This letter presents the development of a dc–20 GHz impedance tuner in a 45-nm silicon-on-insulator complementary metal–oxide–semiconductor (CMOS) process. The proposed tuner consists of several short transmission line sections and tunable capacitors made up of small interdigital capacitors and switches. To increase the power handling capability, a switch topology using triple-stacked transistors and parallel capacitors is developed. The combination of the tuner architecture and the capacitor-switch arrangement enables the new tuner to have wideband operation, good tuning range, low loss, and high-power handling. The fabricated tuner can operate with RF power up to 27 dBm and provide a characteristic impedance tuning range from 56 to $36~\Omega $ over dc–20-GHz bandwidth. In different operating states, the tuner demonstrates an insertion loss of 0.25–1.25 dB, an output 1-dB compression point (OP1dB) of better than 24 dBm, and an input third-order intercept point (IIP3) of better than 40 dBm.

In this paper, we present a wideband DC-67 GHz single-pole-triple-throw (SP3T) switch using a series-shunt topology on 45-nm silicon-on-insulator (SOI) CMOS process from global foundries. For wideband operation, series inductors are implemented to resonate out the parasitic capacitance at drain and source terminals of NMOS transistors used in the switch. Additionally, for shunt branches, a double stack-FET topology is used to handle higher power levels. In all, the switch has an insertion loss from 1.42 dB to 2.62 dB over frequency range from 10 MHz to 67 GHz. The measured input 1-dB compression power (IP1dB) at 20 GHz exceeds 10 dBm while the third order input intercept point (IIP3) at 20 GHz is 20.8 dBm.

This letter demonstrates the application of the second-harmonic injection linearization technique in a field-effect transistor (FET) resistive mixer operating at $Ka$ -band. The single-ended mixer uses a 0.15-$\mu \text{m}$ gallium arsenide (GaAs) pseudomorphic high electron-mobility transistor (pHEMT) at its core. Compared to the same mixer without the second-harmonic injection, the linearized mixer achieves an improvement of up to 8 dB in the third-order intermodulation (IM3) levels while leaving the fundamental output power and noise figure unaffected. To the best of authors knowledge, this is the first demonstration of the technique with a mixer operating at millimeter-wave frequencies.

In this paper, for the first time we develop and benchmark the performance of three down-converting Field Effect Transistor (FE T) resistive mixers at millimeter wave (mm W) frequencies employing \pmb0.15-μm enhancement (E)-mode Gallium Arsenide (GaAs) pseudomorphic high electron mobility transistors (pHEMTs), depletion (D)-mode GaAs pHEMTs, and D-mode Gallium Nitride (GaN) pHEMTs. Our experimental results at 27 GHz demonstrate that the E-mode mixer achieves the highest reported input 3 rd order intercept point (IIP3) of 37.5dBm at mm W frequencies to the best of our knowledge. Also, operating at the same LO drive up to 20 dBm, the E-mode GaAs mixer impressively out performs both the D-mode GaAs and GaN mixers.

This paper demonstrates rapidly reconfigurable, high-fidelity optical arbitrary waveform generation (OAWG) in a heterogeneous photonic integrated circuit (PIC). The heterogeneous PIC combines advantages of high-speed indium phosphide (InP) modulators and low-loss, high-contrast silicon nitride (Si3N4) arrayed waveguide gratings (AWGs) so that high-fidelity optical waveform syntheses with rapid waveform updates are possible. The generated optical waveforms spanned a 160 GHz spectral bandwidth starting from an optical frequency comb consisting of eight comb lines separated by 20 GHz channel spacing. The Error Vector Magnitude (EVM) values of the generated waveforms were approximately 16.4%. The OAWG module can rapidly and arbitrarily reconfigure waveforms upon every pulse arriving at 2 ns repetition time. The result of this work indicates the feasibility of truly dynamic optical arbitrary waveform generation where the reconfiguration rate or the modulator bandwidth must exceed the channel spacing of the AWG and the optical frequency comb.

A very wideband GaAs MMIC Schottky diode frequency doubler has been designed and tested; it has at least a 10-75 GHz output frequency range with indications it works to 90 GHz. Measured in a test fixture with 1.85 mm connectors over an output frequency of 10-67 GHz, it has 9 to 15 dB conversion loss, -15 to -25 dBc fundamental suppression, -22 to -35 dBc third harmonic suppression, and 15 to 20 dB input return loss, all including interconnect and connector losses. Additionally, calibrated wafer probe tests show it has under 16 dB conversion loss up to 75 GHz, and qualitative wafer probe tests indicates it works up to 90 GHz, in good agreement with simulated results. It is singly balanced with a broadside coupled 4: 1 transmission line transformer that provides wideband impedance match and input ground return. This appears to be the widest operating bandwidth reported for a GaAs MMIC diode frequency doubler. A family of frequency doublers and mixers has been designed and built that use this new circuit topology.

Optical arbitrary waveform generation (OAWG) is capable of working in the `line by line' regime to control arbitrary amplitude and phase over individual lines of an optical frequency comb (OFC). To accomplish this, the OAWG waveform shaper separates the optical comb lines from the OFC using a spectral demultiplexer, applies modulation to each line with an array of modulators, and combines the modulated comb lines using a spectral multiplexer. User-defined complex spectral shapes and arbitrary temporal waveforms can be synthesized on demand, thereby facilitating applications like microwave arbitrary waveform generation, agile microwave-photonic filtering, and coherent communications, etc.. The main limitation of current integrated solutions has been high optical insertion loss and a large degree of crosstalk between channels. Instead of attempting a monolithic technology platform, we implement low loss and high contrast Si 3 N 4 arrayed waveguide gratings (AWGs) as spectral de/multiplexer, and InP phase modulator array for nearly pure phase modulation. We heterogeneously integrate Si 3 N 4 and InP chips by active alignment and UV epoxy assembly, and demonstrate dynamic OAWG in a single integrated chipset.

We demonstrate static and dynamic optical arbitrary waveform generation in a heterogeneously integrated Si3N4 AWG-InP modulator array-Si3N4 AWG chip-scale module. High contrast Si3N4 AWGs and nearly pure phase InP modulation provided high fidelity OAWG.

This paper investigates energy savings in data centers with flexible-bandwidth power-aware source-synchronous optical interconnects. Network simulations show ≥ 5× energy savings. Link experiment shows error-free operation from 625 Mb/s to 10 Gb/s.

This paper proposes and demonstrates a flexible-bandwidth optical interconnect architecture for data centers exploiting wavelength routing in arrayed waveguide grating routers and fast tunable lasers. The proposed architecture provides hierarchical all-to-all connectivity with low contention and dynamic interconnection reconfiguration for higher bandwidth provisioning between hot spots. An eight-cluster core network experiment testbed with hierarchical all-to-all interconnection shows 1.77× throughput increase and 1.19× network energy efficiency improvement in the case of intercluster hot-spot traffic, while guaranteeing more than 97% throughput for the portion of the traffic with uniform random distribution.

For the first time, we experimentally demonstrate rapid and flexible bandwidth adjustment in an optically interconnected data center network with hot spots. Experiments show 1.77× throughput increase for hotspot links and 23% improvement in energy efficiency for the entire network.

We designed and experimentally demonstrated a scalable datacenter architecture with intra-cluster all-to-all topology and flat inter-cluster interconnection based on wavelength routing in AWGRs. Experiments show 97% intra-cluster throughput for uniform random traffic, and error-free inter-cluster communication at 10 Gb/s.