Internet access holds great promise for all levels of cable operators — from large MSOs to small independents. Data services represent a new and potentially significant, fast-growing revenue stream. However, use of first-generation cable modem solutions requires operators to upgrade their cable plants from one-way coaxial plants to hybrid fiber/coax (HFC) networks, thus spending up to $200 per home passed. Because 80–85 percent of the U.S. cable infrastructure is still one-way coaxial plant, projections are that it will require several years and a large capital investment to upgrade the cable infrastructure.

In the interim, some cable operators are considering cable modem solutions with telephony-return to capitalize on the market for Internet access, as a means of reducing system upgrade costs and time to market. Others are considering delaying data services until their cable network upgrade is completed.

However, cable modem systems based on S-CDMA technology provide a viable alternative, allowing cable operators to deploy data services over two-way, pure-coaxial plants. Based on their robust upstream capability, these cable modem systems operate effectively over noisy, two-way, pure-coaxial plants, reducing the capital investment for system upgrade down to $11 per home passed. With minimal plant upgrade required, S-CDMA-based systems accelerate time-to-market, enabling operators to offer data services with about the same amount of start-up time as telephony-return solutions.

Furthermore, Quality of Service (QoS) capability enables operators to generate three times the revenue of telephony-return-based service in recurring monthly revenue per subscriber, given a mix of residential, SOHO (small office/home office), and corporate users. It can therefore be argued that S-CDMA-based two-way systems provide a more attractive business proposition for data services than that of telephony-return solutions.

This article compares S-CDMA access systems running over two-way, pure-coaxial plants with telephony-return cable modem solutions. It is first necessary to review several technical issues related to the deployment of high-speed services over coaxial plants, compared to HFC networks, such as downstream channel availability and upstream channel characteristics.

Downstream channel availability

Unlike HFC networks based on 750 MHz systems, coaxial plants operate over systems from 280 MHz to 550 MHz. Depending on the demographics, some of these systems use the majority of downstream channels for video programming, thus limiting data services to a single or a few downstream channels. There are, however, several regions that are not used for video programming because of the high level of interference, but which become available for data services when S-CDMA technology is used (see Table 1). These include:

  • FM band. The FM band, located between 88 MHz and 108 MHz, is used in some systems for RF radio transmission over cable TV, but is not used for video broadcasting because of the high level of interference from RF signals over-the-air. Because many cable systems do not deliver FM radio programming, these channels are often unused.
  • 108 MHz though 121 MHz RF spectrum. The portion of the RF spectrum from 108 MHz to 121 MHz is generally unused for video broadcasting because of the high level of interference from aircraft navigation system signals.
  • Roll-off region. The roll-off region consists of the upper frequencies — the upper 6 to 12 MHz in each system — which are not used for video broadcasting because of severe amplitude tilt distortion, high group delay, and the low signal-to-noise ratio. Amplitude tilt distortion is a result of high attenuation over the cable in high frequencies and the non-linearity characteristics of amplifiers at the edges. The low signal-to-noise ratio is a result of higher attenuation in high frequencies and lower amplifier gains in its non-linearity region.

All of the above unused frequencies occupy 45 MHz of the RF spectrum, or slightly more than seven 6-MHz channels. Utilizing a robust S-CDMA transmission technology, these channels can be used for downstream data services. By spreading the signal over frequency and time, S-CDMA provides high noise immunity against narrowband and impulse noise interference, such as over-the-air interference.

A robust adaptive equalizer compensates for the severe amplitude tilt and group delay presence in the roll-off region. The high code gain enables operation with a signal-to-noise ratio below 15 dB, thus combating the relative high in the roll-off region (see Figure 1).

Upstream channel characteristics

Without an advanced cable modem solution, the upstream channel characteristics of pure-coaxial plants severely limit deployment of high-speed data services. These characteristics include:

  • Low signal-to-noise ratio. Unlike HFC networks with a service area of 500 to 2,000 homes passed and five to six amplifiers in cascade, the service area of traditional coaxial plants covers 5,000 to 20,000 homes passed, with 20–30 amplifiers in cascade, resulting in a low signal-to-noise ratio.
  • High group delay. The high number of amplifiers in cascade increases overall group delay, specifically in the corner channels — 5–11 MHz and 36–42 MHz — thus limiting overall upstream channel availability.
  • High common path distortion. The high number of connectors can cause severe common path distortion, resulting from corrosion or oxidation on connections of dissimilar metals, producing a diode-like effect. When forward-path signals pass through this diode, potentially harmful second- and third-order beats every 6 MHz can be created in the reverse path.
  • High impulse noise. The long antenna created by the coaxial plants may cause severe impulse noise in the return path.

Recent field trials in November 1996 and January 1997, running over cable systems from two MSOs, demonstrate the throughput and error performance of an S-CDMA-based access system under severe channel conditions. Each of the two trials was conducted over a 6 MHz channel — the first between 11 and 17 MHz and the second between 5 and 11 MHz. Cable modem performance in both trials was excellent. The following is a detailed description of the three test scenarios in the first trial:

  • One clean fiber node with 3,400 homes passed and 2,890 cable TV subscribers.
  • Unclean fiber node (installed, but not tuned) with 6,200 homes passed and 5,270 cable TV subscribers.
  • Aggregation of eight fiber nodes with 30,000 homes passed and 25,500 cable TV subscribers.
  • All tests were conducted over a 6 MHz channel between 11 and 17 MHz, with severe channel conditions, including:
  • Signal-to-noise and interference ratio of 13 dB
  • Narrowband interference
  • Shortwave at 11 and 12 MHz
  • Ham radio at 14 MHz
  • Telemetry at 17.5 and 18 MHz
  • Severe ingress noise
  • Severe impulse noise, predominately from power-line signals.

As indicated in Figure 2, the S-CDMA access system operated error-free over 98.3 percent of the time in all cases. In the case of the "unclean node," which was not tuned, S-CDMA provided 99 percent levels, indicating that this system delivers high-performance data services in the return path with virtually no network tuning.

The S-CDMA-based system successfully overcomes noise from severe multiple narrowband interference and impulse sources. This demonstrates the resilience of the system over either pure-coax or HFC cable plants, with no need to install high-pass filters. This provides a new level of flexibility for cable operators, who can now deploy data services rapidly, without the expense or time associated with major plant upgrades.


Figure 3 demonstrates the bit error rate (BER) distribution for a five-second interval over an unclean node. The system operated at zero BER 96.2 percent of the time, and it quickly recovered from several incidents with BER less than E-8.

Based on these rigorous tests, the conclusion is that S-CDMA-based systems can operate error-free:

  • Over both pure-coaxial plants and HFC.
  • In the high-noise portion of the upstream spectrum below 20 MHz, which until now has not been considered practical for high-speed data services.
  • Over noisy, untreated plants.
  • Over systems with large and highly aggregated nodes.
Economic comparison

Among the assumptions,* telephony-return systems require installation of a modem pool at the headend location, at a ratio of 1:3 to 1:5 for the number of simultaneous users vs. the number of subscribers. T-1 lines are used to transmit the multiple dial-up analog circuits from a telco's central office to a cable operator's headend location, with a resulting impact on the net monthly revenue. We have made the conservative assumption that subscribers will use their existing telephony line connections.

In the case of two-way coaxial plants, it is assumed that the cable plants begin as two-way capable, thus requiring return-path module installation, system balancing and changes in the powering systems. In other cases, they may require a complete installation of new housings or new amplifiers. The average return-path activation cost per amplifier is $150 for a return module and $50 for installation and balancing at 1.5 hours. In order to calculate the cost per mile, four and five amplifiers were used in 450 MHz and 550 MHz systems, respectively. The changes for the system powering were calculated at 10 percent of the overall return-path activation cost. The overall return-path activation cost per cable modem subscriber was based on 100 miles per home passed, a 67 percent basic cable penetration rate, and a 10 percent cable modem penetration rate. It is assumed that return-path activation costs are amortized 50/50 between data services and other advanced interactive services.


As for recurring monthly revenue, it is assumed that a two-way solution with QoS controls enables operators to capitalize on the existing demand for broadband access from a mix of residential, small office/home office (SOHO) and mid- to large corporate users. A conservative mix of users is based on 88 percent residential users, 10 percent small office/home office users, and 2 percent corporate users. Furthermore, it is assumed that two-way data service over cable commands a premium over telephony-return because of its service superiority — providing continuous connection and faster access.

The following comparison between two-way, pure-coaxial systems and telephony-return solutions is based on the above assumptions, as well as current prices for telephony-return cable modems, modem pools, and T-1 lines (see Table 2 for detailed cost breakdown). The potential average monthly revenue from two-way service is estimated at $61 to $80, compared to $19 to $26 for telephony-return, because of the ability to capture higher paying users. This translates to three times greater monthly recurring revenue (or up to $720 per cable modem subscriber per year) for two-way systems. The telephony return solution requires an additional $6 per subscriber to provide a connection between the central office and the headend site. If one assumed a requirement for a second telephone line, telephony-return service would be even more expensive — about $35–$40 — yet provide the same upstream speed as current dial-up Internet access services.


An S-CDMA-based solution protects operators' investment in network equipment, because it operates over both coaxial plants and HFC networks. Telephony-return solutions could become obsolete in two years, thus lowering the overall return-on-investment of telephony solutions.

As for the network access issues, telephony-return solutions share similar problems with current analog and ISDN dial-up access systems, including blocking in the network, tying-up subscribers' phone lines, lack of a continuous line connection, and system scalability. Furthermore, TCP/IP protocol has severe limitations in asymmetric configurations such as telephony-return, thus reducing downstream throughput for many applications and limiting system scalability.

In a more global sense, two-way solutions provide a strategic advantage for cable operators, enabling them to build expertise with two-way plants faster — a necessity for success in the competitive telecommunications market. Telephony-return solutions, on the other hand, require cable operators to transfer more than $15 per month to local telephone carriers for each cable modem subscriber to cover the T-1 connection between the central office and the headend site, and an optional second telephone line.

Up-front costs per cable modem subscriber are $346 to $578 for telephony-return, and $475 to $551 for two-way activation, making the initial costs of two-way slightly more expensive, but not significantly. It is important to note that up-front costs for telephony-return are primarily variable costs, whereas two-way coaxial plants have a combination of fixed up-front costs for return-path activation, and variable costs for the customer-premises modem, which scale to provide lower costs as more subscribers are added. In addition, up-front costs for two-way, pure-coax activation can be amortized over a longer period, because the return-path module can be used as part of the overall system upgrade, thus lowering the overall costs.


It should be noted that telephony-return requires no changes to the existing infrastructure, while activating the return-path of two-way, pure-coaxial plants requires about four months to activate a 50,000-home passed system, assuming a crew of five technicians. In the scope of rolling out new services, these delays are not significant, as the service can be rolled out in phases over several areas.

In summary, two-way, pure-coaxial data services are a better investment for cable operators than telephony-return. Return-path activation over pure-coax provides a better solution for cable operators, given the ultimate goal to upgrade to HFC networks. S-CDMA-based systems allow faster activation of two-way systems, thus building a stronger market position in the race to provide broadband data services.

*We have simplified the following economic analysis by excluding costs that are common to both two-way pure-coax and telephony-return systems. These include up-front costs, such as cluster servers for the headend and local content servers, as well as ongoing maintenance costs, such as Internet access charges and customer support.