Patent Application
20120327957
Throughout the Information age, from the 1990's till present day, the worldwide focus on increasing capacity in communication networks has only grown. What spurs this on is the continuous growth in the Internet traffic with more web-applications becoming part of our daily lives. More people get a fibre to the home (FTTH), which paves the way for increased use of bandwidth-heavy content, such as iTunes, YouTube, MySpace, social platforms like Facebook, and video services on the Internet, including video-on-demand and HDTV broadcasting. A conservative estimate today is that the Internet traffic is going to double every 2 years, and that each currently installed fiber pair in the Atlantic region will need to provide 40-Tbit/s capacity by the year 2015. The projected capacity demands together with current upgrades of deployed networks indicate that there is also a need for more efficient and intelligent communication networks based on optical technologies and optical/electrical interfaces. Alongside the efforts to increase the traffic capacity is an increasing focus on doing this while reducing also power consumption. This is motivated by the increasing environmental awareness of the footprint left by the telecommunication industry. In Japan, for instance, the power needed to keep the Internet running is approaching some 10% of the national total electrical power consumption. In Internet exchange stations, there is an abundance of equipment that produces so much heat that active cooling is required, which requires further energy use.
Electronic routers used to switch traffic from one path to another in network nodes are currently among the most power-consuming devices in today's telecommunication systems. It is projected, for instance, that these will consume 9% of the electric power in Japan by 2015 unless there is a paradigm shift in data signal switching technologies.
There are many indications that components operating primarily based on optical signals rather than electrical signals may reduce the power consumption significantly. In October 2007 at a Nature Photonics Conference, it was concluded that “the answer to the problem lies with a move to all-optical switching and routing technologies that should be far more compact, consume less power and potentially be easier to upgrade to higher data rates owing to their transparent nature”.
In serial communications, fewer components will in general be used, and with ultra-short pulses very high bit rates may become available. Historically, increases in serial data rates have lead to cost savings, due to reduced complexity in management, reduced power consumption and a reduced number of components.
There is currently a strong trend towards transitioning the Ethernet standard, traditionally used for computer communications, to also be used in network protocols of future telecommunication systems. In an Ethernet network, the data bits are arranged serially in data packets of certain lengths and each packet has an address tag. Today, there are considerable efforts to push the standard from 10 Gbit/s Ethernet (“GE”) to 100 GE by multiplexing electrical signals in the time domain (electrical time division multiplexing, ETDM). In 5-10 years from now, it might then be expected that Internet routers and Internet exchange stations have several 100 GE lines that need to be transmitted to the same destination, and to avoid congestion in Internet exchange stations it may be necessary to employ an optical Ethernet multiplexing scheme. This could result in an optical 1000 GE, or 1 Terabit/s Ethernet (TE). The 1 TE must be optical (i.e. optical time division multiplexing, OTDM), as there are no signs that electronics will be able to provide these speeds in any foreseeable future, and certainly not so at reasonable power levels.
A great concern with the technologies developed so far is their massive power consumption. Using optical and high-speed serial data transmission technologies, and designing optical networks which rely on optically switchable data paths (circuit switching) rather than on electronic packet switching can reduce the power consumption enormously: A commercial router using 4 wavelengths with 40 Gbit/s ETDM rates uses megawatts of power, whereas an OTDM based 160 Gbit/s circuit switched router uses only about 20 W. Furthermore, 100 Gbit/s ETDM is still a research laboratory state-of-the-art, and there has only been a single basic back-to-back demonstration above 100 Gbit/s, namely an ETDM signal at 160 Gbit/s. In contrast, there have been several full 160 Gbit/s transmission system demonstrations since before 1999 using OTDM. And while ETDM does not appear able to go beyond 160 Gbit/s at present, OTDM does not have any obvious limits. By adding 4 data levels on the phase of each pulse and including polarisation, a data rate of 2.56 Tbit/s on one wavelength has been achieved. Recently, a breakthrough linked the research fields of OTDM with coherent communications, and it was shown that coherent receiver technology could be used for serial data at 640 Gbit/s (4 phase levels and polarisation multiplexing on 160 Gbaud pulse train). Coherent receiver technology offers very good transmission impairment tolerance because it detects the electric field of the data signal and can determine the phase directly and subsequently filter away, to a large extent, causal influences on data quality through digital signal processing.
If 100 GE is to evolve into the next “natural” generation, 1000 GHz pulse trains will thus be needed. It is challenging to generate and handle such short pulses, having pulse widths on the order of 400 fs or less, and having a very low time jitter (˜50 fs). To process very short pulses, one can benefit tremendously from optimising the temporal shape of the optical pulses used in the signal processing. Particularly, pulses with temporally flat tops have been seen to be able to greatly improve the performance of optical switches. With the appropriate dimensions, a flat-top pulse could allow for very uniform switching of optical data pulses in a fast optical switch.
Other systems and methods are described in the below documents. EP 2 034 634, which discloses a system comprising an O/E conversion element converting an input NRZ optical signal into an electric signal. A clock recovery circuit recovers a clock signal from the electric signal obtained by the O/E conversion element. That is, the data is not synchronised to a local master clock. A phase modulator applies phase modulation to the NRZ optical signal, using the recovered clock signal. An intensity modulator applies intensity modulation to the NRZ optical signal, using the recovered clock signal. A dispersion medium compensates for a frequency chirp of an optical signal output from the intensity modulator. As a result, the pulse width of the NRZ optical signal is compressed, and the NRZ optical signal is converted into an RZ optical signal.
Satoki Kawanishi, ‘Ultrahigh-speed optical time-division-multiplexed transmission technology based on optical signal processing, IEEE journal of quantum electronics, vol. 34, no. 11, 1 Nov. 1998.
EP 1 363 420 which discloses an optical modulation/multiplexing circuit that can fabricate a plurality of nonlinear optical waveguide devices and silica optical waveguides through a small number of processes, and achieve the simplification of the fabrication process and stabilization of the operation by hybrid integration with reduced connection loss. It employs lithium niobate domain inversion optical waveguides as nonlinear optical switches, and implements functions necessary for modulation and multiplexing such as input, splitting, multiplexing and timing adjustment of optical modulation signals and an optical clock signal by connecting glass waveguides to the input and output terminals of the domain inversion optical waveguides.
EP 1 819 070 which discloses a clock recovery circuit for a digital signal comprising a clock signal and a high-frequency jitter component due to polarisation scrambling in the optical domain.
It is expected that coming 1 Tbit/s network topologies may well be based, at least partly, on Ethernet frames.
As OTDM keeps the data on a single serial string at the same wavelength, it inherently lends itself conveniently to various types or degrees of format conversion. It has often been stated as being a benefit of OTDM that the single-wavelength nature makes it easy to perform e.g. wavelength conversion or regeneration in a single step, something which is not possible with parallel channels, separated in fibers or by wavelength.
A packet-based OTDM ring network could benefit from the present invention. Such a system would require synchronization of packets sent from one node and received by another. It is not necessarily possible to synchronize a master clock at the sending node with a master clock at the receiving node. Two master clocks will, however, have almost the same oscillation frequency, which is a feature that will allow the present invention to work in networks in which node equipment may originate from different manufactures that use slightly different technologies in their equipment, leading for instance to an difference between clocks in different nodes.
Ethernet packets can have varying lengths and bit rates (repetition frequencies). To enable the conversion from Ethernet packet signal to OTDM signals, the inventors have invented, in a first aspect, a method for synchronizing an incoming data packet signal, such as an Ethernet data packet signal, to a clock signal having a clock signal oscillation frequency, the incoming data packet signal having a first bitrate. The method comprises:
The incoming data packet signal is an optical signal, and the signal is split optically in a first part and a second part. The clock signal is a local master clock signal having a clock signal oscillation frequency to which the data packet signal is synchronised. The clock signal is thus not recovered from the data packet signal for synchronisation purposes, but the data packet is synchronised to the local clock. This allows for numerous incoming asynchronous data packets all being locked to the same local master clock frequency.
In the following, a number of further aspects, preferred and/or optional features, elements, examples and implementations will be summarized. Features or elements described in relation to one embodiment or aspect may be combined with or applied to the other embodiments or aspects where applicable. For example, structural and functional features applied in relation to a method may also be used as features in relation to the synchronizer or converter and vice versa. Also, explanations of underlying mechanisms of the invention as realized by the inventors are presented for explanatory purposes, and should not be used in ex post facto analysis for deducing the invention.
An issue with Ethernet packets is that they may very well have a slight offset in their repetition frequency from a nominal frequency. The frequency of a 10 GE packet signal is thus not necessarily 10 GHz, but rather 10 GHz +Δ/ƒ, where Δƒ is the offset frequency. The incoming data packet signal is split into a first part and a second part, each of which will, unless any processing is performed on them, have the same bitrate as the incoming packet data signal.
The offset frequency signal is provided to a system that can stretch or compress the incoming data packet signal based on the offset frequency signal. The incoming data packet signal is stretched or compressed into a synchronized data packet signal, which is a data packet signal that represents the incoming data packet signal, but is synchronized with the clock signal. Such a system is in some cases appropriately referred to as an optical time lens. The optical time lens functionality can be provided by feeding the offset frequency signal to a field programmable gate array (FPGA) which then controls a modulator, such as a Lithium Niobate modulator, to modify the phase across the signal, in this case the incoming data packet signal.
As opposed to prior art, the time lens in the present invention can stretch or compress a full data packet waveform until the temporal separation between adjacent bits in the packet matches the period of the clock signal. The applied phase modulation is therefore applied to the full packet. A packet envelope detector can determine the waveform extent (i.e. the full packet length) and this information be fed to the FPGA to make the FPGA apply the modulation across the packet. In many situations, a parabolic phase-shape is advantageous for application to the full packet to stretch or compress it.
Other systems for stretching or compressing signals may go under terms other than “optical time lens”. This term is just useful for capturing the functionality that is required.
The stretching or compressing required to obtain a synchronized data packet signal having a bitrate matching the clock signal oscillation frequency can thus advantageously include the following steps:
This is a new use of the optical time lens which, once this method has been provided by the present invention, can be used by a person of average skill in the art to set up a system for stretching/compressing the full packet.
A second aspect of the invention provides a method for converting an incoming data packet signal into a synchronized return-to-zero signal. The method comprises the steps of the first aspect method to provide a synchronized data packet signal based on the incoming data packet signal. The method further comprises:
The result of the sampling is the synchronized return-to-zero signal. In this way, a (non-return-to-zero) data packet, represented by (at least a part of) the incoming data packet signal, has been converted into a return-to-zero signal, in principle ready for transmission over an optical fiber.
The sampling may be obtained by letting the bits in the data packet signal open a gate for the narrow clock pulses, using e.g. a non-linear optical loop mirror (NOLM), a Kerr switch, or four wave mixing.
A third aspect of the invention provides a method for preparing one or more incoming line data packet signals for transmission as a time-division multiplexed return-to-zero signal. The method comprises:
These resulting time division multiplexed return-to-zero signal lends itself to further transmission, for instance in local area networks, building-to-building networks or networks connecting servers and/or connecting servers to internet exchange stations. If there is only one incoming data packet signal, the time division multiplexing does nothing material to the return-to-zero signal.
The time-division multiplexing might advantageously be performed on a stable integrated planar lightwave circuit with fixed accurate delays.
The method enables transmission of multiple data packets lines, such as Ethernet lines, as a single OTDM signal. This design for synchronisation and multiplexing prevents traffic congestion, as each incoming line is guaranteed a time slot into which it can be multiplexed.
At a receiver end, demultiplexing the OTDM signal readily provides the synchronised 10 GE Ethernet data packet signals.
The converting of the lines is performed with reference to a single master clock signal, although different clocks could be used for different lines, if necessary. This will require further processing when multiplexing the lines.
A fourth aspect of the invention provides a synchronizer for synchronizing an incoming data packet signal to a clock signal having a clock signal oscillation frequency, the incoming data packet signal having a first bitrate. The synchronizer comprises:
The splitter for splitting the received incoming data packet signal into a first part and a second part is preferably an optical splitter splitting the signal optically.
A fifth aspect of the invention provides a converter for converting an incoming data packet signal into a synchronized return-to-zero signal. Such a converter comprises a synchronizer in accordance with the fourth aspect of the invention, and further comprises:
It is common for all aspects of the present invention that the incoming data packet signal is an optical signal, and the signal is split optically in a first part and a second part. Furthermore, the clock signal is a local master clock signal having a clock signal oscillation frequency to which the data packet signal is synchronised. The clock signal is not recovered from the data packet signal for synchronisation purposes. Therefore the data packet signals, comprising asynchronous data packets, are synchronised to the local master clock.
In the following, the invention is described by way of examples, with reference to the figures.
In an alternative embodiment it can be provided by an optical or electrical mixer. In the case of using an electrical mixer, photodetectors are needed in front of the mixer—the mixer then multiplies the now electrical data signal with the local clock signal and lowpass filters the output. This yields a slowly varying signal with a sinusoidal dependence on the frequency difference between the clock and the data signal. The optical equivalent could be a demultiplexer-device. In this case, no photodetectors are needed in front of the mixer, if both the clock and the data are optical (the clock would then use the output from the pulse source 323 from
As schematically illustrated in
The packet envelope detector can determine the envelope electrically or optically based on the second part of the incoming data packet signal. At high speeds, a slow photodetector can be used as packet envelope detector, in which case a part of the incoming data packet signal (typically part of the second part) must be provided to the photodetector.
A synchronizer might have two inputs instead of one, whereby the first and the second parts are provided from the outside and the splitter can be avoided. It is, however, practical to include the splitter in the synchronizer because it ensures the correct relationship between the first and second part of the incoming data packet signal, crucial for obtaining the synchronization.
The synchronizer produces the synchronized data packet signal 202a, which has a bitrate that is synchronized with the clock signal frequency. It may often be, that the individual data pulses in the synchronized data packet signal 202a are temporarily too wide to be interleaved with other pulse trains. Therefore, the individual data pulses in the synchronized data packet signal 202a are temporarily compressed in order to provide a fully converted signal. This full conversion, bitrate synchronization followed by pulse compression, takes place in a converter (301) that includes the synchronizer.
Generally the incoming data packet signals are synchronised to a local master clock. Advantageously the incoming data packet signals will therefore, after the conversion, have the same frequency as the local clock. All data signals will therefore be synchronous and they may be multiplexed in the timedomain, whereas if there where deviations on the frequencies of the incoming data signals this would not be possible.
In a preferred implementation, the sampler comprises a nonlinear optical medium such as a nonlinear fibre (e.g. a Highly Non-linear Fibre, HNLF), a non-linear compact waveguide (such as a chalcogenide waveguide or a silcon nanowire type of waveguide), or a non-linear crystal (such as a periodically pooled Lithium Niobate, PPLN). In such nonlinear optical media, the short pulses 325a-325c changes the transmissive properties of the media on an ultrafast time scale, with a response time of e.g. ˜5 fs, to make a window or gate for the longer data pulses from signal 202a. The short pulses are preferably return-to-zero, but may in principle be a modulation on a CW background with too little intensity to “open the gate” in the non-linear medium. Thereby, only the part of the longer data pulses that overlaps with the gate will be transmitted or sampled, while the rest of the longer data pulse will be blocked by the sampler. The result is a return-to-zero signal 302a consisting of a train of short, such as having a duration shorter than 1 ps or 500 fs, data pulses.
It is clear that the sampling pulses 325a, 325b, 325c must fit within the time slots available. In a scenario where ten 10 GE signals are converted to corresponding synchronized return-to-zero signals for transmission in a common OTDM signal, the available time slot is 10 ps. If the number of channels is 100, the available time slot is 1 ps. The pulse source must accordingly be able to produce pulses that can fit well within this time slot, and preferably provide pulses of duration less that 1 ps, such as less than 500 fs. At the same time, the OTDM signal is to be transmitted over some distance, which puts other demands on the shape of the pulses. However, the shorter the distance, the wider the spectral width of the pulses can be. Accordingly, the transmission conditions should be considered when choosing the pulse source. There is no reason to use spectrally very narrow pulses if wider pulses will do just fine, for instance for transmission in a local area environment.
| Number | Date | Country | Kind |
|---|---|---|---|
| 10152266.2 | Feb 2010 | EP | regional |
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/DK11/50025 | 2/1/2011 | WO | 00 | 8/20/2012 |
| Number | Date | Country | |
|---|---|---|---|
| 61300126 | Feb 2010 | US |
