4096-OFDM Implementation on the HFC plant with
Fiber Deep and Distributed Access Architecture
Maxwell Huang
Study on 4096-OFDM Implementation on R-PHY + FD Architecture
4096-OFDM implementation over entire HFC
plant becomes feasible because
• CNR is improved significantly by R-PHY.
• Distortion and noise are significantly
improved by FD.
• MER distribution becomes consistent
over the entire plant by R-PHY+FD.
MER (dB)
Room Temp. Over Temp.
R-PHY
47
46
Power Amplifier
41.5
39.5
CPE
41.5
40.5
EOL MER
37.9
36.5
Device
Remote PHY + Fiber Deep Architecture




4096-QAM
=12 bit / symbol
=58.4Mbps / 6MHz
42dB MER required
Lower to 36dB by
OFDM w/ LDPC
EOL MER Estimate
Study on PAPR Increase
Assume 2dB degradation in the MER
caused by compression due to PAPR
Device
PAPR and Compression
CCDF Comparison
Probability All SC- QAM 6*4096-OFDM
10%
3.58
3.65
1.00%
6.29
6.71
0.10%
7.78
8.3
0.01%
8.79
9.55
0.001%
9.52
10.5
0.0001%
10.17
10.59
Peak
10.7
10.62
OFDM suffers the more
degradation in MER
than QAM when power
amplifier operates close
to its maximum output
power.
R-PHY
Power Amplifier
@ Compression
CPE
EOL MER (dB)
MER (dB)
Room Temp. Over Temp.
47
46
39.5
37.5
41.5
36.9
40.5
35.3
EOL MER Estimate
May not support
4096-OFDM
Partial Band CFR _ A Solution Under Investigation
Adaptive Baseband
Proposed solution to PAPR reduction:
• Applying the Crest Factor Reduction (CFR)
technique such as Adaptive Baseband on the
partial band. e.g. ONLY for highest OFDM
(channel A).
Reasoning:
• Partial band CFR still effectively reduces PAPR
but mitigating the trade-offs in performance
and computational complexity.
Apply the Adaptive Baseband
on OFDM channel A
Study on Power Increase
R-PHY + FD architecture could bring an unprecedented thermal challenges for the node
because
• Roughly 25 watt of DC power increase results from enabling the super high output
capability needed for fiber deep deployment.
• Roughly 20Watt DC power increase results from introducing R-PHY module in the
node.
A proposed power saving solution
The Dynamic Power Saving feature makes the bias current adjustable in field, so that
node can smartly set the bias current according to the actual cable losses and the change
in spectrum loading.
APSIS Compliant
Power saving associated with
cable loss distribution
Power saving associated with
Spectrum Loading
Thank You!
The Capacity of Analog Op3cs in DOCSIS 3.1 HFC Networks Michael He John Skrobko, Wen Zhang, Qi Zhang Calculated DS CNR vs. EIN as a Func3on of OIP 2
4K-­‐qam 2K-­‐qam 1K-­‐qam 4dB ! m
$
I pd &
#
" 2 %
CNR =
2
(2 ⋅ e ⋅ I pd + RIN ⋅ I pd2 + EINtotal
)⋅ B
Constella3on (QAM) DS OFDM Marginal DS Min OIP CNR (dB) (dBm) 4096 44.5 -­‐2 2048 40.5 -­‐8 1024 37 -­‐11 512 33.5 256 30 Note: 1.  OMI (2.1%/6MHz) for this example is chosen for 105-­‐1218 MHz OFDM signal loading. 2.  The Marginal CNR targets for DS OFDM signals are 3 dB above required values. 3.  The Min OIP (Op[cal Input Power) is for the DS Rx with EIN (equivalent input noise) = 3.5 pA/√Hz. < -­‐12 Measured Upstream NPR for Mul3ple Loads 12 dB DR 24 dB DR 12 dB DR Note: 1.  The US Rx op[cal input power is -­‐13 dBm. The EIN of US analog Rx is 1.3pA/√Hz. 2.  Typically the US op[cal link required min. NPR dynamic range (DR) is 12 dB. Thermal Noise Contributes more to CNR at Low OIP US OFDMA Constella3on (QAM) 9.5dB 1024 512 256 -­‐22.5 dBm Required CNR (dB) US Min OIP (dBm) 5-­‐85 MHz 5-­‐204 MHz 35.5 -­‐17 32.5 29 -­‐20 -­‐22.5 -­‐13 (measured) -­‐16 -­‐19 Note: 1.  OMI (5%/6.4MHz) for this example is chosen for 5-­‐204 MHz loading. The EIN is 1.3pA/√Hz. 2.  The US Min OIP results (in table) are in mee[ng the Required CNR with 12 dB of dynamic range, and are extrapolated base on the measured NPR DR at -­‐13 dBm Link Budget vs. Fiber Deep Requirements Rx 1 WDM Mux Tx 2 Fiber link (40km) Rx 2 Rx n Tx n Loss (dB) WDM Demux Tx 1 3 9 3 Total: 15 Link budget (dB) Probable Fiber Deep Op3cal Link 25 20 15 R-­‐phy 2K-­‐qam 512-­‐qam R-­‐phy/EDR 1K-­‐qam 4K-­‐qam 10 5 0 Downstream (up to 1218 MHz) Analog (moderate order) 10G Digital Upstream (5-­‐204 MHz) Analog (high order) Note: 1.  Assuming output power of analog Tx DS(US) is 10 dBm (3 dBm) for link budget calcula[on. 2.  Assuming 10GE op[cal transmission link budgets is with the EML Tx minimum output of 0 dBm, and APD Rx receiving sensi[vity of -­‐21 dBm (w/ 2dB fiber dispersion penalty). Summary q  CNR of analog op[cal links is dominated by EIN of the op[cal Rx at low op[cal input power. q  Analog op[cal link is s[ll workable for 4K-­‐qam OFDM DS (up to 1218 MHz) at -­‐2dBm OIP, while for 1K-­‐qam OFDMA US (5-­‐204MHz) at -­‐13dBm OIP. q  Digital op[cs can support 4K/1K OFDM/OFDMA with 9 dB and 5 dB more link power budget than DS/US analog op[cs, respec[vely. Thank You Michael He Cisco Systems michahe@cisco.com DOCSIS 3.1 Profile Management
Application and Algorithms
Greg White, Karthik Sundaresan
(CableLabs)
D3.1 Profiles & Creation Problem
N-Dimensional Vectors
• Modulation Profile: Vector of modulation orders
• CM MER: Vector of reported signal quality
How to choose best profiles?
• CMTS supports up to 16 profiles per channel
• “Profile A” : lowest common denominator
Dimensionality Problem
• N = 3800 or 7600 subcarriers
• 73800 possible profiles (8 bit - 14 bit Modulations)
• 73800-choose-15 = ~1048000 possible 16-profile sets
• Simplifying assumptions don’t help
D3.1 Profiles : Objective Function
• What function are we trying to maximize?
channel capacity using set of profiles P
• 𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺 𝐽𝐽 =
channel capacity using only profile A
• 𝐽𝐽𝑃𝑃,𝐴𝐴 =
1
Φ
𝐾𝐾𝐴𝐴 ∙∑∀𝑥𝑥∈𝑃𝑃 𝐾𝐾 𝑥𝑥
𝑥𝑥
– Φx = Nx/N (fraction of users assigned to profile x)
– Kx = sum of bit-loading values (all subcarriers) for profile X
Optimization Methods (1): PCA
• Profile Coalescation Algorithm
Optimization Methods (2): K-Means
• Clustering
using K-Means
Optimization Methods (3): KCA
KMeans
PCA
Start with K-Means to
quickly get initial clusters
KCA
Use PCA to reduce to
optimal set of profiles
Algorithm Comparison
• KCA : best choice for fast runtime
Use PCA to
reduce to 16
21 profiles
using K-Means
DOCSIS 3.1 Multicast Profile
Management Mechanism
Evan Sun Ph.D.
Standard Engineer
Multicast Profile Management
• Multicast Profile (MP) Optimization Prerequisite
– CM joins or Leaves multicast group
• Internal Profile Management
– CM joins or leaves procedures
• External Profile Application
– Interfaces between the PMA and CMTS
MP Optimization Trigger Conditions
CMTS
CM A
Multicast
Group One
CM B
Multicast
Group Two
Client 1
CM C
CM D
Client 2
• CM is working on DOCSIS 3.1 mode with OFDM downstream channel;
• First Client connected with CM wants to join the multicast group;
• Last Client connected with CM leaves the multicast group.
Internal Profile Management
CM Joins Multicast Group
PM Module
CM supports the
multicast profile
CM Leaves Multicast Group
Yes
PM Module
End
Find a higher profile for the remaining
group members can support
Choose a lower common profile, and
then test it using the OPT messages
All CMs support
the new profile
Force to replicate the multicast on
multiple profiles
Yes
End
All CMs support
the new profile
Using the current profile
Yes
Using the
new profile
External Profile Management
PMA
Multicast Profile Optimization
Trigger Message
Profile Optimization
Multicast Profile Test REQ
CMTS
CM 1
OPT-REQ
OPT-REQ
OFDM DS Profile Test RSP (CM1)
OFDM DS Profile Test RSP (CM2)
Testing Completed
Multicast Profile Switchover Message
OPT-RSP
OPT-RSP
OPT-ACK
OPT-ACK
DBC-REQ
DBC-REQ
DBC-RSP
DBC-RSP
DBC-ACK
DBC-ACK
CM 2
Interfaces between CMTS and PMA
1. CM Joins/Leaves Descriptor
2. OFDM DS Multicast Profile Test
Request Message
3. OFDM DS Multicast Profile Test
Response Message
4. Multicast Group Information
Request Message
5. Multicast Group Information
Descriptor
HUAWEI
To enrich life through communication
The World Is Flat
Capacity Optimization in a Coaxial
Network, Constrained by Total RF Power
Karl Moerder PhD, Futurewei Technologies Inc.
Fred Harris PhD, San Diego State University
The world is flat
• Capacity Optimization
• What do we think?
• What do we know?
• What can we prove?
• What does it mean?
What do we think?
Modulation Constellation Density
256 QAM
256 QAM
256 QAM
256 QAM
Input
PSD
256 QAM
f
200
400
600
800
1000
1200
Output
PSD
Frequenc y MHz
6dB
18dB
12dB
24dB
30dB
f
200
400
600
800
Frequenc y MHz
1000
1200
What do we know?
Modulation Constellation Density
16384 QAM
4096 QAM
1024 QAM
256 QAM
64 QAM
Input
PSD
6dB
f
200
400
600
800
1000
1200
Output
PSD
Frequenc y MHz
6dB
f
200
400
600
800
Frequenc y MHz
1000
1200
What can we prove?
Power vs Frequency
Power (Arbitrary Scale)
4000
3500
3000
2500
2000
1500
64-QAM
1000
16-QAM
4-QAM
500
0
0
200
400
600
800
Frequency (MHz)
1000
1200
What does it mean?
• It means the closer we come to a flat power spectral
density out of the amplifier and into the coax, the more
efficiently we use our limited RF power.
• For the same total RF power, a nearly flat spectrum at the
amplifier output significantly reduces the distortion from
the amplifier.
• The above points become increasingly important as the
total bandwidth gets wider.
• Pre-emphasis can be approximated with smaller
constellations at higher frequency and boosting the gain for
the smaller constellations.
Thank You
Karl Moerder, karl.moerder@huawei.com
Fred Harris, fred.harris@sdsu.edu
Hi Ho, Hi Ho to a Gigabit We Go
Positioning the HFC Network for the New Gigabit Era
Phil Miguelez
Comcast
What’s driving
the need for a
new network
architecture?
•
•
•
•
Source – www.gig-u.org
Competition
HSD growth
D3.1 / R-Phy
Future FDX
Architecture Migration Goals
• Continue to extend the life of the HFC network
• Provide expanded capacity needed to meet subscriber usage
demands and fend off competitive challenges with D3.1
– Reduce node serving area size to increase data capacity per HHP
• Improve OpEx and network reliability by eliminating RF actives
– Enable a passive coax access link to the home
• Provide a future migration path to an all IP / all fiber network
– Fiber Deep
Distributed Access Architecture
FTTH
Network Migration Options
Drop-In BW Expansion
Fiber Deep
FTTH
− Maintain existing station locations
and HHP
− Upgrade Amps and Node
electronics to 85/1GHz or 1.2 GHz
− Enables 1 Gb DS / 100 - 200 Mb US
peak rates. Lower avg rates
− High HP per node limits HSD tier
rate penetration
− Eliminate all RF Amps and reduce
serving area size to 128 HP per
node max
− Upgrade Node to 85/1218 MHz
− Enables true 1 Gb DS / 200 Mb US
delivered data rates
− Lower HP per node permits
increased HSD tier rate penetration
− Replace HFC network
with RFoG / PON overlay
− Enables 1 Gb DS and US
symmetric data rates
Cost
− Lower, $XX/HHP for avg system
− Large cost variations due to
density and plant condition
− Modest, $XXX/HHP based on
assumed 60/40 aerial / UG split
− High, $XXXX/HHP
− Incremental $XXX to
connect each subscriber
Pros
− Rapid scalability
− Operationally familiar
− Modular transition to R-PHY
− Migration path to FTTH
− Low OpEx cost
− Allows 2Gb to 10Gb HSD
Cons
− No long term network benefit
− Requires continued node splits
− Higher cost solution over time
− Workforce training and scale
− Slower to ramp
−
−
−
Description
/ Benefits
Cost prohibitive
Slowest to scale
Requires all new CPE
Fiber Deep N+0 Architecture Concept
Fiber Deep N+0 Design Challenges
• High output node level and tilt to allow maximum HP reach
– 64 dBmV analog ref output, Linear tilt extension from 1 GHz to 1.2 GHz
– N+0 node expansion ratio is typically 12:1 (Average: 70 HP, Max: 128 HP)
– Express cable used to reach additional taps
• 85 MHz Mid Split migration
– Legacy STB OOB agility issues means changing out older STB’s
• Maintain existing plant power design
– Node power consumption design closely watched
– PS location and size remain unchanged to avoid permit issues
– Added coax power lines, Access Cable bridging
• Network design training, Construction training
Conclusions / Lessons Learned
• Gigabit over builders are an expanding threat to every MSO
• D3.1 plus Fiber Deep N+0 provides the data capacity to meet
competitive challenges and deliver Gb per subscriber rates
• US BW change creates the largest challenge due to the wide
array of deployed legacy STB’s and requires the most planning
• Commercial customers determine the node cut in schedule
• Continual communication with the municipality and customer
is key to a successful, pain free network migration plan