Linearly polarized vector modes: enabling MIMO-free mode-division multiplexing

Transcription

1 Linerl polrized vector modes: enling MIMO-free mode-division multipleing Liin Wng, Rez Mirzei Nejd, Alessndro Corsi, Jichun Lin, Younès Messddeq, Leslie A. Rusch, nd Sophie LRochelle OSA Optics Epress, (Volume 25, Issue 10) (2017) OSA. Personl use of this mteril is permitted. Permission from OSA must e otined for ll other uses, in n current or future medi, including reprinting/repulishing this mteril for dvertising or promotionl purposes, creting new collective works, for resle or redistriution to servers or lists, or reuse of n coprighted component of this work in other works.

2 Linerl polrized vector modes: enling MIMO-free mode-division multipleing LIXIAN WANG, REZA MIRZAEI NEJAD, ALESSANDRO CORSI, JIACHUAN LIN, YOUNÈS MESSADDEQ, LESLIE RUSCH, AND SOPHIE LAROCHELLE * Centre for Optics, Photonics nd Lsers (COPL), Deprtment of Electricl nd Computer Engineering, Université Lvl, Cnd * sophie.lrochelle@gel.ulvl.c Astrct: We eperimentll investigte mode-division multipleing in n ellipticl ring core fier (ERCF) tht supports linerl polrized vector modes (LPV). Chrcteriztion show tht the ERCF ehiits good polriztion mintining properties over eight LPV modes with effective inde difference lrger thn The ERCF further displs stle mode power nd polriztion etinction rtio when sujected to eternl perturtions. Crosstlk etween the LPV modes, fter propgting through 0.9 km ERCF, is elow 14 db. B using si LPV modes s independent dt chnnels, we chieved the trnsmission of 32 Gud QPSK over 0.9 km ERCF without n multiple-input-multiple-output (MIMO) or polriztion-division multipleing (PDM) signl processing Opticl Societ of Americ OCIS codes: ( ) Fier optics nd opticl communictions; ( ) Fiers, polriztion-mintining; ( ) Multipleing. References nd Links 1. D. J. Richrdson, J. M. Fini, nd L. E. Nelson, Spce-division multipleing in opticl fires, Nt. Photonics 7(5), (2013). 2. P. J. Winzer, Mking sptil multipleing relit, Nt. Photonics 8(5), (2014). 3. N. Bozinovic, S. Golowich, P. Kristensen, nd S. Rmchndrn, Control of oritl ngulr momentum of light with opticl fiers, Opt. Lett. 37(13), (2012). 4. C. Brunet, B. Ung, L. Wng, Y. Messddeq, S. LRochelle, nd L. A. Rusch, Design of fmil of ring-core fiers for OAM trnsmission studies, Opt. Epress 23(8), (2015). 5. N. Bozinovic, Y. Yue, Y. Ren, M. Tur, P. Kristensen, H. Hung, A. E. Willner, nd S. Rmchndrn, Teritscle oritl ngulr momentum mode division multipleing in fiers, Science 340(6140), (2013). 6. R. M. Nejd, K. Allhverdn, P. Vit, S. Amirlizdeh, C. Brunet, Y. Messddeq, S. LRochelle, nd L. A. Rusch, Mode division multipleing using oritl ngulr momentum modes over 1.4-km ring core fier, J. Lightwve Technol. 34(18), (2016). 7. C. Simonneu, A. D mto, P. Jin, G. Lroille, J.-F. Morizur, nd G. Chrlet, 450G/s trnsmission over 4.4 km of multimode OM2 fier with direct detection using mode group multipleing, in Opticl Fier Communiction Conference (OSA, 2016), pper Tu2J G. Milione, E. Ip, M.-J. Li, J. Stone, G. Peng, nd T. Wng, Mode crosstlk mtri mesurement of 1 km ellipticl core few-mode opticl fier, Opt. Lett. 41(12), (2016). 9. G. Milione, E. Ip, M.-J. Li, J. Stone, G. Peng, nd T. Wng, Sptil mode nlsis of n ellipticl-core, fewmode, opticl fier for MIMO-less spce-division-multipleing, in Opticl Fier Communiction Conference (OSA, 2016), pper W1F E. Ip, G. Milione, M.-J. Li, N. Cvijetic, K. Knonkis, J. Stone, G. Peng, X. Prieto, C. Montero, V. Moreno, nd J. Liñres, SDM trnsmission of rel-time 10GE trffic using commercil SFP + trnsceivers over 0.5km ellipticl-core few-mode fier, Opt. Epress 23(13), (2015). 11. G. Milione, P. N. Ji, E. Ip, M.-J. Li, J. Stone, nd G. Peng, Rel-time Bi-directionl 10GE Trnsmission using MIMO-less Spce-division-multipleing with Sptil Modes, in Opticl Fier Communiction Conference (OSA, 2016), pper W1F L. Wng nd S. LRochelle, Design of eight-mode polriztion-mintining few-mode fier for multiple-input multiple-output-free sptil division multipleing, Opt. Lett. 40(24), (2015). 13. H. Yn, S. Li, Z. Xie, X. Zheng, H. Zhng, nd B. Zhou, Design of PANDA ring-core fier with 10 polriztion-mintining modes, Photonics Res. 5(1), 1 (2017). 14. L. Wng, R. M. Nejd, A. Corsi, J. Lin, Y. Messddeq, L. A. Rusch, nd S. LRochelle, MIMO-Free trnsmission over si vector modes in polriztion mintining ellipticl ring core fier, in Opticl Fier Communiction Conference (2017), pper Tu2J.2.

3 15. Y. Jung, S. R. Hn, S. Kim, U. C. Pek, nd K. Oh, Verstile control of geometric irefringence in ellipticl hollow opticl fier, Opt. Lett. 31(18), (2006). 16. L. Wng, P. Vit, B. Ung, Y. Messddeq, L. A. Rusch, nd S. LRochelle, Chrcteriztion of OAM fiers using fier Brgg grtings, Opt. Epress 22(13), (2014). 17. V. Arrizón, U. Ruiz, R. Crrd, nd L. A. González, Pielted phse computer hologrms for the ccurte encoding of sclr comple fields, J. Opt. Soc. Am. A 24(11), (2007). 18. S. Kwkmi nd M. Iked, Trnsmission chrcteristics of two-mode opticl wveguide, IEEE J. Quntum Electron. 14(8), (1978). 19. P. Gregg, P. Kristensen, nd S. Rmchndrn, 13.4km OAM stte propgtion recirculting fier loop, Opt. Epress 24(17), (2016). 20. F. Ymn, N. Bi, B. Zhu, T. Wng, nd G. Li, Long distnce trnsmission in few-mode fiers, Opt. Epress 18(12), (2010). 1. Introduction As technique to implement sptil division multipleing, the ke dvntge of mode division multipleing (MDM) in few-mode fiers (FMFs) lies in its ilit to chieve ver high chnnel densit, propert due to the fct tht mode fields hve significnt sptil overlps [1]. However, FMFs re susceptile to mode coupling, especill mong the modes of the sme mode group. Multiple-input-multiple-output (MIMO) processing is therefore usull required to recover dt, which consequentl increses compleit of the receivers. In short-rech dt communiction pplictions, where power consumption nd cost re serious concerns, MIMO-free trnsmission is highl desired [2]. In order to reduce mode coupling nd simplif MIMO processing, novel fier designs hve een proposed to eploit modl sis sets other thn the trditionl linerl polrized (LP) mode groups of wekl guiding step inde fiers. For emple, n pproch tht hs recentl received much ttention is the use of ring core fiers (RCFs) with high refrctive inde contrst (~10 2 ) [3,4] to trnsmit oritl ngulr momentum (OAM) modes. In these fiers, the effective indices of OAM modes with opposite spin-orit hndedness (ssocited to the HE nd EH vector modes respectivel) re seprted >10 4 so s to suppress the mode coupling. Eperimentl demonstrtions of km length dt trnsmission with MDM over few OAM modes were reported in [5,6]. However the two polriztion sttes of OAM modes ( ± l order right-circulr polrized nd the l order left-circulr) remin two times degenerte. Chnnel chnnel opticl polriztion demultipleing ws therefore used in [5] while it ws ccomplished through digitl signl processing (DSP) in [6]. These results thus show tht, ecuse of the circulr smmetr of these fiers, OAM trnsmission over RCF cnnot e completel MIMO-free nd 2 2 MIMO is still needed. Another tpe of modl sis tht cn simplif or llevite MIMO processing is multipleing over mode groups, or over degenerte mode groups. An emple of the ltter is the four chnnel trnsmission chieved in [7] selective ecittion t the trnsmitter of onl one mode in ech of the degenerte mode groups nd susequent direct detection t the receiver of ll sptil modes of the si lowest order mode groups (LP 01 to LP 12 ) t the receiver to void power fding. In this sstem, mode demultipleing remins comple nd the full sptil dimension of ten ville chnnels (including polriztion) is not eploited. As for the former, n emple is multipleing over Hermit-Gussin sptil modes supported ellipticl core FMFs (EFMFs). EFMFs rek the circulr smmetr of wekl-guiding stepinde FMF so tht, over short propgtion distnce, coupling etween the sptil modes with different orienttions is voided seprting their effective indices [8,9]. Similr to the OAM modes in RCFs, the Hermit-Gussin modes in EFMFs lso consists of two stronglcoupled orthogonl polriztions. Although MIMO-free trnsmission hs een demonstrted over 500m EFMF, it ws t the cost of scrificing the polriztion chnnels [10,11]. Bsed on the ove discussion, sptill efficient nd complete MIMO-free MDM requires trnsmission over non-degenerte vector modes, i.e. the effective indices of ll modes must e well seprted, including ll sptil nd polriztion sttes, in order to chieve

4 sufficientl low inter-modl coupling. Recentl, we proposed novel ellipticl ring core fier (ERCF) design tht is le to stilize/seprte ll the higher-order vector modes [12] nd more recentl pnd fier design ws proposed for the sme purpose [13]. The vector modes of the ERCF re ll linerl polrized, which gretl fcilittes the multipleing nd demultipleing of the sptil chnnels. Becuse of their liner polriztion sttes, we nme these modes linerl-polrized vector modes (LPV modes). The field profiles show similrities to the LP modes of FMFs, ut the LPV modes re vectoril eigenmodes of the fier, not mode groups. In this pper, following our initil report in [14], we now present in detil the eperimentl results of recentl fricted ERCF design with complete description of our chrcteriztion techniques, mesurement results nd trnsmission eperiments. In Section 2, the prmeters of the fricted ERCF re introduced. The mode properties, including the modl effective indices, the mode field profiles, the mode stilit, the polriztion etinction rtio (PER) nd the impulse response re reported. Section 3 is devoted to demonstrtion of MIMO-free dt trnsmission over si LPV modes in 0.9 km ERCF t ud rtes reching the limit of our eperimentl cpilit (32 Gud QPSK). Section 4 summrizes the results nd presents the conclusions. Fig. 1. ) Fier geometr nd prmeter definition, ) design prmeters, c) photogrph of the fricted ERCF. 2. Modes in Ellipticl Ring Core Fier Through numericl simultions [12], the refrctive inde profile of the ERCF, shown in Figs. 1() nd 1(), ws designed to support ten linerl polrized vector modes (LPV modes), nmel: LPV 01, LPV 01, LPV 11, LPV 11, LPV 11, LPV 11, LPV 21, LPV 21, LPV 21 nd LPV 21. The effective inde seprtions (Δn eff ) etween the eight higher-order vector modes is lrger thn , which is the tpicl vlue of irefringence in conventionl polriztion mintining fiers. For this design, the two orthogonl polriztion sttes of the fundmentl mode, LPV 01 nd LPV 01, remin degenerte with clculted Δn eff ~ The effective res re: 48.6 μm 2 (LPV, 01), 35.9 μm 2 (LPV, 11), 46.8 μm 2 (LPV, 11), 47.3 μm 2 (LPV, 21) nd 46.9 μm 2 (LPV, 21). We use technique similr to tht descried in [15] to fricte the ERCF strting with clindricl preform, fricted with modified chemicl vpor deposition (MCVD), whose core hs circulr ring-shped profile. Two slices of the preform re cut long its length on two opposite sides resulting in two prllel surfces long its longitudinl is. Afterwrds the preform is heted so tht the flt surfces dispper due to the surfce tension nd the flow of mteril nd, consequentl, the ring core ecomes ellipticl during the process. The preform, fter the heting nd rounding process, hs dimeter of 25.6 mm. A length of pproimtel 30 mm of this preform, now with clindricl cldding, ws drw resulting in 1 km long fier from which 100 m ws reserved for tests nd chrcteriztion. Figure 1(c) shows the picture of the fier end fce (the cldding dimeter is 125 μm). We chrcterized the effective indices of the fricted ERCF using the FBG writing technique descried in [16]. The uniform FBG ws written in ~40 cm long ERCF smple loded with deuterium. Ech reflection pek on the FBG spectrum, corresponding to i th

5 guided fier mode, hs Brgg wvelength, B eff i λ B, relted to the modl effective inde, i n eff, i i the usul eqution λ = 2n Λ, where Λ is the grting period. The FBG reflected spectrl response mesurement, shown in Fig. 2(), displs totl of nine peks corresponding to self-coupling of the guided modes supported the ERCF. The spectrum lso shows crosscoupling peks etween i th nd j th, modes t intermedite wvelengths λ i j = ( n i + n j ) Λ. B eff eff Fig. 2. ) Mesured mode intensit profiles t the output of 5 m long ERCF (top row) compred to numericll clculted ones (oth mplitude nd phse). ) Reflection spectrum of uniform FBG written in the ERCF. Using the spectrum, the modl effective indices re clculted. B mesuring the spectrl spcing of the reflected peks, the vlues of effective inde seprtion etween guided fier modes, Δn eff, re esil deduced nd the results re listed in Tle 1. The reflected peks of LPV 01 nd LPV 01 (on the most right-hnd side of Fig. 2) re superimposed since these two vector modes re lmost degenerte nd their Δn eff is eond the resolution of this mesurement technique. Δn eff mong the other eight higher order vector modes re ll > 10 4, in greement with the trget specifictions for this design. Figure 2() lso gives the intensit mode profiles tht were eperimentll mesured (top row) nd numericll clculted (ottom row) fter propgting in 5 m long ERCF. The modl intensit profiles re otined eciting one mode t time in the ERCF using the free spce mode ecittion setup descried in detil in the following section. LPV 21 nd LPV 21 modes showed high loss nd cler imges of their mode profiles could not e otined nd neither could the e used for dt trnsmission over km-long distnce. However, the reflection peks elonging to LPV 21 nd LPV 21 modes cn still e detected in the FBG mesurement eperiment due to the short fier smple (~40 cm) nd the ver high dnmic rnge of the frequenc-swept opticl interference technique.

6 LVP01 01 Tle 1. Comprison of design nd mesured Δn eff. LVP LVP11 LVP11 LVP11 LVP11 LVP21 LVP21 LVP21 LVP21 Δn eff design Δn eff mesured 1.31E E E E E E E E E E E E E E E E E-4 3. Mode Stilit in Ellipticl Ring Core Fier The effective inde seprtion (Δn eff >10 4 ) ws designed to reduce coupling etween orthogonl polriztions of the vector modes during propgtion. For the proposed ERCF design, this gol ws chieved for the eight higher order vector modes nd, in this section, we eperimentll investigte the stilit of the si vectors modes tht cn e used for dt trnsmission. The results re compred to the mesurement of the fundmentl mode tht does not ehiit polriztion mintining propert. In prgrph 3.2, we descrie how we tested the stilit of the mode field profiles nd the polriztion sttes of the si higher order vector modes under eternl perturtions. 3.1 Mode ecittion nd detection In the eperiments, two different techniques re used for eciting the vector modes in ERCF. The first one (Fig. 3()) genertes n opticl comple field tht mtches the desired fier mode progrming phse onl sptil light modultor (SLM) [17] nd then coupling it into the fier. Figure 3() gives the tpicl hologrm ptterns corresponding to the fier modes. Since the free-spce comple field cn hve the sme mplitude nd phse distriutions s those of the fier modes, it cn e used to ecite n vector mode in ERCF with reltivel high mode purit. The mesured mode profiles in Fig. 2() (top row) re mesured using this technique. However, mode multipleing of mn vector modes with this first technique, possil requiring severl SLMs, would e oth complicted nd epensive. In comprison to the SLM encoding of the comple mode field, the second technique, shown in Figs. 3(c) nd 3(d), is simpler to implement. In this second technique, conventionl opticl phse pltes re used, insted of the epensive SLM, producing n incident lser em in free spce tht hs similr phse distriutions s the fier modes lthough the mplitude distriution is different (Fig. 2() gives the clculted modl phse profiles). The required phse ptterns for the LPV 11 nd LPV 11 modes re ectl the sme s those used for circulr-core step-inde FMFs. However, the phse plte needed to ecite the LPV 21 modes in ERCF is different from tht used for LP 21 modes in conventionl step-inde FMFs s it requires uneven re of the 0 nd π phse shift regions. Through eperiments, it ws estimted tht the optiml ngles of these two sections were 76 nd 104 in order to otin the est mode ecittion purit. Since we did not hve such specil phse plte, we emulted this second technique for the LP 21 mode using phse plte progrmmed on SLM. The gol of using such n emultor ws to evlute whether sufficient mode purit could e chieved in prctice with this simpler coupling technique tht presents greter eperimentl convenience when doing multipleing for trnsmission eperiments. We found tht the mode ecittion purit of the second technique is s good s tht of the first one ecept for the ecittion of the LPV 01 nd LPV 21 modes, which lws ccompn ech other, resulting in lower purit. However, the LPV 01 will not e used in the dt trnsmission eperiment discussed in Section 4 nd, consequentl, the unwnted ecittion of the LPV 01 mode cn e considered s dditionl coupling loss for the LPV 21 mode, ut it will not otherwise impct sstem performnce. In oth techniques, Figs. 3() nd 3(c), hlf-wve plte (HWP) is inserted efore the fier to lign the incident polriztion long the proper ERCF is while qurter-wve plte (QWP) is used to compenste for n polriztion ellipticit introduced mirror nd lenses.

7 Once djusted, the setting of the HWP nd QWP re kept constnt for ll or polrized modes. Fig. 3. Eperimentl techniques used for ecittion of linerl polrized vector modes in ERCF showing () setup with phse onl sptil light modultor (SLM) (1st coupling technique), () the hologrms progrmmed on the SLM in ) (see [17] for detils), c) setup with regulr opticl phse plte (2nd coupling technique) nd d) schemtic of the used phse pltes in c). The collimtor is F240APC-1550 (Thorls) nd the SLM is PLUTO- TELCO (HOLOEYE). To filter the LPV modes t the ERCF output we use the setup shown in Fig. 4 tht minl consists of HWP, QWP, polrizer nd SLM. Depending on its setting, this receiver detects one given vector mode t time. The polrizer is ligned to mimize the response of the polriztion sensitive SLM. Adjustment of the HWP then llows one polriztion stte, or polriztion, to e selected nd pss through the polrizer. The QWP is djusted to compenste for polriztion distortion nd improve cross tlk s discussed ove. Menwhile, the phse pttern displed on the SLM converts the desired mode field to Gussin-like em (however with n ellipticl intensit profile) which cn e coupled into single mode fier (SMF28) nd finll detected high-speed photodetector (PD). Fig. 4. Schemtic digrm of the setup for the filtering nd detection of the vector modes coming out from the ERCF.

8 3.2 Test of modl stilit in 5m ERCF In order to emine stilit of the power injected in the ERCF modes under eternl perturtions, we ecited the fier vector modes, one mode t time, in 5 m long ERCF using the first technique descried in Section 3.1. As shown in Fig. 5, drop-in polriztion controller device (PC) is pplied on the ERCF to induce pressure nd twist. A polrizer is plced t the output of the fier fter which oth the mode profiles nd the power fluctutions cn e monitored CCD cmer nd power meter respectivel. Firstl, we fied the position of the polrizer nd induced perturtion rotting the PC while recording the mode profiles nd mesuring the power fluctution of ech mode. If the vector mode is stle, its mode profile should e constnt to indicte tht there is no coupling to other sptil modes (mode profile mesurement on the CCD cmer) nd its polriztion stte should e fied (mesured detecting power fluctutions fter the polrizer). As shown,,, in Fig. 6() to (d), LPV 11, LPV 11, nd LPV21 modes hve such stle polriztion sttes with ver smll power fluctutions ( P) of ± 0.04 db, ± 0.01 db nd ± 0.05 db respectivel when mesured on photodetector. Moreover, the mode profiles cptured the CCD cmer show rel visile chnges, which qulittivel illustrtes tht the coupling mong the different sptil different modes is smll. Videos of the mode profiles recorded when the PC is rotted from 0 to 180 cn e found in the complementr files. Similr chrcteristics were oserved for oth polriztion modes, or, so tht onl one mesurement is displed in Fig. 6 for ech mode. As discussed in Section 2, the two polriztions of the fundmentl LPV 01 mode re nerl degenerte nd, therefore, these modes couple to ech other under eternl perturtions resulting in the lrge power fluctutions s oserved in Fig. 6(). Fig. 5. Schemtic digrm for the test of mode stilit in 5 m ERCF. Secondl, we fied the PC nd rotted the polrizer from 0 to 180 nd monitor the opticl power fter the polrizer. We define the polriztion etinction rtio (PER) s the rtio etween the mimum nd minimum opticl powers. The lrger the PER, the closer the mode is to pure liner polriztion. The si higher-order vector modes coming out from the ERCF,,,, LPV 11, LPV 11, nd LPV 21, hve high PERs of 16.5 db, 22.2 db nd 20.5 db respectivel, s indicted in Fig. 6 nd Tle 2. Since the polriztion stte of the LPV 01 mode is unstle, its PER ws not mesured.

9 Fig. 6. Power fluctutions ( P) nd emples of mesured mode intensit profiles in 5 m ERCF under eternl perturtions for one polriztion of ech mode () LPV 01 (see Visuliztion 1), () LPV 11 (see Visuliztion 2), (c) LPV 11 (see Visuliztion 3) nd (d) LPV 21 (see Visuliztion 4). Polriztion etinction rtios for ech mode re lso indicted on the grphs. The polriztion etinction rtio (PER>16.5 db) nd the power fluctutions (DP<0.05 db) confirm tht the higher order vector modes in ERCF cn e considered s linerl polrized nd re quite stle under eternl perturtions. 3.3 Impulse response mesurement of 0.9 km ERCF As the fier length increses, coupling etween djcent vector modes, which is inversel proportionl to their effective inde difference [18], cn ccumulte nd limit the miml distnce of MIMO-free dt trnsmission. In this section, we chrcterize the chnnel impulse response using the time-of-flight technique [19] to estimte modl crosstlk during propgtion in the fier. The setup, shown in Fig. 7, consists of pssivel mode-locked fier lser, with 20 MHz repetition rte nd pulse width of ~1ps, injected in 0.9 km long ERCF fier using the second technique, descried in Fig. 3(c) nd (d), for selective ecittion of vector modes. For eperimentl convenience, the phse pltes used for ll modes were progrmmed on SLM (SLM1) tht did not perform n other em shping function. The trnsmitted opticl pulses were detected using the receiver in Fig. 4 nd mesured n electricl smpling oscilloscope. Fig. 7. Eperimentl setup for the time-of-flight mesurement. We mesured the crosstlk inside the mode pirs {LPV 11 nd LPV 11}, {LPV 11 nd LPV 11}, nd {LPV 21 nd LPV 21}. For these three mode pirs, Δn eff is of the order of 10 4, which could results in nontrivil crosstlk fter propgtion over few kilometers in the fier.

10 Fig. 8. Chnnel impulse response mesurements otined when sending nd receiving the sme mode: ) LPV 11, ) LPV 11, nd c) LPV 21. Impulse response mesurement when receiving the orthogonl polriztion: d) sending LPV 11, receiving LPV 11, e) sending LPV 11, receiving LPV 11, nd f) sending LPV 21, receiving LPV 21. In comprison, Δn eff etween the mode pirs re of the order of 10 3, which should result in much smller crosstlk tht cn e neglected [20] so tht it is not mesured here. Figure 8(), () nd (c) re the mesured output pulses when sending nd receiving the sme vector mode. The fluctutions on the right-hnd side re due to the impulse response of the PD. We chose to send the vector mode which trvels slower in the mode pir so tht the impulse response of the PD will hve less impct on the crosstlk mesurement. It ws oserved tht, in the three mode pirs, the modes with the slower group velocities re LPV 11, LPV 11 nd LPV 21 respectivel, in good greement with the numericl clcultion of the fier design [12]. Figures 8(d), 8(e), nd 8(f) refer to the detected signls when sending one polriztion of the mode pir ut receiving the orthogonl one. The pedestls etween the peks re the distriuted mode coupling tht hppens during propgtion in the fier while the peks on ech side re the mode coupling induced the imperfections in the MUX/DeMUX setup t the trnsmitter (leding edge on the left-hnd side) nd the receiver (triling edge on the righthnd side). Tle 2. Summr of the mode stilit in ERCF. LVP 11 & 11 LVP LVP 11 & LVP11 LVP 21 & LVP21 Δn eff ΔP (5m ERCF) ~ ± 0.04 db ~ ± 0.01 db ~ ± 0.05 db PER (5m ERCF) 16.5 db 22.2 db 20.4 db Crosstlk (0.9 km ERCF) 14.2 db 25.7 db 18.3 db Received energ is clculted integrting the power (colored regions in Fig. 8) nd mode crosstlk in deciel is estimted tking the rtio. The mesurement noise floor, determined verging the signl outside the impulse response region, is sutrcted to the energ clcultions. Tle 2 lists ll the mesurement results otined in Section 3 with respect to the chrcteriztion of the si higher-order vector modes. It cn e clerl seen tht the power fluctutions ( P), the polriztion etinction rtios (PER) nd the propgtion crosstlk, which re indictive of the mode stilit, ll inversel scle with n eff. The distriuted mode coupling of the si modes re ll less thn 14 db fter propgting in

11 ERCF up to 0.9 km long, which indictes tht it should e possile to use the higher-order vector modes in ERCF s independent signl chnnels in MIMO-free mode division multipleing pplictions. 4. MIMO-Free Dt Trnsmission over Si Vector modes In this section, we report complete eperiment of MIMO-free dt trnsmission of QPSK signls over 0.9 km ERCF (the longest spool ville t this time). The schemtic digrm of the eperimentl setup is shown in Fig. 9. A single polriztion non-return-to-zero (NRZ) QPSK signl is generted n IQ modultor driven two pseudo-rndom inr sequence (PRBS) signls ( nd ). The tunle lser hs linewidth of 100 khz nd is set to 1550 nm. The QPSK modulted signl is split into 6 chnnels which re then time deled (τ 1 to τ 6 in Fig. 9) to emulte 6 independent QPSK signls. The signls corresponding to LPV 11 nd LPV 11 re comined through fier-sed polriztion em cominer (PBC) nd then re sent in free spce through collimtor nd shped phse plte efore eing coupled into the ERCF. The fier-sed polriztion controller (PC) nd the hlf wve plte (HWP2) re djusted so tht the two polriztions lign with the mjor or minor is of the ERCF. Qurter wve pltes (QWP1 nd QWP2) re used t the fier input nd output to correct the polriztion distortion induced free spce optics such s lens nd mirrors. A similr signl pth s tht of the LPV 11 is used for the two LPV 11 chnnels ut the phse plte is rotted 90. As discussed in section 3.1, since we did not hve the required specil phse plte for LPV 21 modes, we used sptil light modultor (SLM1) to emulte it. The required mirrors nd SLM, in similr rrngement s in Fig. 4, re not shown on the schemtic, onl the function phse plte lock is represented. After the SLM, the LPV 21 nd LPV 21 chnnels re comined in free spce using HWP1 nd polriztion em splitter (PBS) nd coupled into the fier. After propgting through 0.9 km ERCF, the signls enter the DeMUX stge which receives onl one mode t time s lred descried in Section 3.1. Fig. 9. Eperimentl setup of MIMO-free trnsmission of QPSK signls over si vector modes in 0.9 km ERCF. BPG: it pttern genertor; PBC: fier-sed polriztion em cominer; PBS: free-spce polriztion em splitter; BS: free-spce em splitter; HWP: hlf-wve plte; QWP: qurter-wve plte; ATT: opticl ttenutor; EDFA: erium-doped fier mplifier; OBF: opticl ndpss filter; CR: coherent receiver; LO: opticl locl oscilltor; RTO: rel-time oscilloscope. The free-spce optics in MUX nd DeMUX require creful lignment. A given chnnel of the MUX is first djusted nd serves s the reference, while the other ones re susequentl ligned. For the lignment of the sptil orienttion of the modes, QWP1 nd QWP2 re first removed while HWP2 nd HWP3 re djusted to lign the input polriztion long the fier nd SLM2 is. We djust the MUX nd DeMUX simultneousl while monitoring the em profile fter SLM2 using cmer. As long s the sptil modl cross tlk is smll, SLM2 converts the fier output ck to fundmentl Gussin-like em when sending/receiving the sme mode, while it converts it to higher order em, whose center is drk, when sending/receiving different modes. Figure 10 shows the mesured em profiles fter SLM2 for different comintions of the phse ptterns in MUX nd DeMUX. For further optimiztion, we remove the cmer nd djust MUX nd DeMUX monitoring the opticl power t the SMF output of the DeMUX until the smllest possile mode crosstlk is otined.

12 Fig. 10. The mode ptterns mesured fter the SLM in the DeMUX ccording to the different comintions of the MUX nd DeMUX phse ptterns. For the finl polriztion lignment, we first set the MUX nd DeMUX to send/receive the sme sptil mode nd then we rotte HWP2/HWP3 to minimize the received power so tht MUX nd DeMUX re sending/receiving orthogonl polriztions. QWP1 nd QWP2 re then instlled ck in the setup nd rotted for further minimiztion of the received opticl power. The polriztion lignment is then completed nd fied for ll sptil modes, i.e. QWP1 nd QWP2 re left untouched. To receive or polrized of n LPV modes, we just need to rotte the HWP3 ± 45 in the DeMUX. The MUX nd DeMUX re stle for periods of few hours t time during which it error rte (BER) mesurements cn e performed. We find tht it is mostl the 6-is stge tht is sensitive to the mient temperture nd it usull needs to e redjusted ever d. All the other optics nd mechnicl mounts re ver stle nd cn e left untouched, even for whole week. Mode Received Tle 3. Power trnsfer mtri (in db) of the MDM sstem. LVP11 11 Mode Sent LVP LVP11 LVP11 LVP21 LVP21 Totl Crosstlk (db) LVP LVP LVP LVP LVP LVP For BER mesurement, the signl coming out of the DeMUX is firstl ttenuted n opticl ttenutor (ATT) nd mplified n erium-doped fier mplifier (EDFA) so tht the opticl signl-to-noise rtio (OSNR) cn e djusted. The EDFA is followed n opticl ndpss filter with ndwidth of ~1 nm. The received signl is then directed to coherent receiver (CR) with ndwidth of 22 GHz in which it is mied with n opticl locl oscilltor (10 khz linewidth nd 13 dbm power). Finll, the demodulted I nd Q signls re cptured rel-time oscilloscope (RTO). We use offline processing tht include conventionl digitl signl processing (DSP) for single-polriztion single-mode coherent detection

13 sstems. The DSP simpl consists of locks of retiming, single equlizer using constnt modulus lgorithm (CMA) (11 tps needed), frequenc offset estimtion, crrier phse recover using 4th power lgorithm nd finll BER clcultion. No MIMO or PDM processing is used. The input opticl power efore HWP2 is out 7 dbm for LPV, 11 nd LPV, 11, nd 10 dbm for LPV, 21. The received opticl power, t the SMF output of the DeMUX is round 14 dbm for ll si chnnels. The totl losses induced the input mode coupling, the DeMUX nd propgting in the 0.9 km long ERCF, re therefore estimted to e 21 db for LPV, 11 nd LPV, 11, nd 24 db for LPV, 21. Opticl time domin reflectometr mesurement ws performed t 1550 nm on the ERCF with commercil instrument, without mode selective ecittion, nd led vlue of 2 db/km. The input opticl powers efore entering HWP2 re out 7 dbm (for LPV,11 nd LPV,11 ) nd 10 dbm (for LPV,21 ). The received opticl powers coming out from the SMF of DeMUX re round 14 dbm for ll the si chnnels. So the totl sstem loss, induced the input mode coupling, the DeMUX nd the 0.9 km ERCF, is estimted to e 21 db (for LPV,11 nd LPV,11 ) nd 24 db (for LPV,21 ) respectivel. Fig. 11. ) BER vs OSNR trnsmission curves t 24 Gud; ) BER t different ud rte (the si vector mode chnnels re ll on); c) the est nd the worst constelltion digrms t 16 Gud nd 32 Gud respectivel.

14 Tle 3 shows the mesured power trnsfer mtri of the MDM sstem, indicting oth the crosstlk etween chnnels nd the totl crosstlk contriuted ll the other chnnels. The vlues of ech column re otined sending onl one mode nd receiving the powers of ll the modes one one. The verged crosstlk in the LPV 11&LPV 11, LPV 11&LPV 11 nd LPV 21&LPV 21 mode pirs re 14.3 db, 22.8 db nd 16.3 db respectivel which coincide with the trend in their n eff (Tle 2). Note tht crosstlk in Tle 3 now include crosstlk t the MUX nd DeMu. The crosstlk etween the different mode pirs re elieved to originte from the imperfect mode ecittion nd filtering in the MUX nd DeMUX. We trnsmitted dt in ll the si vector modes nd mesured it error rte (BER) over 10 6 its to evlute the performnce of the MDM sstem. BER versus opticl signl-to-noise rtio (OSNR) t ud rte of 24 Gud is depicted in Fig. 11() for the trnsmission of one single mode t time (colored solid lines) nd the simultneous trnsmission of ll si modes (dshed lines with mrkers), showing n OSNR penlt round 8 db t the forwrd error correction (FEC) threshold of corresponding to 7% overhed. The sme simple single polriztion DSP for dt recover is used in ll cses, whether sending onl one mode or sending ll 6 modes simultneousl. Furthermore, we swept the ud rte from 16 to 32 Gud t n OSNR of ~25 db nd clculted the BER s displed in Fig. 11(). Figure 11(c) gives tpicl received constelltions t 16 Gud nd 32 Gud for the est nd worst chnnels. BER vlues elow the FEC threshold ws chieved for ll si chnnels for ud rtes up to 32 Gud. The worst chnnel in Fig. 11() is LPV 21 while it is LPV 11 in Fig. 11(). We elieve tht this is due to slight vritions in the lignment of the MUX nd DeMUX. 5. Discussion nd Conclusions In conclusion, we demonstrted tht few-mode polriztion mintining fier sed on n ellipticl ring core hs sufficientl lrge effective inde seprtions etween its orthogonll polrized vector modes to llow stle propgtion over km-length scle. The linerl polrized vector modes were found to mintin oth their sptil orienttions nd their polriztion sttes, even under significnt eternl perturtions such s pressure nd twists. The fct tht these vector modes ehiit orthogonl liner polriztion sttes further simplif mode division multipleing nd demultipleing. After propgting through the km-long ERCF, mesured crosstlk ws sufficientl low to llow completel MIMO-free dt trnsmission over si vector modes up to 32 Gud QPSK. We conclude tht few-mode polriztion mintining fiers re good cndidtes for short rech (<2 km) MIMO-free MDM dt trnsmissions, for emple in opticl interconnects of future high-cpcit dt centers, either t the inter-ord or inter-rck levels. At this time, crosstlk is still the min limittion in the trnsmission performnce. Crosstlk is strongl linked to the effective inde seprtion etween modes s well s to the chrcteristics of the sptil mode multipleers nd demultipleers. Further improvement to oth these spects, the fier design nd the multipleing sstem, re needed to demonstrte solutions supporting trnsmissions over higher numer of stle linerl polrized vector modes. Acknowledgments We cknowledge the help of Mr. N. Grégoire, Mr. S. Morenc, Mr. P. Chrétien nd Mr. P. Lrochelle. This work ws supported the Cnd Reserch Chir in Advnced Photonics Technologies for Communictions (APTEC), the Cnd Reserch Chir in Communictions Sstems Enling the Cloud, the Cnd Ecellence Reserch Chir in Enling Photonic Innovtions for Informtion nd Communictions (CERCP), nd the Nturl Sciences nd Engineering Reserch Council of Cnd (NSERC).