Motivation and Background
The exponential growth of bandwidth
demand in emerging applications such as artificial intelligence (AI) training,
inference, and cloud computing is pushing traditional electrical interconnects
to their physical and power-efficiency limits. At very high data rates,
conventional pluggable optics struggle with increasing power consumption, degraded
signal integrity, and limited form factor scalability.
Co-packaged optics (CPO), which
integrate optical transceivers onto the same substrate (or eventually
interposer) as the ASIC, provide a promising solution by significantly reducing
electrical interconnect length to the optics, thereby minimizing insertion
losses and lowering system power consumption. However, this paradigm introduces
complex design challenges, requiring holistic co-design across electronics,
photonics, and packaging domains.
Problem Statement
The realization of next-generation
co-packaged optical transceivers requires simultaneous support for ultra-high
data throughput and low-loss RF signal routing on a unified platform. Key
challenges include:
- Heterogeneous integration platforms that
enable high-density, low-loss interconnects between electronic ICs (EICs)
and photonic ICs (PICs) while preserving bandwidth scalability.
- Wireline architectures optimized for
high-speed digital I/O, demanding low crosstalk, high bandwidth density
(high Terabit/s/mm), and energy-efficient operation (low pJ/bit).
- RF/mixed-signal driver EICs capable of efficient
interfacing between the ASIC and the optics.
Current solutions either escalate
system cost, introduce insertion loss and crosstalk, or fail to achieve the
required balance between performance, scalability, and manufacturability. A
co-optimized heterogeneous integration framework is therefore essential to meet
the demands of next-generation optical interconnects.
Research Objectives
This PhD will pursue the following
objectives:
- Optimize Imec RF/mixed-signal interposers enabling
high-density, low-loss electrical links design among ASICs, driver ICs,
and PICs.
- Design high-Q passive components (capacitors,
resistors, inductors and matching networks) within IMEC RF/Mixed-Signal interposers
to reduce active chip area and system cost.
- Investigate hybrid integration strategies that
combine on-chip active circuits with interposer-embedded passive networks,
enabling co-optimization for performance and manufacturability, and
benchmark against alternative integration technologies.
- Design, simulate, and experimentally validate:
- High-speed, low-power SerDes with minimized jitter.
- Efficient EICs for optical modulation across
different integration approaches.
- Interposers optimized for interconnect with low
insertion loss, crosstalk, and parasitic.
- Fabrication of a demonstrator based on a
co-simulation and co-optimization framework that integrates photonic and
electronic subsystems with dense, low-parasitic interconnects, thereby
enabling enhanced energy efficiency and scalable data rates.
Methodology
- Modeling and simulation: Multi-domain co-simulation
of SerDes, RF drivers, photonic modulators, and interposer structures.
- Co-design of EICs and RF/mixed-signal interposers
with embedded passive components and optimized high-frequency routing.
- Prototype fabrication: Fabrication of test structures
and integrated demonstrators on imec’s RF/mixed-signal interposer
platform.
- Experimental characterization: S-parameter
measurements, eye-diagram analysis, and system-level validation of energy
efficiency and data rate.
Expected Outcomes
- Demonstration of low-power, high-speed SerDes
interfaces optimized for co-packaged optics (CPO).
- RF/mixed-signal interposer demonstrators with
minimized parasitics, insertion loss, and crosstalk.
- A unified co-simulation and co-optimization framework
integrating EIC, PIC, and packaging design considerations.
- Experimental proof-of-concept of a heterogeneous,
co-packaged optical transceiver achieving good energy efficiency and
bandwidth scalability relative to current state-of-the-art solutions.