John M. Wilson
Research Subjects
John M. Wilson has a total of 21 co-authors. Here are the top co-authors by number of shared publications:
- C. Thomas Gray
NVIDIA Corporation
4 shared publications Stephen G. Tell
NVIDIA, Inc., Durham, NC, USA
4 shared publicationsJohn W. Poulton
Circuits Research, NVIDIA, Inc., Durham, North Carolina United States
3 shared publicationsNikola Nedovic
NVIDIA, Inc., Santa Clara, CA, USA
3 shared publicationsSanquan Song
NVIDIA, Inc., Santa Clara, CA, USA
3 shared publications- Walker J. Turner
NVIDIA, Inc., Durham, NC, USA
3 shared publications
Latest Publications
- 13 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 59 (4) , 1146-1157
A 0.190-pJ/bit 25.2-Gb/s/wire Inverter-Based AC-Coupled Transceiver for Short-Reach Die-to-Die Interfaces in 5-nm CMOS
This article presents an inverter-based short-reach ac-coupled toggle (ISR-ACT) link targeted for short-reach die-to-die communication over silicon interposer or similar high-density interconnect. The ISR-ACT’s transmitter (TX) sends non-return-to-zero (NRZ) data through a small series capacitor to inject low-swing pulses into the line. These pulses are amplified and latched by a two-stage receiver (RX), where the 1st-stage transimpedance amplifier (TIA) amplifies the pulse data and positive feedback around both stages captures the data and maintains the dc level on the line. Achieving low signal swing through a capacitor divider, the ACT driver consumes less than half the power compared to regular rail-to-rail CMOS drivers while providing dc voltage isolation between the driver and the RX. Fabricated in a 5-nm standard CMOS process, the ISR-ACT link, operating on a 0.75-V supply, shows 0.66 UI margin at 25.2 Gb/s/wire over a 1.2-mm on-chip channel and demonstrates an energy efficiency of 0.190 pJ/bit.
- 13 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 59 (4) , 1146-1157
A 0.190-pJ/bit 25.2-Gb/s/wire Inverter-Based AC-Coupled Transceiver for Short-Reach Die-to-Die Interfaces in 5-nm CMOS
This article presents an inverter-based short-reach ac-coupled toggle (ISR-ACT) link targeted for short-reach die-to-die communication over silicon interposer or similar high-density interconnect. The ISR-ACT’s transmitter (TX) sends non-return-to-zero (NRZ) data through a small series capacitor to inject low-swing pulses into the line. These pulses are amplified and latched by a two-stage receiver (RX), where the 1st-stage transimpedance amplifier (TIA) amplifies the pulse data and positive feedback around both stages captures the data and maintains the dc level on the line. Achieving low signal swing through a capacitor divider, the ACT driver consumes less than half the power compared to regular rail-to-rail CMOS drivers while providing dc voltage isolation between the driver and the RX. Fabricated in a 5-nm standard CMOS process, the ISR-ACT link, operating on a 0.75-V supply, shows 0.66 UI margin at 25.2 Gb/s/wire over a 1.2-mm on-chip channel and demonstrates an energy efficiency of 0.190 pJ/bit.
- Top Cited Papers
- 9 January 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 58 (4) , 1062-1073
A 0.297-pJ/Bit 50.4-Gb/s/Wire Inverter-Based Short-Reach Simultaneous Bi-Directional Transceiver for Die-to-Die Interface in 5-nm CMOS
This article presents a clock-forwarded, inverter-based short-reach simultaneous bi-directional (ISR-SBD) physical layer (PHY) targeted for die-to-die communication over silicon interposers or similar high-density interconnect. Short-reach links of this type are increasingly important to support larger systems built with chiplets and multiple die and to facilitate the shift to medium- and long-range optical communication based on silicon photonics. This project explores the advantages of simultaneous bi-directional signaling (SBD) over other bandwidth-doubling techniques (e.g., PAM4). Fabricated in a 5-nm standard CMOS process, the ISR-SBD PHY demonstrates 50.4 Gb/s/wire (25.2 Gb/s each direction) and 0.297 pJ/bit on a 750-mV supply over a 1.2-mm on- chip channel.
- 9 January 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 58 (4) , 1062-1073
Top Cited PapersA 0.297-pJ/Bit 50.4-Gb/s/Wire Inverter-Based Short-Reach Simultaneous Bi-Directional Transceiver for Die-to-Die Interface in 5-nm CMOS
This article presents a clock-forwarded, inverter-based short-reach simultaneous bi-directional (ISR-SBD) physical layer (PHY) targeted for die-to-die communication over silicon interposers or similar high-density interconnect. Short-reach links of this type are increasingly important to support larger systems built with chiplets and multiple die and to facilitate the shift to medium- and long-range optical communication based on silicon photonics. This project explores the advantages of simultaneous bi-directional signaling (SBD) over other bandwidth-doubling techniques (e.g., PAM4). Fabricated in a 5-nm standard CMOS process, the ISR-SBD PHY demonstrates 50.4 Gb/s/wire (25.2 Gb/s each direction) and 0.297 pJ/bit on a 750-mV supply over a 1.2-mm on- chip channel.
- 19 June 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Solid-State Circuits Magazine
- Vol. 11 (2) , 54-68
Ground-Referenced Single-Ended Signaling: Applications for High-Density, Short-Haul Communication Systems
As the end of the Moore's law era approaches, increases in computational perfor mance can no longer rely on the historical scaling of integrated circuit feature sizes that the industry has depended on for the last half century. This reduction in scaling trends, combined with manufacturing and packaging limits to maximum die size, will constrain single-die computational performance. Therefore, high-performance computers of the future will need to rely more heavily on architectural and system-level improvements. One way to provide these improvements is through the use of advanced packaging techniques, such as multichip modules (MCMs), where systems too large for a single die consist of multiple semiconductor dice, connected as shown in Figure 1, using on- and offpackage high-speed, short-reach links. These short-reach links are essential to scaling computational performance in systems where several processing units must communicate at data rates approaching on-chip total bandwidth while consuming a small fraction of total device power. Another key component to scaling aggregate performance in large computational systems is the use of optical links for long-haul communication. In these systems, processors and networking switches use short-reach links to connect to special purpose electrical/optical (E/O) interface chips [1]. Even memory interfaces (e.g., high-bandwidth memory) have begun the shift toward energy-efficient short-reach links to meet market demands [2]. Overall, our goal is to enable massively scalable parallelism, where it is crucial to understand that the energy-cost of communication is a tax on computational performance. In other words, every Joule saved in communication is a Joule that can be used for computation.
- 19 June 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Solid-State Circuits Magazine
- Vol. 11 (2) , 54-68
Ground-Referenced Single-Ended Signaling: Applications for High-Density, Short-Haul Communication Systems
As the end of the Moore's law era approaches, increases in computational perfor mance can no longer rely on the historical scaling of integrated circuit feature sizes that the industry has depended on for the last half century. This reduction in scaling trends, combined with manufacturing and packaging limits to maximum die size, will constrain single-die computational performance. Therefore, high-performance computers of the future will need to rely more heavily on architectural and system-level improvements. One way to provide these improvements is through the use of advanced packaging techniques, such as multichip modules (MCMs), where systems too large for a single die consist of multiple semiconductor dice, connected as shown in Figure 1, using on- and offpackage high-speed, short-reach links. These short-reach links are essential to scaling computational performance in systems where several processing units must communicate at data rates approaching on-chip total bandwidth while consuming a small fraction of total device power. Another key component to scaling aggregate performance in large computational systems is the use of optical links for long-haul communication. In these systems, processors and networking switches use short-reach links to connect to special purpose electrical/optical (E/O) interface chips [1]. Even memory interfaces (e.g., high-bandwidth memory) have begun the shift toward energy-efficient short-reach links to meet market demands [2]. Overall, our goal is to enable massively scalable parallelism, where it is crucial to understand that the energy-cost of communication is a tax on computational performance. In other words, every Joule saved in communication is a Joule that can be used for computation.
- 17 May 2019
- Published by Elsevier
- p. 260-266
Caver communities and organizations: Cultural and historical considerations
- 17 May 2019
- Published by Elsevier
- p. 260-266
Caver communities and organizations: Cultural and historical considerations
- 9 November 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 54 (1) , 43-54
A 1.17-pJ/b, 25-Gb/s/pin Ground-Referenced Single-Ended Serial Link for Off- and On-Package Communication Using a Process- and Temperature-Adaptive Voltage Regulator
This paper describes a short-reach serial link to connect chips mounted on the same package or on neighboring packages on a printed circuit board (PCB). The link employs an energy-efficient, single-ended ground-referenced signaling scheme. Implemented in 16-nm FinFET CMOS technology, the link operates at a data rate of 25 Gb/s/pin with 1.17-pJ/bit energy efficiency and uses a simple but robust matched-delay clock forwarding scheme that cancels most sources of jitter. The modest frequency-dependent attenuation of short-reach links is compensated using an analog equalizer in the transmitter. The receiver includes active-inductor peaking in the input amplifier to improve overall receiver frequency response. The link employs a novel power supply regulation scheme at both ends that uses a PLL ring-oscillator supply voltage as a reference to flatten circuit speed and reduce power consumption variation across PVT. The link can be calibrated once at an arbitrary voltage and temperature, then track VT variation without the need for periodic re-calibration. The link operates over a 10-mm-long on-package channel with -4 dB of attenuation with 0.77-UI eye opening at bit-error rate (BER) of 10⁻¹⁵. A package-to-package link with 54 mm of PCB and 26 mm of on-package trace with -8.5 dB of loss at Nyquist operates with 0.42 UI of eye opening at BER of 10⁻¹⁵. Overall link die area is 686 μm x 565 μm with the transceiver circuitry taking up 20% of the area. The transceiver's on-chip regulator is supplied from an off-chip 950-mV supply, while the support logic operates on a separate 850-mV supply.
- 9 November 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 54 (1) , 43-54
A 1.17-pJ/b, 25-Gb/s/pin Ground-Referenced Single-Ended Serial Link for Off- and On-Package Communication Using a Process- and Temperature-Adaptive Voltage Regulator
This paper describes a short-reach serial link to connect chips mounted on the same package or on neighboring packages on a printed circuit board (PCB). The link employs an energy-efficient, single-ended ground-referenced signaling scheme. Implemented in 16-nm FinFET CMOS technology, the link operates at a data rate of 25 Gb/s/pin with 1.17-pJ/bit energy efficiency and uses a simple but robust matched-delay clock forwarding scheme that cancels most sources of jitter. The modest frequency-dependent attenuation of short-reach links is compensated using an analog equalizer in the transmitter. The receiver includes active-inductor peaking in the input amplifier to improve overall receiver frequency response. The link employs a novel power supply regulation scheme at both ends that uses a PLL ring-oscillator supply voltage as a reference to flatten circuit speed and reduce power consumption variation across PVT. The link can be calibrated once at an arbitrary voltage and temperature, then track VT variation without the need for periodic re-calibration. The link operates over a 10-mm-long on-package channel with -4 dB of attenuation with 0.77-UI eye opening at bit-error rate (BER) of 10⁻¹⁵. A package-to-package link with 54 mm of PCB and 26 mm of on-package trace with -8.5 dB of loss at Nyquist operates with 0.42 UI of eye opening at BER of 10⁻¹⁵. Overall link die area is 686 μm x 565 μm with the transceiver circuitry taking up 20% of the area. The transceiver's on-chip regulator is supplied from an off-chip 950-mV supply, while the support logic operates on a separate 850-mV supply.
- 1 January 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 156-157
8.6 A 6.5-to-23.3fJ/b/mm balanced charge-recycling bus in 16nm FinFET CMOS at 1.7-to-2.6Gb/s/wire with clock forwarding and low-crosstalk contraflow wiring
Signaling over chip-scale global interconnect is consuming a larger fraction of total power in large processor chips, as processes continue to shrink. Solving this growing crisis requires simple, low-energy and area-efficient signaling for high-bandwidth data buses. This paper describes a balanced charge-recycling bus (BCRB) that achieves quadratic power savings, relative to signaling with full-swing CMOS repeaters. The scheme stacks two CMOS repeated wire links, one operating in the Vtop domain, between Vdd and Vmid=Vdd/2, the other, Vbot, between Vmid and GND. Unlike previous work [1], we require no voltage regulator to maintain the Vmid voltage at Vdd/2, to compensate for differences in data activity in Vtop and Vbot domains. The BCRB also uses simple single-ended signaling, to achieve higher bandwidth per unit bus width than differential buses [2] and lower signaling energy than precharging schemes [3], since we take full advantage of low switching activity and bus-invert coding.
- 1 January 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 156-157
8.6 A 6.5-to-23.3fJ/b/mm balanced charge-recycling bus in 16nm FinFET CMOS at 1.7-to-2.6Gb/s/wire with clock forwarding and low-crosstalk contraflow wiring
Signaling over chip-scale global interconnect is consuming a larger fraction of total power in large processor chips, as processes continue to shrink. Solving this growing crisis requires simple, low-energy and area-efficient signaling for high-bandwidth data buses. This paper describes a balanced charge-recycling bus (BCRB) that achieves quadratic power savings, relative to signaling with full-swing CMOS repeaters. The scheme stacks two CMOS repeated wire links, one operating in the Vtop domain, between Vdd and Vmid=Vdd/2, the other, Vbot, between Vmid and GND. Unlike previous work [1], we require no voltage regulator to maintain the Vmid voltage at Vdd/2, to compensate for differences in data activity in Vtop and Vbot domains. The BCRB also uses simple single-ended signaling, to achieve higher bandwidth per unit bus width than differential buses [2] and lower signaling energy than precharging schemes [3], since we take full advantage of low switching activity and bus-invert coding.