Open Access
Add to BibSonomyAbstract
We have performed a detailed study of four Perfluorinated Graded Index Polymer Optical Fiber (50/490, 62.5/490 and 120/490) issued from several manufacturers and designed to be used in short length (100m) small office/home office network operating at 850nm. We have used commercially available 850nm low cost XFP transceivers and have compared the power dispersion penalties measurements with the ones realized using a photoreceiver including the Electronic Dispersion Compensation technology to increase the range of the transmission without impairments. It will be demonstrated the potential of the polymer fibers coupled to the electronic dispersion compensation technology to successfully transmit a 10GBase-SR signal with a Bit Error Rate of 10-12 over 100m.
©2008 Optical Society of America
1. Introduction
During the last few years, a significant increase of the data rate through the corporate Network or the Internet has been observed, due mainly to the explosion of the media exchange (music, Video on Demand (VOD), big data files,…). The IEEE802.3 Gigabit Ethernet Standard is becoming the commonly installed standard in an office network. In order to satisfy the growing demand of the 10Gbps’ applications, some manufacturers of optical fiber such as Draka Comteq [1] or Corning [2] have developed new Extended Bandwidth Glass Multimode Fiber allowing to transmit 10Gbps signal through more than 500m. These fibers – principally optimized for a 850nm wavelength operation – exhibit a high Bandwidth-Length product (>2GHz.km) and the well known (more than 30 years old) fabrication process of the Glass multimode fiber allows to obtain glass fibers with high purity. These glass multimode fibers enable to reach a good tradeoff between cost and performance. Nevertheless, in order to decrease the connection cost of an optical LAN, novel optical fibers, based on polymer material, have been developed by Chromis Fiberoptics [3] and Asahi Glass [4] In fact, contrary to the silica based fiber, the mechanical flexibility of the polymer fiber enhances the easy handle and connection of this kind of fiber: clip on connector has been also developed to be fixed directly on the external coating of the polymer fiber without the use of expensive cutting tool like in the glass fiber’s world. Even if these polymer optical fibers exhibit some drawbacks – a well known one is its high attenuation – they are designed to be used in a Small Office/Home Office high bit rate network and open the road to the “Do It Yourself” concept. The graded index polymer optical fiber (GIPOF) based on an amorphous fluoropolymer (CYTOP®) have been designed to operate in high bit rate network at 850nm. These polymer fibers present a core diameter from 50μm to 120μm [3, 4] and need to be tested for the deployment of a 10Gbps optical network.
The aim of this paper is to investigate the potential of the PF GIPOF commercially available for 10Gbps application. These PF fibers have been designed to be implanted in a high speed Small Office/Home Office with easiness and to provide high performance. The improvement of the fabrication processes of these new PF GIPOF will allow to decrease the spectral attenuation and so to increase the Optical Power Budget (OPB). In fact, two different fabrication processes (perform method and co-extrusion process) have been used to pull the PF GIPOF, with each its advantages and its drawbacks which affected the shape of the fiber transfer function as well as the attenuation profile.
The high attenuation of the polymer based optical fibers as well as their intermodal dispersion induce link length’s limitation particularly at 850nm. In fact, Pedrotti, et al., [5] report the state of art of the multi-gigabit transmission over POF: most of the presented works exhibits a bit rate-length product less than 1Gbps-km. Li, et al., [6] report a high bit rate-length product but with the use of APD receivers not compatible with low cost systems. Giaretta, et al., [7] have realized an 11Gbps transmission over a range less than 100m by using a 1300nm FP laser and coupling optics which are not suited for a low cost 10Gbps optical network.
We propose in this paper to investigate the use of conventional low cost devices such as VCSEL (Vertical Cavity Surface Emitting Laser) and PIN photodiodes with plastic fiber in order to have a pragmatic approach of the future high speed optical network. With the use of the Phyworks technology [8] based on the Electronic Dispersion compensation devices (EDC) placed at the electrical output of a conventional photoreceiver (here, a GaAs photodiode), we will describe the potential of the full system “low cost devices/PF GIPOF/EDC component” to increase the performance (link length, power dispersion penalties,…) of the 10Gbps Local Area Networks.
The EDC technology allows overcoming the modal dispersion of the multimode polymer fibers but this technology can be used with multimode glass fibers with the same result. Phyworks company wants to produce 10Gbps TX/RX module including VCSEL TOSA (Transceiver Optical SubAssembly) and driver for the TX part and photodiode ROSA (Receiver Optical SubAssembly), EDC and transimpedance amplifier (TIA) chips for the RX part with target price of less than 200$ per module. The coupling pair EDC/TIA cost should be around 55$ (low cost devices contrary to the ones used in the previous studies).
Thus, the first part of this study deals with the physical parameters characterization of the PerFluorinated Graded Index Polymer Optical Fiber (PF GIPOF) under test (50/490, 62.5/490 and 120/490). Bandwidth measurements as well as optical backscattering measurements have been performed in order to understand the PF GIPOF’s behavior at 850nm, particularly fiber attenuation and modal bandwidth behaviors.
The second part of this study consists in the achievement of a 10.3125Gbps transmission complying with the IEEE 802.3ae standard through 100m of the PF GIPOF under test. We show the Optical Power Budget (OPB), the power dispersion penalties and the 10GBase-SR mask testing of the PF GIPOF/EDC devices designed to be installed in a Small Office/Home Office (SOHO) environment.
2. Description of the perfluorinated graded index polymer optical fibers under test
A comparative study between 4 different PF GIPOF has been performed in order to understand the modal behavior of the new 50μm (called 50A) and 62.5μm (called 62.5A) core diameter PF GIPOFs in comparison to the 120μm ones provided by two different manufacturers (called 120A and 120B). This study is, to the author’s knowledge, the first comparative study of 4 types of PF GIPOF.
All the following data are given for 850nm wavelength and are issued from manufacturers’ datasheets. The different fibers issued from the manufacturer #A have a high attenuation (45dB/km) and provide a bandwidth – length (BL) product greater than 500MHz.km with a Numerical Aperture (NA) of 0.18. They are fabricated using a co-extrusion process. The fiber from manufacturer #B has a lower attenuation (23dB/km) and a slightly lower BL product (400MHz.km) and its NA is 0.19. They are fabricated using an interfacial-gel process. All fibers have a 490μm external coating and we choose to work with fiber length of 100m. This length is representative of the main part of preinstalled multimode fiber paths (<300m) for indoor applications [9].
The backscattering technique (Fig. 1) has been used at 850nm to localize the defects in the fiber core and to determine its length. The main optical intensity peak which is localized around 100m for all the fibers allows to determine the fiber length by the Fresnel reflection of the propagated light at the end of the fiber (polymer/air interface at the connector). Backscattering echoes materialize the defects along the fiber. The PF GIPOF issued from the manufacturer #A (more particularly concerning the 50A sample) exhibits some defects as opposed to the PF GIPOF from the manufacturer #B. The aim of this study is to achieve a 10Gbps transmission with the best Bit Error Rate (BER) through 100m of PF GIPOF: use of 110m 62.5A PF GIPOF induces too many penalties (intermodal dispersion) in the link and does not allow obtaining a 10-12 BER (as discussed in the next section) so the length of this fiber has been reduced to 100m. Then, we have performed fiber bandwidth measurements in the frequency domain at 850nm to check the potential of PF GIPOF to transmit high bit rate signals through the required length. High speed 850nm VCSEL (10Gbps) has been modulated with a small amplitude sinusoidal waveform and the envelop of the propagated waveform (through the fiber under test) is directly detected in a high speed/large area photoreceiver compatible with the multimode fiber. These E/O and O/E devices have been used in this experiment as reference devices in order not to induce penalties on the bandwidth measurements. We determine the transfer function of the PF GIPOF in order to obtain the 3- dB optical bandwidth. The modulus of H(f) defined in (1) allows us to obtain the 3-dB optical bandwidth of the PF GIPOF at 850 nm by measuring the 6-dB cut-off frequency in the electrical domain.
S21-100m and S21-2m represent, respectively, the measurement of the S parameters of the PF GIPOF under test and a short length of PF fiber used as a reference length. The measurement is shown in Fig. 2 and the PF GIPOFs 62A, 120A and 120B exhibit a BL product of 400MHz.km as expected [3, 4]. The 50A PF GIPOF present a 2.5GHz modulation bandwidth and we suspect a non concentricity default of the fiber core regarding to the position of the fiber cladding. In fact, the output of the laser (50μm Glass optical fiber pigtail with FC connector) has been directly coupled to the fiber under test with the use of a FC adapter where the alignment in between the lead in fiber and the fiber under test has been done regarding to the external coating of the fibers fixed onto an 2.5mm ferrule. A misalignment of the PF GIPOF core with respect to the lead in glass optical fiber induces a bad optical coupling and so some impairment in the bandwidth’s measurement. By using a test setup allowing to offset the launch injection (measured offset # 10μm) of the lead in fiber regarding to the fiber under test and owning a camera viewing system, it has been shown that the 50A fiber exhibits a BL product closed to the 62A fiber as expected in the datasheets.
3. Design of the test setup used in the 10 Gbps application
We have developed a specific experimental test setup performing the BER measurement of a 10Gbps optical link based on low cost devices as well as the mask testing of the 10GBase-SR standard (Fig. 3). This test bench is composed of a 12.5Gbps Bit Error Rate analyzer divided into two parts: a pattern generator (PG) and an error detector (ED). A Pseudo Random Bit Sequence (PRBS) of 220-1 bits is provided by the PG with a 10.3125Gbps data rate. This sequence feeds the baseband input of a commercially available XFP transceiver. Then, the PRBS sequence directly modulates the current of a high speed 850nm multimode Vertical Cavity Surface Emitting Laser (VCSEL) used as the TX in the picolight XFP transceiver. The optical link is composed of a 50μm core PF GIPOF patchcord, a 50μm Glass optical fiber We have developed a specific experimental test setup performing the BER measurement of a 10Gbps optical link based on low cost devices as well as the mask testing of the 10GBase-SR standard (Fig. 3). This test bench is composed of a 12.5Gbps Bit Error Rate analyzer divided into two parts: a pattern generator (PG) and an error detector (ED). A Pseudo Random Bit Sequence (PRBS) of 220-1 bits is provided by the PG with a 10.3125Gbps data rate. This sequence feeds the baseband input of a commercially available XFP transceiver. Then, the PRBS sequence directly modulates the current of a high speed 850nm multimode Vertical Cavity Surface Emitting Laser (VCSEL) used as the TX in the picolight XFP transceiver. The optical link is composed of a 50μm core PF GIPOF patchcord, a 50μm Glass optical fiber based variable optical attenuator (VOA) and the PF GIPOF under test. Two different photoreceivers have been used to recover the PRBS sequence sent by the PG:
- - the RX part of the XFP transceiver,
- - a conventional GaAs LC Receiver Optical SubAssembly photodiode (LC-ROSA) coupled with an EDC device provided by Phyworks
The recovered sequence is then analyzed by the ED part of the BER analyzer and the BER has been measured as a function of the optical power adjusted by the use of the VOA in front of the receiver (Fig. 3). This measurement allows us to analyze the power dispersion penalties of an optical link composed either by a 50A, 62.5A, 120A, 120B PF GIPOF patch cord or the fiber under test (length closed to 100m).
The use of a sampling oscilloscope allows the eye diagram measurement and the mask testing of the 10GBase-SR standard.
Fig. 3. Measurement test setup of the 10Gbps transmission over the PF GIPOF (electrical connection (dashed lines) and optical connection (solid lines)).
The EDC technology is a cost effective solution to struggle the intermodal dispersion inherent to the optical link based on the multimode fiber. The phyworks serial EDC replaces the existing Clock Data Recovery (CDR) of a classical XFP transceiver.
4. Review of the measurement results: optical power budget, power dispersion penalties and mask testing
We have performed the transmission of a 10.3125 Gbps baseband signal over the 4 fibers with the use of the test setup described in the previous section. Some studies on high speed data transmission over PF GIPOF have been already published [7]. As mentioned previously, this study [7] has been achieved at 1300nm and with a 130μm core diameter PF GIPOF. This fiber has been excited in restricted mode launch conditions with the use of a single mode fiber coupled to the 1300nm laser which increases the fiber bandwidth. Moreover, in order not to induce mode filtering at the receiver and to reduce the coupling losses due to the large core diameter of the fiber, Giaretta, et al., have used a collimating lens and a focusing lens inducing 4.8dB coupling losses. Our study is realized at 850nm (low cost 10G XFP transceiver) and without adaptive optics due to the fiber core diameter.
The variable optical attenuator based on the 50μm glass multimode fiber and localized at the beginning of the optical link has been used firstly to adjust the optical power in front of the photoreceiver. On the other hand, this attenuator do not induce a numerical aperture mismatch when we connect the fiber under test (NA=0.18). In fact, the 50μm core based glass optical fiber presents a NA=0.2 and the mismatch of the numerical aperture is then reduced: the use of a 62.5μm glass multimode fiber based VOA (NA=0.27) will increase the coupling losses with all the 4 PF GIPOF owning a NA=0.18.
4.1 Optical power budget
The optical power budget has been firstly measured for the different fibers under study.
It refers to the allocation of available optical power among various loss-producing mechanisms such as launch coupling losses, fiber attenuation and connector losses, in order to ensure that adequate signal strength is available at the photoreceiver. Figure 4 explains the measurements and the results are summarized in the table 1. Pin-SL represents the optical power coupled in a patchcord and measured before the VOA. Pout-SL designs the output optical power of the combined system patchcord/VOA/short length of PF GIPOF under test. Pout is related to the optical power measured at the end of the desired length (# 100m).
Table 1. Measurement of the optical power (dBm) at 850nm (SL designs a Short Length of PF GIPOF)a
The insertion losses (PF GIPOF patchcord+VOA) change between 2.1 up to 3.6dB for all the used short length of fiber under test. The available optical power at the output of the fibers under test is much higher for the 120B fiber than for the fiber provided by the manufacturer #A. The TX part of the XFP module is able to launch -1.5dBm optical power. The stressed eye sensitivity of the receiver is - 7.5dBm inducing an optical power budget of 6dB [10]. The available output power is then much higher than the sensitivity of the XFP module at 850nm and power penalties will be expected during the BER measurement.
4.2 Power dispersion penalties measurements
In order to exhibit the power dispersion penalties for all the fibers under test, we have measured the BER as a function of the received optical power in the back-to-back case and by inserting the required length corresponding to the 4 fibers under test.
The BER measurements realized on the 50A PF fiber do not provide goods results (no error free in all case even if we remove the VOA due to a default of concentricity of the fiber core regarding to the cladding) as described in the section 2. We summarize the power dispersion penalties for a BER=10-9 in the Table. 2.
Table 2. Measurement of the power dispersion penalties (dB) of the optical link
The PF GIPOF 62.5A, 120A and 120B coupled with an EDC device exhibit power dispersion penalties around 4dB. These penalties are inherent to the modal bandwidth of the fiber. As described in the datasheets of the Picolight XFP module, the worst case distance range of a 10.3125Gbps signal over a 50/125 glass multimode fiber with a Bandwidth-length product of 400MHz.km (500MHz.km) is 66m (82m). These BL products have been given in order to have a comparative reference with the measured modal bandwidth of the 4 PF GIPOF under test. As mentioned previously, the 50A PF fiber cannot transmit such kind of high bit rate signal. The two 120μm core diameter PF GIPOF exhibit a power dispersion penalties around 3.5dB with the use of an EDC device and 9dB without dispersion compensation.
The number of propagation modes in an optical fiber depends on the core diameter. The 120μm PF GIPOF own a larger core diameter than the 62.5A fiber. By the way, the propagated modes in the 120μm - core fiber are much numerous in that case. The intermodal dispersion in the large core PF GIPOF is then much higher than in the 62.5A fiber. These results has been demonstrated by the power dispersion penalties measurement in between the two fibers without the use of an EDC device (5dB for the 62.5A compared to 9dB for the 120μm PF fibers). Nevertheless, all these power dispersion penalties have been induced not only by the modal bandwidth limitations of the fiber but also by a mode filtering at the receiver (active area # 70μm).
The transmission of a 10Gbps signal has been achieved over 100m of the 62.5A PF GIPOF with a BER=10-12. The length of the optical link based on the 62.5A fiber is higher than the range of the optical link based on glass multimode fiber provided in the datasheets of the picolight module (BL product # 400-500MHz.km – length 66m/82m). The additional range gained with the use of an EDC device at the photoreceiver is equal to 18m which demonstrated the potential of the dispersion compensation devices coupled with the 62.5μm core diameter based PF GIPOF.
4.3 Mask testing
The 10GbE template for the characterization of the shape of an eye diagram by the mask testing method is defined by the IEEE 802.3ae standard and particularly in the physical medium dependent section. This measurement allows to investigate the bandwidth limitations, the intersymbol interference (ISI) inherent to the overall system (TX, RX and optical fiber) and the signal jitter. The mask testing indicates the impact of the optical link on the quality of the signal. In the Fig. 6, we present the mask testing measurement realized on the 62.5A PF GIPOF (100m length) and particularly a comparative measurement between EDC and no EDC configurations. In the two cases, the optical power measured at the output of the 100m length 62.5A fiber is closed to -11dBm. By using the conventional photoreceiver (RX part of the low cost picolight XFP), the BER is closed to 10-7 and set around 3.10-12 with the use of the EDC device. The eye opening is then higher with the use of the EDC device. It has been shown that the EDC device enhances the signal quality by reducing the intersymbol interference phenomena which demonstrates without any doubt the potential of this kind of device.
5. Conclusion
We have investigated in this paper the use of a fluorinated based polymer optical fiber in a short range (100m) high bit rate optical network (10Gbps). The shape of 4 PF GIPOF has been firstly analysed as well as the modal bandwidth measurement at 850nm. The bandwidth length product of the fibers under test is close to 450MHz.km except for the 50A fiber. Then, we have designed an experimental test setup to investigate the potential of the PF GIPOF coupled with an electronic dispersion compensation device localized at the electrical output of the photoreceiver. The optical power budget, the power dispersion penalties and the mask testing of the optical link to the 10GBase-SR standard (10Gbps at 850nm on multimode fibers) have been displayed. It has been shown that the 62.5A PF fiber/EDC device coupling allow the achievement of a 10Gbps transmission with good signal quality (BER#10-12 with a high eye opening). In this study, we have used low cost devices (XFP, plastic fiber, no lens) to keep a pragmatic approach of the 10Gbps transmission in a short range SOHO network.
6. Acknowledgments
The authors want to thank André Goffin from the Royal Military Academy of Brussels for the fruitful discussions, B. Bareel from Nexans for the supply of the clip-on connectors, N. Weiner from Phyworks for the supply of the EDC devices coupling with 850nm commercially available photodetectors, Y. Watanabe from Asahi Glass for the supply of the Lucina fiber and W. White from Chromis fiberoptics for the supply of the GigaPOF fibers.
References and links
1. Datasheets of the Draka Fibers, http://www.draka.com/draka/Drakafiber/Languages/English/Navigation/Markets&Products/Technical_Support/Datasheets/DCOF_MaxCap-550.pdf
2. Datasheets of the Corning Fibers, http://www.corning.com/opticalfiber/products__applications/products/infinicor.aspx
3. Datasheets of the Chromis Fibers Perfluorinated GIPOF “GigaPOF” , http://www.chromisfiber.com/products.htm
4. Datasheets of the Asahi Glass Fibers Perfluorinated GIPOF “Lucina” , http://www.agc.co.jp/english/rd_e/e_lucina.html
5. K. D. Pedrotti, R. P. Dahlgren, J. A. Wysocki, S. E. Ralph, R. Gandhi, and A. Polley, “Multi-gigabit transmission over POF,” in Proceedings of POF World West, (Santa Clara, June 21–22, 2006).
6. W. Li, et al., “Record 2.5 Gbit/s 550 m GIPOF Transmission experiments at 840 and 1310 nm wavelength,” in Proceedings of POF1999, (China, July 13–16, 1999), pp 60–63.
7. G. Giaretta, et. al., “High-Speed (11 Gbit/s) Data Transmission Using Perfluorinated Graded-Index Polymer Optical Fibers for Short Interconnects (<100m),” IEEE Photon. Tech. Lett. 12, 347–349 (March 2000). [CrossRef]
8. Description of the Phyworks Technology, http://www.phyworks-ic.com/technology.asp
9. M. Bennett, A. Flatman, and B. Tolley, “Broad Market Potential of 10Gb/s Ethernet on FDDI grade MM Fiber,” IEEE 802.3 , (2004), http://www.ieee802.org/3/tutorial/mar04/10GMMF_SG_v031604a.pdf
10. Datasheets of the Picolight XFP module, http://www.picolight.com/UserFiles/PDF/05001778R3_RoHS_SW__XFP_.pdfhttp://www.picolight.com/UserFiles/PDF/05001778R3_RoHS_SW__XFP_.pdf
