Access Incognito's library of Guides, Tips / Tutorials and more.

A Technical Guide to DOCSIS 3.1 and Beyond – Part. 2: Technical Advantages 

Published on 20 Oct 2015


Error Correction — All previous versions of DOCSIS used FEC (forward error correction) as a mean to deal with the BER (bit error rate), the rate at which errors occur in the signal due to degradation However, DOCSIS 3.1 increases efficiency with additional error correction methods. In DOCSIS 3.0 the FEC algorithm used was Reed Solomon (RS); providing a decent coding gain of approximately 6dB compared to the un-encoded stream. In North America this was coupled with Trellis Coded Modulation (TCM), which provided an extra 2dB gain.

During the development of the DOCSIS 3.1 standard, the committee evaluated additional error correction methods. One suitable method was called Low Density Parity Check (LDPC); currently used in WiMax, WiFi (802.11n) and Digital Video Broadcasting. Although LDPC is more computationally expensive, though less intensive as Turbo Codes, modern application-specific integrated circuit and DSP / SoC are sufficiently powerful to easily add this as the FEC mechanism, providing options at both the headend and CPE. If the DSP or FPGA is sufficiently powerful this FEC capability could be enabled via a software upgrade; which several CCAP vendors have already taken advantage of. This added approximately 6dB gain over DOCSIS 3.0 Reed Solomon method, shrinking the gap to the theoretical Shannon Limit at only 1dB for the spectral efficiency vs signal to noise ratio. The efficiencies of LDPC encoding has a direct impact on the modulations. The efficiencies provided by LDPC saves about 2bps/Hz; which means that a 6MHz downstream channel can effectively transmit an additional 12Mbps.

Modulation — Before DOCSIS 3.1, the maximum modulation available was 64-QAM and 256-QAM for upstream and downstream respectively. With the introduction of LDPC FEC, what used to require a 27dB signal to noise ratio now only requires a 22.5dB signal to noise ratio. With these increased efficiencies comes the availability to add some higher order modulation. An alternative way to look at this is that with a higher order modulation (like 1024-QAM), the required signal-to-noise ratio (SNR) using LDPC is the same as 256-QAM using RS+TCM. DOCSIS 3.1 has added support for up to 4096-QAM with future optional 8192-QAM and 16384-QAM in the downstream; while the standard now supports 1024-QAM and future optional 2048-QAM and 4096-QAM in the upstream. The important factor is that with LDPC and OFDM combined we can lower the required signal budget at higher order modulations to fit within where the cable plant commonly reaches most modems, which makes 2048-QAM quite attainable as the SNR threshold is only 36dB. As you can see in the graph below in the Variable Modulation Profiles of DOCSIS 3.1, this captures the majority of devices in today’s plant.

Given a uniform distribution of devices in your network, SNR through DOCSIS 3.1 is able to use higher order modulations for higher throughput. This model has been validated in a production environment with 20mm CMs, courtesy of Dave Urban of Comcast.

Variable Modulation Profiles — Augmenting the modulation story is a very key attribute of the 3.1 standard: multiple modulation profiles. Previously, operators were required to choose a modulation profile that would provide service to all the cable modems in the plant, which usually meant choosing a modulation based on the cable modems who had the lowest SNR. Even if the outside plant supported it, operators would still have been limited to 256-QAM. With DOCSIS 3.1, it is possible to use multiple modulation profiles, providing lower QAM modulation for those with lower SNR and higher order modulations for modems with better SNR. Assuming a typical gaussian distribution with a mean of 36dB and 2dB deviation, we would see an improvement of 35.8% network efficiency improvement when using multiple modulation profiles with DOCSIS 3.1.

Multicarrier Transmission — One of the most important contributions in DOCSIS 3.1 was the introduction of Orthogonal Frequency Division Multiplexing (OFDM). OFDM is based on the idea of frequency-division multiplexing, but the multiplexed streams are considered parts of a single stream. The bitstream is split into parallel data streams that are each transferred over its own sub-carrier; which are summed to form an OFDM signal. OFDM may be new to DOCSIS, however it has been used for many application including PLC, WiFi, and cell networks. What this meant for DOCSIS is that subcarrier spacing could be say 25kHz wide instead of 6MHz; which equates to a 960-QAM sub-carriers only requiring 24MHz of bandwidth. In DOCSIS, an adaptive equalizer with 24 taps covers 4.5 µs per 6Mhz channel. OFDM is simpler: a single guard interval is added which is generally set as the longest expected echo. Assuming as high as 2% bandwidth per guard band pair, we see we will save approximately 6% per 24Mhz of OFDM space. Additionally, the OFDM subcarriers can be bonded inside a block spectrum that could be as large as 192MHz which theoretically support 1.99Gbps. This has the added benefit of being able to leverage up to 10% efficiencies at the band edge on digital channel. Combining OFDM and multiple modulation profiles, channel bandwidths can be assigned to match exact real-time subscriber demand and/or channel conditions. Beyond the spectral efficiencies of OFDM with an LDPC FEC (COFDM) rather than mapping 6 or 8 bits to a symbol, the 4096-QAM maps 12 bits to each data subcarrier in the OFDM symbol.

Increased Spectrum Utilization — We have discussed higher order and variable modulation profiles, improvements in forward error correction and spectrum efficiency. DOCSIS 3.1 also has increased the RF domain resulting in additional capacity. The upstream frequency range is from 5MHz to 204MHz, growing from the prior limit of 42MHz.  While the downstream starts at 258MHz and spans to 1218MHz, optionally growing to 1794MHz. This increased spectrum utilization does have an impact on MoCA which previously was operating at frequencies higher than DOCSIS used. As a result, operators will need to ensure filters in the range of 860MHz to 1.7GHz are properly in place at the customer premises otherwise the MoCA signal will interfere with DOCSIS 3.1 service in the tap if the operator is using that frequency range. Some have even suggested that two F-connectors might be in the future for the gateway to enable full capacity for the respective frequency bands.

One of the tenants of DOCSIS 3.1 is to provide a backwards compatible framework; as such the committee has chosen to preserve downstream channels below 258MHz for legacy mode. This provides a mechanism for operators to deploy a hybrid SC-QAM and OFDM without having to service to an all OFDM deployment in one step.

Stay tuned for my next blog, where I’ll explore some of the flexibility operators have with DOCSIS 3.1, the impact of fiber, and what the future holds for the standard.