DOCSIS 3 0 Upstream Channel Bonding in the

Agenda § ATDMA Refresher § Frequency Stacking – CM Upstream (US) max output – Channel placement § Plant Effect – Power/Hz & laser clipping – US expansion to 85 MHz? § Options to “KISS” Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 2

Why Bother § Need more speed to retain customers § Faster US speed needed for faster DS TCP speed § Allow more chs for 1. x & 2. 0 CMs to load balance § Limit or reduce node splits § Introduce new US service of 50 to 100 Mbps § Allow migration of existing customers to higher tier and DOCSIS 3. 0 capabilities § Better Stat Muxing of medium tiers Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 3

US Best Practices § Assuming one or more ATDMA chs and one TDMA ch – Provide quarantine ch of TDMA for 1. x CMs and/or settop boxes – Load balancing 2. 0 CMs to TDMA ch complicates things • Explained later and may require DOCSIS restricted LBGs § Keep US freqs relatively close & same phy parameters when freq stacking § Bond US channels of similar parameters/speed § Use caution with rate adapt – refer to rate adapt paper § Logical channel 1 intended for lower/more robust modulation § Best to use 6. 4 MHz if possible instead of 2, 3. 2 MHz wide channels § Recommended not to bother with odd constellations like 8 or 32 -QAM – MER requirements not much different than those for next higher mod § Spectrum allocation has many variables such as; group delay, known ingress freqs, etc. Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 4

ATDMA General Deployment Recommendations § After increasing CW to 6. 4 MHz, measure & document unequalized MER near CMTS US port – Transmit at multiple test points in plant § 25 d. B or higher Unequalized MER is recommended – Less than 25 d. B reduces operating margin – Check US MER as well as per-CM MER § Pick freq below 30 MHz – Away from diplex filter grp delay § Turn on Pre-Equalization § Get rid of D 1. x CMs or create “quarantine ch” Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 5

US MER(SNR) Issues § Doubling ch width can drop MER > 5 -8 d. B! § Equalized vs unequalized MER readings § Modulation profile choices – 64 -qam for data – 16 -qam for maintenance & Vo. IP? – Max output for 64 -QAM is 54 d. Bm. V • “cab up n power-adjust continue x” can help § Pre-EQ effect – Great feature in 1. 1 & > CMs – Could mask issues Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 6

D 3. 0 US Issues § Frequency Stacking – What freqs to use, level considerations, laser clipping? § Diplex Filter expansion to 85 MHz – If 1 GHz amp upgrades planned, think about pluggable diplex filters – Potential CPE overload – RFo. G could be perfect scenario (maybe even 200 MHz) § How graceful is resiliency (partial mode) § Monitoring, Testing, & Troubleshooting – Test equipment may need to have D 3. 0 capabilities Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 7

US Level Issues § Max Tx for D 2. 0 64 -QAM for 1 ch is 54 d. Bm. V § D 3. 0 spec raised US ch max power by 3 d. B – Max of 57 d. Bm. V for 32 & 64 -QAM – Max of 58 d. Bm. V for 8 & 16 -QAM – Max of 61 d. Bm. V for QPSK § Max output is limited when stacking & drops 3 d. B every time you double – If not maxed out, levels don’t change – Four freqs stacked at 64 -QAM maxes out at 51 d. Bm. V/ch § CM ranging is different for 3. 0 vs legacy – CM will have some dynamic range to allow specific chs to be a few d. B different vs. other chs Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 8

Total Power § One US ch now could be 4 chs Txing at same time – Possibly 6. 4 MHz each; nearly 26 MHz US ch loading § Lots of power hitting US laser Tx § Probability of laser clipping increases, especially if using legacy FP lasers – DFB lasers have significantly more dynamic range – BDR (baseband digital return) is another option § Monitor systems above 42 MHz to see 2 nd & 3 rd order harmonics – Any burst noise above diplex filter coming out of US receiver is usually indicative of laser clipping Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 9

Laser Clipping distortion above diplex filter roll-off frequency § Green line represents flat US noise floor with no clipping § Blue trace indicates laser clipping Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 10

US Frequency and Channel Placement § Freq assignments; 5 to 42, 55, 65, 85 MHz? – Diplex filters, line EQs, step attenuators, equalized taps – CPE overload (TV IF between 41 -47 MHz) – Placement of different ch types § Each US ch used for bonding is an individual ch – Don't have to be contiguous – Can have different phy layer attributes like; • Modulation, ch width, tdma or scdma, etc. – But, may be wise to bond “like” channels § May be wise to keep freqs relatively close so plant problems like loss and tilt don’t cause issues Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 11

Sample Upstream Spectrum Usage # BW Modulation Style Primary Usage 61. 4 64. 6 3. 2 64 -QAM ATDMA D 3. 0 9 54. 8 61. 2 6. 4 64 -QAM ATDMA D 3. 0 8 48. 2 54. 6 6. 4 64 -QAM ATDMA D 3. 0 7 41. 6 48 6. 4 64 -QAM ATDMA D 3. 0 6 35 41. 4 64 -QAM ATDMA D 3. 0 & 2. 0 5 28. 4 34. 8 6. 4 64 -QAM ATDMA D 3. 0 & 2. 0 4 23. 6 26. 8 3. 2 16 -QAM TDMA D 1. x 3 20. 2 23. 4 3. 2 QPSK TDMA D 1. 0 DSG 2 13. 6 20 6. 4 64 -QAM SCDMA D 3. 0 1 Presentation_ID To 10 Euro Split From 7 13. 4 64 -QAM SCDMA D 3. 0 © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential TV IF 12

Reasons CM Does Not Bond on Intended USs § § § CM not in w-online mode or using 1. 0 cm file Mtc-mode off Mtc-mode required-attribute & no attribute in cm file No BG configured or incorrect fiber node config CM not set for bonding or firmware issue All US chs not “sta” – US(s) shut – Max or Min Tx issues – Poor MER, plant issues, mis-wired § Oversubscribed CIR – Call signaling (n. RTPS), min US guaranteed speed, – Could have multiple single ch bonding groups § Note: US service flows like UGS & RTPS are assigned to single ch bonding Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 14

US Load Balance & Isolation Example CMTS US 0 @ 24 MHz 4 -Way Fiber Optic Rx 1 Filter CMTS US 2 @ 31 MHz Amplifier 4 -Way Fiber Optic Rx 2 CMTS US 1 @ 24 MHz § Attempting to “share” one US port across two other US ports – Can cause isolation issues – Load balance issues (ambiguous grouping) – Low Tx CM in HE can overcome isolation and show up on wrong ports § Note: D 3. 0 CMs in mtc mode do not load balance on US Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 15

System Levels Reverse • 17 d. B at 5 MHz & 32 d. B at 1 GHz • Eliminates max transmit CMs CS(CEQ) tap • Eliminates high DS tilt to TV 26 350’ 1. 5 d. B 23 500’ 2 17 FEQ w/ US pad 600’ 2. 5 Input 17 Reverse 43 d. Bm. V transmit level @ the tap 42 39. 5 4 Step Attenuator or EQ tap 29 38 X PIII. 5” cable. 40 d. B @ 30 MHz A total design variation of ~14 d. B! Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 16

Transmit Level Possibilities § Use D 3. 0 CM in 2. 0 mode – Single frequency on D 3. 0 CM offers 3 d. B higher power § Running D 3. 0 CM in low mod scheme allows higher Tx power § SCDMA with more codes may allow higher Tx power depending on implementation § Min level could cause issues in lab or HE test CM – – Presentation_ID Pmin = +17 d. Bm. V for 1. 6 MHz Pmin = +20 d. Bm. V for 3. 2 MHz Pmin = +23 d. Bm. V for 6. 4 MHz Note: This is much worse with SCDMA plus, CMs with low Tx have huge range for potential laser clipping or “bleed-over” © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 17

Summary § Long term D 3. 0 service planning – Ensure optimized frequency allocation – Enable seamless upgrade to higher D 3. 0 tiers – Wire once § Account for phy connectivity, not just ch capacity – Not advantageous to combine noise to satisfy connectivity § Fix Max Tx issues now – Design for tight “bell-curve” (43 -48 d. Bm. V), if possible § Good News – ECR to increase US Tx levels – 61 d. Bm. V max, with 3 d. B typical Presentation_ID © 2009 Cisco Systems, Inc. All rights reserved. Cisco Confidential 18