Sync Tested Mesh Microwave System

Transcription

1 Sync Tested Mesh Microwave System Billy Marshall Pre-sales Engineer International Telecom Sync Forum November 2013

2 CCSL Microwave Solution CCSL have developed a self-organising mesh microwave solution for smallcell backhaul Chronos assisted with characterising elements of the system for synchronisation performance System Features Compact design for street furniture No alignment. Very easy/quick to deploy 480Mbps Ethernet with QoS Simple planning, easy organic growth Synchronisation Features Integrated GPS, SyncE Low latency, low transient, fast switching Auto switch to secondary sync source if GPS fails

3 Sync Performance Testing Characterise equipment SyncE output performance under 3 operational modes GPS Lock GPS Fail, recover clock from neighbouring radio GPS via one hop Radio Link GPS Fail on two nodes, recover clock from remote SyncE line via one hop Radio Link Characterise processing latency and Packet Delay Variation under various traffic and topology conditions Equipment used CCSL Umbra v3 SyncWatch (SyncE, PTPv2), Paragon (SyncE Master) Traffic Generator (Sunrise 10G)

4 SyncE Test Setup GPS GPS SyncE Master SyncE Ethernet Tester SWITCH CCS Wired Node 28GHz SyncWatch EMS Loopback tester Traffic load 40% of maximum throughput, packet size 1518 bytes Remote Node SyncE at 1G Electrical to be measured for TIE and MTIE Measurement reference GPS

5 SyncE Modes Performance Overview 120ns 30ns/div -180ns 1.5 hrs GPS Locked GPS Locked Neighbour Network Locked Neighbour

6 SyncE output when GPS Lock (Baseline) Testing normal operation (GPS Lock) performance of a single node 320ns 24h TIE Plot Oscillator is low-noise VCXO 40ns/div Performance just outside PRC mask Well under G.823+1ppb mask -120ns 24 hours MTIE Plot GPS GPS G ppb SyncE Master SyncE Ethernet Tester SWITCH CCS Wired Node 28GHz SyncWatch ETSI PRC EMS Loopback tester

7 Reference from GPS Locked Neighbour Switch off GPS on measured node Node recovers sync from GPS locked neighbour Switching transient low Medium-term wander increases, long term performance remains the same 25ns Input Switch Transient TIE Plot 20ns over 26s 0ns 38 seconds Long Term MTIE of Input Mode SyncE Master GPS GPS LOST G ppb SyncE Ethernet Tester SWITCH CCS Wired Node 28GHz SyncWatch ETSI PRC EMS Loopback tester

8 External SyncE Line Locked Neighbour Switch off GPS on neighbour node also Node recovers sync from neighbour that is locked to external 1G SyncE line Switching transient low 240ns 20ns/div Input Switch Transient TIE Plot 220ns over 22s Switch BACK to local GPS Long-term wander increases slightly Switch BACK to GPS (normal operation) caused largest transient in testing (220ns) 0ns 67ns over 22s Switch TO Network 38 seconds Long Term MTIE of Input Mode SyncE Master GPS LOST G ppb SyncE Ethernet Tester SWITCH CCS Wired Node 28GHz SyncWatch ETSI PRC EMS Loopback tester

9 SyncE Output Performance Overlays 1 us 50ns div G /240ns jump 0.05ppm slope Input Switch Transients Switch back to local GPS Plot of SyncE performance of ALL switchovers against mask merge mask From GPS via neighbour to External 1G SyncE via Neighbour From local GPS to Neighbour locked to GPS 0ns 38 s Masks plotted are from G.8262 switching transients these seem loose and actual performance likely to be driven by specific operator downstream requirements Largest transient believed caused by wander between SyncE external line source and the wander from local GPS No standards masks for switching between PRC sources G ppb ETSI PRC Overlay of Input Mode MTIE 1G SyncE via Neighbour GPS via Neighbour Performance consistent across all 3 modes Medium-term wander most adversely affected by input mode however effect is small Again, tough to know which masks are applicable here, will be driven by operator downstream requirements Local GPS

10 PDV Test Setup GPS Reference GPS Reference 1588 Grand Master CCS Wired Node 1588 PDV Tester Packet generator SWIT CH Loopback tester GPS Reference Loopback tester SWIT CH 1588 PDV Tester Moved the PDV tester around to measure at different points in the mesh, changed configuration to force packet path GPS at both ends allows PDV and absolute one-way link(s) latency to be calculated Traffic loading was done with small (64 byte) and maximum (1518 byte) packets at 40%-60%

11 PDV Performance Measured Floor Packet Percentage (10µs) to characterise the PDV added by this system In a deployed network this is important as PTPv2 Slave requirements detailed in G In deployment this equipment will be at the edge and PTPv2 packets may already have large PDV Blocks of packets nearer floor are an effect of software set 150µs packet scheduling period Most packets make this schedule, a smattering of packets miss and are put into next schedule 400µs 40µs /div µs FPP = 49.5% Latency = 41.2µS 10µs FPP = 21.5% Latency=42.3µs 10µs FPP = 12.7% Latency = 95.1uS GPS Reference 1 0µs 1588 Grand Master CCS Wired Node Packet generator SWIT CH 3 2 Loopback tester 68 byte 40% traffic - 2 node chain - single hop 64 byte 60% traffic - single node feeding 2 nodes - 1 hop 1518 byte 50% traffic - 4 node chain, 3 hops GPS Reference Loopback tester SWIT CH 1588 PDV Tester

12 Conclusions / Thoughts Synchronous Ethernet SyncE performance under normal and automatic switching scenarios was just over PRC quality G.8262 input switch transient masks are easily achievable Standards masks for wander transfer where DUT is two devices over a proprietary link rather than through a single piece of equipment? PTPv2 PDV Performance Latency was low (~42µs per link) FPP good, future inclusion of transparent clocking should reduce PDV across links. Are there likely to be specific operator requirements for PDV transfer or limits for a network segment i.e. hand off from one transport medium to another?

13 Thanks