Broadband deployments in dense city areas Comparing G.fast and VDSL/35b from street cabinets Rob F.M. van den Brink TNO 1
Different broadband migration strategies: FttH G.fast Hybrid FttH Curb 20-200m FttCAB 3GBB (up to 100Mb/s) VDSL2 < 1km copper 4GBB (up to 1 Gb/s) (c) TNO Full FttH Basement Meter closet Disruptive copper approach From new nodes, near/at the home Aims at 1 Gb/s (>500Mb/s) Typically within 20-200 range Reverse powered New developments: G.Fast (up to 106MHz) Gradual copper approach From existing nodes, e.g. street cabinets 50 100 200 400 800Mb/s VDSL VDSL/vec VDSL/vec/bond Typically within 300-1000m range Locally (or remotely) powered New developments: VDSL/35b (up to 35 MHz, 8192 carriers) branded as Vplus, Super Vectoring, When to use G.fast and/or when VDSL/35b? measurements 2
Comparing technologies: G.fast Results with first G.fast prototypes (Q4/2014) from different vendors First prototypes Q4/2014: 2 vendors, different implementations Up to 4 vectored lines Retransmission fully supported Improvements are ongoing Adtran at TNO Sckipio at TNO 3
Comparing technologies: G.fast TNO measurements on first G.fast prototypes (Q4/2014) 800 700 Vendor 1 Vendor 2 Aggregate bitrate compared to standard vectored single G.fast standard 800 700 Bitrate compared to standard single vectored G.fast standard Agg. Bitrate in Mbit/s 600 500 400 300 Agg. Bitrate in Mbit/s 600 500 400 300 200 200 100 50 100 150 200 250 Looplength in m 100 50 100 150 200 250 Looplength in m 4 Observations so far (4 vectored lines): Vectored bitrate (4 lines) close to single-line performance Transmission feels already as pretty robust (for a prototype) Starts up rapidly (within seconds) G.fast performs much better as expected at loops above 250m
Comparing technologies: G.fast TNO measurements on improved G.fast prototype (Q2/2015) 1000 900 800 700 Sckipio april 2015 16 lines vectored GA1 single link GA1 single link GA2 bitrate in Mbps 600 500 400 300 200 100 0 0 50 100 150 200 250 300 350 400 450 500 loop length Observations so far (16 vectored lines): Again: vectored bitrate close to single-line performance Again: Very fast startup times G.fast outperforms VDSL/17a up to 450m Product upgrade during test significant improvement 5
Comparing technologies: G.fast TNO measurements on G.fast prototype (Q2/2015) First conclusion: G.Fast has significant potential in Cabinet deployments as well 6
Comparing technologies: G.fast and VDSL/17a TNO measurements on G.fast prototype, compared with VDSL/17a 180 trainingtime DSL systems 160 140 120 time in s 100 80 60 40 20 0 1 2 3 4 VDSL2 Bonded VDSL2 G.fast(cold) G.fast(hot) Start-up times (cold starts or after severe interuptions) G.fast starts-up in seconds! Much faster then we have seen for VDSL/17a Seems to be a fundamental difference between G.fast and VDSL 7
Comparing technologies: G.fast and VDSL/35b VDSL/35b equipment not available yet 8
Comparing technologies: G.fast and VDSL/35b 800 600 [Mb/s] Aggregate (dn+up) bitrates VDSL/35b (ALU) VDSL/17a (ALU) 400 200 These are Aggregate bitrates!!! (C) TNO 2015 Length [m] 0 0 200 400 600 800 1k Aimed bitrates for VDSL/35b Bitrates found on ALU website (Russell, Spruyt, VanHastel; Oct 2014) Not yet tested by TNO 9
Comparing technologies: G.fast and VDSL/35b 800 600 [Mb/s] Aggregate (dn+up) bitrates VDSL/35b (ALU) VDSL/17a (ALU) VDSL/17 (Survey.NL) 400 200 These are Aggregate bitrates!!! (C) TNO 2015 Length [m] 0 0 200 400 600 800 1k Comparison with deployed VDSL/17a Survey of many operational VDSL/17a lines (vectored) in the Netherlands, as reference Aimed VDSL/35b rates plausible below 300m Aimed VDSL/35b rates optimistic above 350m 10
Comparing technologies: G.fast and VDSL/35b 800 600 [Mb/s] Aggregate (dn+up) bitrates Gfast (16 vec, proto#2) VDSL/35b (ALU) VDSL/17a (ALU) VDSL/17 (Survey.NL) 400 200 (C) TNO 2015 Length [m] 0 0 200 400 600 800 1k Comparison with G.fast measurements @ TNO G.fast easily outperforms VDSL/35b up to 350-400m Q: How often do short loops occur in practice? 11
Comparing technologies: G.fast and VDSL/35b Cumulated length distribution of 1465 manipulation points Cumulated length distribution of 381 manipulation points 100 100 90 90 CDF [%] CDF [%] 80 70 60 50 40 30 20 10 0 0 100 200 300 400 500 600 700 800 900 1000 Length (m) 100 90 80 70 60 50 40 30 20 10 TNO Amsterdam Cumulated length distribution of 246 manipulation points TNO Rotterdam 0 0 100 200 300 400 500 600 700 800 900 1000 Length (m) CDF [%] CDF [%] 80 70 60 50 40 30 20 10 Den Haag 0 0 100 200 300 400 500 600 700 800 900 1000 Length (m) 100 90 80 70 60 50 40 30 20 10 TNO Cumulated length distribution of 395 manipulation points TNO Utrecht 0 0 100 200 300 400 500 600 700 800 900 1000 Length (m) Length distribution beyond cabinets in city centers Significant differences among cities Ranges from 95% to 37% within 300m for major cities in the Netherlands Technology optimum even different per street cabinet 12
Comparing technologies: G.fast and VDSL/35b Bitrate coverage when deploying G.fast from cabinets Large vector groups required if delivered to all homes 13
Comparing technologies: G.fast and VDSL/35b Bitrate coverage when deploying VDSL/35b from cabinets Large vector groups required if delivered to all homes 14
Comparing technologies: G.fast and VDSL/35b Optimum technology is location and ambition dependent and relies on how each of these technologies perform G.Fast from current street cabinets up to 106MHz, TDD Coexistence issues with VDSL Rapid start-up times (5..10 sec) First standard in dec 2014 2014: prototypes (different vendors) Potential for many improvements Can extend the value of copper for many years Technology for large vector groups still to be developed (>96 should be feasible) VDSL/35b from current street cabinets up to 35MHz, FDD Aims to be compatible with VDSL/17 Slower start-up times (VDSL>1min), bonded VDSL even >2.5 min observed First standard still in progress Extension on mature technology Close to its final performance Quick solution for the short term, but restricts further frequency usage Technology for large vector groups available (>200 should be feasible) 15
Comparing technologies: G.fast and VDSL/35b Optimum technology is location and ambition dependent and relies on how each of these technologies perform G.Fast from current street cabinets up to 106MHz, TDD Coexistence issues with VDSL Rapid start-up times (5..10 sec) First standard in dec 2014 2014: prototypes (different vendors) Potential for many improvements Can extend the value of copper for many years Technology for large vector groups still to be developed (>96 should be feasible) VDSL/35b from current street cabinets up to 35MHz, FDD Aims to be compatible with VDSL/17 Slower start-up times (VDSL>1min), bonded VDSL even >2.5 min observed First standard still in progress Extension on mature technology Close to its final performance Quick solution for the short term, but restricts further frequency usage Technology for large vector groups available (>200 should be feasible) 16
Benchmarking DSL: TNO G.fast Test bed Extension units 80m sections 80m sections 240m (planned) 400m section Fan-out unit 30.. 100m sections 10 m sections High level overview Based on real cables Currently: 30-500m range, in discrete steps 12 quads per cabel = 24 wirepairs All cables are shielded TNO G.fast-testbed Mimics reality in several ways: Crosstalk from customers at different locations Reflections from splices can be added Reflections from waterstops included (typically 2-4m) Reflection from manipulation boxes can be added Others (like bridgetaps) may be added in future Manipulation Box 10m 17
Benchmarking DSL: TNO G.fast Test bed Extension units 80m sections 80m sections 240m (planned) 400m section Fan-out unit 30.. 100m sections 10 m sections Manipulation Box 10m 18
Benchmarking DSL: TNO G.fast Test bed Fan-out unit 30-100m in 10m steps Crosstalk from customers at different locations 19
Benchmarking DSL: TNO G.fast Test bed Fan-out unit, addition of splices 10m sections, with/without splices Different splice types 20
Benchmarking DSL: TNO G.fast Test bed Extension units 80m sections 80m sections 240m (planned) 400m section Fan-out unit 30.. 100m sections 10 m sections TNO G.fast-testbed Manipulation Box 10m Manipulation box Cross connects near customers 21
Benchmarking DSL: TNO G.fast Test bed Extension units 80m sections 80m sections 240m (planned) 400m section Fan-out unit 30.. 100m sections 10 m sections TNO G.fast-testbed Manipulation Box 10m Extension unit Multiple cascadable sections 12 quads, crosstalk from 11 or 23 customers Reflections from waterstops 22
Benchmarking DSL: TNO G.fast Test bed Extension units 80m sections 80m sections 240m (planned) 400m section Fan-out unit 30.. 100m sections 10 m sections 0-5 KPN cable #1, 100m, waterstops every 4.0m of 30cm measurement model Manipulation Box 10m -10 23 Transmission [db] --> -15-20 -25-30 -35-40 -45 (c) TNO, 2013-50 0 20 40 60 80 100 120 140 160 180 200 freq [MHz] --> Testbed: presence of waterstops Transmission dips from multiple reflections at every 2-4m Caused by the protection against water in case of cable leakage Typical for Dutch cables (an maybe others as well)
Benchmarking DSL: TNO G.fast Test bed Testbed: inclusion of waterstops Transmission dips from multiple reflections at every 2-4m Caused by the protection against water in case of cable leakage Typical for Dutch cables (an maybe others as well) 24
Benchmarking DSL: TNO G.fast Test bed Extension units 80m sections 80m sections 240m (planned) 400m section Fan-out unit 30.. 100m sections 10 m sections 0 TRAN:[dB] Balanced - Transmission Manipulation Box 10m -20-40 -60-80 Many combinations: 30m 40m 50m 260m (c) TNO 2015 freq:[mhz] -100 0 50 100 150 25
Benchmarking DSL: You also need to inject noise Copper loops alone are not enough, You also need to generate noise (full G.fast band) Stationary noise (to mimic Alien disturbance) RFI noise (to mimic radio interference) Impulse noise (real world measurements) Noise injection unit TNO has developed this all for G.fast frequencies We offers benchmarking as a service Our G.fast noise generation technology can also be obtained from Spirent (DLS 5900) 26
Conclusions Discussed options for broadband deployment in dense city areas - Results from labtrials with first G.fast prototypes: - Aggregate bitrates allready above 700-500Mb/s, within 0-200m. - Fast start-up times, vectoring already up to 16 lines, progressing well. - G.fast performs much better than expected: usable up to even 450m. - G.fast bitrate outperforms VDSL/35b bitrate within 350m. - Loop length from cabinets are often short in dense city areas: - Ranges from 95% to 37% within 300m for a few major cities in NL. - Both G.fast and VDSL/35b are good candidates within this reach. - Preference depends on much more than just bitrate: max vector size, compatibility with legacy, investments vs ambition, robustness, start-up time, etc - Benchmarking essential part of choosing what technology: - To tress a modem under realistic and different conditions (testbed, noise,..) - To identify the maturity, limitations and true capabilities of a technology. Using G.fast from existing street cabinets has significant potential 27
For more information about G.fast, join us at ultrafastbroadband.nl 28
29