IEEE-82.3 Plenary Session July 15-2, 212 San Diego, California USA Measurements of VCSEL Mode Delays & Implications for System Performance Dr. Abhijit Sengupta CommScope Labs 13 E. Lookout Drive, Richardson, TX 7582 +1-972-792-3198 asengupta@commscope.com
Motivation Pattern Dependent BER shoulders have been observed for transmission over MMF links with Multimode VCSELs BER shoulder related penalty divergence is caused by RIN and MPN In multimode VCSEL, MPN is related to spatial hole burning The question is: does spatial dependence of gain switching cause observable VCSEL modal delays (LMD) If LMD is pattern dependent, the deterministic jitter has to be accounted for in the penalty Since MPN in VCSEL is related to transverse modes, is there a spatial dependence of spectral width? How to account for the spatial dependence of spectral width and LMD in the traditional MPN model? Is it possible to design VCSEL with low MPN and characterize MPN performance from spatial measurements? 2
Patter Dependent Penalty Tranceiver Effect BER 1.E+ 1.E-1 1.E-2 1.E-3 1.E-4 1.E-5 1.E-6 1.E-7 1.E-8 1.E-9 1.E-1 1.E-11 1.E-12 1.E-13 1.E-14 1.E-15 1.E-16 BW related Noise related -17-16 -15-14 -13-12 -11-1 -9-8 -7-6 -5-4 -3-2 RxPower (dbm) D_Stand_ F14_PRBS31 Stand_F14_PRBS31 D_Stand_F14_PRBS23 OFS_BIMMF_F6_PRBS23 OFS_BIMMF_F6_PRBS31 D-TMS1_F8_PRBS23 BIMMF_F8_PRBS23 BIMMF_F8_PRBS31 D-TMS1_F8_PRBS31 Both the Standard and BI OM4 fibers show the pattern dependent BER shoulder with 1G PRBS What is the source of pattern dependent penalty? Whatever the cause is, the errors occur rather infrequently (BER<1E-8) 3
RIN and MPN related Penalty Divergence Reference: A. Sengupta, Proc. IWCS, Vol. 57, p. 32 (28) MMF Link penalty diverges at different characteristic critical lengths for VCSELs with different RIN and MPN Simulations show that the point of penalty divergence can be pushed out to longer length by reducing the RIN and MPN 4
Measurement Methods OSA Spectral width λ (r) RIN (ω,r) PG VCSEL MMF SMF PD Sampling Oscilloscope LMD(r) Jitter(r) Pattern trigger 5
Radial Dependence of VCSEL Spectra r (um) 26 24 22 2 18 16 14 12 1 8 6 4 2-2 -4-6 -8-1 -12-14 -16-18 -2-22 -24-26 QSFP-28-1 QSFP-34-1 QSFP-34-2 QSFP-34-3 QSFP-34-4 SFP+123 SFP+C15 SFP+C16 SFP+J13 SFP+J14 SFP+AT17 SFP+AT18 846 847 848 849 85 851 852 853 center wavelength_prms (nm) Clearly, there is an observable radial dependence of the average spectral width r (um) 26 24 22 2 18 16 14 12 1 8 6 4 2-2 -4-6 -8-1 -12-14 -16-18 -2-22 -24-26 QSFP-28-1 QSFP-34-2 QSFP-34-4 SFP+C15 SFP+J13 SFP+AT17 QSFP-34-1 QSFP-34-3 SFP+123 SFP+C16 SFP+J14 SFP+AT18.5 1 1.5 2 prms width Spectral width may also vary from one bit to another but the bit level spectral width measurement is difficult 6
VCSEL Mode Delay or LMD measurements In absence of chromatic dispersion over a short length of fiber, we should see no temporal changes in the bits even though the spectral width has a radial dependence. If the spatial dependence of gain switching cause observable VCSEL modal delays (LMD), can we observe a radial dependence of the LMD? Is LMD pattern dependent? 7
Pattern Dependent Spatio-Temporal Effect Normalized Amplitude (arb. unit) power (linear).25.2.15.1.5 Error @ bit Fiber length =1m 2ps +25 um +15 um um -25 um 1 2 3 4 5 6 7 8 9 time / ps Bits 1 1 1 8
MMF-SMF Radial Scan (Short Length).25 Higher order modes are delayed relative to low order VCSEL mode Fiber length =1m um radial offset 1um radial offset Normalized Amplitude Amplitude (arb. unit) (arb. unit).2.15.1.5 2ps 2ps 1 2 3 4 5 6 time (ps) 9
MMF-MMF Radial Scan (Short Length) Bits 1 1 Normalized Amplitude (arb. unit).1.9.8.7.6.5.4.3.2 Fiber length =1m -17.5um -2.5um +7.5um +17.5um.1 1 15 11 115 12 125 13 135 14 145 15 time (ps) No significant changes are noticed due to limited mode filtering by the MMF 1
Angle and Radial Offset Dependence Fiber length =1m.35.35.3 angle high angle low.3 um 1um Normalized Amplitude Amplitude (arb. unit) (arb. unit).25.2.15.1 Normalized Amplitude Amplitude (arb. unit) (arb. unit).25.2.15.1.5.5 2 4 6 8 1 12 14 16 18 2 5 1 15 2 25 time (ps) time (ps) The temporal delay or phase lag is similar for higher radial offsets and higher angular offsets 11
Pattern A (Angle Dependence) Bits 1 1 1 1 1 1 1 Normalized Amplitude (arb. unit) Amplitude (arb. unit).3.25.2.15.1.5 Fiber length =1m angle low angle high 1ps 1ps 2ps 1ps 1 2 3 4 5 6 7 8 9 1 time (ps) The relative delay or phase lag of the VCSEL modes depends on the bit sequence 12
Pattern A (Radial Offset Dependence) Bits 1 1 1 1 1 1 1.3 Fiber length =1m um 1um Normalized Amplitude (arb. unit) Amplitude (arb. unit).25.2.15.1.5 1ps 2ps 1ps 1ps 1 2 3 4 5 6 7 8 9 1 time (ps) The relative delay of the VCSEL modes show similar dependence on radial and angular offsets. There is a finite and noticeable delay between when the VCSEL modes are turned on and off 13
Pattern B (Angle Dependence) Bits 1 1 1 1 1 1.35 Fiber length =1m angle low angle high Normalized Amplitude (arb. unit) Amplitude (arb. unit).3.25.2.15.1 1ps 4ps 5ps.5 Noise at bit 1 2 3 4 5 6 7 8 9 1 time (ps) 14
Pattern B (Radial Offset Dependence) Bits 1 1 1 1 1 1.35.3 Fiber length =1m um 1um Normalized Amplitude (arb. unit) Amplitude (arb. unit).25.2.15.1 1ps 4ps 5ps.5 Noise at bit 1 2 3 4 5 6 7 8 9 1 time (ps) 15
Summary The relative delay of the VCSEL modes show similar dependence to radial offsets and angular offsets. The relative delay of the VCSEL modes depends on the bit sequence The occurrence of the observable LMD is limited to a few bits in a long PRBS, and therefore the BER shoulder related to the effect remains below BER~1E-8 There is a finite and noticeable delay in VCSEL modes on and off and there is a spatial dependence The relative delay can be up to 2-4ps and impact the signal irrespective of the fiber length MPN has a square law dependence on the source spectral width, and radial dependence of spectral width may translate into increased modal noise in the different mode groups of the MMF 16
Further Measurements PG VCSEL MMF RIN L= (ω) M((ω) = RIN L= (ω) RIN L=x (ω) PG VCSEL MMF RIN L=x (ω) RIN L= (ω) is measured at with a very short length of MMF, such that the anticorrelation of the noise in individual competing VCSEL modes are maintained RIN measurement is repeated with sufficiently long length (X) of MMF, so the anticorrelation of the noise in the modes are lost due to the temporal separation caused by the DMD M can be used as a measure of MPN 17
Concluding remarks Recent research shows spatial hole burning effect increases the MPN among modes that overlaps spatially Simulations based on spatial hole burning also confirm that there are significant delays in the gain switching of different transverse modes and the delays depend on electrode configurations May be it would be possible to design the electrodes and cavity to reduce the overlap of the transverse modes and possibly reduce the MPN In the frequency domain, radial dependence of RIN can be applied to characterize mode power fluctuations in individual modes Radial dependent measurements of RIN, LMD, spectral width may be used to characterize VCSEL for MPN performances Possibly, the radial dependence can be incorporated in the existing model of MPN or k-factor for better prediction of link performance. 18
References A. Sengupta, Proc. IWCS, Vol. 57, p. 32 (28) J. Y. Law, G. H. M. van Tartwijk and G. P. Agrawal, Quantum Semiclass. Opt. Vol. 9, 737 (1997) Dellunde et al., IEEE J. Quant. Elec. Vol. 33, p. 1197 (1997) J. Y. Law and G. P. Agrawal, IEEE Phot. Tech. Lett. Vol. 9, p.437 (1997); A. Valle, IEEE J. Quant. Elec. Vol. 34, p1924 (1998) A.Valle and L. Pesquera, IEEE Phot. Tech. Lett. Vol. 13, p. 272 (21); Tayahi et al., Proc. SPIE Vol. 5737, p.129 (25); 19