24 February 2015 in IEEE Transactions on Microwave Theory and Techniques
An 8-bit 100-GS/s digital-to-analog converter (DAC) using a distributed output topology in 28-nm low-power CMOS for optical communications is presented. The DAC can convert 1-k symbols stored in the 1-kbyte design-for-test on-chip memory cyclically. By interleaving four 25-GS/s return-to-zero DACs, the highest signal frequency of the 100-GS/s DAC is about 25 GHz and the output image is located beyond 75 GHz. The 3-dB bandwidth exceeds 13 GHz at 100 GS/s. The effective number of bits and spurious-free dynamic range ranges from 5.3 bit and 41 dB to 3.2 bit and 27 dB from dc up to 24.9 GHz at 100 GS/s, respectively. Transmission rates of 120 and 45 Gb/s are obtained in an electrical and an optical back-to-back experiment, respectively. The DAC test chip consumes 2.5 W from a power supply with multiple outputs of 1.1, 1.5, and 2 V.
17 October 2023 in IEEE Journal of Solid-State Circuits
To enhance sampling rates of CMOS digital-to-analog converters (DACs), analog multiplexing of several DAC output signals in the time domain provides a solution. In this article, a full CMOS integration of two sub-DACs and an active analog multiplexer (AMUX) on a single chip in 28-nm fully depleted silicon-on-insulator (FD-SOI) CMOS technology is presented for the first time for sampling rates of 100 GS/s and beyond. Sampling rates up to 108 GS/s for broadband pulse-amplitude modulated (PAM) signals and up to 118 GS/s for oversampled signals are shown outperforming previously reported data in terms of sampling rate or data rate, respectively. Two 8-bit sub-DACs up to 59 GS/s with CMOS inverter-based output drivers and pseudo-segmentation provide the analog input data for the 2:1 AMUX realized in current-mode topology. An additional on-chip memory of 256 kB completes the system to a universal arbitrary waveform generator (AWG). At 100 GS/s, the total power consumption is about 4 W. Generally, an AMUX is able to shift the limits of DACs in well-established CMOS technologies toward higher frequencies independent on technology advances and opens a second, conceptual path for achieving higher sampling rates with an additional benefit of a principle bandwidth extension by the AMUX operation.
2 May 2023 in IEEE Transactions on Circuits and Systems I: Regular Papers
Mixed-signal AI accelerators offer the possibility of higher energy efficiency for moderate resolution computations compared to their digital counterparts. All-analog implementations, where all operations are performed in the analog domain, can further improve this energy advantage. An energy efficient multiply-accumulate cell for all-analog neural layer processing macros is presented. The proposed analog two-quadrant multiplier circuit consists of two complementary MOSFETs where the pulse width modulated input activation is applied to the gates and the weight signal to the isolated back-gate. The analog multi-bit resolution weight is dynamically stored on a memory capacitor. The multiply-accumulate operation result is represented by charge accumulated on a summation line and drawn from or put onto a computation capacitance. Simulation results based on a 22 nm FD-SOI CMOS technology show that the cell consumes about 0.67 fJ for a circuit-level multiply-accumulate operation. An area efficiency of 166 $\times$ 10 $^{12}$ MAC/s/mm $^2$ is achieved.