MMF Channel Characteristics

Transcription

1 MMF Channel Characteristics J. Ewen, E. Borisch JDS Uniphase P. Pepeljugoski, A. Risteski IBM 1

2 Motivation / Outline Fiber impulse response Critical importance of launch conditions, connectors, etc. Variability of a single fiber Initiate discussion of channel model, worst-case fiber, etc. Fiber measurements Intended as illustration, not exhaustive study Typical performance metrics: OFL bandwidth, DMD Impulse response vs. launch

3 Test Setup: Fiber Impulse Response Pat. Gen. trigger Offset patch cord SMF center launch ROFL launch 131nm EML SMF 6.5 µm MMF (7m 11m) Oscilloscope 6.5µm, 131nm O/E EML source Measure step response Differentiate for impulse response FFT for frequency response Measurement bandwidth ~9.5GHz Step Response x 1 9 Impulse Response Frequency Response Time (ns) f 3dB = 9.7 GHz Frequency (Hz) Calibration m fiber 3

4 Fiber Characteristics 6.5µm MMF.km spool cut back: 11m, 55m, 75m, 139m, 81m, 7m individual lengths Off-the-shelf fiber No particular characteristics requested Overfilled Bandwidth Measurement CPR ~ 33dB BW corrected for measurement system bandwidth Multiple fiber lengths measured & compared for consistency BW ~.GHz-km Step Response Frequency Response Impulse Response x Time (ns) f 3dB = 3.7 GHz Raw Response Calibrated 55m fiber Frequency (Hz)

5 Fiber Response vs. Launch Condition (1/) Simulated Eye Diagram Measured Fiber Response Fiber Impulse Response (arb) Mode-Conditioning Patch Cord Time (ns) Fiber Impulse Response (arb) Center Launch 1.5dB ROFL Time (ns) Fiber Impulse Response (arb) Time (ns) Offset Patch Cord Center Launch 1.5dB ROFL 139m, 6.5µm MMF 5

6 Fiber Response vs. Launch Condition (/) Fiber Impulse Response (arb) dB ROFL Center Launch MCP # Time (ns) 139m, 6.5µm fiber Three launch conditions Mode Conditioning Patch Cord Center launch 1.5dB ROFL Multiple peaks in response Significant differential delay among mode groups t ~ 15ps Poor eye diagrams 6

7 Mode Conditioning Patch Cord Fiber Impulse Response (arb) m, MCP#1 55m, MCP# 55m, MCP# Time (ns) 55m, 6.5µm fiber Good measurement stability over time (blue, green) Dramatic difference with different patch cords 7

8 Test Setup: DMD Measurement Laser (131 nm) Driver Pigtail Short 9 µm SMF (-3m).5-1µm precision stages X-Y-Z Short MMF (-3m).5-1µm precision stages X-Y-Z Test MMF (L) SM tolerance connectors Motion controllers Short MMF (-3m) Photo Diode (1 Ghz) GPIB Coaxial cable Signal Analyzer fiber core diameter laser beam 131nm, gain-switched FP source ~5ps pulse width (FWQM) -8 to +8 µm offset range 1 µm step scans scanning step fiber center 8

9 3 DMD Measurements (1/3) DMD= 1. ps/m 5 Offset ( µm) Time (ps) 55m, 6.5µm MMF 9

10 DMD Measurements (/3) 6.5µm MMF, 55m, DMD=1.ps/m µm offset.5µm offset Impulse Response (arb) Time (ps) Results consistent with previous MCP measurements 1

11 DMD Measurements (3/3) 6.5µm MMF, DMD=.6ps/m 6.5µm MMF, DMD=.36ps/m Impulse Response (arb) µm offset µm offset µm offset Impulse Response (arb) µm offset µm offset µm offset Time (ps) Time (ps) Starting to look at other fibers Preliminary results suggest previous measurements are not atypical 11

12 EQ Simulation Example Ideal EQ (i.e. textbook simulation) Infinite length, T/-spaced linear equalizer 8-tap, T/ fractional-spaced linear equalizer Optical SNR penalty relative to 1 1 BER Measured impulse response 55m, 6.5µm fiber 7.5GHz, th -order BT receiver 35ps rise/fall transmitter 1

13 Summary Wide variation in impulse response from a single fiber Wide variation within tolerance specification of MCP Connectors mode mixing can have significant effect on BW, without modal noise Significant precursor or postcursor ISI depending on launch condition Typical fiber performance metrics offer little insight OFL bandwidth & DMD correlation to impulse response? 13

14 Observations Bandwidth is not a useful performance metric Details of the impulse response determine actual performance DMD is not a useful performance metric Scanning across the entire core of the fiber not representative of expected launch conditions Controlling the channel response seems impractical Almost any mode power distribution (MPD) seems possible with reasonable tolerances on launch conditions Channel model requirements? Assume worst case MPD will depend on proposed solutions Identify simple channel metrics, e.g. restricted DMD? Time extent of impulse response for defined launch (with tolerance) 1

15 Backup 15

16 Connectors: Effect on Bandwidth 75m 55m 75m vs. 55m (same fiber pull) Gives significantly worse output eye Total insertion loss ~1.5dB Connector quality poor 16