1. Introduction

Since optical code (OC) label [1–4] can be easily identified by the encoding and decoding method, and its en/decoder, such as the super structured fiber Bragg gratings (SSFBG) en/decoder [5], the passive planar full en/decoder [6] and the multiple label decoder based on four-wave mixing (FWM) [7], becomes more and more powerful, it is considered as one of the most promising candidates for label processing in optical packet switching (OPS) networks [8]. Recently OC label has been employed in packet-based optical multicasting [9] so as to provide scalable multicast networking. However, as each OC label can only offer 1 bit of information (matched or unmatched) during the decoding process, it still has some limitations for this application. In a multicast network, each OC label denotes a multicast group (i.e., all the destination nodes of a multicast service). Since a multicast group will expire when its corresponding service is finished, at each node the fast tunable OC label decoders are needed and the multicast forwarding table also needs to be updated frequently, both of which will heavily increase the node’s complexity. In [10], OC labels are multiplexed in the time domain by delay lines and couplers, and then multi-bit information can be carried by these multiplexed labels. In this paper, a stacked OC label, which can integrate multi-bit information into a single label, is proposed and employed in optical multicasting to simplify the node’s structure. With our fiber Bragg gratings (FBG) fabricating techniques [11], the en/decoder of the stacked label is implemented by a single FBG with high label processing speed. The rest of the paper is organized as follows: in Section 2, the stacked OC label en/decoder based on FBG is explained and constructed. Then in Section 3 an experimental setup of five nodes is used to validate the stacked label’s practicability for optical multicasting. Finally Section 4 concludes this paper.

2. Stacked OC label en/decoder based on FBG

The principle of constructing an FBG-based stacked OC label en/decoder can be explained as below. Firstly, assume Φ is the code space of M orthogonal N-chip bipolar optical codes, which can represent M nodes’ addresses in an OPS network, and ∀C _i,C _j ∈ Φ, the maximum of each chip’s absolute value in the convolution result of C _i*C _j ^* satisfies:

where C _j ^* is the reverse of C _j, λ is the cross-correlation limit and λ ≪ N. Suppose Φ _s is an arbitrary proper subset of Φ, then a new stacked code C_s can be constructed as below:

In the following, C _k (C _k ∈ Φ _s) will be called as the basic code of C_s. Based on (1), (2), we obtain:

where m=|Φ _s|. It should be noted that C_s is a multi-level code and does not belong to Φ, and (3) shows that due to this multi-level characteristic, C_s can be quasi-matched with all the codes in Φ _s simultaneously and be nearly orthogonal to the other codes out of Φ _s as well. Assume C_s=[c₁, c₂…c_N], then the refractive index’s spatial modulation function of the FBG-based stacked en/decoder can be given:

where Λ is the period of the grating, Z₀ is the chip period and A(z) is the profile of each chip’s amplitude and satisfies that |A(z)|=0 (z<0 or z>Z₀/2), i.e., the chip’s duty cycle is below 50%. If |A(z)| is very small (e.g., less than 2e-4) when 0≤z≤Z₀/2, the en/decoder’s impulse response can be approximately given by [12]:

where K is a constant coefficient, n_e is the fiber’s effective refractive index and c is the speed of light. In practice, a short optical pulse x(t) is first launched into an encoder to generate an OC-label, and then the label is decoded by a decoder for recognition. Suppose the impulse response of the encoder and decoder are h_e(t) and h_d(t), and C_s and C_i ^* (i=1, 2, … M) are the encoder’s and decoder’s code sequence respectively. Then based on (4) and (5), the input optical short pulse x(t) and the decoded signal y(t) have the relation:

where $B\left(t\right)=x\left(t\right)*A\left(\frac{c}{2n_e}t\right)*A\left(\frac{c}{2n_e}t\right)$ and D=[d ₁,d ₂,…d _2N-1]=C_s*C _i ^*. Since x(f) is very short, it can be approximately regarded that |B(t)|=0 (t<0 or t>2n_eZ₀/c), therefore,

where $Y_{max}=\underset{t}{\max }(\mid K^{2}B\left(t\right)\cos \left(\frac{2\mathrm{ct}}{\Lambda n_e}\right)\mid ).$. Then based on (3) and (7), we obtain:

Since λ ≪ N , for a moderate m, we have: (N −(m−1)l)Y _max ≪ mλ Y _max . So (8) denotes that a quasi-auto-correlation peak can be get in y(t) if and only if the decoder is coded by C _i ^*(C _i ∈ Φ _s). Assume each node k(k=1, 2, … M) of the OPS network is equipped with a decoder coded by C _k ^* to denote its own address, then the label encoded by C_s, which will be called as stacked label in the following, can simultaneously carry the address information of m nodes, and only the decoders of those m nodes can recognize this stacked label. So with this stacked label, multicasting to those m nodes can be realized. When the ratio of the quasi-auto-correlation peak to the cross-correlation peak is given, m can be increased by optimizing the structure of the code space Φ or increasing the code length N.

Since bipolar multi-level code sequence is introduced in (4), besides high accurate phase control, precise amplitude control is also needed during the fabrication of the stacked en/decoder. Both of these two demands can be met with our scan-exposure technique. In our fabricating process, the UV laser (Coherent’s frequency-doubled argon laser) is focused so that its FWHM is about 70μm. A uniform phase mask is adopted. The whole label is written on a single FBG in the form of superimposition and only once scan-exposure is needed, so the phase error can be minimized, and the en/decoder can have a very compact structure and high label processing speed. The relation between the FBG’s refractive index modulation amplitude and the exposal time is first calibrated. With the help of this relation, a 31-chip, 200-Gchip/s stacked encoder with the code OC-A+B+C and four normal 31-chip, 200-Gchip/s decoders with the code OC-A, OC-B, OC-C and OC-D are fabricated, where OC-A, B, C, D are four gold sequences. Figure 1 gives the measured and calculated reflection spectrum of the stacked encoder OC-A+B+C, which contains bipolar three-level code sequence. It can be seen that very good agreement is get, so the stacked en/decoder can be fabricated with very high precision.

 

Fig. 1. Measured and calculated reflection spectrum of the stacked encoder OC-A+B+C

Download Full Size | PDF

3. Experiment and results

 

Fig. 2. The experimental setup for evaluating the stacked OC label’s performance in a multicast network

Download Full Size | PDF

The setup given in Fig. 2 is used to evaluate the stacked label’s performance in a multicast network. It consists of a source node and four destination nodes: A, B, C and D. Node A, B and C compose a multicast group. In the source node, firstly, Calmar Optcom’s Pico-second Fiber Laser generates a 10-GHz, 2-ps-wide optical pulse train at 1552 nm (since the chip speed of the en/decoder is 200-Gchip/s, the required bandwidth of the OC label is about 200GHz, so the bandwidth of the 2-ps optical pulse is wide enough to support the OC label’s en/decoding process). Then the pulses’ repetition rate is reduced through a LiNO₃ intensity modulator (IM) that controlled by a pulse pattern generator and a microwave phase shifter, so as to guarantee that adjacent optical pulses’ decoded results will not interfere with each other at the destination nodes. The output of IM is launched into an erbium-doped fiber amplifier (EDFA), a circulator and an FBG encoder with code OC-A+B+C, to generate a stacked label that denotes the multicast group. Since we only focus on the stacked label’s recognition performance, payload is omitted at the source node for simplification. Then the output packet (i.e., the generated stacked label with an empty payload) is sent to the four destination nodes, each of which has a label recognition processor (LRP) that consists of an EDFA, a circulator and an FBG decoder. Finally the LRPs’ output results are observed by a digital sample oscilloscope (DSO).

 

Fig. 3. The LRP’s output results at the destination nodes (50 ps/div)

Download Full Size | PDF

The LRP’s output results at each destination node are given in Fig. 3. At Node A, B and C, a quasi-auto-correlation peak can be obtained since the decoder’s code matches with one of the stacked label’s basic code; this quasi-auto-correlation peak can be used to open an optical switch placed after the LRP, so as to route the corresponding packet to the node’s local network. While at Node D, only cross-correlation floor is get for none of the stacked label’s basic codes matches with OC-D, then the packet will be discarded since the cross-correlation floor cannot change the following optical switch’s state. Therefore only the nodes in the multicast group can accept the stacked label’s packet and optical multicasting is realized in our setup. From Fig. 3, it can also be seen that the time needed for label recognition at each node is less than 310 ps, which is fast enough in practice since the packet’s length in OPS is at the order of tens to hundreds of nanoseconds; Besides, for all the destination nodes, the ratio of quasi-auto-correlation peak to cross-correlation peak is about 8 dB, which is high enough to control an optical flip-flop switch such as that in [13]. Finally, it can also be noted that if the multicast group changes, we only need to generate a new stacked label at the source node, and all the LRPs at the destination nodes can be kept unchanged, so the node’s structure is simplified a lot. All the results above show that the stacked label is practical for optical multicasting application.

4. Conclusion

We have proposed and experimentally demonstrated the stacked OC label and its en/decoder based on FBG. The en/decoders have compact structure and fast processing speed, and can be constructed with high precision by our FBG fabricating techniques. The experimental results showed that with the stacked OC label, optical multicasting can be well supported without tunable decoder or updating multicast forward table at each node, so the node’s structure can be simplified a lot.