Wing Ko

The full adder (FA) and full subtractor (FS) are key operations used in the arithmetic and logic units of digital processors. Performing the FA and FS optically may enable the direct line-rate processing of high-speed optical data. Previous demonstrations of optical FA and FS have not been shown for phase-encoded input and output data signals. Such phase-encoded data formats could offer improved noise tolerances and longer distances of transmission compared to amplitude-encoded data formats. In this work, we experimentally demonstrate an optical FA and FS for binary-phase-shift-keying (BPSK) data signals at 2.5- and 5-Gbit/s. Our approach utilizes input BPSK data signals and continuous-wave pumps that are encoded at different frequencies, in which one of the input BPSK data signals is negated for the FS case. The data and pumps undergo nonlinear wave mixing in a periodically poled lithium niobate (PPLN) waveguide to generate the FA or FS BPSK outputs at different frequencies. The experimental results show error-free logic operation for both the FA and FS outputs at 2.5- and 5-Gbit/s over 1536 captured bits.

An optical-tapped-delay-line (OTDL) can implement various processing functions on data that is transmitted over an optical network, including format conversion, complex filtering, and pattern recognition. Previous demonstrations of OTDL functions have either processed: (i) one data channel with one or more OTDL functions, or (ii) multiple data channels with one OTDL function per channel. In this work, we experimentally demonstrate three OTDL functions simultaneously on two optical data channels to obtain six optical data outputs. The inputs include: (i) two quadrature-phase-shift-keying (QPSK) channels, and (ii) two on-off-keying (OOK) channels. In each case, delayed data copies and pumps encoded with the OTDL function weights undergo nonlinear wave mixing in a periodically poled lithium niobate (PPLN) waveguide. Six optical data outputs are simultaneously generated at six different frequencies. In the QPSK case, the average symbol correctness over 3072 symbols is: (i) ~99.5% at 7.5-Gbaud and ~98.8% at 10-Gbaud for format conversion, (ii) ~98.1% at 7.5-Gbaud and ~94.3% at 10-Gbaud for complex filtering, and (iii) 100% at both baud rates for pattern recognition. In the OOK case, the average symbol correctness at 7.5-Gbaud is ~99.8% for format conversion, ~99.9% for complex filtering, and 100% for pattern recognition.

Free-space optical (FSO) communication in the mid-infrared (mid-IR) region has attracted increasing interest due to its lower atmospheric attenuation compared to near-infrared (near-IR) wavelengths. To further enhance link capacity, mode-division multiplexing (MDM) can be employed to transmit multiple orthogonal spatial channels simultaneously. Previous schemes for mid-IR MDM have converted Gaussian data beams from the near IR into the mid-IR before changing the beams into orthogonal modes. However, such a scheme employs a separate nonlinear wavelength converter for each mode (i.e., N converters for N channels), potentially causing high system complexity. In this paper, we demonstrate using a single wavelength converter from near-IR to achieve a mid-IR mode-division multiplexing (MDM) link. Two orbital angular momentum (OAM) channels with tunable mode orders are generated using spatial light modulators (SLMs) at 1550 nm. The channels are multiplexed and simultaneously converted to 3.4 μm via difference frequency generation (DFG) in a single PPLN crystal. We demonstrate mid-IR intensity modulation/direct detection (IM/DD) and coherent heterodyne communication. A total of 2-Gbit/s on-off keying (OOK) and 4-Gbit/s 16-quadrature amplitude modulation (QAM) MDM links are achieved, all with bit error ratios (BERs) below the 7% forward error correction (FEC) threshold.

We experimentally demonstrate a >Tbit/s mid-infrared (mid-IR) free-space optical (FSO) communication system by combining wavelength-division multiplexing (WDM), polarization-division multiplexing (PDM), and mode-division multiplexing (MDM). We transmit 16 multiplexed data channels simultaneously by employing 4 wavelengths at ∼3.4 μm, 2 polarizations, and 2 orbital angular momentum (OAM) modes with orders of ℓ = +1 and ℓ = +3. The channels are generated in the C-band at the transmitter and converted to mid-IR through difference frequency generation (DFG) nonlinear process. At the receiver, they are converted back to C-band through DFG and detected in the C-band. With each channel carrying a 32-Gbaud QPSK signal, a total data rate of 1.024 Tbit/s is achieved. Our experimental results show that (i) the channel crosstalk between different wavelengths, polarizations, and modes are less than -19 dB, -16 dB, and -15 dB, respectively, (ii) bit-error rates of all multiplexed channels can reach below the 20% forward error correction threshold, and (iii) there is <1-dB optical signal-to-noise ratio (OSNR) penalty induced by channel multiplexing in a single domain (WDM, PDM, or MDM alone), while the OSNR penalty increases to ∼2 dB when combining the three multiplexing schemes.

Optical signal processing can provide high-speed operation on optical data and avoidance of optical-to-electrical-to-optical conversion. In many networking scenarios, the optical data may be encoded as quadrature-phase-shift-keying (QPSK) for spectral efficiency and noise tolerance. One signal processing function is pattern matching between two data streams, which directly compares their data patterns and can be useful for network security and improving data transmission. In this work, we experimentally demonstrate optical pattern matching between two independent QPSK data inputs using nonlinear wave mixing. The QPSK channels and their delayed copies are combined at the input of a highly nonlinear fiber (HNLF) along with a continuous wave pump. The data channels and pump undergo four-wave mixing in the HNLF to generate a 9-ary quadrature amplitude modulation (9-QAM) output. The output symbols correspond to the matched, partially matched, and mismatched 2-symbol patterns between the two QPSK inputs. The 9-QAM output constellation has an error-vector-magnitude of 8.4% and 8.9% at 3-Gbaud and 5-Gbaud rates, respectively, resulting in error-free optical pattern matching between the two QPSK inputs over 1536 symbols.

Optical communication networks employ various modulation formats depending on the performance capabilities and desired functionalities of their different zones. Different modulation formats may be used in different parts of a network, such that an edge node may aggregate multiple channels of lower-order formats into a single channel of higher-order format. In this work, we experimentally demonstrate the optical data aggregation of two independent channels into a single higher-order format channel using nonlinear wave mixing. By transmitting each data channel with two coherent pilot tones, we apply complex weights to the input data channels and shift the output constellation by a bias vector. These functions are achieved in a periodically poled lithium niobate (PPLN) waveguide. We show that our system is format reconfigurable, allowing for both amplitude- and phase-encoded inputs and that the two inputs can be in different formats. Moreover, we demonstrate the tunability of our approach by performing aggregation at two separate baud rates, 3 and 5 Gbaud. Bit error rate (BER) measurements for our cases show a power penalty is < = 1 dB for the aggregated quadrature phase shift keying (QPSK) output. Overall, we have shown the following format aggregations: (i) two on-off keying (OOK) inputs to one QPSK, (ii) two binary phase shift keying (BPSK) inputs to one QPSK, (iii) one OOK and one BPSK inputs to one QPSK, (iv) two 4-level pulse amplitude modulations (PAM4) inputs to one 16-ary quadrature amplitude modulation (16QAM), (v) two QPSK inputs to one 16QAM, and (vi) one QPSK and one 16QAM inputs to one 64QAM.

Optical signal processing (OSP) has the potential to enable line-rate processing functionalities in optical communication networks through direct operation on optical data and avoidance of optical-to-electrical-to-optical (OEO) conversion. One such functionality is optical pattern recognition, in which the location of target patterns in the transmitted optical data stream is identified; common optical data formats include quadrature-phase-shift-keying (QPSK) data. Whereas prior approaches have performed optical pattern recognition to locate a single pattern in a QPSK optical data stream, it might be advantageous to optically recognize multiple patterns simultaneously. In this paper, we experimentally demonstrate optical recognition of multiple QPSK patterns using two approaches based on nonlinear wave mixing, one using time multiplexing and one using wavelength multiplexing. In the intra-symbol time multiplexing (IS-TM) approach: (i) the multiple target patterns are encoded on the phases of auxiliary patterns at different time instances, and (ii) the nonlinear wave mixing generates a single output representing the multiple pattern occurrences at different instances. Using IS-TM, we recognize (i) 2 patterns with 2 or 3 symbols for 2- and 3-Gbaud rates; and (ii) 3 patterns with 2 or 3 symbols for 1-Gbaud input QPSK signals. In the wavelength multiplexing approach: (i) the multiple target patterns are encoded on the phases of separate continuous wave (CW) pumps at different wavelengths, and (ii) the nonlinear wave mixing generates multiple outputs at different wavelengths corresponding to different pattern recognitions. Using wavelength multiplexing, we recognize (i) 1 pattern with 2 or 3 symbols at a 15-Gbaud rate and (ii) 2 patterns with 2 or 3 symbols at 15-Gbaud and 20-Gbaud input QPSK data. Both approaches have error-free correlation results across 1536 transmitted QPSK symbols.

Recent interest has grown in processing matrix operations within the optical domain due to its potential advantages in bandwidth, latency, and parallelism. While optical implementations of matrix-vector multiplication (MVM) have been demonstrated, they still partially rely on the electronic domain to perform the matrix operation. In this paper, we demonstrate MVM using nonlinear wave mixing, enabling the matrix operation to be performed in the optical domain. Each column of the matrix is modulated onto different subcarriers, in which the column elements are serialized in time. The vector elements are encoded onto separate CW tones at different wavelengths. The MVM operation is carried out using nonlinear wave mixing in a single periodically poled lithium niobate waveguide, in which the elements of the output vector are serialized in time. We show MVM at two different clock rates (3 and 5 GHz) and for different input vectors. The resulting output vectors have error standard deviations of 2.89–3.55%, corresponding to 4.8–5.1 bits of precision.