered HFC an architecture class with several many approaches for upstream spectrum technologies and architectures that may be expansion and performance as well as employed, the same could be applied for the optical transport in HFC style architectures. DFC architecture class. We wish to consider a distributed access layer architecture approach the digital fiber To simply summarize the delta coax (DFC) style architecture. between HFC and DFC Architecture Classes: The differences have been defined already between HFC and DFC. In addition, HFC is a―Mediaor several technology and architecture choices ‖ DigitalConversion Architecture that could be grouped under the DFC Class DFC is a―PHY or MAC/PHY of Architecture were also covered. This Processing Architecture‖ section examines the use of DFC style architecture. Technology Optionsfor DFC The DFC class of architecture could The DFC Architecture selected as an use several optical transport technologies example in figure 32, illustrates 10G/10G to/from the Headend link to the node; this is EPON as the optical transport placing the called ―Ethernet Narrowcast‖. The optical optical MAC/PHY in the optical node and technologies could employ PON, Active the second selection is DOCSIS as the RF Ethernet, G.709, or others to carry data and technology. The architecture is an upstream management communications to/from the only PHY in the node. node. Consider then the use of HFC and The DFC class of architecture could DFC to support the legacy and new use several coaxial-based MAC/PHY architecture simultaneously, this approach technologies such as DOCSIS, Edge QAM may be referred to as the DFC Split Access MPEG TS, MoCA, BPL, HomePlug, Model, as illustrated in figure 32. The HFC HPNA, G.hn, HiNOC, WiFi over Coax. is usedto support legacy transport technology, services, and most importantly the centralized access architecture for the
downstream such as the very high capacity layer systems, like a CMTS have highly CMTS/UEQ and the plus side is the massive redundant systems. and existing downstreamoptical transport is leveraged. There is no need to place The DFC style architecture is used downstream RF MAC or PHYs in the node for two-way high capacity optical in most configurations. transmission to the node (like 10G Ethernet or 10G EPON) however initially this In the DFC split access model, the architecture will just consider using the HFC upstream optical transport is leveraged upstream for the expanded coax upstream, as well, which may include the Sub-split 5- see figure 32. The Partitioned CMTS (aka 42 MHz band and even perhaps Mid-split. P-CMTS) using upstream only is examined The HFC upstream optical transport will in this paper, however additional support a centralized access layer. The HFC downstream capacity could use the existing with centralized access layer may be optical connection and the placement of a considered as the high availability future P-CMTS for the downstream could be architecture, because the OSP performs just added later if needed, perhaps over 1 GHz media conversion and the centralized access spectrum range.
Fiber MAC/PHY Layer Optical Systems Performs Carries Separate than Media or Digital Conversion Analog or Optical Optical Layer is Simple and Digital Optical Node Performs Transport with no MAC/PHY Layer and Modulated Media or Digital Layer limited or no Optical OAM Signals Conversion
Layer 2+ Media Layer 1 Optical Transport Media or Digital MAC/PHY Conversion Carries RF Signals Conversion
Optical Xmtr Optical Rcvr
Broadcast s
Services s r
a C e p
t o l i DS MAC/PHY h m
F g g DS MAC/PHY Optical Xmtr Optical Rcvr i b n H i i X DS MAC/PHY h n
c e
t - DS MAC/PHY i r
w p
DS MAC/PHY S l US MAC/PHY n Optical Rcvr Optical Xmtr o
i e US MAC/PHY t Coax c
e x
US MAC & PON t
o e US MAC & PON r P (Future) r US MAC & PON Fiber DS PHY 1 – 10G CER Ethernet or US PHY PON, etc Open Ethernet Narrowcast Overlay (PON, AE, other) PHY or MAC/PHY Coax Carries RF Signals Processing Optical MAC/PHY Optical Layer is Optical Optical carries Digital packet is the Optical Intelligent with Connection streams and the Coaxial Distribution MAC/PHY Layer Carriers Digital portion contains the MAC/PHY Transport Layer and with Optical packet or PHY Layer (Not Analog or for DFC OAM&P streams not Digital Conversion of Spectrum BDR style blocks ADC/DAC) FIGURE 32: DIGITAL FIBER COAX (DFC) SPLIT ACCESS
Defining an architecture that placed the flexible, economical and will keep the downstream PHY in the node was not outside plant (nodes) as simple as possible. financially prudent since the HFC forward The P-CMTS using just the PHY and just already exists and has massive optical the upstream was selected to keep costs capacity. Centralized access layer down as much as possible and may only be architecturesover HFC have proven to be used ifconditions would require a distributed
architecture. An example could be in same as the advantages of BDR with the situations where the use of an extremely exception of begin agnostic to the traffic. Its long distancelink above 50 km or much performance would be slightly better than higher is required. We believe the MSOs BDR since the A to D and D to A will want to keep the outside plant (OSP) as conversions are eliminated. simple as possible for as long as possible; this has proven to be a very The DFC style architecture would valuablecharacteristicfor 50 + years. use optical technology that would have less Placing the entire CMTS in the node was not configuration and two-way transport considered prudent because of the very high capacity with optical monitoring. The other capital cost and it is least flexible for the advantage is distance and performance when future. compared with HFC style optical transmission. The DFC style architecture Description of P-CMTS (upstream only) could be a QAM narrowcast overlay This paper has created a few new competitor, where typically optical links are labels for cable networking architectures just long, costly, and challenging to optically for use within the paper. Another term that configure. was created in a 2008 whitepaper but perhaps not well known is the Partitioned Cons CMTS (aka a P-CMTS) [6]. The P-CMTS The disadvantages of the DFC style proposes to remote the DOCSIS PHY sub- architecture in this case using P-CMTS system from the core CMTS chassis. The approach are that the solution would only primary reason for the P-CMTS is that it can work with DOCSIS returns (or the specific potentially permit the PHY sub-systems to demodulations for which it was ‖ of be located a long distance from the MAC programmed) and would be ―unaware sub-systems, such as the node, and use the any other traffic that may exist on the optical transport options defined above for network. Digital Fiber Coax to/from the node. The Placing MAC/PHY or just PHY digital packet streams carrying the DOCSIS- functionsin the node may be difficult to encapsulated packets are transmitted across change as new technology becomes the optical link between the external PHY available. This should be a very important (node) and the core CMTS chassis. consideration for MSOs as they reflect on Pros the MAC/PHY technology changes in just The advantages of this approach are the last 10-15 years, as described in the that the system would perform as if there preceding HFC architecture section. The were no optical link at all. The RF plant thought of touching every node to make a would essentially connect directly to the MAC or PHY change may be unthinkable to CMTS upstream port in the node as if it some operators. would have been in the headend. The Costs is an additional concern, when advantages of this approach would be the placing MAC/PHY or just PHYs in the node
this means each node may need to be across many service groups expanding and configured with enough capacity to meet the contracting service group size at the servicetier offeringand traffic capacity DOCSIS layer where and when needed. estimates up front. Another option is to Placing the CMTS functions of any kind or make the node configurable to add capacity, any MAC/PHY or PHY technology may this would mean visiting each node when have higher start-up costs and total cost of additional capacity is needed.The MSOs ownership could be a challenge. More study have typically allocated capacity in the in needed to determine the viability of this largestserving area possible, to gain type of architecture. Placing the intelligence economies of scale and additional capacity. in the node and distributing the architecture may limit future flexibility. Additional There may also be performance concerns of power and space will need to be concerns with TCP latency with distributed explored if remote DOCSIS P-CMTS architectures. The reliability and Upstream is explored. redundancy is also a consideration, there are more active components in the field.
Overall Assessment of DFCStyle NETWORK MIGRATION ANALYSIS AND Architectures STRATEGIES The HFC Architecture enables centralized access layer architectures; and This section provides analysis of someof the migration strategies. It is very DFC enables distributed access layer important to note that the starting points of architecture. As discussed above, the the MSO will greatly influence the selection industry since the 1990s has examined the of a particular network technology and placement of intelligence in the nodes; like architecture path. Additionally, the network CMTS and since then different part of the utilization and capacity planning forecast in CMTS has been considered for the node, the local market will be a major driver for Edge QAM, PON, and many other the migration strategy selected and timing technologies have been considered. by the cable operator. These are some of the The main consideration for DFC factors that may influence the operator’s style architecture may be the very long links selection path: of QAM overlay in the forward and in the future the new high capacity upstream. In Competitive and user consumption levelHigh-Speed data those markets, there may be benefits from a Video services offering DFC style architecture, but more study is Deployment level of STB which use needed and there are trade-offs. Proprietary OOB or DSG Deployment level of DTA in market MSOs using HFC and DOCSIS have All Digital Offering benefited from leveraging DOCSIS capacity Desire to offer analog service tier
Current system spectrum capacity The Top-split analysis for a 500 level (750 MHz or 1 GHz) HHP node has 64QAM as possible for Top- split (900-1050) and QPSK for Top-split Downstream Migration Analysis (1250-1550) but neither achieve the capacity The findings of this report illustrate of High-split. Top-split (900-1050) with the MSOs existing downstream capacity is Mid-split in a 500 HHP node may reach the sufficient for this decade and beyond at the capacity of High-split. spectrum level 750 or equivalent assuming The Top-split migration to a 125 the upstream split options. There will be HHP Service Group, reduces the funneling reclamation of the analog service tier and noise by 6 dB, which permits 256QAM for spectrum, reduction in the service group Top-split (900-1050) and 16QAM for Top- size, and reallocation of the distribution split (1250-1550) and both of these Top-split network from MPEG TS to IP.The options coupled with Sub-split will yield the downstream migration will be managed capacity of High-split. mainly from the headend systems and CPE migration to IP. Additionally, investment in Below are some high-level full spectrum to each node will be need this considerations for the migration and split decade and an additional forward transmitter selection. to the node service group based on our service and capacity projections. Consider Mid-split first (This buys about a decade and churns out old Upstream Migration Analysis STBs to avoid the OOB) Consider an eventual High- The cable industry has existing splitupgrade path capable of 200-300 spectrum capacity and channel bonding MHz capability in the upstream that will meet If there are concerns with video service, STB OOB, and want to their service and capacity needs for many avoid touching the passives, consider years to come, perhaps 2015-2017. Top-Split (900-1050)with Mid-split and a 500 HHP node as this has same A migration to Mid-split first in the capacity as High-split (200). 2015 timeframe or when the capacity of Consider reserving the above 1 GHz Sub-split is exhausted by the technical and for the next decade. business analysis a good first step. A migration to Top-split (1250- 1550) with Sub-split will require a The next choice is High-split or a 125 HHP service group to have the Top-split option. All of these solutions may capacity of High-split. start with 500 HHP node service group, but this will depend on the capacity requirement and split option selected.The service tier, customer traffic, and the node service group network elements will influence the service group size.
COST ANALYSIS which supports a smooth and economical transition while assuring The report has covered some of the revenue targets per 6 MHz channel key inputs and levers to begin to assess the for the MSO are met costs of the upstream spectrum options. The High-Split 200 may requirea forward underlining requirementsof the network laser upgrade to 1 GHz to gain the architecture to meet the capacity targets spectrum lost to upstream have been documented. In this section some Top-split (900-1050) or Top-split of those key technical assumptions will be 1250-1550 will leverage existing covered to provide perspective as to the lasers for downstream drivers and considerations in the cost Full Spectrum to a 500 HHP Node to analysis. The cost analysis makes some 250 HHP segmentation is expected estimates to cost for products that have not within the decade yet been invented, so these are rough QAM overlay solutions may need to estimates used for discussions purposes migrate to full spectrum only. The actual relationship between the upstream migration options may vary. Passive changes may be avoided for entire decade and beyond The HFC network allows operators FTTLA may be avoided for entire to employ a number of methods to manage decade and beyond the abundant downstream spectrum. Some of these options include analog reclamation, UpstreamCost Analysis switched digital video delivery of multicast The following analysis is focused on content and service group size management, a number of upstream options. The mainly achieved by node segmentation or feasibility and relative cost of each path is node splitting to achieve the desired ratios. compared. The analysis assumes a ―typical‖ The relative costs and benefits of these HFC node has the following characteristics downstream augmentations are well shown in Figure 33. understood and used extensively today. As shown in the previous sections, the upstream Beginning with a 500 home passed spectrum has been sufficient to meet the node, the first approach was to determine demand, but before the end of the decade, what might be possible without having to additional spectrum may be required, in disturb the layout of the physical plant. some MSO markets. Figure 34 shows what the gain Downstream Cost Analysis requirements (excluding port, EQ losses) would be for an upstream amplifier at the Converged Edge Router ranges of operating frequencies reviewed DOCSIS/Edge QAM device will earlier in this paper. For this analysis, 0.75‖ enable an effective migrations PIII class cable was assumed for express The Downstream may leverage a 6MHz by 6MHz channel investment
amplifier spans and 0.625‖ PIII class cable was assumed for tapped feeder spans. Typical Node Assumptions Homes Passed 500 Home Passed Density 75 hp/mile Node Mileage 6.67miles Amplifiers/mile 4.5 /mile Taps/Mile 30 /mile Amplfiers 30 Taps 200 Highest Tap Value 23 dB Lowest Tap Value 8 dB Largest Amplifier Span 2000 ft Largest Feeder Span 1000 ft Largest Drop Span 150.0 ft Home Split Loss to Modem 4 dB Maximum Modem Power 65 dBmV FIGURE 33: GENERAL NODE ASSUMPTIONS
Sub-Split Mid-Split High-Split 200 Top-Split (900-1050)Top Split (1250-1550) Upper Frequency MHz 42 85 200 1050 1550 Typical Maximum Cable Loss (Amp to Amp 70 deg F) dB 7.1 10.1 15.5 35.5 43.1 Aditional Gain Required for Thermal Control (0 to 140 deg F) +/-dB 0.5 0.7 1.1 2.5 3.0 Total Reverse Amplifier Gain Required dB 7.6 10.8 16.6 38.0 46.1
FIGURE 34: RETURN AMPLIFIER GAIN CALCULATION
It is worth noting that the Sub-split, split solutions. Reverse RF AGC systems Mid-split and High-split gain requirements do not exist today, and could be complex can be satisfied with commonly available and problematic to design. Thermal components that are currently used in equalization would be sufficient to control amplifier designs today and would likely the expected level changes at 200 MHz and involve no cost premium. However, the below, but it is not certain that thermal Top-Split options would likely require equalization alone will provide the required multistage high gain amplifiers to overcome control above 750MHz. This needs more predicted losses, which would be more study. costly. It is also important to note that thermal control would likely become a Figure 35 is a summary of path loss major issue in the Top-split designs. comparisons from home to the input of the Figure 34 shows seasonal temperature first amplifier, which will ultimately swings of 5 to 6 dB loss change per determine the system operation point. It is amplifier span would be likely in the top interesting to note that as soon as the upper
frequency is moved beyond the Sub-split expected loss from 42 to 200 MHz, and a limit, the maximum loss path tends toward very large loss profile at 1000 MHz and the last tap in cascade as opposed to the first above. The expected systemperformance tap. There is a moderate increase in can be calculated for each scenario.
Top-Split (900-1050) Top Split (1250-1550) Sub-Split Mid-Split High-Split 200 with Sub-split with Sub-split Upper Frequency MHz 42 85 200 1050 1550 Worst Case Path Loss dB 29.4 30.8 35.6 53.1 59.8 Hardline Cable Type 0.625 PIII 0.625 PIII 0.625 PIII 0.625 PIII 0.625 PIII Cable Loss/ft dB/ft 0.0042 0.0060 0.0092 0.0211 0.0256 Drop Cable Type Series 6 Series 6 Series 6 Series 6 Series 6 Cable Loss/ft dB/ft 0.0134 0.0190 0.0292 0.0669 0.0812 Path Loss from First Tap dB 29.4 30.5 32.3 39.1 41.7 Hardline Cable to First Tap ft 100 100 100 100 100 Cable Loss dB 0.42 0.60 0.92 2.11 2.56 Tap Port Loss dB 23 23 23 23 23 Total Drop Loss dB 2.0 2.9 4.4 10.0 12.2 In Home Passive Loss to Modem dB 4 4 4 4 4 Path Loss from Last Tap dB 28.2 30.8 35.6 53.1 59.8 Hardline Cable to Last Tap ft 1000 1000 1000 1000 1000 Cable Loss dB 4.21 5.99 9.19 21.06 25.59 Tap Insertion Loss dB 10 10 10 10 10 Tap Port Loss dB 8 8 8 8 8 Total Drop Loss dB 2.0 2.9 4.4 10.0 12.2 In Home Passive Loss to Modem dB 4 4 4 4 4 FIGURE 35: PATH LOSS FROM HOME TO FIRST AMPLIFIER
Figure 36 shows the compared Armed with the input level and performance calculations for the 500 home station noise figure, the single station passed node outlined in Figure 33. The amplifier C/N is calculated and then desired performance target is 256QAM for funneled through the total number of each scenario; if it can be achieved, the distribution amplifiers serving the node to throughput per subscriber will be yield the C/N performance expected at the maximized. For each approach, it is input of the node. Worst case performance assumed that a CPE device is available with and lowest cost for return optical links was upstream bonding capability that can use the assumed to be obtained from analog DFB entire spectrum available at a reasonable lasers up to 200 MHz which suggests cost. The number of bonded carriers staying with Analog Return; but cost parity transmitting must not exceed the maximum between analog and broadband digital return allowable modem transmit level, so the (BDR) systems at 1000 MHz and above is maximum power per carrier is calculated not now possible. to exceed 65 dBmV total transmitted power. The maximum power, along with the worst- The results show that the solutions case path loss, yields the input level to the up to 200 MHz have sufficient performance reverse amplifiers in the HFC Network. If to support 256QAM modulation at a 500 the return level was greater than 15 dBmV, HHP node.The top split options suffer from it was assumed that it would be attenuated to cable loss, not to exceed +65 dBmV, and 15 dBmV. noise funneling.The Top-split (900 -1050) may operate at 64QAM and Top-split (1250- 1550) may operate at QPSKand stay within
margin budget using a 500 HHP node. Cost highest throughput per subscriber at the projectionsfor the various solutions show lowest relative cost. that High-split 200 MHz return delivers the
Top-Split (900-1050) Top Split (1250-1550) Sub-Split Mid-Split High-Split 200 with Sub-split with Sub-split Upper Frequency MHz 42 85 200 1050 1550 Homes Passed 500 500 500 500 500 HSD Take Rate 50% 50% 50% 50% 50% HSD Customers 250 250 250 250 250 Desired Carrier BW MHz 6.4 6.4 6.4 6.4 6.4 Modulation Type 64-QAM 256-QAM 256-QAM 64-QAM QPSK Bits/Symbol 6 8 8 6 2 Desired C/N dB 33 40 40 33 20 Number Carriers in Bonding Group 3.5 11 29 23 47 Max Power per Carrier Allowed in Home dBmV 60 55 50 51 48 Worst Case Path Loss dB 29.4 30.8 35.6 53.1 59.8 Maximum Return Amplifier Input dBmV 30 24 15 -2 -11 Actual Return Amplifier Input dBmV 15 15 15 -2 -11 Assumed Noise Figure of Amplifier dB 7 7 7 7 7 Return Amplifier C/N (Single Station) dB 65 65 65 48 39 Number of Amplifiers in Service Group 30 30 30 30 30 Return Amplifier C/N (Funneled) dB 50.4 50.4 50.4 33.7 23.9 Optical Return Path Technology DFB DFB DFB BDR BDR Assumed Optical C/N dB 48 45 41 50 50 System C/N dB 46.0 43.9 40.5 33.6 23.9 Expected Maximum Datarate after Overhead Mbps 91.8 370.2 975.9 603.2 455.3 Extra Datarate from Sub/Mid Bands 91.8 91.8 Total Datarate from all Bands 91.8 370.2 975.9 694.9 547.1 Throughput/Customer Mbps 0.37 1.48 3.90 2.78 2.19 Cost Scale 100% 100% 139% 218% Solution Figure of Merit (Throughput/Cost Scale) 1.48 3.90 2.00 1.00 FIGURE 36: PERFORMANCE OF 500 HP NODE
FIGURE 37: RELATIVE COST AND THROUGHPUT COMPARISON 500 HP NODE SOLUTIONS
Further analysis of the Top-split a four way node split. The noise funneling options concludes that reducing the node is reduced to a level where higher order size, and thereby the funneled noise in the modulations are possible. The costs of the serving group could yield higher modulation additional node splits seem to scale capability. Figure 38 shows the comparison appropriately with the additional throughput again with the Top-split scenarios including per subscriber.
Top-Split (900-1050) Top Split (1250-1550) Sub-Split Mid-Split High-Split 200 with Sub-split with Sub-split Upper Frequency MHz 42 85 200 1050 1550 Homes Passed 500 500 500 125 125 HSD Take Rate 50% 50% 50% 50% 50% HSD Customers 250 250 250 62.5 62.5 Desired Carrier BW MHz 6.4 6.4 6.4 6.4 6.4 Modulation Type 64-QAM 256-QAM 256-QAM 256-QAM 16-QAM Bits/Symbol 6 8 8 8 4 Desired C/N dB 33 40 40 40 27 Number Carriers in Bonding Group 3.5 11 29 23 47 Max Power per Carrier Allowed in Home dBmV 60 55 50 51 48 Worst Case Path Loss dB 29.4 30.8 35.6 53.1 59.8 Maximum Return Amplifier Input dBmV 30 24 15 -2 -11 Actual Return Amplifier Input dBmV 15 15 15 -2 -11 Assumed Noise Figure of Amplifier dB 7 7 7 7 7 Return Amplifier C/N (Single Station) dB 65 65 65 48 39 Number of Amplifiers in Service Group 30 30 30 7 7 Return Amplifier C/N (Funneled) dB 50.4 50.4 50.4 40.0 30.2 Optical Return Path Technology DFB DFB DFB BDR BDR Assumed Optical C/N dB 48 45 41 50 50 System C/N dB 46.0 43.9 40.5 39.6 30.2 Expected Maximum Datarate after Overhead Mbps 91.8 370.2 975.9 774.0 863.8 Extra Datarate from Sub/Mid Bands 91.8 91.8 Total Datarate from all Bands 91.8 370.2 975.9 865.8 955.6 Throughput/Customer Mbps 0.37 1.48 3.90 12.38 13.82 Cost Scale 100% 128% 240% 302% Solution Figure of Merit (Throughput/Cost Scale) 1.48 3.06 5.17 4.58 FIGURE 38: COMPARISONS WITH TOP SPLIT ONLY AT 125 HP NODE
FIGURE 39: COST AND THROUGHPUT COMPARISON WITH TOP SPLIT ONLY AT 125 HP NODE
Summary of Cost Analysis these may represent ―initial costs‖ to provide capacity above Mid-split. The cost analysis As stated in the opening of this material in figures 38 and 39 may represent section, the cost analysis providesestimates the ―end state cost analysis‖ to achieve the for discussion purposes only. The actual capacity requirements for this decade and cost relative to one another may vary widely beyond (approximately up to 2021 to 2023 when products have been developed and timeframe). Moreover, these last released to market. analysesalso provide apple-to-apples cost estimates of High-split and the two Top-split Perhaps a reasonable way to consider options to reach similar network capacity the data found in figures 36 and 37is that targets.
The cost analyses capture the These are the findings of the cost equipment and labor cost estimates for the analysis comparing four spectrum split optical transport and HFC systems. The cost options. The costs of the spectrum option to analysis predicts that Mid-split and High- yield similar data capacity consideringHigh- split will share similar costs on the return split (200) and both Top-split options are path systems, however the loss of illustrated in figure 40. downstream spectrum will have to be replaced if High-split is selected, captured in Cost Scale 350% figures 38 and 39. The costs to solve the 300% 250% STBOut of Band (OOB), impacts to analog 200% 150% Cost Scale service tier, and loss of video spectrum for 100% 50% STBs and TVs was not calculated in the 0% Mid-Split High-Split 200 Top-Split Top Split analysis of High-splits. (900-1050) (1250-1550) FIGURE 40: COST ANALYSIS The two Top-split options do not account for the cost ofMid-split this The Top-split (900-1050) with Mid- isassumed to be the first upstream split given a 500 HHP may reach the augmentation selection by the MSOs. capacity of High-split. The analysis Additionally, the High-split option uses the estimates that Top-split (900-1050) is about spectrum of Mid-split thus we considered 1.39 times the costof High-split (200), in this a sunk cost. The use of either Top-split this HFC topology. Again the Mid-split cost (900-1050) option with Mid-split should is not considered. achieve similar or greater upstream capacity The Top-split options that avoid compared to High-split. Either Top-split using Mid-split spectrum but does use the optionwith Sub-split will need to move to a existing Sub-split spectrum will need to 125 HHP node service group to achieve the move to a 125 HHP node service group to capacity of High-split. The Top-split achieve similar capacity to High-split. The options do not account for the additional analysis estimates that Top-split (900-1050) DOCSIS capacity because more channels with Sub-split is about 2.4 times the costof are needed to achieve the capacity level of High-split (200). The Top-split (1250-1550) High-split (200). with Sub-split is about 3 times the costof The cost analysis intentionally High-split (200). excluded the items related to High-split and Top-split mentioned above in order to illustrate the relative cost for the spectrum split. We have additionalanalyses, which includesome of the items which were excluded from High-split and the Top-split options.
EXECUTIVE SUMMARY AND CONCLUSIONS compatibility and the drivers behind this methodology. The cable industry will transition and evolve their existing networks from largely Video and High-Speed Data Service broadcast video to unicast, likely using IP- Estimates based delivery technology. The High-Speed Internet Service and networks will continue Unicast services like On-Demand to expand, meeting the needs for higher Video and High-Speed Internet will service tiers and more capacity downstream dominate the MSO service offering and upstream. and spectrum allocation for the coming decade. The paper provides assumptions and Today, less than 4% of the MSO’s predictions for the evolution of advanced downstream spectrum is allocated to video and high-speed Internet services. The High-Speed Data; however, it is report assesses the timing and drivers of forecasted that this might be 50% network change, possible technologies, within the next 10 years. architectures, migration options, and a cost An industry accepted modeling tool vs. performance analysis. based on a thirty-year history of Data Service offerings and capabilities The comparisons found in the report predicts a ~ 2.5 Gbps Downstream are at nearly every layer of the MSO’s Next and 500 Mbps Upstream Internet Generation Cable Access Network. These service offering in 10 years. areas include spectrum splits, data network The recent announcements from the technology, and network architecture United Kingdom’s cable provider options. The examination will include a Virgin Media of a 1.5 Gbps Down / look at the traditional centralized access 150 Mbps Up trial and Verizon’s layer data architecture used over cable HFC FiOS reports of upgrades to 10 Gbps networks. Down / 2.5 Gbps Up, is disrupting We also define a possible alternative this thirty-year industry benchmark architecture to HFC, which we refer to as study of data service growth. Digital Fiber Coax (DFC). This is a distributed access layer architecture using Cost Analysis & Performance Summaries EPON or Ethernet style optical transport to The upstream spectrum split options the node where a coax based MAC/PHY or reviewedhave vastly different HFC topology PHY resides supporting downstream and and data network layer requirements, which upstream; or just one direction (upstream). have significant impact to the cost estimates. Additionally, as the industry considers the The cost analyses capture the equipment and future evolution of the network the report labor cost estimates for the optical transport examines the importance of backward and HFC systems.
The cost analysis predicts that Mid- excluded from High-split and the Top-split split and High-split will share similar costs options, that were not included in this report. for return path electronics, however the loss of downstream spectrum will have to be Top-split costs are driven by the replaced if High-split is selected. The High- network characteristics including: cable loss split option will have some impacts to that progressively increases as frequency network functions and services. The costs to increases, modem maximum power output solve the STB Out of Band (OOB), impacts composite not to exceed +65 dBmV and the to analog service tier, and loss of video funneling effect of a large number of return spectrum for STBs and TVs was not path amplifiers. These critical factors calculated in the analysis of High-split. andkey findings of the report illustrate the impacts to the HFC network topology, such The two Top-split options do not as the size of the node service group. account for the additional DOCSIS channels Additionally, the Top-split options must use needed to achieve the capacity level similar lower order modulations, resulting in more to High-split (200). The Top-split (900- spectrum and more CMTS ports needed to 1050) with Mid-split (given a 500 HHP sustain equivalent capacity of High-split. service area) may reach the capacity of High-split. The analysis estimates that Top- Key Network Performance Factors split (900-1050) is about 1.39 times the Network data capacity parity costof High-split (200), in this HFC between High-split and the Top-split topology. Again, the Mid-split cost is not Options have vastly different considered. network topologies and costs. The Top-split options that avoid The major reasons why lower using Mid-split spectrum, but does use the frequency return (Sub-, Mid-, and existing Sub-split spectrum, will need to High-split) and higher frequency move to a 125 HHP node service group to return (Top-split options) have such achieve similar capacity to High-split. The performance differences which analysis estimates that Top-split (900-1050) impact network architecture, with Sub-split is about 2.4 times the costof capacity, and cost are as follows: High-split (200). The Top-split (1250-1550) First Major Reason: Cable loss with Sub-split is about 3 times the cost of progressivelyincreases as frequency High-split (200). increases, thus a major factor when considering higher frequency return. The cost analysis intentionally Second Major Reason: Modem excluded the items related to High-split and maximum power output composite Top-split mentioned above in order to not to exceed +65 dBmV (to illustrate the relative cost for the spectrum minimize power and cost, split. We have additional analyses, which andmaintain acceptable distortion) include some of the items that were
Third Major Reason: Funneling need to support full spectrum per effects of a large number of return node service group within the decade path amplifiers. This is not a factor at Mid-split is an excellent first step: low frequency because the cableloss low cost, small spectrum, high data is low enough that a cable modem capacity which lasts about a decade can provide adequate power level to Mid-split in place of a node split maintain high C/N. may enable a 500 HHP node to last Existing 5-42 / 750 MHz system about a decade with a 500 HHP node may remain HFC Conclusions 1: architecture unchanged until 2015 and then a remains viable well through this series of network migration steps decade and beyond that are defined in this paper may HFC Conclusions 2: existing occur technology, costs, flexibility and Existing 1 GHz Passives do not have versatility to support transport of to be touched until perhaps the year virtually any new MAC/PHY 2023 technology remains a core benefit Existing Passives may support up to and value 1050 MHz for additional upstream or HFC Conclusions 3: Optical downstream capacity transport supported with DFB analog A limitation of power passing taps is lasers and BDR last throughout the the AC power choke resonance. This decade is an important finding when HFC Conclusions 4: HFC allows the leveraging the existing passives; outside plant to remain simple, just therefore the use above 1050 MHz performing media and/or digital may not be predictable or even conversion (for BDR), thus no MAC possible or PHY layer processing is required Passives represent approximately in the node 180-220 devices per 500 HHP node HFC Conclusions 5: enables a group centralized access layer for Small Nodes and FTTLA are not economies of scale and just in time required until perhaps the second investment in capacity half of the 2020’s decade. Digital Fiber Coax (DFC) Downstream spectrum recovery DFC Conclusion 1: A distributed methods will support the transition MAC or PHY architecture that from broadcast to unicast service would compete or complement HFC delivery and will help solve the is not a viable replacement in nearly downstream capacity challenge all cases, except extremely long Downstream (assuming an distance to the node,as discussed in equivalent 750 MHz system) will the report, however more study is required.
DFC Conclusion 2: A distributed allowing just in time investment architecture and its risks include: while converting to IP stranding capital, low flexibility and DOCSIS Conclusion 3: DOCSIS limited versatility will support new upstream spectrum, DOCSIS Conclusion 1: enables the increase modulation schemes, may full spectrum migration to IP in both add MACand PHY layer the downstream and upstream improvements (perhaps OFDM) DOCSIS Conclusion 2: enables a Figure 41:DOCSIS QAM Estimates smooth channel-by-channel for:HFC Topology, Spectrum Split migration, this is key for the MSO to and PHY Capacity Comparison maximize revenue per MHz and Top-split (900- Top split Top-split (900- Top split Top-split (900- Top split Top-split (900- Top split DOCSIS QAM Sub-split Mid-split High-split 200 1050) with (1250-1550) 1050) with (1250-1550) 1050) with (1250-1550) 1050) with (1250-1550) Sub-split with Sub-split Mid-split with Mid-split Sub-split with Sub-split Mid-split with Mid-split Node Service Group 500 500 500 500 500 500 500 125 125 125 125 Capacity from Sub-split 92 92 92 92 92 92 92 92 92 92 92 Capacity from Mid-split 225 225 225 225 225 225 Capacity from High-split 599 Capacity from Top-split 617 410 617 410 782 820 782 820 Total PHY Channel Bond 92 316 916 708 502 933 726 874 912 1,099 1,136 Capacity (Usable) in Mbps FIGURE 41: HFC TOPOLOGY, SPECTRUM SPLIT AND PHY CAPACITY COMPARISON
Network Evolution Prediction Summary Year 2019: Upstream 500 HHP Node with Mid-split DOCSIS is Year 2015:Upstream 500 HHP Node exhausted (reduce SG or add Split/Segment ―or‖ add new spectrum) spectrumupstream& keep node size Year 2020: Mid-split (5-85) with Year 2017: North America 5-42 is either DOCSIS QAM or OFDM is exhausted because of High-Speed approaching capacity and in year Internet Service Tier, more upstream 2021 would be out of capacity driven spectrum required. by Service Tier growth; thus more Year 2017-18: Downstream 500 upstream Spectrum is Required HHP Node is at capacity (assuming (High-split or Top-split) 250 Subs with 0.9-1.4 Gbps of Year 2021: HSD Max Service Tier Traffic Plus 3.6 Gbps of MSO Video prediction of 2.2 Gbps + 1.8 Gbps of Traffic)(reduce SG or add spectrum) MSO Videoapproaching capacity of Year 2018-19:Euro 5-65 is 750 MHz system exhausted because of High-Speed Year 2021: Downstream 250 HHP Internet Service Tier,more upstream Node is at capacity (assuming 125 spectrum required. HSD Subs with 2.3 Gbps of Traffic
Plus 1.8 Gbps of MSO Video MHz should all share the same Traffic) change SG size or service channel bonding group, this Year 2021-22: Upstream 500 HHP maximizes the use of existing Node nears capacity with High-split spectrum and delays investment 200 or Top-split (900-1050)+Mid- Sharing channel bonding groups with split Using DOCSIS QAM or OFDM DOCSIS 3.0 and Any Successor (use of OFDM bought a few months creates ―one‖ IP Network (cap and of capacity) grow networks hang around awhile) Year 2023:Upstream HSD Max Sharing the same bonding group Service Tier Prediction of 995 Mbps assures previous and future Upstream Serviceconsumes High- investment may be applied in split 200 or Top-split (+Mid-split) creating larger IP based bandwidth Using DOCSIS QAM or OFDM and not stranding previous capital Year 2023: Downstream HSD Max investment Service Tier Prediction Consumes Backward Compatibility has All Downstream Spectrum thus benefitted industries like the IEEE additional Spectrum Above 1 GHz is Ethernet, WiFi, and EPON saving required or FTTx the entire eco-system money Year 2023 - 2024: MSOs touch the Backward Compatibility simply passives to increase spectrum allows the MSOs to delay and above 1 GHz to achieve higher perhaps avoid major investment to Downstream and Upstream the network such as adding more capacity based on HSD predictions spectrum or running fiber deeper. If these do not materialize (i.e. All of our analysis in this report High-speed data service prediction assumes backward compatibility over 4 Gbps Down and 1 Gbps Up in with DOCSIS 3.0 QAM and any the year 2023) nodes splits/node successor, like DOCSIS OFDM; thus segmentation will solve the traffic creating a larger and larger IP growth projections for many more bonding group with each year’s years investment. If this is not the case the Finally,if neither the Service Growth investment in HFC upgrades will Rates nor Traffic Utilization Growth pull forward. It is uncertain of the Rates are maintained at a 50% exact level of financial impact but CAGR, then the timing and drivers the total cost of ownership may be for investment will change and the higher when deploying two separate HFC will last far longer. IP based network technologies.
Importance of Backward Compatibility These are the major takeaways of this paper, however additional information is DOCSIS 3.0 QAM based and any contained in the report providing perspective successorshould consider that every and details to the conclusions cited above.
The examination of the next generation cable access network spans several disciplines and this report is not a complete analysis of all of the possibilities.
REFERENCES
[1] Communications Technology, ―Virgin Media Trials 1.5 Gbps Broadband‖, April 21, [2] CED Magazine, ―Verizon open to 10G PON bids in 2011‖, June 23, 2010 [3] 2011Tom Cloonan, ―On the Evolution of the HFC Network and the DOCSIS® CMTS - A Roadmap for the 2012-2016 Era,‖ SCTE Cable Tech Expo, 2008.
[4] ANSI/SCTE 55-2 2008, ―Digital Broadband Delivery System: Out of Band Transport Part 2: Mode B‖, SCTE
[5] ANSI/SCTE 55-1 2009, Digital Broadband Delivery System: Out of Band Transport Part 1: Mode A.
[6] MoCA, "The Only Home Entertainment Networking Standard In Use By All Three Pay TV Segments—Cable, Satellite, IPTV, MoCA Presentation 2010, www.mocalliance.org.
LIST OF ABBREVIATIONS AND ACRONYMS
BPON Broadband PON CAGR Compound Annual Growth Rate CBR Constant Bit Rate DBS Digital Broadcast System DFC Digital Fiber Coax DOCSIS Data Over Cable Service Interface Specifications DSG DOCSIS Set-top Gateway DTA Digital Terminal Adapter EoC Ethernet over Coax EPON Ethernet Passive Optical Network FTTH Fiber To The Home FTTLA Fiber to the Last Active FTTP Fiber to the Premise FTTx see (FTTH, FTTP, etc) Gbps Gigabits per Second GPON Gigabit PON
HFC Hybrid Fiber Coaxial HHP Households Passed HPNA HomePNA Alliance HSD High Speed Data IP Internet Protocol IPTV Internet ProtocolTV MAC Media Access Layer Mbps Megabit per Second MoCA Multimedia over Coax Alliance MSO Multiple Systems Operator OFDM Orthogonal Frequency Division Multiplexing OSP Outside Plant OTT Over The Top P2P Peer-to-peer PHY Physical Layer QAM Quadrature Amplitude Modulation QoS Quality of Service RFoG RF Over Glass SDV Switch Digital Video UHF Ultra High Frequency US Upstream VoD Video on Demand