Transcript Radiation Effects on SRAM FPGA designs, Ukube SEE rates

Semiconductor Modeling And
Test Chip Design for
Characterization of Radiation
Effects
Sotiris Athanasiou
National Trainee, TEC-EDM
Supervisors:
Boris Glass, Richard Jansen
SCOPE
Radiation Effects
Semiconductor Modeling
•TCAD modeling and Doping problem
•Compact Models and extensions on verilogA
Test Chip
•Muxes
•Test Structures
•WideBand Amplifier
•Layout
Rad Effects on SRAM FPGA designs| Sotiris Athanasiou | ESTEC,NL| Data doc | Presentation| Pag. 2
Radiation Effects
Space is a difficult place for electronics to operate!
Ionized Particles and protons hit the device:
TID
SEE
SEU,MBU,SEFI,SEL,SEB,SEGR,SET
Space engineering,
Techniques for Radiation Effects Mitigation in
ASICs and FPGAs
European Space Agency
Rad Effects on SRAM FPGA designs| Sotiris Athanasiou | ESTEC,NL| Data doc | Presentation| Pag. 3
Space engineering,
Calculation of radiation and its effects and
margin policy handbook
European Space Agency
Total Ionizing Dose
TID creates trapped charge in the MOS structure in MOSFET:

Vth Shift

TID leakage currents
Single Event Transient
More complex phenomena:
a)A plasma track with e-h pairs is created
b)A path of free carriers between p and n areas – funnel
c)Charge generated outside diffuses to junction
Why this work
Usually testing is done afterwards more on gate level, and less on transistor
level.
Radiation effects not available in simulation of space electronics.
Same concept at ESD design
•
Requires Technology information for simulation
TCAD model towards electrical behavior
under radiation 1/3
TCAD commercial software extended with Geant4 (high energy physics
simulator)
Technology
parameters
GDSII
3d
Mesh
Geant4
simulations
Electrical behavior
under radiation
TCAD model towards electrical behavior
under radiation 2/3
Correctness of Technology parameters:
Technology
parameters
TCAD
model
/home/soathana/Desktop/TCAD/IdVd.pn
g
Drift diffusion
Spice
simulation
results simulation
Foundry spice
model
TCAD model towards electrical behavior
under radiation 3/3
TCAD modeling:
Technology parameters extracted from DESMICREX study
And foundry PDK info
a)Major problem remains Doping profiles 1D vs 3D
Without foundry access it is difficult to obtain Doping.
Methods exist to extract from electrical behavior doping, with some accuracy
rely on development of custom code.
b)Luck of methodology on parameters to match
Compact Models 1/4
Compact Model approach to overcome these issues
Umc 90 nm PDK BSIM4 Compact Model
Use of Silvaco VerilogA BSIM4 implementation
a) Model card extraction into VerilogA model
b) Fix errors/ remove unused code
c) Implement SET, TID leakage into modified model
d) simulate
Compact Models 2/4
VerilogA electrical behavior against cadence implementation.
Tested against dedicated Matlab model.
A lot of interesting issues arise regarding to what
extent parameters need to be matched IV, CV,
Vth, Idiode e.t.c.
Compact Model Approach offers a lot of
advantages:
1)Access to equation parameters as well as device parameters
2)Better behavior with respect to paracitics
3)No circuit modifications
4)Foundry sensitive parameters not required
Compact Models 3/4
compact model SET simulation
compact model TID leakage simulation
Compact Models 4/4
•
Parameters extracted from relevant available publications
•
This does gives an indication but not specific values.
•
Data can lead to predictions but in some cases can lead to over/under
estimating the behavior
•
Need experimental results to fine tune the equations (parameter extraction)
Characterization Test Chip and compact
model modification
Umc180 example, substrate resistor network.
rbodyMod = 0 (Off)
rbodyMod = 1 (On)
rbodyMod = 2 (On : Scalable Substrate Network): {5R/3R/1R}
Does rbodyMod 0
means no res?
NO!
Characterization Test Chip Design
Must have a large number of devices to characterize
Minimum number of pins possible
Must facilitate a large number of measurements
MUX architecture
Characterization Test Chip Design
1K devices of each type, placed in 5 rows of 200
•
Investigate charge sharing
Possible types of measurements
•
CV
•
IV
•
Noise
•
Matching
•
TID leakage
•
SET pulse
•
4 terminal (Kelvin) measurements
Characterization Test Chip Design
Wide Band Amplifier to test behavior under high frequencies.
Extract SET pulse shape for first time.
Characterization Test Chip Design
Building test benches to test functionality
A.D. OP27 behavioral model for simulation purposes
Characterization Test Chip layout
Test Structures
-generated automatically through cadence SKILL
-Each block contains 14K DUTs
Characterization Test Chip layout
MUX layout
Future Work
Finalize Test Chip->Tape out
Measurements
Fitting Measurement data to Models-> extract parameters
Insert parameters back to models
Once proof of concept verified, can easily generate new test structures
and fabricate next characterization chips.
Rad Effects on SRAM FPGA designs| Sotiris Athanasiou | ESTEC,NL| Data doc | Presentation| Pag. 21
THANK YOU
Sotiris Athanasiou
[email protected]
[email protected]