System level behaviour of VCSEL-based optical interconnects: a circuit-level simulation approach

1
System level behaviour of VCSEL-based optical
interconnectsa circuit-level simulation approach

• Michiel De Wilde
• Electronics and Information Systems dept.
• Olivier Rits
• Information Technology dept.
• Ghent University IMEC, Belgium
• IST project Interconnect by Optics

2
Optical interconnect rationale
(www.intel.com)
3
Optical interconnect rationale (2)

• Packet routers
• Parallel and distributed processing systems

4
Optical interconnect rationale (3)
Wire capacitance resistance skin effect?
bit-rate limit
(Miller-Özaktas)
5
Optical interconnect rationale (4)

• Physical solution use of optics
• No electromagnetic wave phenomena (crosstalk)
• Losses barely sensitive to distance frequency
• Potential to scale much better than wires
• Advantage in using optical interconnectsfor ever
  decreasing distances
• From chip to chip onboard, and board to board

6
VCSEL-based parallel optical I/O
Multi-waveguide connectors (e.g. overmoulded POF
bundles)
Parallel waveguides (e.g. POF flexes)
Hybridised 2D photodetector array
Hybridised 2D VCSEL array
optical devices
silicon
7
Design space
8
Design space (2)

• Continuously-valued design parameters as well
• ?, operating currents, numerical aperture

• Choices affect system-level characteristics
• Technological feasibility (interoperability,
  yield)
• Timing characteristics (delay, skew)
• Reliability(spikes, power ? temperature,
  misalignment)
• Monetary cost

9
Design space exploration

• Tradeoffs between choices
• Multi-objective
• Counteracting effects

• Find systematic way of making choices design
  methodology

• Goal formalized into a design tool
• The designer states system-level characteristics
• The design tool assists in making product and
  parameter choices

10
Design methodology development
11
Design methodology focus step 1Predicting
effect of design alternatives

• Issues for direct estimation(e.g. from tabular
  data)
• Difficult prediction of noise/variation
  propagation
• Dynamic multi-domain interactions (electrical,
  optical, thermal)

• Implement framework for time-domain link
  simulation to
• Estimate system-level properties for various
  setups
• Verify behaviour of optical interconnect within a
  digital system (mixed-signal simulation)

12
Simulation frameworklink building block models

• Circuit-level behavioural modelsinstead of
  physical models
• Only time-dependent equations
• No spatial dependency

• Mixed-signal Verilog-AMS (or VHDL-AMS)instead of
  SPICE
• Direct expression of differential equations
• Native support for signals in different domains

13
Example photodiode model
module pin_photodiode(in,anode,cathode)
input in inout anode, cathode power
in electrical anode, cathode parameter
real Cdep0, Cbo0, Rbas0, Resp0,
Id0 parameter real pole-1/(CdepRbas)
parameter real laplace_coeff_0CdepCbo
parameter real laplace_coeff_1CdepCboRbas
charge rc analog begin
I(cathode,anode) lt laplace_zp(RespPwr(in)Id,,
pole,0) Q(rc) lt laplace_np(V(cathode,a
node),laplace_coeff_0,laplace_coeff_1,pole,0)
end endmodule

• Terminals
• Model parameters
• Equations describing internal state and outputs

14
Simulation frameworkmodel hierarchy
15
Driver/receiver model

• Normal analog electrical circuits
• IP protection no real circuit provided

16
VCSEL model

• Nonlinear 1st order differential equation system

(M.X. Jungo)

• VCSEL characterisation is hard

17
Fiber-based optical path

• Abstraction of dispersion (short distance)

• TBD Statistical modelling of misalignment

18
Fiber bend losses

• TBD Take into account that the mode
  distributions do not directly stabilise after a
  bend

19
Simulation features

• Process corner simulation
• Best-case or worst-case value for all model
  parameters
• Lower and upper boundaries are not very tight

• Statistical simulation (partly TBD)
• Time-invariable statistics
• Inter-process intra-process variations
• Misalignment
• Dynamic statistics
• Noise (e.g. VCSEL RIN noise)
• Effects like power supply spikes

20
Simulation example transient
1 link with 10dB attenuation in the optical path
(exaggerated VCSEL model parameters)
21
Conclusion

• Framework for simulation of guided wave optical
  interconnect systems
• Design methodology development enable prediction
  of system-level interconnect characteristics
• Mixed-signal simulation of optical interconnect
  within a digital system

• Operational, but not yet mature
• Simulation is doable, characterisation is
  hard(especially statistical characterisation)

22
Acknowledgements

• IST Interconnect by Optics Project
• Helix AG driver/receiver block diagrams
• Avalon Photonics VCSEL measurements
• Hannes Lambrecht (Ghent University, IMEC-INTEC)
  macrobend losses
• Fund for Scientific Research Flanders (Belgium)
  (F.W.O.)
• Research assistantship

23
Thermal effects

• Temperatures are difficult to predict
• VCSELs are very temperature sensitive
• Operating temperature of up to 85C

• Temperature distribution in the simulation
• Simulator implementation not difficult
• TBD Estimate expected temperature differences in
  a VCSEL array