10G Full Duplex DOCSIS Implementation

Exceeds Expectations

A Technical Paper prepared for SCTE by

Richard S Prodan, Ph.D.

Engineering Fellow
Comcast Cable
1401 Wynkoop Street #300
720-512-3742
[email protected]

© 2021, SCTE® CableLabs® and NCTA. All rights reserved. 1

 Table of Contents
Title Page Number
1. Introduction.......................................................................................................................................... 3
2. DOCSIS 3.1 – The Basis of DOCSIS 4.0 Full Duplex Modulation...................................................... 3
3. DOCSIS 4.0 - Full Duplex Spectrum Sharing ..................................................................................... 4
4. FDX Node Evaluation in the Comcast Lab Cable Plant ...................................................................... 6
5. Echo Cancellation Performance Exceeded ........................................................................................ 7
6. The Future Evolution of FDX Technology ......................................................................................... 10
7. Conclusion......................................................................................................................................... 10
Abbreviations .............................................................................................................................................. 11
Bibliography & References.......................................................................................................................... 11

List of Figures
Title Page Number
Figure 1 - DOCSIS 3.1 OFDM Transmission and Reception ....................................................................... 4
Figure 2 - Example DOCSIS 4.0 FDX Spectrum Utilization ......................................................................... 5
Figure 3 – Node + 0 Cable Plant (Comcast Lab - Network Model 1) ........................................................... 5
Figure 4 – FDX Node with the RPD containing the FDX SoC ...................................................................... 6
Figure 5 – Node Port Upstream and Downstream Levels ............................................................................ 6
Figure 6 – FDX Node Downstream Signal Echo Paths and Levels .............................................................. 7
Figure 7 – Node Signal to Echo Interference (Pre and Post Echo Cancelation) .......................................... 8
Figure 8 – Upstream bit-loading vs. relative echo power.............................................................................. 9

List of Tables
Title Page Number
Table 1 – FDX Cable System Data Rates .................................................................................................... 9
Table 2 – DOCSIS Bit-Loading Specifications .............................................................................................. 9

© 2021, SCTE® CableLabs® and NCTA. All rights reserved. 2

1. Introduction
The advent of DOCSIS 4.0 Full Duplex (FDX) technology has arrived after several years of industry
collaboration and the publication the DOCSIS 4.0 FDX specifications. Implementation of an FDX node
reference design incorporating the remote PHY ASIC SoC has been initially evaluated in the Comcast
labs. The key technology to enable simultaneous transmission and reception in the same spectrum
requires real-time removal of the high-power cable plant reflections of the transmitted downstream signal
from the low-power upstream signal received using “echo cancellation”.

The successful performance of this enabling technology demonstrates the potential for greatly increasing
upstream throughput in an additional nearly 600 MHz of spectrum shared concurrently with the
transmission of the downstream signal. Evaluations indicate increased spectral efficiency beyond the
minimum performance requirements in the FDX specification. This paper will analyze the implied
capacity limitations in the specs. A comparison to the measured performance is made which demonstrates
exceeding these limits. This comparison will conclude that the performance limits in the FDX
specification should be reexamined to update the expected efficiency gains enabled by full duplex echo
cancellation technology as currently realized.

2. DOCSIS 3.1 – The Basis of DOCSIS 4.0 Full Duplex Modulation

DOCSIS 3.1 provides higher throughput and greater flexibility for cable data systems. The PHY layer
based on Orthogonal Frequency Division Multiplexing (OFDM) differs significantly from that in
previous generations based on Single Carrier Quadrature Amplitude Modulation (SC-QAM).

OFDM is a modulation scheme where many closely spaced, complex valued, harmonically related
(orthogonal) QAM data subcarriers of various modulation orders are transformed into a time domain
waveform or OFDM symbol. The resulting OFDM symbol in the time domain can be reversibly
transformed back into the original QAM data subcarriers in the frequency domain. This reversible process
uses the mathematically efficient implementation of the Discrete Fourier Transform (DFT) called the Fast
Fourier Transform (FFT) to accomplish this reversible transformation. The N-point FFT transforms a
group of N time samples into an equal number of N frequency domain complex valued subcarriers. The
inverse FFT reverses this transformation from the frequency domain into the time domain.

As shown in Figure 1, forward error correction (FEC) coding and interleaving is applied to the input data
bits. The error protected data bits are mapped into N frequency domain QAM symbols modulated onto N
orthogonal subcarriers. These QAM subcarriers are serial-to-parallel converted and input into an inverse
Fast Fourier Transform (iFFT). This transformation produces a single time domain OFDM symbol
comprised of the summation of all N subcarriers. A portion of the end of each OFDM symbol known as a
cyclic prefix (CP) is prepended to the beginning of the same symbol to prevent inter-symbol interference
(ISI) due to signal micro-reflections or “echoes”.

Any contiguous group of samples within each OFDM symbol can be used to recover the subcarriers using
the periodicity property of the DFT. If the latter part of the symbol past the prepended cyclic prefix is
used, and the micro-reflections are confined in time to the duration of the cyclic prefix, then the
remaining samples in the OFDM symbol are free of inter-symbol interference. Consecutive groups of
QAM subcarriers are thus transformed into OFDM symbols and transmitted successively over the cable
plant.

© 2021, SCTE® CableLabs® and NCTA. All rights reserved. 3

The reverse process at the cable modem receiver synchronizes the ISI free portion of each OFDM
symbol, performs an FFT and parallel-to-serial conversion to recover the QAM modulated data
subcarriers which are then demodulated, deinterleaved, and FEC decoded to recover the error-corrected
data bits. Details of the DOCSIS 3.1 OFDM/OFDMA transmission and reception are described in a prior
paper [1].

3. DOCSIS 4.0 - Full Duplex Spectrum Sharing

Full Duplex DOCSIS 4.0 uses the same OFDM/OFDMA modulation as DOCSIS 3.1 but overlaps
downstream transmissions from node to modem and upstream transmissions from modem to node within
the same frequency spectrum in the 108 MHz to 684 MHz FDX band. This greatly increases the upstream
bandwidth yielding a total channel data rate up to 5.7 Gbps with 1024 QAM subcarriers and over 4 Gbps
total data payload rate without overhead.

Legacy upstream DOCSIS 3.0 and 3.1 signals remain in the 5 to 85 MHz (mid-split) band. Legacy
downstream DOCSIS 3.0 and 3.1 signals occupy the 804 MHz to 1002 MHz band, and only DOCSIS 3.1
OFDM signals occupy 1002 MHz to 1218 MHz. An example of such spectrum allocation is shown in
Figure 2 with downstream (DS) and upstream (US) frequency bands.

© 2021, SCTE® CableLabs® and NCTA. All rights reserved. 4

The FDX band is divided into three 192 MHz sub-bands in the FDX band containing three downstream
OFDM channels overlapped with six upstream OFDMA channels. The FDX node simultaneously
transmits and receives these channels within the FDX band. Cable modems operate in a dynamic FDD
mode with the direction of each sub-band upstream transmission or downstream reception dynamically
assigned with a Resource Block Allocation (RBA) message allocating the upstream or downstream
capacity within the FDX band in a flexible manner as capacity demands require.

IG1 v IG2 v IG3 v IG4 v

TG1 TG2

Full Duplex DOCSIS 4.0 was originally specified for a Node + 0 plant without amplifiers beyond the
node. The Node + 0 cable plant architecture is shown in Figure 3. When a modem transmits in an FDX
sub-band, modems on the same tap (Iso-Tap) cannot receive on the downstream in the same sub-band due
to the high upstream interference introduced across tap ports into the low downstream receive level of the
other modems on the same tap. These modems are said to belong to the same interference group (IG).
However, modems separated sufficiently apart across different taps can have sufficient isolation to
interference. Such separated modems can belong to different IGs where the transmission of a modem in a
sub-band does not significantly interfere with the reception of other modems in other IGs within the same
sub-band. Identification of each modem IG is performed in the “sounding” process in which each modem
in turn transmits a test upstream transmission while all other modems measure the received level relative
to the downstream received level in a modulation error ratio (MER) triggered measurement window. Low
received MER modems are placed in a common IG while higher MER modems remain in other IGs that
are isolated from the test modem interference.

© 2021, SCTE® CableLabs® and NCTA. All rights reserved. 5

Modems in different IGs can be grouped into a single Transmission Group (TG) with common sub-band
RBA transmit/receive assignments. Other TGs can operate with complementary RBA assignments
providing full duplex simultaneous transmission and reception across all TG sub-bands at the node. A
detailed treatment of FDX PHY layer operation of interference management is described in [2].

4. FDX Node Evaluation in the Comcast Lab Cable Plant

The operation of the simultaneous transmission and reception in the FDX node reference design that was
evaluated in our lab. The 7 tap Node + 0 cable plant design used in this evaluation is shown in Figure 3.
The FDX node reference design is depicted in Figure 4. The Remote PHY device (RPD) containing the
FDX system-on-chip (SoC) in the FDX node receives legacy upstream below 85 MHz and legacy
downstream from 804 MHz to 1218 MHz (not shown). These legacy bands are needed to initialize the
node as well as cable modems in DOCSIS 3.1 mode prior to adding DOCSIS 4.0 FDX functionality in the
FDX band.

The downstream continuous OFDM signal is launched from the node port at a level from 37 to 48
dBmV/6 MHz with a 10 dB uptilt in the FDX band (see Figure 5). The downstream level determines the
maximum reach (number of equalized taps and feeder plus drop cable lengths) for a modem receive level
of 0 dBmV/6 MHz across the entire FDX and legacy bands.

© 2021, SCTE® CableLabs® and NCTA. All rights reserved. 6

The combined upstream burst OFDMA signals of all granted modem minislots arrives at the node port
without significant tilt at a low level near 0 dBmV/6.4 MHz. The upstream receive level is limited by the
65 dBmV total composite power (TCP) of the cable modems in the highest loss path from the modem to
the node. Thus, the upstream path being essentially equal path loss will transmit with the same power
spectral density resulting in the same 0 dBmV/6.4 MHz across the FDX band at the node port.

The high-level downstream signal launched into the cable plant will result in reflected signal energy back
toward the node due to tap return losses causing impedance mismatches with the cable characteristic
impedance in the cascade of connected tap-to-tap transmission lines. The node echo paths and the
resulting reflected signals or echoes are shown in Figure 6.

Note that the downstream echo power is below the downstream launch power at the node port but the
echo is higher than the received upstream signal at the node port. Taking the inverse FFT of the reflected
signal power shows the distribution of the relative echo level and delay of the echo path components from
the node port and each tap relative to the 0 dB node launch signal reference. This echo level being higher
than the received upstream signal at the node port results in a negative signal-to-noise ratio (SNR). The
upstream signal would not be recoverable with a negative SNR. Hence echo cancelation technology of
sufficient cancelation depth to suppress this interference and increase upstream received SNR is required.

5. Echo Cancellation Performance Exceeded

The RPD SoC in the FDX node reference design of Figure 4 reduces the echo level below the upstream
receive level using self-interference and echo cancelation of the downstream echo. The node downstream
signal is sampled and digitized as a reference for echo cancelation. The upstream signal with the

© 2021, SCTE® CableLabs® and NCTA. All rights reserved. 7

downstream echo through and the leakage across the FDX node port directional coupler of Figure 4
(replacing the frequency division duplex diplexer) is also digitized. Using the downstream reference
signal, the echo cancelation operation within the FDX SoC subtracts the echo plus leakage signals from
the upstream input plus interference.

The evaluated performance of the echo cancelation is shown in Figure 7. The resultant bit-loading
achieved was 1024-QAM subcarriers in the received OFDMA symbols with zero codeword errors after
LDPC decoding. The echo is substantially cancelled. The residual post EC signal to echo ratio is
increased to 35 dB to support 1024-QAM an shown in Table 2, Upstream CNR Performance for D3.1 – a
significantly positive upstream receiver SNR!

Figure 7 – Node Signal to Echo Interference (Pre and Post Echo Cancelation)

The upstream and downstream data rates per channel type and the total aggregate data rates for the
channel plan of Figure 2 are calculated in Table 1.

© 2021, SCTE® CableLabs® and NCTA. All rights reserved. 8

 Table 1 – FDX Cable System Data Rates

Table 2 – DOCSIS Bit-Loading Specifications

Upstream CNR Performance for D4.0 FDX Upstream CNR Performance for D3.1
Modulation CNR Minimum Modulation CNR Minimum
QAM Order Efficiency Threshold Power QAM Order Efficiency Threshold Power
(bits/subcarrier) (dB)* (dBmV/6 MHz) (bits/subcarrier) (dB)* (dBmV/6 MHz)
QPSK 2.0 12.5 0.0 QPSK 2.0 11.0 -4.0
8-QAM 3.0 15.5 0.0 8-QAM 3.0 14.0 -4.0
16-QAM 4.0 18.5 0.0 16-QAM 4.0 17.0 -4.0
32-QAM 5.0 22.0 0.0 32-QAM 5.0 20.0 -4.0
64-QAM 6.0 25.5 0.0 64-QAM 6.0 23.0 -4.0
128-QAM 7.0 29.0 1.0 128-QAM 7.0 26.0 0.0
256-QAM 8.0 32.0 3.0 256-QAM 8.0 29.0 0.0
512-QAM 9.0 36.0 5.0 512-QAM 9.0 32.5 0.0
1024-QAM 10.0 44.0 7.0 1024-QAM 10.0 35.5 0.0
2048-QAM 11.0 39.0 7.0
4096-QAM 12.0 43.0 10.0

The upstream bit-loading is a function of the echo level present relative to the minimum upstream receive
level. This relationship is illustrated in Figure 8 where lower echo levels yield higher bit-loading post
echo cancelation and vice versa. The assessment of this trade-off for a given minimum upstream receive
level is ongoing for a more complete characterization of the echo cancelation performance as
implemented.

Figure 8 – Upstream bit-loading vs. relative echo power

© 2021, SCTE® CableLabs® and NCTA. All rights reserved. 9

6. The Future Evolution of FDX Technology
The demonstrated results of this technology in our labs indicates a path forward to extend the reach of
echo cancelation beyond all-passive Node + 0 cable systems. The depth of echo cancelation with the
system designed transmit and receive signal levels and resulting echo levels shows 1024-QAM OFDMA
modulation is achievable in a Node + 0 cable system. The use of this technology in FDX amplifiers can
extend the reach into Node + N system designs.

Managing downstream output and upstream input levels in amplifier cascades such that the upstream
signal to echo ratio is reduced below that of the node can increase the achievable SNR at the upstream
amplifier output for the same echo cancelation depth in the node. Longer amplifier cascades will
accumulate the reduced echo cancelation residual interference lowering the received SNR and achievable
bit-loading at the node. But as shown in the Table 2 SNR vs. bit-loading, a 6 dB decrease in SNR to 29
dB lowers the modulation efficiency from 10 to 8 bits per subcarrier. This compromise could enable the
use of FDX amplifiers with echo cancelation in limited cascade depths for a 20 percent decrease in
overall upstream capacity over the 576 MHz wide FDX band. Such a compromise may be entirely
acceptable in exchange for the still very significant increase in wideband upstream capacity with amplifier
cascades extending the reach of FDX technology to Node + N cable systems.

7. Conclusion
The initial evaluation of echo cancelation technology of an ASIC SoC enabled RPD in a DOCSIS 4.0
FDX node reference design was described. The performance in a representative Node + 0 cable system
design in our labs demonstrated the successful implementation of FDX echo cancelation in a
representative but challenging Comcast Node + 0 cable system design. The measured 1024-QAM
OFDMA error-free performance exceeded the upstream capacity of 64-QAM OFDMA in the DOCSIS
4.0 FDX PHY specification – a 67 percent capacity increase.

The echo cancelation performance utilized in a future FDX amplifier can extend the reach of FDX beyond
all-passive Node + 0 cable systems. A trade-off of amplifier cascade depth vs reduced upstream spectral
efficiency could provide significant increases in total upstream capacity in existing 1.2 GHz Node + N
cable systems.

© 2021, SCTE® CableLabs® and NCTA. All rights reserved. 10

 Abbreviations
CP cyclic prefix
DFT Discrete Fourier Transform
DS downstream
FDX Full Duplex
FEC forward error correction
FFT Fast Fourier Transform
iFFT inverse Fast Fourier Transform
IG interference group
ISI inter-symbol interference
MER modulation error ratio
OFDM Orthogonal Frequency Division Multiplexing
RBA Resource Block Allocation
RPD Remote PHY device
SC-QAM Single Carrier Quadrature Amplitude Modulation
SNR signal-to-noise ratio
SoC system-on-chip
TCP total composite power
TG transmission group
US upstream

Bibliography & References

[1] Demystifying the DOCSIS 3.1 PHY, R. Prodan, L. Montreuil, A. Kliger, BZ Shen, SCTE Cable-Tec
Expo 2014

[2] Full Duplex DOCSIS PHY Layer Design and Analysis for the Fiber Deep Architecture, R. Prodan,
SCTE Cable-Tec Expo 2017

© 2021, SCTE® CableLabs® and NCTA. All rights reserved. 11