Analog and Mixed Signal Center, Electrical and Computer Engineering Department, Texas A&M University, College Station, TX 77843 USA. He is now with Apple Inc., Cupertino, CA 95014 USA
Auto-generated profile, based on publicly available information
Shengchang Cai
Analog and Mixed Signal Center, Electrical and Computer Engineering Department, Texas A&M University, College Station, TX 77843 USA. He is now with Apple Inc., Cupertino, CA 95014 USA
Increased bandwidth per pin is necessary to support the growing data throughput in mobile systems. This article presents a low-power simultaneous bidirectional (SBD) transceiver employing a novel voltage-mode driver with a resistor-transconductance (R-gm) hybrid subtractor to enable concurrent transmission and reception of data on a single differential channel. A finite impulse response (FIR) filter-based adaptive echo cancellation system that cancels both near- and far-end echoes allows for the support of a wide range of channels, while a continuous-time linear equalizer (CTLE) efficiently compensates for channel loss. The quarter-rate source-synchronous transceiver utilizes a low-complexity phase interpolator-based clock and data alignment system to set the optimum data sampling point. Fabricated in 28-nm CMOS, the 32-Gb/s SBD transceiver achieves 1.83 mW/Gb/s and compensates 10.2-dB loss at a bit error rate (BER) <10⁻¹².
Wireline input/output (I/O) bandwidth demand within networks in large data centers has increased rapidly over the last decade because of the explosion in data generation from cloud computing, mobile devices, and Internet of Things applications. To support this, both the number of I/O lanes per component and the per-lane data rate have continuously scaled up. For example, Figure 1 shows that the aggregate bandwidth of a typical large Ethernet switch application-specified integrated circuit (IC) has increased by an order of magnitude roughly every five years, starting at about 240 Gb/s with 24 channels operating at 10 Gb/s in 2008 and projected to near 25.6 Tb/s with 256 channels operating close to 100 Gb/s in 2019 [1]. This has led to the development of the current serial I/O standards at 56 Gb/s and planned future links operating in excess of 100 Gb/s.
A reference switching pre‐emphasis (RSP) technique is proposed for successive approximation register (SAR) ADC in order to improve the SAR DAC settling time, which limits the conversion speed of conventional SAR ADCs. While SAR ADCs that employ redundancy techniques relax DAC settling requirements by overlapping search spaces, this requires more conversion steps which offsets the benefit from the DAC settling relaxation, especially for low–medium resolution SAR ADCs. The proposed RSP technique improves DAC settling by breaking its dependence on ADC resolution and achieves up to 71% speedup of DAC settling relative to a conventional DAC switching scheme without introducing additional conversion steps or significant hardware overhead.