1. Field of the Invention
The present invention relates generally to optical telecommunication, and particularly to all-optical header processing and packet switching.
2. Background of the Invention
With the increasing popularity of the World Wide Web, the Internet protocol which is by nature a packet-switching technology has become the de facto data transmission standard. However, current optical networks are still based on circuit-switching technology and therefore may not be suitable for handling heavy data traffic. Processing and routing of packet-formatted signals optically are hampered by the lack of practical optical buffers and the limited capability of available optical logic gates.
In the conventional design, a typical packet switch generally includes a header processing unit (HPU) and a packet routing unit (PRU). The HPU processes the header of an incoming packet, determines which output port the input packet should be sent, and set the PRU accordingly. During the packet transmission, either physical or logical, a data path is then provided by the PRU between the related input and output ports. Most of the current optical packet switches are in fact hybrid optical packet switches, i.e., while the packet remains in the optical domain, a copy of the packet header is converted into electrical signals for processing in the HPU. The decision of the HPU is then used to set the PRU to route the packet.
Therefore, it is an object of the present invention to provide an all-optical switching of an optical packet, or at least provide the public with a useful choice.
It is another object of the present invention to provide all-optical header processing, or at least provide the public with a useful choice.
It is a further object of the present invention to provide an all-optical add/drop node, or at least provide the public with a useful choice.
According to an aspect of the present invention, a process for all-optically switching an incoming optical signal having at least a data packet is provided. The data packet including at least an address bit in its header, and the process includes
According to a second aspect of the present invention, an optical signal switch, which receives an incoming optical signal, is provided. The incoming optical signal includes at least a data packet having at least an address bit in its header, and the switch includes:
According to a third aspect of the present invention, an optical header processor for processing a header of an incoming a data packet is provided. The data packet has at least an address bit in its header, and the header processor includes:
According to a further aspect of the present invention, an optical add/drop node for downloading an incoming data packet from an optical signal to a local network and for uploading a local data packet onto the optical signal includes
Other aspects and advantages of the invention will become apparent from the following detailed description, taken in conjunction with the accompanying drawings, which description illustrates by way of example the principles of the invention.
a illustrates the format of an input data signal suitable for being used in an exemplary header processor of the invention;
b illustrates the format of an input control signal suitable for being used in an exemplary header processor of the invention;
c illustrates the processing by the exemplary header processor of the invention;
a-c respectively illustrate an input control signal, a first input data packet in which its “1” address bit temporally overlaps with the trigger of the control signal, and a second input data packet in which its “1” address bit does not overlap with the trigger of the control signal;
a-c respectively illustrate synchronized temporal profiles of the input data packets, the input control signal and the output control signal;
a-d illustrates another synchronized temporal profiles of the input control signal, the input data packets, the output control signal, and the output data packets.
It is observed that the presence of a signal at a wavelength λd of a Fabry-Perot laser diode (FP-LD) can lower the injection locking threshold at another wavelength λc. The generation of a packet long output control signal from interaction at a single bit is based on the bistable nature of injection locking in the FP-LD, i.e., it takes less power to maintain injection-locking than to initiate one.
a-c illustrate an exemplary scheme for all-optical header processing and output control signal generation using a single FP-LD 205. The exemplary scheme requires an input control signal 201, 202 at wavelength λc that has a trigger 215 at power PH and a long trailer at power PT where PH>PT (
Furthermore, the wavelengths of the input control signal 201, 202 and incoming data packets 207, 208, λc and λd, are located at the longer wavelength side of two different longitudinal modes of the FP-LD 205. The power of the input control signal trigger PH is chosen to be Pth2≦PH<Pth1 where Pth1 and Pth2 are the injection-locking thresholds of the FP-LD 205 at wavelength λc in the absence and presence of the data packets 207, 208 at λd respectively. In other words, Pth1 and Pth2 are respectively the single-mode and two-mode injection locking thresholds of the FP-LD 205 at λc. Thus the input control signal trigger will injection-lock the FP-LD 205 and experience power gain if it matches a ‘1’ in the data packet 207, 208, but the input control signal trigger 215 cannot initiate injection locking alone, i.e., if it matches with a ‘0’=0 in the data packet 207, 208. The power of the input control signal trailer PT is chosen to be Pth3≦PT<Pth2, where Pth3 is the power at which the FP-LD 205 returns to the unlocked state from the locked state, such that once the input control signal trigger 215 initiates injection-locking in the FP-LD 205 the trailer 217 can sustain the injection-locking state at wavelength λc till the end of the input control signal 201 because of the bi-stable property of injection-locking. The input control signal trailer 217, however, cannot initiate injection locking even in the presence of the ‘1’ bits in the data packet 207, 208. From
c shows the signal processing schematically. The FP-LD 205 is set in the locked state at wavelength λd in the beginning. For data packet pk_1207, the input control signal trigger 215 of the input control signal 201 matches with the ‘1’ bit 209 in the header 211 in the time domain. Injection locking at wavelength λc is initiated and sustained through the duration of the input control signal 201. In other words, FP-LD is switched from the locked state to the unlocked state at wavelength λd, and from the unlocked state to the locked state at wavelength λc. The output control signal 221 of the FP-LD 205 at λc is therefore high (ON state), and the output data signal 223 may be relatively significantly suppressed as could be understood in the art.
During the guard band period 219 of the input control signal, the FP-LD 205 returns back to the unlocked state at wavelength λc in the exemplary embodiment due to the zero power of the input control signal during this period.
Afterwards, data packet pk_2208 arrives, and the control signal trigger 215 of the input control signal 202 does not match with the ‘1’ bit 209 in the header 211. Injection locking at wavelength λc is not initiated either at the control signal trigger 215 or at other part of the input control signal 202. The output control signal of the FP-LD 205 at wavelength λc is low (OFF state), while the FP-LD 205 may transmit data packet pk_2208 in its output data signal 223 as shown in
a shows the temporal profile of the input control signal.
Alternatively, the input control signal can be generated by direct modulation of a distributed feedback (DFB) semiconductor laser diode with square electrical pulses (not shown) as generally understood in the art. The peaks of the natural relaxation oscillations initiated by the modulation can function as the trigger while the steady state output of the laser serves as the trailer of the required control signals.
Other types of laser diodes may be also suitable for generating the input control signal, for example, a vertical cavity surface emitting laser (VCSEL) or a Fabry-Perot laser diode.
In this demonstration of one output port of the 1×N optical switch, a data packet indicates its intended output port at a node by setting the corresponding address bit in the header to ‘0’ and the rest of the address bits to ‘1’s. The input control signal header contains the complement of the address, i.e., the bit in the control signal header corresponding to the output port is set to ‘1’, all other bits in the control signal header are set to ‘0’. With this arrangement, when the address the data packet matches with that of the input control signal, the FP-LD will be injection-locked at wavelength λd and will therefore transmit the data packet. Otherwise, the FP-LD will be injection-locked at wavelength λc and blocks the data packet. For simplicity, it is assumed in the following that the header of a data packet contains only the address of an output port of a single node. In the embodiment, four different data packets are encoded with different header bits which indicate four different output ports. The data packet header is therefore only 4 bits long. The bit period at the header is 200 ps long which means the header rate is 5 Gb/s. The packet headers of the four different packets are arranged in the order of ‘0111’, ‘1011’, ‘1101’, and ‘1110’ for pk_1 to pk_4 respectively. Thus the data packets are arranged such that they are intended for output port 1 to 4 in consecutive order. The data packet payload is 48 bits long. The bit period at the payload is 100 ps long corresponding to a payload rate of 10 Gb/s. The guard period is 800 ps long. The payload length is chosen for convenience and in principle can be hundreds of thousands of bits long. The guard period is limited by the rise-time and fall-time of the control signal and can be further reduced. The header of the control signal is ‘0100’ which corresponds to the complement of the address of output port 2 of the 1×4 switch. When the header of the data packet matches with that of the input control signal, for example, pk_2 only, the FP-LD transmits the data packets and otherwise blocks the packets. However, because of the finite response time of the FP-LD, part of the address headers of the blocked data packets, i.e., pk_1, pk_3, and pk_4, is able to pass through the FP-LD before injection locking by the control signal header at wavelength λc can take place. The switching ratio is about 7.5 dB measured in the time domain with zero level at around −20 dBm.
The functions of the add/drop node are (i) to add a packet to the network (denoted by thick solid line in
The node is constructed with five all-optical logic devices. HPU serves as an all-optical header processor. INV functions as a signal inverter, i.e. the output of INV will be high if the input is low and vice versa. SW1, SW2, and SW3 are simple all-optical on/off switches. SW1 controls whether a transit packet in the ring is allowed to continue on in the ring, SW2 determines whether a packet in the ring can be sent to the local drop port, and SW3 determines whether a packet from the add port of the node can be placed in the ring.
The exemplary node operates as follows. If the address header of a transit data packet in the ring entering the node through the transit port matches with the address of the node carried by the control signal input to HPU, HPU will generate a control signal that will set SW1 to block the packet while setting SW2 and SW3 to transmit. As a result, the data packet is removed from the ring and passed to the drop port. Thus the packet drop function of the node is achieved. At the same time SW3 allows a packet from the local add port, if any, to be placed in the ring to fill the vacated time slot. If the address header of the transit data packet in the ring does not match with that of the node, SW1 is set to transmit while SW2 and SW3 are set to block. In this case, the packet in the ring is transmitted to the next node while no packet is allowed to be added to the ring.
In order to avoid differentiating an empty time slot from an occupied time slot, the address of empty packets is defined such that they are accepted by all the nodes in the network. In addition, a node will continue to transmit empty packets even if it has nothing to send. As a result, each node continuously accepts empty packets from the node upstream and sends empty packets to the node downstream even if there are no user data packets in the network. By this construction, the task of separating empty slots from those containing user information sent from other nodes is pushed to the local node, thus relieving the optical signal processing requirement of the proposed add/drop node. The operation of the proposed all-optical add/drop node is summarized in Table 1.
Various alternatives can be made to the embodiment(s) described above. For example, other types of laser devices can be used to replace the Fabry-Perot laser diode. Preferably, such laser devices is also a multimode laser and can work in bistable states. When such laser device works as a switch or header processor, two or three external optical signals at different wavelengths are injected into the laser device. The laser device can be injection-locked by one of the externally injected optical signals, and the processing is achieved when the wavelength at which the laser device is injection locked is changed from one wavelength to another or when the laser diode changes from one of the locked and unlocked states at a particular wavelength to the other, preferably at the wavelength of the input control signal. It could be understood in the art that the processing can also be achieved when the laser diode changes from one of the locked and unlocked states at the wavelength of the input data signal. In the scenario, when the laser diode is in the locked state at the wavelength of the input data signal, the intensity of output data signal may experience a significant gain; whereas when the laser diode is in the unlocked state at the wavelength of the input data signal, the output data signal may be suppressed to a relatively significant extent.
In addition, the input control signal may have various formats, and
Furthermore, the powers and wavelength detunes of both the control signal and the data signal can be chosen to determine at which situation—whether the address bit of the data packet matches or does not match with the trigger of the control signal—the laser device should transmit or block the data packet. For example, the address bit can be implemented by the absence of optical intensity in the corresponding bit window. In such a scenario, the output at λc is switched off if the trigger matches the address bit.