An evolutionary approach to Gigabit-class DOCSIS
Three influential technology executives have mapped the options.
That cable companies would evolve their networks from hybrid fiber/coax (HFC) technology to fiber-to-the-home (FTTH), with its attendant gigabit transmission rates, was never in doubt. The questions have always been about the timeframe and what the intermediary steps would be.
The timing remains a question, but the intermediary steps? In a rare – if not unprecedented – collaboration, three of the most influential technology executives in the cable vendor segment have mapped the options, analyzed their feasibility and produced a paper – at 185 pages longer than some novels – evaluating the implementation of each.
The authors of the paper are John Chapman, CTO of Cisco’s Cable Access Business Unit, Mike Emmendorfer, Arris’ senior director of solution architecture and strategy, and Robert Howald, fellow of technical staff for customer architecture at Motorola Mobility, along with Shaul Shulman, system architect at Intel, another somewhat sizeable company that does a bit of business with cable.
The paper, titled “Mission is possible: An evolutionary approach to Gigabit-class DOCSIS,” predicts that cable might be able to ride DOCSIS technology to 2 Gbps rates (the authors refer to it as DOCSIS Next-Generation, or DOCSIS NG).
At The Cable Show, Chapman went further than that – much further. “We think we can get 10 gigabits on DOCSIS someday,” he said.
The proximate issue addressed by the authors is the lack of bandwidth on the upstream, now in the range of 5 MHz to 42 MHz. That the upper end of the range needs to be extended to 85 MHz has been a common expectation, but the authors suggest that that will eventually prove to be insufficient, as well. Ultimately, cable will find it useful to go (in stages) to 300 MHz, and perhaps even 400 MHz.
Such action at the bottom of the spectrum range will certainly necessitate a reaction at the upper end.
Most cable systems have at least 750 MHz of spectrum, while some have gone to 860 MHz, and a few others have expanded to 1 GHz.
If cable operators gradually reallocate the spectrum up to 300 or 400 MHz for use on the upstream, however, that would consume a big chunk of the spectrum currently dedicated to the downstream, even for those that have already gone to 1 GHz. So another measure the paper’s authors recommend is the eventual expansion of the upper limit of the range to 1.4 GHz or 1.5 GHz, and maybe even to 1.7 GHz.
DOCSIS technology would have to be developed further to accommodate the changes. As the technology advances, the single-carrier (SC) modulation techniques used today will almost certainly have to give way to new multi-carrier modulation technologies, which in turn are likely to necessitate the use of more advanced error correction techniques than those currently in use.
That’s a lot of change to manage, and the paper dives deep into the details of what’s technologically feasible.
In the paper and in The Cable Show session in which the paper was presented, the authors assured that their roadmap maintains enough backward compatibility to protect MSOs’ previous investments in equipment. They also provided carefully calculated estimates of when each evolutionary technological step might be necessitated.
In short, the paper proposes a lot of equipment purchases, which the authors try to address and soft-pedal at the same time. Kevin Leddy, Time Warner Cable’s executive vice president of technology policy and product management and the moderator of The Cable Show session in which the paper was presented, wasn’t going to let them off that easy.
“I’d like to see the price tag for all of this for our company. It would be $5 billion for us,” Leddy interjected.
Acerbic commentary aside, cable operators will eventually all be on the hook for substantial capital investments, and everyone knows it. The issue is managing the process.
The four authors did not agree on all issues, but there were many points on which they did see eye to eye – including a few points on which all agree they need to gather more information before they agree on anything.
What follows is our summary of the paper’s summary. The full paper includes highly detailed arguments behind all of the following points.
1. Maintain backwards compatibility
The proposal is that whatever form DOCSIS NG takes, it would use separate spectrum but coexist on the same HFC plant. Backwards compatibility would refer to the sharing of spectrum between current DOCSIS and DOCSIS NG.
One example of this is where a 5 to 42 MHz spectrum is used for four carriers (or more) of DOCSIS 3.0. At the same time, a DOCSIS NG cable modem (CM) would be able to use the same four channels (or more), plus any additional bandwidth that a new PHY might be able to take advantage of.
The initial goal is to advance CMTS and CM technology to extend the life of the current HFC plant (using 5 to 42 MHz). The next step would be what the authors refer to as the mid-split, extending the upstream spectrum to 85 MHz, which they assert can be achieved with today's DOCSIS 3.0 technology.
If an HFC plant upgrade strategy could be defined that would allow a cost-effective, two-stage upgrade – first to mid-split, and then later to high-split (anywhere from 200 to 500 MHz) – then the advantage of higher data rates could be realized sooner.
The high-split, the authors said, offers the best technical solution likely to lead to the highest-performance product at the best price in the long term. The logistical challenges that high-split encounters are not to be underestimated, they wrote, but they are both solvable and manageable and significantly less imposing than the one other option – the “top-split” approach. The top-split differs from the other splits in that the additional downstream bandwidth is carved out of the upper end of the expanded spectrum range, somewhere above 1 GHz.
The short-term goal is to make use of any and all available tools to manage downstream spectrum congestion, such as analog reclamation, SDV and H.264, and to deploy 1 GHz plant equipment whenever possible, the group said. The long-term goal is to utilize spectrum above 1 GHz and push toward 1.7 GHz.
Field measurements have shown that the spectrum up to 1.2 GHz is available in the passive RF link. Measurements also show that up to 1.7 GHz is available with what the authors characterize as “modest plant intrusiveness.”
Spectrum above 1 GHz is unspecified and is inherently more challenging than the standard HFC band, which leads to the conclusion that advanced modulation techniques such as orthogonal frequency division multiplexing (OFDM) will be useful, if not necessary.
4. A new upstream PHY layer
The recommendation for DOCSIS NG upstream is to add orthogonal frequency division multiple access (OFDMA) with a new forward error correction (FEC) scheme more efficient than the Reed-Solomon FEC commonly in use today, specifically a low-density parity check (LDPC) code.
There is a considerable amount of new spectrum available with DOCSIS NG that only requires a single modulation. Although advanced time division multiple access (ATDMA) and synchronous code division multiple access (SCDMA) could be extended, now is a unique time to upgrade the DOCSIS PHY to include the best technology available, which the authors of the report said they believe is OFDMA and LDPC FEC.
5. A new downstream PHY layer
The recommendation for DOCSIS NG downstream is to add OFDM with LDPC FEC.
Using the spectrum above 1 GHz requires an advanced PHY with a more complex modulation such as OFDM. To minimize the cost impact on CMs, a cap could be placed on the number of QAM channels required within the existing spectrum. OFDM will also be used below 1 GHz and will likely supplant legacy QAM bandwidth over time.
Peak-to-average power ratio (PAPR) for multi-carrier technologies such as OFDM is worse than for a single, isolated QAM carrier, but the authors said they do not anticipate PAPR issues for either the upstream or the downstream when compared with single-carrier, channel-bonded DOCSIS.
Through testing, the authors determined that as the number of single-carrier QAMs in a given spectrum increases, multiple SC-QAMs and OFDM exhibit similar Gaussian characteristics. Further, there are shaping techniques available for OFDM that can mitigate the impact of PAPR.
7. Higher orders of modulation
The recommendation is to study the option to define up to 4K-QAM for OFDM in both the upstream and downstream.
All of these new modulation schemes may not be usable today. The new LDPC FEC provides the equivalent of 5 to 6 dB of performance improvement, which, for the same noise floor, allows two orders increase in modulation. Thus, an upstream that runs at 64-QAM would run at 256-QAM with a new FEC, the authors determined.
Also, as fiber goes deeper, coax runs become shorter and other possible architectural changes are considered (the point of entry home gateway, digital optics with remote PHY), there may be opportunities to use higher orders of modulation. The DOCSIS NG PHY will define these options.
8. SCDMA support in cable modems
The recommendation is to not require SCDMA in a DOCSIS NG CM that employs OFDMA. It is generally agreed that OFDMA with LDPC will be able to replace the role that SCDMA and ATDMA perform today. Thus, in a DOCSIS NG CM, SCDMA would be redundant.
9. Upstream MAC layer baseline
The recommendation is to use the SCDMA MAC functionality as a basis for designing the OFDMA MAC layer.
10. Legacy issues
The legacy migration concerns with mid-split and high-split – such as analog TV, RF interference, ADI and out-of-band (OOB) – have workable solutions.
Analog TV is being reduced or eliminated in many leading markets. A smaller collection of analog channels could be remapped to higher channel numbers.
The extended upstream and downstream spectrums overlap various critical over-the-air carriers such as aeronautical frequencies. Either the HFC plant can be improved to reduce leakage or certain critical frequencies can be skipped.
Adjacent device interference (ADI) should only occur in dense HFC plants and can be managed with a legacy mitigation adaptor (LMA), which is defined in the paper.
The OOB channel can be recreated locally with an LMA, either through down-conversion or a DSG conversion. However, the majority of STBs should not need this.
11. High-split crossover frequencies
Further study is required to determine the upper frequency of the high-split upstream spectrum and the lower frequency of the downstream spectrum.
12. ATDMA in the upstream
Further study is required to determine how many ATDMA channels a CM and a CMTS should support in the upstream.
For example, the authors point out that many cable operators are already deploying three full-width carriers or four carriers of mixed widths between 20 MHz and 42 MHz. In order to fully exploit a 5 to 42 MHz spectrum, a DOCSIS NG CM would need to support these channels, so four is the minimum. Newer DOCSIS 3.0 CMs promise eight upstream channels. It depends on the market penetration of these CMs as to the impact on backward compatibility.
13. SCDMA in the CMTS
Further study is required to determine if SCDMA should be retained. It is generally agreed that SCDMA does offer better performance below 20 MHz than ATDMA. For DOCSIS 3.0, SCDMA may be required to get that extra fourth full-size carrier and is an important component for maximizing the throughput available in the 5 to 42 MHz band.
On the other hand, the authors noted, retaining SCDMA in addition to ATDMA and OFDMA potentially adds product cost, development cost and testing cost. This has to be weighed against any significant market penetration of SCDMA prior to DOCSIS NG being available.
One possible approach is to specify a small number of channels of SCDMA as mandatory and more channels as optional. However, an overall objective is to limit PHY technologies in the CMTS silicon to one or two, and that would imply the elimination of SCDMA.
Early deployment of mid-split would also help alleviate the need for SCDMA, as that would provide the extra spectrum to relieve the congestion in 5 to 42 MHz.
14. FEC for SC systems
Further study is required to determine if LDPC FEC functionality should be added to enhance the existing upstream and downstream PHY.
The argument supporting the advanced FEC is that it increases capacity and that the existing SC-QAM will benefit from this investment to optimize efficiency in systems that will be operating single-carrier mode for many more years. The contrary position is to cap the legacy design and only expand capability with OFDM.
15. Upstream ATDMA capabilities
Further study is required to determine if ATDMA functionality should be extended with wider channels, more channels, higher-order modulation formats and improved alpha.
The argument supporting this work is that the changes represent simple extensions of DOCSIS 3.0, and field experience and RF characterization of ATDMA tools suggests a high probability of success. The argument against this approach is to cap the legacy design and only expand capability with OFDM, and also that an OFDM implementation would be less complex.
16. Downstream QAM capabilities
Further study is required to determine if downstream QAM functionality, currently defined by ITU-T J.83, should be extended with wider channels and higher-order modulation formats.
The argument for doing this work is that it represents simple extensions of DOCSIS 3.0, and field experience and characterization of ATDMA SC tools suggests a high probability of success. Again, the alternative is to cap the legacy design and focus on expanding capability only with OFDM, contending that an OFDM implementation would be less complex.
17. Upstream MAC improvements
Further study is required to determine if any changes not directly related to OFDM are worth pursuing.
Current suggestions include changing the request mechanism from grant-based to queue-based, elimination of 16-bit mini slots and eliminating request slots on each upstream carrier. Modifications need to be weighed against increases in performance, a decrease in cost and the need for backward compatibility.