THESIS
2022
1 online resource (xvii, 111 pages) : illustrations (some color)
Abstract
Entities such as telecommunications enterprises, online retailers, and
social media generate huge amounts of data, and most of these entities use
cloud computing to process and store the data in data centers having
communication speeds of 200 Gbps or higher. Although advanced
communication equipment for 200 GbE can produce a greater data rate with
smaller size and lower cost and power, the size reduction can cause reliability
issues in data centers, as electromagnetic interference (EMI) can become a
significant problem in these highly dense facilities.
Optical transceivers supporting four-level pulse amplitude modulation
(PAM-4) signaling are widely used in high-speed communication links in data
centers, and while it is true that the optical communication channel is immune
to EM radiati...[
Read more ]
Entities such as telecommunications enterprises, online retailers, and
social media generate huge amounts of data, and most of these entities use
cloud computing to process and store the data in data centers having
communication speeds of 200 Gbps or higher. Although advanced
communication equipment for 200 GbE can produce a greater data rate with
smaller size and lower cost and power, the size reduction can cause reliability
issues in data centers, as electromagnetic interference (EMI) can become a
significant problem in these highly dense facilities.
Optical transceivers supporting four-level pulse amplitude modulation
(PAM-4) signaling are widely used in high-speed communication links in data
centers, and while it is true that the optical communication channel is immune
to EM radiation, the transmitter in the transceiver package consists of
electronic components which can generate and radiate EMI. Additionally, to
compensate for the bandwidth (BW) limitation of the laser and channel,
different types of equalization are utilized. It is intuitive to believe that a
fractional-spaced asymmetric feed forward equalizer (FSA-FFE) circuit could
further increase the EM emission from the transmitters. Also, advanced technology nodes like 14 nm FinFETs are extensively used for implementation
of high-speed communication systems, and the technology scaling can
significantly affect the EM emission from the transmitter. However, a
systematic study of how asymmetric equalization and technology scaling affect
the EM emission is still lacking. To fill this gap, this thesis analyzes the effect
of technology scaling and the sources of EMI-related CM noise in PAM-4
transmitters, and then proposes an on-chip active technique to mitigate
EMI-related CM noise.
In the first part of the thesis, the effects of the PAM-4 current-mode logic
(CML) driver, different FFE configurations and technology scaling on
EMI-related CM noise are analyzed mathematically, and then further evaluated
by behavioral and transistor-level simulations in 40 nm CMOS and 14 nm
FinFET technologies. It is observed that the CM current in differential circuits
generates CM noise which radiates in the environment and causes EMI with
electronic devices in proximity. This noise is generated due to the rising and
falling time mismatch created by the driver circuit and asymmetric
equalization. It is demonstrated that the intrinsic impedance variations of the
CML driver circuit, amplitude of the equalization current, and offset in the
FSA-FFE circuit are the main sources of EMI-related CM noise in
transmitters. Additionally, the technology scaling is used to achieve higher data
rates, which results in increase of the CM noise in advanced technology nodes.
In the second part of the thesis, we present a novel on-chip active circuit
technique to mitigate EMI-related CM noise in high-speed PAM-4 transmitters.
An automatic CM noise cancellation (CMNC) system architecture in 40 nm
technology is designed for a 56 Gbps PAM-4 optical transmitter. This solution
provides the benefits of small size and low cost by eliminating the need for discrete
components to suppress EMI. It is demonstrated that the CM noise cancellation
system efficiently mitigates EMI by suppressing the CM noise up to 90% while
consuming 5 mW, with a core area of 17 μm × 9 μm.
Post a Comment