The deployment of high-speed data services over all-coax and hybrid fiber/coax (HFC) networks has always been anticipated to follow a relatively straightforward implementation strategy. However, recent industry experience deploying cable modem termination systems (CMTSs) has uncovered an unforeseen and challenging system engineering issue when deploying new services over HFC systems.

The problem is a mismatch in CMTS upstream "ports" to the large number of return path "ports" on HFC systems. Adding more CMTS upstream ports places an operator in an undesirable upfront capitalization situation because the upstream costs would be well beyond the revenue stream during initial sparse deployment. This article illustrates the engineering problem that has been discovered and suggests how to avoid spending too much money before the subscriber revenue stream is in place.

The incremental growth story

In general, there are two engineering and business rules to follow when deploying high-speed data services on all-coax plants: first, the new service must be available to any and all subscribers in the headend serving area, typically a city or collection of adjoining towns; and second, the business model for deployment must incrementally add high-speed data CMTS equipment in concert with subscriber demand (capacity) and the associated gains in revenue.

All-coax cable television distribution plants are well-suited to the deployment of high-speed data services in that there are one to few downstream distribution coax trunks leaving the headend, and there are one to few upstream coax trunks entering the headend. For initial deployment scenarios, a single downstream transmit channel (e.g. single 64-QAM 30-Mbps digital data channel) can service the entire all-coax cable plant by distributing the same downstream signal to all downstream trunks.

Similarly, one or more CMTS upstream channels can share the same CMTS upstream port, which can also be coupled to more than one upstream trunks via an RF combiner. One or more upstream data channels can be supported by the CMTS gear, each separated by frequency. An upstream port is the F-connector which makes the 5–42 MHz upstream spectrum available to the CMTS gear.

The use of an RF combiner in the upstream to combine several trunks into one port is of limited use because the known noise funneling problem raises the noise floor at the port. The rise in noise floor is a combination of both background system thermal noise and of externally generated ingress noise.

These noise sources collectively form the impairment noise that must be overcome by the upstream data channel for any interactive service, including high-speed data services, impulse PPV and others. The number of upstream trunks that may combined is chiefly limited by the noise characteristics of the return plant. Some all-coax return plant trunks may be noisier than others.


Initial deployment of high-speed data services on all-coax plants can typically be accomplished using one CMTS for the entire plant. Existing CMTS equipment typically comes in one of two scalability architectures: a "fixed" scale configuration with one downstream port and only one upstream port; or a "flexible" scale configuration with one or more downstream ports and one or more upstream ports. Incremental growth to meet new subscriber demand or capacity is different for fixed vs. flexible architecture.

Fixed-scale CMTS equipment requires an additional CMTS when subscriber growth calls for additional data capacity. With a flexible-scale CMTS, when more upstream capacity is needed, the cable operator can add an upstream channel demodulator (demod) card to the CMTS. When more downstream capacity is needed, the cable operator can add a downstream channel or purchase a new CMTS box.

There is a large difference between fixed- and flexible-scale CMTS systems. With a fixed configuration, the operator must recombine upstream trunks into as few ports as possible to avoid having to purchase a large number of fixed-scale CMTS boxes. With a flexible configuration, each upstream channel may be connected to a different upstream trunk, eliminating any need for the recombination of trunks. This has two benefits: first, the cost of an additional upstream channel is generally less than the cost of a fixed configuration CMTS box; and the noise floor is reduced at the upstream port, making the operator free to distribute to trunks and combine upstream trunks with a flexible scale system.

At some point in the growth of service deployment, more downstream capacity will be required to meet subscriber demand. In a fixed-scale CMTS, a new downstream channel is required for every upstream channel added and vice versa, regardless of whether the downstream or upstream channel capacity has been filled by demand. In a flexible-scale CMTS, the relationship of the downstream channels to the upstream channels within a single CMTS box are separately scalable, allowing the addition of downstream or upstream channels to follow demand.

In addition, this flexible scalability allows for capital expenditures to more closely match revenue growth, and also allows for noise impairment to be better controlled by use of more upstream ports per downstream channel. This latter point is very important in that the cable operator has much more flexibility in managing the recombination of upstream trunks and subsequent noise funneling issues.

When the downstream channel capacity has been exceeded, and not enough RF spectrum is available in the cable plant, the operator has the option of upgrading the plant to HFC. The upgrade will produce more upstream and downstream trunks. In an incremental HFC upgrade scenario, the operator has the option to do upgrades only where high-speed data capacity is needed, i.e., where the active subscribers and revenue are coming from. Upgrading the entire plant to HFC is not required.

Another option is to postpone the HFC upgrade for as long as possible. When ingress noise management is a motivation for upgrades, the cable operator has the ability to manage the noise. Options include keeping the plant tuned, using high-pass filters, or using more sophisticated, soon-to-be-available and affordable, intelligent switching filters. Combinations of these tools allow the cable operator to increase revenue growth before spending capital to upgrade the plant.

The cable modem handling capacity of CMTS equipment is also an issue and affects how and where cable operators must grow capacity to meet demand and capital spending. Some thoughts on how to get the most out of CMTS equipment are: use the highest available modulation available in the downstream and the upstream (i.e. 30 Mbps 64 QAM in the downstream, 2.56 Mbps QPSK or greater in the upstream); use FEC (forward error correction) on the downstream and upstream as bit errors reduce the number of modems supportable by the channel; use the data channel effectively by looking for superior performance and low delay; and use flexible-scale CMTS systems only, avoiding fixed, one-to-one scale CMTS systems.

An ideal world for high-speed data service deployment, business and growth is to upgrade to HFC only after having established a revenue stream from high-speed data penetration. Because the world is not ideal, new high-speed data services must be deployed on existing all-HFC plants. The issue that arises is the matching of CMTS upstream ports to the large number of cable plant upstream ports. The number of return ports is a direct function of node size (the smaller the node, the more ports). The ability to recombine upstream trunks is directly influenced by thermal noise issues of return path lasers and by ingress noise management.


The following example illustrates the return port abundance problem. Assume a small, 20,000 homes-passed plant is converted to all-HFC with a node size of 500 homes passed. This yields 40 separate returns. Assume that Fabry-Perot (FP) lasers have been used for the upstream returns based on their affordability. Typical FP lasers allow a recombination of four upstream trunks into one upstream port.

With fixed-scale CMTS equipment, 10 boxes are required. Worst-case economic impact would be that where one box might have supported the entire previous all-coax plant, nine additional boxes are now required.

With flexible-scale CMTS equipment, one box is required, provided it supports 10 upstream return ports. The one box might have supported the previous all-coax plant and just rolls over to support the new HFC plant. Capital may be needed to purchase additional upstream channel demodulator support for the CMTS.

Smaller node sizes increase the number of upstream return trunks. In the above example, if the node size was 2,000 homes passed instead of 500 homes passed, then the number of returns trunks would have been 10, not 40. Ten return trunks could be recombined into three upstream return ports.

The port mismatch problem gets worse with larger systems. A typical 50,000- or 200,000-homes-passed system greatly multiplies the number of upstream return ports. In the 20,000-home example, a system which passes 200,000 homes has 10 times the number of return trunks (i.e. 400!). Recombination yields 100 return ports, which is still a large number of ports that must be supported by CMTS equipment.

The design goal for node sizes for HFC plants is on the order of 500 HHP. This goal is still valid in light of this port mismatch problem; however, it is difficult to support in initial deployment of interactive services. A better idea would be to start with 2,000 HHP or 4,000 HHP- sized nodes, while laying sufficient fiber to downsize to 500 HHP per node in incremental steps, as service demand and revenue grow.

Initial sparse deployment problem

Recall that service must be made available to the entire serving area. From the previous example, the worst subscriber support scenario would be one sub per upstream return port. The available revenue from 100 subscribers is not sufficient to purchase CMTS equipment with 100 upstream return channels. Note that in this scenario, one downstream data channel is sufficient to supply services to any subscriber in the serving area until such a time as when demand exceeds the capacity of that single channel.

Recombining return trunks at a greater ratio than four-to-one causes noise funneling contribution and reduces the carrier-to-noise ratio below a 25 dB margin at the upstream return port. This ratio is being used by several cable operators. Converting the upstream lasers from FP to direct feedback (DFB) lasers allows the upstream return trunks to be recombined at a ratio of up to 10-to-one, which is attractive. If the plant currently has FP lasers, the cost differential to go to DFB is substantial, and in most cases prohibitive.

A high noise floor interrupts all upstream modulation schemes in an HFC plant. The ability to recombine upstream return trunks is limited by the lowest capable interactive service; for example, impulse pay-per-view, interactive two-way node management protocols, etc. The recombination problem affects more than just high-speed data services for Internet.


Solutions for the initial sparse deployment scenario are few. Either buy sufficient CMTS equipment to cover the upstream return ports, or look into solutions that recombine data but do not recombine noise. Look toward CMTS solutions that support a large number of upstream return ports per downstream port.


A port mismatch problem exists when deploying high-speed data services on new HFC plant. This problem results in the economic reality that more CMTS boxes must be purchased than the initial sparse subscriber capacity requires, or the revenue stream comfortably allows. Upstream trunk port combiner techniques that recombine high-speed data streams but do not recombine noise impairments are both clearly and desperately needed in the industry. When available, these devices and techniques will help cable operators better match capital growth with revenue growth in the initial deployment of high-speed data services on HFC systems.

This article suggests holding off doing HFC upgrades for as long as possible, using techniques for managing noise impairments, such as high pass filters or newer technology intelligent switching filters. When upgrading to HFC, do so in an incremental fashion by only upgrading plant where there is a service demand and a revenue stream. Start with large node sizes and then, if needed, incrementally decrease toward 500 HHP as service and revenue increase.