Monday, January 3, 2011

HFC Vs FTTh

Since the early launches of Fiber to the Home (FTTH), advocates of this system architecture have touted its superior bandwidth as one of the main justifications for naming it the broadband network architecture of the future. However, while it is indisputable that fiber information capacity is vastly greater than that of a coaxial cable entering the home, today’s FTTH implementations yield no clear advantage over the near term possibilities for existing coaxial infrastructure. There are several points of confusion that arise when comparing cable’s Hybrid Fiber/Coax (HFC) architecture to a telco’s FTTH implementation.

Much of the confusion begins with comparing the number of subscribers in a given FTTH system to the number in a given HFC service group. Some writers on the subject begin with the assumption of 500 subscribers per cable TV service group, based on the late 90’s architecture statements from some of the largest cable system operators at the time. For HFC, a service group is usually defined by the number of subscribers served by a single optical node. However, the 500 subscriber figure is rapidly becoming a dated one. Much of the US HFC infrastructure was designed with segmentable optical nodes so that an optical node feeding 500 homes at the time could be reconfigured into 4 service areas of roughly 125 subscribers each. When those systems were designed, the optical wavelengths available to CATV system designers were limited to either 1310 or 1550 nm for purposes of economy, so eight separate fibers and 2 spares to each node were often part of the fiber network’s design to provide for unique content on each pre-planned, segmented branch.

Today, with the availability of more optical wavelengths made for cable television distribution (CWDM or even DWDM), the fibers that were once planned for use at a single node station could be extended to other newly installed, deeper node stations, thus further reducing the number of subscribers in a service area.
In fact, with the addition of just a single 2x2 node in each branch of the originally installed big 4x4 node’s service area, it is possible to cut the original 500 subscriber service area model down to 8 service areas of little more than 32 subscribers each- the magic number that happens to coincide with the telco GPON definition of 32 subscribers per service area.

Once the HFC service area is down to 32 subscribers, the coaxial system approaches the capacity of current GPON systems. This is simply because, whether in the coaxial or fiber part of cable’s HFC plant, the system carries digital information using radio-centric signaling, whereas GPON FTTH adaptation uses basic line-code (on-off) centric signaling.
Radio-centric signaling commonly separates different services into discrete silos, referred to as channels, a method known as frequency division multiplexing. In HFC digital radio signaling, each channel is configured to transport digital information by a specific amplitude modulation method known as Quadrature Amplitude Modulation (QAM).

A common QAM level in CATV is 256 QAM, which amounts to about a 38 MBPS data rate per each channel. HFC systems are migrating toward 1,002 MHz upper frequency limit, so at 6 MHz per channel, there’s about 158 channels of capacity on these upgraded systems. Simply put, that’s 6 GBPS full spectrum that fits on a single wavelength using the radio-centric cable signaling method. While this capacity is no big secret, it is almost never presented this way in the HFC vs. FTTH forum because cable employs so many different radio formats over the typical system that it’s hard to make a direct comparison given the state of a cable system today, with its mix of analog channels, digital channels, cable modem channels, etc.

One of the tricky things about having this kind of capacity, until very recently, has been the difficulty in accessing the available data rate. Since each channel is 38 MBPS, the maximum bandwidth available to one user would be 38 MBPS under DOCSIS 1.X and DOCSIS 2.0. But with the adaptation of channel bonding, a new feature in DOCSIS 3.0 a cable modem can access 8 channels simultaneously, with plans for 10 channels at a time. So, with current technology, it is possible for a radio-centric network to provide access of up to 1 GBPS per radio.
So far we’ve covered the downstream capacity of an HFC system and have found, that using radio-centric signaling, the capacity flowing from the headend to the subscriber is quite large given a subscriber group size comparable to a GPON system and a spectrum filled exclusively with digital carriers.
The next consideration is what to do with the upstream performance.
US Cable systems are divided downstream and upstream by frequency splitting passives, known as diplex filters, usually located in the line equipment- the amplifiers and nodes that deliver the signal to the subscribers via coax and fiber. The upstream path occupies spectrum from 5 to 42 MHz and the downstream occupies 54 to as high as 1,002 MHz currently. In short, the upstream path is just a tiny sliver compared to the downstream path. This legacy frequency split is due exclusively to the legacy frequency allocations of broadcast stations in the US. Since channel 2 is designated as the lowest carrier in television broadcast, and it occupies the spectrum from 54 to 60 MHz, then cable also currently begins its downstream spectrum at 54 MHz so that legacy analog television tuners can be used to receive cable signals without the need for a converter box. However, with the end of analog broadcast television upon us, this utility will soon be obsolete, even though cable companies can continue to carry analog stations beyond broadcast’s conversion to all digital transmission. Once consumers settle into an all digital world, the legacy frequency split will no longer serve any meaningful purpose. So, those networks that have certain modular equipment in the outside plant will be able to reclaim the lower end of the formerly analog spectrum for upstream signal path use by changing the frequency split (diplex) modules in their line equipment, thereby expanding the bandwidth of the return path up to a level nearly symmetrical with that used for downstream radio access. That means potential bidirectional access of 1 GBPS per subscriber radio (modem).
The scenario described above is somewhat arbitrary. That is, the 1 GBPS access could be any other rate that is or will be provided for in current or future multiple access radio standards. However, we will extend the 1 GBPS bidirectional scenario into the remainder of this part of the discussion.
We’ve taken up 2 GBPS of the roughly 6 GBPS of capacity providing for a pretty seriously high data rate, bidirectional data service. That leaves us 4 GBPS for the last but still very important component of signal carriage- the video services. Put another way, we’ve taken 20 channels of frequency spectrum to provide for data services (just ignoring the extra bandwidth in the 5 to 42 MHz range). That leaves us 138 of 158 channels for video service, assuming a 1,002 MHz system. Now, these remaining channels can be allocated for analog, standard definition video or high definition video, but continuing on the scenario of a premium, all-digital network we should look at providing for a good number of high definition stations. 5 HD services take up about 2 channels, so 160 HD services would take up 64 channels, leaving 74 channels for standard definition. At 7 services per carrier, there’s room enough left for 518 standard definition services- far more if switched digital video is part of the solution.
So, by matching service area size, migrating to all digital, and using multiple access radios for data access, an HFC system can rival the performance of today’s FTTH implementations, and even beat some of the older implementations. In the next installment of this discussion, some ideas for an even more robust implementation of a radio-centric network will be offered. Stay tuned

2 comments: