1. DOCSIS 3.0 DS Planning & Bandwidth Management
John Downey, Consulting Network Engineer – CMTS BU

2. Agenda Objectives Terms M-CMTS I-CMTS Optional Architectures
Adapters and connectivity, linecards, timing servers, e-qams, ..
I-CMTS
Optional Architectures
Frequency Stacking Levels
Frequency Placement
Isolation Concerns

3. DS Questions & Potential Concerns
Why it’s Needed
Competitive pressure, offering higher tiers of service, more customers signing up
Frequency Stacking Levels & Placement
What is the e-qam max output with four channels stacked
Do channels have to be contiguous?
Isolation Concerns
Applications w/ different service grps lead to overlaid networks
Signals destined for one node could “bleed” over to another
DS Frequency Expansion to 1 GHz
Amplifier upgrades are occurring now. It’s best to make the truck roll once, so think about diplex filters, spacing, taps, etc.
We need to add more DS capacity now in the form of DOCSIS With it come some considerations specific to the DS such as:
Why it’s Needed – This can range from competitive pressure, to higher tiers of service, to more customers signing up.
Frequency Stacking Levels & Placement – What is the e-qam max output with 4 channels stacked and do the channels have to be contiguous?
Isolation Concerns – Whenever applications have different service groups, we have overlaid networks. Signals destined for one node could “bleed” over to another.
DS Frequency Expansion to 1 GHz – Amplifier upgrades are occurring now. It’s best to make the truck roll once. Think about diplex filters, spacing, taps, etc.

4. Business Objectives Allow more BW for DOCSIS 1.x & 2.0 CMs
Limit/reduce more node splits
Introduce new HSD service of 50 to 100 Mbps
Allow migration of existing customers to higher tier and DOCSIS 3.0 capability
Better Stat Muxing
The business objectives for many cable system operators today is to provide faster speeds to compete with FioS. These speeds are for DS and US. DOCSIS 2.0 can provide approximately 37 Mbps on the DS and 27 Mbps on the US, aggregate speed.
Some per-CM speeds are approaching these peak rates and exceeding them. The only way to offer higher rates than what DOCSIS 2.0 can offer is to upgrade to DOCSIS 3.0.
Instead of reducing node sizes, which can exceed $10,000 per node split, it may be more economical to add more DS frequencies and logically bond traffic across multiple DS channels. This will require new CMTS equipment and cable modems.
This allows more aggregate speed for higher tier services to share and/or to offer per-CM peak rates in excess of what a single channel can offer.
At the expense of more spectrum allocation, a DOCSIS 3.0 CM can offer 150 Mbps on the DS and 100 Mbps on the US. This may be necessary for 100x25 Mbps service or to allow many modems at 20x5 service to share the bigger “pipe”.

5. DOCSIS 3.0 Terms Local DS = CMTS DS Remote DS = E-QAM DS
Primary = DOCSIS messaging
Secondary = Bonding
“Wideband” generically used to describe D3.0 DS bonding
Channel Grouping Domain (CGD) is MxN mac domain
Mac Domain = 1 DS & N USs
Service group = CMTS Fiber Node config
M-CMTS is an architecture, not necessarily D3.0
Provides DS load balance within MxN domain
I-CMTS allows MxN and bonding within linecard
D3.0 CM supports minimum of 4 US and 4 DS chs
CMs on market are 4x4 (TI-based) & 8x4 (Brcm-based)
To level set, it’s best to understand terminology. DOCSIS 3.0 has added some new terms such as:
Local DS = CMTS DS
Remote DS = E-QAM DS
Primary = DOCSIS messaging
Secondary = Bonding
Combiner Group Domain (CGD) is an MxN mac domain where the old terminology for a Mac Domain was 1 DS and N USs. Keep in mind that more USs in a mac domain require more maps and eats into the DS throughput. Approximately every US port uses .25 Mbps of DS capacity for maps.
DOCSIS 3.0 CMs support a minimum of 4 US and 4 DS channels even though it could be more.
“Wideband” is generically used to describe D3.0 DS bonding even though it was originally a proprietary term.
M-CMTS stands for modular CMTS and is an architecture, not necessarily DOCSIS This architecture requires a DTI server and e-qam with DTI capability. This provides DS load balance of legacy CMs within an MxN domain. This means a D1.x / 2.0 CM could move between multiple DSs without forcing a move to a new set of USs.

6. DOCSIS 3.0 DS Ch Bonding over SPA Overview
Uses M-CMTS compliant Edge-QAM
Cisco RFGW-1D and RFGW-10
Harmonic NSG-9000
Increase legacy DS port density of uBR10K
Uses DTI timing source for DS channels
Enables legacy DOCSIS [1.x/2.0] CMs to use external QAMs for operation
Allows MxN mac domains
Eliminates need for PC from 5x20 card
Allows bonding on all channels in a BG
Hits the 100 Mbps BW mark on a 3-channel CM
The downstream Wideband channel is created by taking DOCSIS frames, putting them into MPEG-TS packets, and placing those MPEG-TS packets onto QAM carriers. However, instead of placing those MPEG-TS packets “horizontally” in time along a single QAM carrier as is done in traditional DOCSIS, the Wideband protocol places those MPEG-TS packets “vertically” across the QAM carriers assigned to a Wideband channel. A DOCSIS frame is literally tipped on its side and striped our across a group of QAM channels.

7. Narrowband Network Topology

8. DOCSIS 3.0 DS Ch Bonding over SPA Requirements
CMTS
Cable linecard
US ports needed and host
DS bonding adapter
1 Gbps fiber or copper ports
DTI timing card
DTI Server (Symetricom)
M-CMTS compliant EQAM
Cisco RFGW-1D or RFGW-10
Harmonic NSG9000
Arris
WCM300 (Linksys BroadLogic)
SA 3-Ch DPC 2505 (BCM 3381)
Cisco/SA DPC x4 TI
Cisco/SA DPC x4 Brcm
While the WB jacket card is similar to a SIP, the processors on the card are not in the data path and are more of a control system for the SPAs

9. DOCSIS 3.0 M-CMTS System D3.0 BG Primary uBR10K CMTS D3.0 DS + Primary
DTI Server
D3.0 DS + Primary
Jacket
Card
DTI Card
DTI Card
D3.0 BG
WB DS
SPA
M-CMTS EQAM
WB DS
SPA
Primary Ch
CPU/RP
Primary
MC5x20H
SA DPC 3000
SA DPC-2505 CM CM CM CM
Backhauls
uBR10K
CMTS
NB DS

10. Questions Any rules-of-thumb to estimate service group size?
What recommendations are there for HE combining/splitting to avoid intrusive changes later?
Any impairments in HE or plant that will affect DOCSIS 3.0 more compared to earlier versions?
Isolation
Off-air Ingress
Attenuation
Freq assignments, spectrum allocation, plant limits
If small amount of extra BW needed, is it possible to split 4 chs from DS port to use 2 chs in one SG and other 2 in another SG?

11. M-CMTS: 100 Mbps Tier & 2 Ch US Bonding
5 DS frequencies
1 I-DS
4 M-DS
(5 Primary)
2 US frequencies
2 channel bonding

12. M-CMTS = 100 Mbps Service Tier
5 DS freqs
2 US freqs
Remote Bonding
Remote Primary
Local Bonding
Local Primary
16-QAM
64-QAM
3.2 MHz
6.4 MHz
5, 5x4 MAC domains with ATDMA & TDMA USs
E-qam overlaid for 2 nodes 70/2 = 35 connectors
3 e-qam chassis with 6 modules each
4 freqs * 35 = 140 QAMs = 6 SPAs = Spumoni & PRE4

13. I-CMTS: 100 Mbps Tier & 2 Ch US Bonding
4 DS frequencies (4 I-DS)
2 US frequencies
2 channel bonding

14. I-CMTS = 100 Mbps Service Tier
4 DS freqs
2 US freqs
Local Bonding
Local Primary
16-QAM
64-QAM
3.2 MHz
6.4 MHz
5, 4x4 MAC domains with ATDMA & TDMA USs
DS connector overlaid for 2 nodes, 35 connectors*2 = 70 nodes
4 freqs * 35 = 140 QAMs = PRE4

15. M-CMTS Overlay 5 DS freqs 3 US freqs 2x5 domain Remote DSs Local DSs
3.2 MHz
6.4 MHz
This diagram graphically shows fiber nodes vs frequency allocation with pros and cons of this scenario. This allows load balance of legacy modems between 2 DS frequencies and bonding on 4 DS frequencies. This requires 5 DS frequencies on the plant and 3 US frequencies.
We can use dcc tech 4 to load balance Basic subs across the local DS and one e-qam primary. If modems have a DS frequency in their config file, use “load balance exclude static strict” so only dynamic load balance takes place.
Pros
Four bonding freqs / e-qam connector
Only 1 e-qam connector per 8 nodes
Basic = 2 DS/2 nodes with DCC support
US load balance of 2.0 CMs
One US connector shared across 2 nodes for diminishing D1.x CMs
Cons
Requires M-CMTS architecture
Requires five DS & three US freqs
Must push 3.0 CMs to remote DS
Bonding group must be same IP bundle
CM ranging overlap with “real” data?

16. I-CMTS with DS Overlay 8 DS freqs 3 US freqs 5x5 domain Local DSs Pros
3.2 MHz
6.4 MHz
This diagram graphically shows fiber nodes vs frequency allocation with pros and cons of this scenario. This allows load balance of legacy modems between 2 DS frequencies and bonding on 4 DS frequencies. This requires 5 DS frequencies on the plant and 3 US frequencies.
We can use dcc tech 4 to load balance Basic subs across the local DS and one e-qam primary. If modems have a DS frequency in their config file, use “load balance exclude static strict” so only dynamic load balance takes place.
Pros
8 bonding freqs per 2 connectors
Only 5 connectors per 8 nodes
Can provide 8 ch DS bonding
US load balance of 2.0 CMs
One US connector shared across 2 nodes for diminishing D1.x CMs
Cons
Requires 8 DS & 3 US freqs

17. M-CMTS Wiring Diagram
This diagram displays how DSs are potentially combined with e-qam modulator DSs and then split to feed multiple service groups or fiber nodes. Downstream four is not used and its associated USs are used in the four mac domains. Upstream connectors use internal frequency stacking on even connectors 0-14 and external stacking on connectors

18. M-CMTS Narrowcast 5 DS freqs 3 US freqs 5x5 domain Remote DSs Local
3.2 MHz
6.4 MHz
This diagram graphically shows fiber nodes vs frequency allocation with pros and cons of this scenario. This allows load balance of legacy modems between 5 DS frequencies and bonding on 4 DS frequencies. This requires 5 DS frequencies on the plant and 3 US frequencies.
We can use dcc tech 4 to load balance Basic subs across the local DS and all 4 e-qam primaries.
The biggest hurdle will be number of e-qam connectors needed and total e-qam channels supported for 1 chassis. If the chassis only supports 48 total e-qam channels and each linecard uses 16, then only 3 linecards would hit the 48 ch limit for e-qams.
Pros
Four bonding freqs / e-qam connector
One e-qam connector per 2 nodes
Basic = 5 DS/2 nodes with DCC support
US load balance of 2.0 CMs
One US connector shared across 2 nodes for diminishing D1.x CMs
Cons
Requires M-CMTS architecture
Requires five DS & three US freqs
Must push 3.0 CMs to remote DS
Four e-qam connectors and 16 e-qam chs per 8 nodes

19. DOCSIS 3.0 DS Considerations
Frequency assignments
CMTS may be limited to 860 MHz or 1 GHz
Legacy CMs (1.x & 2.0) limited to 860 MHz bandedge
E-qam limited to contiguous 24 MHz or 4 channel slots
Annex A may only be 3 chs vs 4 for annex B
CMs may be limited to 50 or 60 MHz passband
M-CMTS architecture requires DTI and local USs
Distance limitation, time offset differences, level differences
Resiliency is another topic to address
If one DS frequency goes bad in the field, how will CMs recover or react?
DOCSIS 3.0 has some inherent issues to address. One of which is frequency assignments. The CMTS may be limited to 860 MHz or 1 GHz and spectrum allocation may be scarce in the actual cable plant.
The E-qam is typically limited to a contiguous 24 MHz passband or 4 ch slots. Keep in mind that Annex A may only be 3, 8 MHz chs vs 4 annex B, 6 MHz chs. The CM limitation may be 50 or 60 MHz passband. D3.0 CMs require at least 60 MHz passband. Example; having the local freq at 117 MHz & the e-qam remote at 740 MHz will not work since the CM only has a 60 MHz passband.
The M-CMTS architecture requires a DTI server and US ports from the CMTS. The spec has a distance limitation of 200 meters between the CMTS and e-qam. There are ideas of utilizing GPS to sync multiple time servers to allow the e-qam to be in a hub site and the CMTS in the HE.
One potential issue with a DS being remote and a DS being local is the difference in time offsets. Load balancing using DCC tech 4 is difficult when the time offset delta is greater than 10 time offset ticks from one DS to another. This may require e-qam “tweaking”. Also, D1.1 config files are required for dcc to work. Another concern could be level differences if the DS frequencies are too far apart. Never use an S card for a modular host and only 3 SPAs are allowed to be hosted per linecard.
There’s no need for D3.0 CMs to load balance since they share a bigger pipe.
Resiliency is another topic to address. If one DS frequency goes bad in the field, how will CMs recover or react? Today it may be the case that the bonding CM will come back up as online and act as a D2.0 CM in regards to speed. Work is currently underway by many CMTS vendors to create dynamic bonding groups and allow a CM to bond across the frequencies it determines to be good, but that information needs to be relayed back to the CMTS.

20. DOCSIS 3.0 DS Considerations (cont)
E-qam licensing?
CM requires 1.1 config file
More DS = more US
Testing and maintaining multiple DS channels
Physical chs have not changed for DOCSIS 3.0
Test equip with built-in CMs need to support bonding
DS ch bonding max power with 4 freqs stacked
Four chs stacked on 1 connector limited to 52 dBmV/ch
DOCSIS 1.x/2.0 DS is 61 dBmV max output
DS isolation issues
One issue concerns economics. How do you deal with frequency stacked DSs if not using them all? Some vendors may offer software licensing to pay when you use.
Testing and maintaining multiple DS channels may not a big issue since the physical channels have not changed for DOCSIS Although, handheld test equipment with built-in modems will have to be updated to support bonding.
Another overlooked issue is DS isolation and E-qam max output power per channel. These will be expanded upon in the next slide. DS channel bonding max power with 4 frequencies stacked on one connector is limited to 52 dBmV per channel. The DOCSIS 1.x/2.0 DS specification is 61 dBmV max output with only 1 channel per connector. This needs to be addressed in the HE combining and splitting so all channels that are combined reach the fiber optic transmitter at the proper level and flat.
For RFGW-1…Single Mode: 52 dBmV to 62 dBmV, Dual Mode: 48 dBmV to 58 dBmV, Tri Mode: 46 dBmV to 56 dBmV, Quad Mode: 44 dBmV to 55 dBmV.

21. DS Ports with Edge-QAM for DS Bonding
61?
52?
DS0 U0
U1 U2 U3
1x4
DS1
DS2
DS3
DSs 0-3 = 603 MHz
Edge-QAM
E-QAM = 609, 615, 621, & 627 MHz
Potential Isolation Path
DS Combiner
DS Splitter
DS Tx
This figure depicts possible DS isolation and level issues that can occur when DS frequency is narrowcast and another DS frequency is introduced across multiple nodes.
Isolation amplifiers can be used to prevent too much loss from the 4-way splitter and keep the signal from backfeeding when architectures like this are implemented. Keep in mind that isolation is related to the combiner devices and how close of a match to 75 ohms the common leg is and where it is connected. The 2-way splitter could have poor isolation because of poor return loss of the fiber transmitter or the CMTS US port. The problem with an isolation amplifier is an active device could be a single point of failure and also, can the isolation amp handle a high input of power of 50 dBmV or so?
It may be best to combine the e-qam modulator DSs closer to the transmitter and use RG6 cable to alleviate some headend (HE) loss. Also, if 16 or more frequencies are stacked on one connector, the device may need to incorporate internal tilt to overcome HE cabling tilt. Sixteen frequencies contiguous will be ~ 100 MHz of passband. If a typical allocation is MHz, this could be 1 dB of tilt. Higher frequency selections will have less of an effect. Also, 16 frequencies stacked will have less power per channel, ~ only 44 dBmV! It may be wise to design the HE combining and splitting with a worst case scenario now.
Requires:
5 DS freqs
3 US freqs in each node
Isolation amp

22. Isolation Amp W Can this device handle 50 dBmV inputs?
This is one example of an isolation amplifier
Can this device handle 50 dBmV inputs?

23. Design Rules & Restrictions
D3.0 spec goes to 1 GHz, some equipment may not
SA DPC2505 speced to 930 MHz
DPC2505, 3 ch CM needs all 3 DSs for 111 Mbps
Can do annex B &/or A; but requires more spectrum
D3.0 CM spec requires 60 MHz capture window
DPC3000 capture of 96 MHz over most spectrum
82 MHz max window supported over entire spectrum
TI 4x4 CM (60 MHz window)
Brcm 8x4 CM (2, 32 MHz bands or 1, 96 MHz band)
DS freqs must be contiguous within tuner block unlike 4x4 CMs
Can use RCC templates to setup both tuners
New feature called Split Tuner will analyze RCC & create 2 Rx modules and move tuners automatically without RCC templates
The DOCSIS 3.0 DS specification indicates 1 GHz support, but some equipment may not support that. For example, before the specification was finalized, a DS bonding modem was introduced with 3-channel bonding capability. The SA DPC2505 is widely used in the industry, but only uses 3 separate DS tuners and speced to 930 MHz. This allows the channels to be almost anywhere in the DS spectrum, but not fully DS 3.0 certified.
The DPC2505 needs all three DSs for 111 Mbps. This can be done with annex B &/or annex A chs, but requires more spectrum if annex A is used.
BCM 3380 chip (8x4) supports 2 tuner windows of 32 MHz each. This is the current generation chip that is used for our products (3010, 3212, 3825, 3925).
BCM 3382 chip (8x4) supports 1 tuner window of 96 MHz. This is a cost reduced chip only for CM and EMTA, not gateway. Based on what I’ve heard from TW so far, they are only interested in D3.0 gateways so I don’t expect them to want any 3382 based products.
The 3383 (8x4) and 3384 (16x4) will use full spectrum tuners?
A new feature called Split Tuner will analyze the RCC and create two receive modules and move the tuners automatically without RCC templates. This was first added in Torrey Pines 4 on mp3 and is in current mainline source.

24. Modem Steering Restrict legacy eMTAs to Local DS
Enforce legacy CMs to only register on Primary-only DS
Enforce legacy CMs to move to specific DS freq
Force 3.0-capable CMs to initialize on bonding capable primary DS
Can specify UCDs sent for each DS
Put voice call service flows on a primary DS
cable docsis30-voice downstream req-attr-mask 0 forb-attr-mask
EDCS explains this feature
Directing CMs is a topic that can not be overlooked. It may be necessary to restrict legacy eMTAs to the Local DS for linecard redundancy purposes. The e-qams on the market today may not support redundancy. For overlay architectures, it may be needed to enforce legacy CMs to only register on Local DS or move to a specific DS frequency. Because D3.0 modems today only support four channels of bonding, it may be necessary to force a 3.0-capable CM to initialize on remote/e-qam DS for channel allocation issues and speed requirements. Allowing a 3.0 CM to use the local DS for primary traffic may only leave 3 channels for bonding and not the intended speed. Having the local frequency at 117 MHz and the e-qam remote at 740 MHz will not work since the CM has a 60 MHz passband. Using a Cisco CMTS, attribute commands are used to achieve the above criteria. Refer to other CMTS vendors for similar functionality and commands.
To restrict voice services to the MC5x20;
cable service attribute voice-enabled downstream-type HA-capable
To keep all D3.0 CMs on the e-qam modulator primary;
cable service attribute ds-bonded downstream-type bonding-enabled enforce
To force Basic subs to register on a Primary-only DS or a specific DS frequency, respectively;
cable service attribute non-ds-bonded downstream-type bonding-disabled
cable service attribute non-ds-bonded legacy-ranging downstream-type frequency <Hz>
These options may be warranted when overlaying D3.0 service with another service from the same CMTS or a different CMTS. We can assign specific US ports to be sent as UCDs for specific DSs. So, the local DS can be a 2x5 mac domain/combiner grouping domain (CGD), but only specific US ports sent as UCDs for the local DS or e-qam channel primary.
Example CLI: interface Cable5/0/0
downstream Modular-Cable 1/0/0 rf-channel 0 upstream 1 3
downstream local upstream 0 2 4

25. Node Splitting Edge-QAM 2x4 domain appears as 2, 1x2 domains 1x4 DS0U0
U1 U2 U3
This diagram is an example of splitting a mac domain into 2 service groups and specifying particular USs to specific DSs.
Example CLI: interface Cable5/0/0
downstream Modular-Cable 1/0/0 rf-channel 0 upstream 2 3
downstream local upstream 0 1
Edge-QAM
2x4 domain appears as 2, 1x2 domains

26. Summary Cost effective and faster time to market
Decrease DS costs – deploy D3.0 later with no additional CMTS investment!
Targeted insertion of D3.0
Leverage existing US chs while adding more DS capacity
Load balance 1.x/2.0 and enable D3.0 when needed
Minimizes capex & opex
Leverage D3.0 bonding for D2.0 tiers & services
Better stat-mux efficiency & improved consumer experience

27. Summary (cont) Long term D3.0 service planning
Insure optimized frequency allocation
Enable seamless upgrade to higher D3.0 tiers
Wire once & add QAM chs as tiers or service take-rates go up
End-to-end solution minimizes risk
CMTS, QAM, and CPE
Can also disable DS bonding
No cable mrc-mode
Per-CM exclude with vendor specific MIB or TLV
One way to have a D3.0 CM not do DS bonding is using a vendor specific MIB (for example for SA CMs its), saCmDsBonding OBJECT-TYPE
SYNTAX INTEGER {disable(0), enable(1)}
MAX-ACCESS read-write
STATUS current
DESCRIPTION
"The MIB will choose whether to enable downstream channel bonding for bonding-capable modems.
Non-bonding modems will ignore this MIB object.
0 : Disable downstream channel bonding.
1 : Enable downstream channel bonding.
This MIB will take effect at the next reboot. This MIB value will be stored to NonVolatile memory(NVM) and will persist across reboots. If the MIB is set via the config file, the CM will store the new setting and reboot if a change is necessary. Removing the setting from the config file will not change the value stored in NVM: the CM will continue to operate using the previously stored value. If an SNMP SET is used to modify the value, then the CM will not use the new setting until the next reboot occurs or is commanded. A factory reset of the CM will set the stored value back to 1."
DEFVAL { 1 } ::= { cmInterfaceInfo 15 }
Or you can set the DOCSIS Downstream Service Flow Forbidden Attribute Mask TLV (25.32) to 0x which will prevent the SF from using a bonding group even when the modem is in the w-online(pt) state.