Optical Power Budgets. Session Chair: Tony Frisch, Xtera

Transcription

1 Optical Power Budgets Session Chair: Tony Frisch, Xtera

2 Power Budgets Key to the Supply or Upgrade of any Submarine System? Presenter: Company: Tony Frisch, Priyanth Mehta, Tim Stuch Xtera Ciena Microsoft

3 TF Presenter Profile Tony started at BT's Research labs and then moved to Alcatel Australia, becoming involved in testing submarine systems. A move to Bell Labs gave him experience in terminal design and troubleshooting, Contents after which he went back to Alcatel France, where he worked in Alcatel Submarine Networks Technical Sales before moving to head Product Marketing. He is now SVP, Repeaters and Branching Unit for Xtera Communications. Name: Title: Tony Frisch SVP Repeaters and Branching Units tony.frisch@xtera.com

4 Presenter Profile Priyanth Mehta is an optical systems designer for the submarine research and development team at Ciena in Ottawa, Canada. He received his B.Sc. (Hons) (2007) and M.Sc. (Hons) (2009) in Optical Physics from the University of Auckland, New Zealand. He then obtained a PhD in the nonlinear properties of semiconductor optical fibres from the Optoelectronics Research Centre, University of Southampton, in At Ciena, his primary fields of research are focussed on improving transmission capacity, reach, and user operability through modem and line terminal enhancements. Priyanth also serves as a contributing delegate in standardisation on the International Telecommunication Union (ITU) for Optical Transport and Access. Name: Title: Priyanth Mehta Submarine Optical Systems Designer prmehta@ciena.com

5 Presenter Profile Tim Stuch is a Principal Network Engineer for Microsoft s Azure WAN transport team, where he is the engineering lead for all of Microsoft s Subsea engagements from contract through to operations. Currently he is working on extending Microsoft s cloud globally with a focus on making open cable concepts into real world deployments. Working on subsea for the last five years, Tim has been designing and deploying IP and transport networks for 20 years with companies like Bay Networks, Nortel, and Ciena. Tim received his B.S. Physics from Guilford College and his Physics Ph.D. from the University of Denver. Contents Name: Tim Stuch Title: Principal Network Engineer tstuch@microsoft.com

6 Other contributors P Murphy AJC O Ait Sab Alcatel-Lucent L Moskowitz AT&T J Gaudette Microsoft N Brochier Orange D Welt Tata C Mott Telstra P Booi Verizon E West Vodafone

7 Contents 1. Introduction and uses 2. Fundamentals with a few simplifications 3. How to construct a Power Budget 4. How it feeds into Acceptance 5. Refinements removing the simplifications handling ROPA or DRA 6. Discussion comparisons / sanity checks open systems

8 1. On-Off OOK Amplitude ASK A brief history DEC FEC 2. Phase DPSK Differential encoding UI DEC FEC Local oscillator ~ 3. Multi-level 16QAM λ/4 ADC Coherent transmission Soft Decision FEC Pulse-shaping Polarizing beam splitter/combiner ADC ADC ADC DSP FEC Framing A lot has changed

9 Specified in ITU-T Rec. 977 "Characteristics of optically amplified optical fibre submarine cable systems" Annex A provides templates in the latest version there are two Example 1 The original Example 2 This (2015) edition introduces a new power budget template for the implementation of coherent systems same Looks different, but the principles are mostly the

10 TS Where it's used Presenter: Company:

11 Bidding Defines key Requirements System Length Traffic capacity Repair and Ageing Commissioning limit i.e. start of life performance to guarantee end of life margin Error performance Any constraints e.g. keep some/all existing traffic when upgrading Should make it easy to compare different proposals

12 System design May not use ITU-T G.977 template Used to explore possible system parameters Fibre loss / effective area Number of repeaters, Power, Bandwidth Transmission formats, Encoding, FEC

13 Acceptance Verify budget in particular: Measure average Q-factor and time-varying Q-factor (5 sigma) Estimate worst case Q-factor over time (5 sigma calculation) Worst performing wavelength and compare with Commissioning limit calculation of required start of life margin to obtain end-of-life performance commitment

14 Track system performance Maintenance and long-term operation Q measured by SLTE, delivered Error Performance Track repairs Additional cable Repair joints Can affect tilt Can measure impact on Q a single repair generally has a small impact, but they accumulate Can determine remaining margin and risk

15 TF Fundamentals Presenter: Company:

16 OSNR, Q OSNR Optical Signal to Noise Ratio [ db/0.1 nm ] Q Quality factor [ db ] Frequency Speed of light / Wavelength [ GHz ] BW Bandwidth [ GHz ] or [ nm ] Important to use consistent units, ideally All SI Linear or db Useful approximations 1.0 nm ~125 GHz (at 1550 nm)

17 Definition OSNR Power in one channel / Noise Spectral Density (NSD) Power [ mw ] / NSD [ mw per 0.1 nm ] OSNR per 0.1 nm 10 log (Sig/NSD) db per 0.1 nm Obtained by: Calculation Measurement using an Optical Spectrum Analyser (OSA) care needed to get accurate measurements

18 Constellation diagrams Represents signals in phase space I = In phase, Q = Quadrature Q I Amplitude Phase Receiver can display in same format Gives an idea of the signal quality Q ~12 db Q ~6 db

19 Definition Q The key parameter used in budgets Q = 6 db Q original defined a Quality factor at the detection point Q = Signal / RMS noise (Linear) Q = 20 log(sig / Noise) Assuming noise is Gaussian Error Rate before FEC = erf(q) = 1-NORM.DIST(Q,0,1,TRUE) Q = -NORM.INV(ERBF,0,1)

20 Calculating noise simplified Amplifier behaves as if there were an input noise of NF x hv [ W/Hz ] NF hv BW G G NF hv BW NF Noise Figure depends on the amplifier h Plank s constant ~ 6.6E-34 v frequency = speed of light / wavelength ~ 3E8/1.55E-6 = 1.93E14 = 193 THz hv ~ 1.28E-19 BW Bandwidth ~ 1.25E10 Hz for 0.1 nm

21 Amplifier characteristic Operating point with 16 db loss 16 Output (dbm) 15 Partial restoration after 1 amplifier Input (dbm) Is self-stabilising: low input produces higher gain Amplifier gain tends to be close to fibre loss

22 A good simplification for long systems Noise accumulation identical sections NF hv BW G G NF hv BW 2 NF hv BW 2 G NF hv BW G After N amplifiers (don t forget that the SLTE contains an amplifier) Noise fibre loss ~ G = N G NF hv BW = N L NF hv BW where L is OSNR = Pch / (N L NF hv BW) LINEAR

23 More commonly used Noise accumulation logarithmic units NF hv BW G G NF hv BW 2 NF hv BW 2 G NF hv BW G NF = 10 log(linear Noise Figure) db Pch = 10 log(power in mw ) dbm/channel L = Section length x fibre attenuation db/km OSNR = Pch 10 log(n) L NF + 58 db/0.1 nm

24 Q dependence on OSNR Simple theory suggests Q = Coefficient x OSNR Coefficient depends on: Receiver optical bandwidth Q(OSNR) electrical bandwidth Modulation format 13 QPSK 9 Example for illustratiom Depends on implementation Practice is more complex Best to measure OSNR (db/0.1 nm)

25 Different Formats 1 bit/symbol 2 bit/symbol 3 bit/symbol 4 bit/symbol BPSK QPSK 8 QAM 16 QAM -3 db 0 db 4 db 7 db More power / OSNR than QPSK Higher line-rates need significantly higher OSNR Theoretical penalties reality will be a little worse particularly as the number of levels increases

26 Different Formats Practice QPSK 11.0 Q db 8 db 8QAM 16QAM Example for illustratiom Depends on implementation OSNR

27 Constructing a Power Budget Presenter: Company:

28 Optical budget (simplified) 1 Q from OSNR may include effect of SLTE Propagation effects non-linearity, dispersion etc. Imperfections pre-emphasis etc. Supervisory small effect Time variations 5 sigma i.e. worst case Effects of terminal if not already included = Q Line at Beginning Of Life (BOL / SOL)

29 Optical budget (simplified) 2 Need to calculate for Beginning of Life (BOL) AND (EOL) Budget must consider End of Life Cable repairs Ageing effects Pump failures Must be some Operating margin, typically 1.0 db (Q), at EOL Consider reducing EOL margin Fewer amplifiers in a new build OR More traffic capacity in an upgrade

30 Margins for ageing and repairs Extra loss due to repairs, typically 0.4 db every 40 km (shallow) 3 db every 1000 km (deep) 0.4 db per land cable repair Extra loss due to fibre ageing, typically db/km Pump failures, repairs typically Pump repair repair 5% of amplifiers; db failure depends on amplifier

31 Gain Tilt OAS Impact of extra attenuation e.g. due to repair or ageing is Gain Tilt 3 db repair 1.64 db 4400 km transmission example Non-linearity? 32 nm Low OSNR Nominal conditions With all repairs and EOL ageing

32 Possible solutions OAS 1 Include Active Tilt Equalizer (ATEQ) in initial system design ATEQ ATEQ Includes sufficient adjustment to maintain gain flatness Drawback:Adds loss, so more repeaters 2 Add a fixed Tilt Equalizer (TEQ) or repeater during repair repair TEQ Reduces losses and simplifies the system Drawback:More complex repair

33 ITU-T G.977 Annex A Example 1 PM Line Parameter BOL EOL 1 Mean Q value (from a simple OSNR calculation) 1.1 Propagation impairments: chromatic dispersion, non-linear effects etc. 1.2 Gain flatness impairments 1.3 Non-optimal optical pre-emphasis impairment 1.4 Wavelength tolerance impairment 1.5 Mean PDL penalty 1.6 Mean PDG penalty 1.7 Mean PMD penalty 5 Segment Q value (computed from 3 and 4) 5.1 BER corresponding to segment Q without FEC 5.2 BER corresponding to segment Q with FEC 5.3 Effective segment Q value with FEC 6 Q limit [ Q at which post-fec objective is met ] 7 Repair margins 8 Segment margins 9 Unallocated supplier margin 10 Commissioning limits

34 ITU-T G.977 Annex A Example 2 Line Parameter BOL A BOL OSNR at full loading1 (XX dbm channel power) B EOL OSNR at full loading1 (XX dbm channel power) 1 Back-to-back Q at BOL OSNR 2 Propagation impairments 3 Other impairments 3.1 Non-optimal optical pre-emphasis impairment 3.2 Wavelength tolerance impairment 3.3 Mean penalty due to polarization-dependent effects EOL 5 BOL segment Q 6 Repair and Ageing Impairments 6.1 Cable repair and ageing 6.2 TTE ageing 7 EOL segment Q 8 FEC limit [ Q at which post-fec objective is met ] 9.1 Customer Segment EOL margin 9.2 Extra margin 10 Commissioning limit

35 Example 1 compared with Example 2 Example 1 Determine B-B OSNR 1 Q (from simple OSNR) 1.2 Gain Flatness Impairment 4 Specified TTE Q value (back-to-back) 5 Segment Q value 5.1 BER without FEC 5.2 BER with FEC 5.3 Effective segment Q value 7 Repair margins Component and fibre-ageing penalty Pump(s) failure penalty Non-optimal decision threshold Example 2 A OSNR (at full loading) 1 Back-back Q at BOL OSNR includes modem (RX) effects 3.6 Unspecified Impairment 5 BOL Segment Q 6 Repair and ageing impairments 6.1 Cable repair and ageing 6.2 TTE Ageing

36 Power Budget Tables (Non-Coherent TTE) Propagation Penalties Propagation Penalties TTE Penalty Q 2 [db] Q 2 [db] Q 2 [db] Repairs/Aging/Extra Margin FEC Limit FEC Limit FEC Limit OSNR [db/0.1nm] OSNR [db/0.1nm] OSNR [db/0.1nm] SLTE Wet- Plant SLTE TTE/Modem implementa6on penalty very low Mean Q 2 = B o OSNR / B e Linear rela6onship Propaga6on penal6es due to the wet- plant are subtracted Modem implementa6on penalty subtracted 1/ Q seg 2 = 1/ Q prop 2 + 1/ Q TTE 2 Repairs and aging done on OSNR

37 Power Budget Tables (Coherent TTE) Q 2 [db] Delivered OSNR Q 2 [db] Propagation Penalties Q 2 [db] Propagation Penalties Repairs/Aging/Extra Margin FEC Limit FEC Limit FEC Limit OSNR [db/0.1nm] OSNR [db/0.1nm] OSNR [db/0.1nm] SLTE Wet- Plant SLTE TTE/Modem implementa6on penalty maeers Actual Q: Q 2 = EC/ B e / B o OSNR + 1/ SNR TTE Propaga6on penal6es due to the wet- plant are subtracted Repairs and aging margin calcula6ons performed in OSNR

38 Power Budget Tables (Coherent) higher modulation Q 2 [db] Example 2 Q 2 [db] EOL Margin (OSNR budgeting) Example 2 EOL Margin (Q budgeting) Example 2 FEC Limit Example 2 FEC Limit OSNR [db/0.1nm] OSNR [db/0.1nm]

39 How it feeds into Acceptance Presenter: Company:

40 1. Q measured by FEC correction rate Acceptance Budget Verification Compare with commissioning limit (Line 10) Should also consider effect of Q variations BOL Margin = Q Line (average) Time variations (5 sigma) Q Limit OR = Q Line (worst measured) Q Limit The period over which average / worst case is measured is important Long periods remove the effect of short-term variations 2. Error performance after FEC

41 In practice require much better than G.828 often 0 errors in 7 days G.8201 for OTN (up to 100G) System Acceptance : G.828 and G.8201 G.828 for SDH, where performance is assessed by measurement of: BBER : Background Block Error Ratio SESR : Severely Errored Second Ratio (>15% errored blocks in 1 sec.) OAS Depends on link distance: e.g. 5,000 km link at 100G BBER <3.50E-4 BER < 3.75E-12 SESR <4.38E-7 Limit Q usually specified at 1E-13 to 1E-15

42 Refinements Presenter: Company:

43 1. Extra losses e.g. Equalizers, Branching Units Refinements 2. Correct Assumptions that are not quite true typical correction db Amplifier gain = Cable loss All sections the same All wavelengths the same System is symmetrical 3. Special cases ROPA Distributed Raman gain Terminal upgrades

44 Extra Losses Equalizer Units Branching Units OADM filters Example: 1 Equaliser every 10 sections Equaliser loss is 5 db Section loss was 20 db not including equalizer units Increase normal section length Decrease length of section with equaliser Section loss is now 20.5 db including equalizer units Adds a penalty of 0.5 db to OSNR

45 Repeater positions are not quite regular Branching unit is not in the optimum position Non-identical sections Examples Mix of submarine and terrestrial different NF and spacing

46 Sections are not identical Calculation OSNR(1) 1 2 OSNR(2) SLTE 1/OSNR = 1/OSNR(1) + 1/OSNR(2) + 1/Q 2 = 1/Q 2 Line + 1/Q2 SLTE Adds complexity, but can be important Example: 10 sections at 20.5 db 9 sections at 20.0 db + 1 section at 25.0 ~0.4 db difference

47 Wavelengths are not identical Noise figure, loss, non-linearity... generally vary with wavelength Channel power needs to be adjusted to suit Noise Figure Average Worst case Average Need to ensure worst case wavelength has adequate margin Must consider non-linear effects

48 Amplifier Gain not equal to Section Loss Pi Ni G G Pi G Ni Po = G (Pi + Ni) G = Po / (Pi + Ni) not Po / Pi H = Pi / (Pi + Ni) actual gain / ideal gain close to 1 After N amplifiers Signal = Po H N Noise = Ni G (1 + H + H H N ) Penalty = H N / (1-H N ) / (1-H) A small effect (<0.5 db in general)

49 Not usually an issue, but worth considering System symmetry Example: one section is long Solution : increase TX power T R Works in one direction Not in the other direction cannot increase amplifier R power T Did the budget consider the worst case?

50 Most often used in unrepeatered systems Systems with ROPA or DRA May use Remote Optically Pumped Amplifiers (ROPA) P TX RX ROPA Distributed Raman Amplification (DRA) TX P RX

51 Budgets involving ROPAs and DRA Raman / ROPA can be treated as an optical amplifier with a defined gain and noise figure Note, however, that a loss in the pump section reduces both the Raman / ROPA gain and adds loss to the signal path Affects repair margins more than with other amplifier types P 1 2 RX Impact is position-dependent Loss at (1) less significant than at (2)

52 Terminal Upgrades Often necessary to retain some old wavelengths Important not to take too much power from old WLs Old WLs New WLs Budget must consider both old and new WLs, Idlers/Loaders Getting budget parameters not always easy Sometimes possible to obtain OSNR measurements Often includes a trial/demonstration allows the budget to be tested May need to demonstrate inter-working of ASK and PSK

53 Other topics / discussions Presenter: Company:

54 Typical characteristics of different systems Can help to spot anomalies Other topics Open systems What is different? How is acceptance handled?

55 Short systems Repeater separation is large Gain flatness easy Larger repair allowances Usually have some extra margin Characteristics of different systems Length (km) Repair margin Extra margin 1, , , , Long systems Repeater separation is smaller Gain flatness hard Smaller repair allowances Usually have little extra margin

56 Open systems Why TS Separation of SLTE from Cable System offers: Transparency of Cable System performance Separates cable specification and performance-monitoring from SLTE Integration of terrestrial and subsea network infrastructure under one management system Allows purchasers to integrate preferred terrestrial solutions where possible for most who run networks, subsea is a small percentage of total capacity Flexibility in the operations model, protecting Purchasers from supply chain uncertainty and/or disparate technology cycles

57 Open systems how they could look SLTE Clear Demark Tx Rx North Bound Open Cable Interface Tx Rx Share Alarms and Supervisory Data with SLTE Supervisory & Management Terminal Station Equipment Open Cable System.. Supervisory & Management Terminal Station Equipment North Bound Tx Rx Open Cable Interface Tx Rx Clear Demark SLTE Provides broadband access for 2 (or more) SLTE Incorporates Cable Supervisory and PFE functionality Choice of SLTE

58 Specification and Acceptance Performance and Acceptance defined on line system characteristics, Most notably OSNR, Power, and Power Tilt Difficult set of measurements Potential for standardization Open systems Name Segment Landing Sites Length Quantity of C hannels at Full Loading 1. System Specification (Open C able System) 1.1 Slope of Tilt [db/th z] 1.2 Gain Deviation from tilt [db] 1.3 Power per carrier [db] 1.4 Span Loss [db] at 1550nm 1.5 Span Length [km] 1.6 Equalized OSNR [db/0.1nm] across the Passband at full loading 1.7 Passband Start/Stop [THz] 1.8 Average DGD across the Passband [ps] 1.9 Mean PDL [db] Open Cable Performance Specification 1.10 Total accumulated C hromatic Dispersion [ps/nm] at 1550nm CLS1 xxx xxx Nominal Exam Exam Start- of- L

59 Thank you for listening Any Ques6ons?