Sarah Azimi Boyang Du Luca Sterpone - Indico at ESA ... · SET: mitigation radiation test results !...
Transcript of Sarah Azimi Boyang Du Luca Sterpone - Indico at ESA ... · SET: mitigation radiation test results !...
-
Politecnico di Torino CAD Group – Dipartimento di Automatica e Informatica Torino - Italy
Sarah Azimi Boyang Du
Luca Sterpone
-
Goal ¨ Analysis of Single Event Transients (SETs) occurrency ¨ Effective SET mitigation
2
-
Outline ¨ SET effects on Flash-based FPGAs ¨ Single Event Transient Analysis (SETA) tool
¨ Analysis ¨ Mitigation ¨ Experimental results
¨ Conclusions and future activities
3
-
SET effect 4
¨ A Single Event Transient (SET) is generated by the injunction of charge collection ¨ A charged particle crosses a junction area ¨ It generates an amount of current, provoking a “glitch” ¨ SET can be indistinguishable from normal signal and
exist for notable distances
SET width SET amplitude Rise ΔV/ΔT Fall ΔV/ΔT
-
Circuits on Flash-based FPGAs 5
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
Flash configura/on memory FPGA array
-
Circuits on Flash-based FPGAs 6
Flash configura/on memory FPGA array
Configura/on process
0 1 0 0 0 0
1 0 0 0 0 0
1 1 0 0 0 0
0 0 0 0 1 0
0 1 0 1 0 0
0 0 0 0 0 0
-
SET scenario 7
¨ Considering a place and route design on FPGA ¨ Fixed logic cells
¨ Defined number of routing segments
Source of SET
-
SET scenario 8
¨ Considering a place and route design on FPGA ¨ Fixed logic cells
¨ Defined number of routing segments
Source of SET Propagation through Gates
Propagation through Routing
-
SET scenario 9
¨ Considering a place and route design on FPGA ¨ Fixed logic cells
¨ Defined number of routing segments
Source of SET Propagation through Gates
SET Classification
FFs / IOs
Propagation through Routing
-
SET Propagation through gates 10
For a 1→0→1 transi,on Δtp is defined as: Δtp = tpHL – tpLH For a 0→1→0 transi,on Δtp is defined as: Δtp = tpLH – tpHL
Fist Region: If(τn < k*tp ) then τn+1 = 0 Second Region: If (τn > (k+3)*tp) then τn+1 = τn + Δtp Third Region: If ((k+1)*tp < τn < (k+3)*tp) then τn+1 = (τn2 -‐ tp2 )/ τn + Δtp Fourth Region: If (k*tp < τn < (k+1)*tp) then τn+1 = (k+1)*tp(1 -‐ e(k – ( τn / tp )) ) + Δtp
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
[Wirth et al, NSREC 2008]
-
SET Propagation through gates 11
For a 1→0→1 transi,on Δtp is defined as: Δtp = tpHL – tpLH For a 0→1→0 transi,on Δtp is defined as: Δtp = tpLH – tpHL
Fist Region: If(τn < k*tp ) then τn+1 = 0 Second Region: If (τn > (k+3)*tp) then τn+1 = τn + Δtp Third Region: If ((k+1)*tp < τn < (k+3)*tp) then τn+1 = (τn2 -‐ tp2 )/ τn + Δtp Fourth Region: If (k*tp < τn < (k+1)*tp) then τn+1 = (k+1)*tp(1 -‐ e(k – ( τn / tp )) ) + Δtp
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V
[Wirth et al, NSREC 2008]
-
SET Propagation through routing 12
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V
[Sterpone et al, RADECS 2014]
-
SET Propagation through routing 13
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V
[Sterpone et al, RADECS 2014]
Propagation Induced Pulse Broadening
-
SET Propagation through routing 14
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V
[Sterpone et al, RADECS 2014]
Propagation Induced Pulse Broadening
Gate to Gate Characterization
-
SET Propagation through routing 15
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V
[Sterpone et al, RADECS 2014]
V
Propagation Induced Pulse Broadening
Gate to Gate Characterization
-
SET classification on FFs and IOs 16
¨ A tool has been developed: ¨ Single Event Transient Analyzer (SETA)
HDL
Synthesis and Implementation
(Place and Route)
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V V
Netlist
Constraints
IO/FF expected Pulses
SETA Tool
-
SETA tool 17
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V V
Netlist
Constraints
Physical Design Description (PDD)
Gate PIPB Characterization
-
SETA tool 18
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V V
Netlist
Constraints
Physical Design Description (PDD)
Sensitive nodes selection
Gate PIPB Characterization
-
SETA tool 19
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V V
Netlist
Constraints
Physical Design Description (PDD)
Sensitive nodes selection SET propagation
Gate PIPB Characterization
-
SETA tool 20
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V V
Netlist
Constraints
Physical Design Description (PDD)
Sensitive nodes selection SET propagation
Gate PIPB Characterization
-
SETA tool 21
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V V
Netlist
Constraints
Physical Design Description (PDD)
Sensitive nodes selection SET propagation
Gate PIPB Characterization
-
SETA tool 22
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V V
Netlist
Constraints
Physical Design Description (PDD)
Sensitive nodes selection SET propagation
Propagation is performed up to “terminal” nodes
(IOs / FFs)
Gate PIPB Characterization
-
SETA tool 23
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V V
Netlist
Constraints
Physical Design Description (PDD)
Sensitive nodes selection SET propagation
SET classification on FFs or IOs
Gate PIPB Characterization
-
SETA tool 24
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V V
Netlist
Constraints
Physical Design Description (PDD)
Sensitive nodes selection SET propagation
SET classification on FFs or IOs
V
Gate PIPB Characterization
-
SET classification on FFs and IOs 25
¨ The classification identifies the number of SET: ¨ Totally filtered
¨ Partially filtered ¨ Equally propagated ¨ Broadened
-
SETA results – EUCLID project (WP1) 26
0"5000"10000"15000"20000"25000"30000"35000"40000"45000"
0,1"ns" 0,3"ns" 0,45"ns" 0,7"ns" 1"ns" 1,5"ns" 2,0"ns" 2,5"ns" 3,0"ns" 4,0"ns"
Combina2onal"Path"7"Single"Event"Transient"sensi2vity"
Filtered" Par2ally"Filtered" Equal" Broadened"
!
-
SETA results – EUCLID project (WP1) 27
0"5000"10000"15000"20000"25000"30000"35000"40000"45000"
0,1"ns" 0,3"ns" 0,45"ns" 0,7"ns" 1"ns" 1,5"ns" 2,0"ns" 2,5"ns" 3,0"ns" 4,0"ns"
Combina2onal"Path"7"Single"Event"Transient"sensi2vity"
Filtered" Par2ally"Filtered" Equal" Broadened"
!Total number of analyzed SET
Type of SET per injected pulse
-
SET: mitigation 28
¨ Selective guard gate (GG) mapper ¨ Inserting a GG logic structure in the input of the
selected FF
[Sterpone and Du, IEEE ETS 2014]
-
SET: mitigation solution 1 29
¨ Selective guard gate (GG) mapper ¨ Inserting a GG logic structure in the input of the
selected FF
[Sterpone and Du, IEEE ETS 2014]
Filtering estimated on the basis of the SETA
report
-
SET: mitigation solution 2 30
¨ Accurate placement acting on the critical paths ¨ Distance between gates is modified in order to
maximize the electrical filtering effect
[Sterpone and Du, IEEE ETS 2014]
-
SET: mitigation solution 2 31
¨ Accurate placement acting on the critical paths ¨ Distance between gates is modified in order to
maximize the electrical filtering effect
[Sterpone and Du, IEEE ETS 2014]
-
SET: mitigation solution 2 32
¨ Accurate placement acting on the critical paths ¨ Distance between gates is modified in order to
maximize the electrical filtering effect
[Sterpone and Du, IEEE ETS 2014]
-
SET: mitigation results 33
Different place and route constraints
Average SET sensitivity
-
SET: mitigation results 34
Different place and route constraints
Average SET sensitivity
Original RISC circuit
Post SETA RISC circuit
-
SET: mitigation results 35
Different place and route constraints
Average SET sensitivity
Effective mitigation of SET
-
SET: mitigation radiation test results ¨ Heavy ions test performed at the Cyclotron of the
Université Catholique de Louvain (UCL) ¨ Kripton ion with a fluence of 3.04E8 (particles) ¨ Average flux 1E4 (particles/sec) ¨ RISC working frequency of 20MHz on ProASIC3
A3P250 RISC processor version SEE Cross-section
[MeV cm2/mg]
Unhardened 1.45E-9
Full TMR + GG 6.37E-10
Our Approach 3.12E-12
36
-
SET: in conclusion… 37
¨ SETA tools are available ¨ Effective analysis of SET propagation
¨ Effective overall SET mitigation
-
SET: in conclusion… 38
¨ SETA tools are available ¨ Effective analysis of SET propagation
¨ Effective overall SET mitigation
-
SET: in conclusion… 39
¨ SETA tools are available ¨ Effective analysis of SET propagation
¨ Effective overall SET mitigation
-
SET: in conclusion… 40
¨ SETA tools are available ¨ Effective analysis of SET propagation
¨ Effective overall SET mitigation
Source of SET Propagation through gates Propagation through routing SET classification on FFs or IOs
V V V
X
-
Physical Design Description 41
¨ The circuit is modeled as a graph ¨ Cell functionality
¨ Routing model
AND
OR INV
-
SET generation phenomena 42
¨ Particle hitting a sensitive node ¨ Generate a SET pulse
¨ Propagates through the logic
AND
OR INV
-
SET generation phenomena 43
[Azimi and Sterpone, IEEE DDECS 2016] [Azimi, Du, Sterpone, Micro Rel, 2015]
¨ SET generation is related to ¨ Linear Energy Transfer (LET) ¨ VersaTile architecture ¨ Technology
-
Why SET generation ? ¨ The type of source SET is mandatory to understand
the exact type of propagation ¨ Mitigation GG insertion is related to SET length
¨ It is necessary to establish the absolute SET count ¨ Calculation of the realistic IOs/FFs error rate for
the whole space mission duration
44
-
Why SET generation ? 45
0"5000"10000"15000"20000"25000"30000"35000"40000"45000"
0,1"ns" 0,3"ns" 0,45"ns" 0,7"ns" 1"ns" 1,5"ns" 2,0"ns" 2,5"ns" 3,0"ns" 4,0"ns"
Combina2onal"Path"7"Single"Event"Transient"sensi2vity"
Filtered" Par2ally"Filtered" Equal" Broadened"
!
-
Identification of source SET length 46
0"5000"10000"15000"20000"25000"30000"35000"40000"45000"
0,1"ns" 0,3"ns" 0,45"ns" 0,7"ns" 1"ns" 1,5"ns" 2,0"ns" 2,5"ns" 3,0"ns" 4,0"ns"
Combina2onal"Path"7"Single"Event"Transient"sensi2vity"
Filtered" Par2ally"Filtered" Equal" Broadened"
!Effective source SET designer must care
-
Identification of effective SET counts 47
0"5000"10000"15000"20000"25000"30000"35000"40000"45000"
0,1"ns" 0,3"ns" 0,45"ns" 0,7"ns" 1"ns" 1,5"ns" 2,0"ns" 2,5"ns" 3,0"ns" 4,0"ns"
Combina2onal"Path"7"Single"Event"Transient"sensi2vity"
Filtered" Par2ally"Filtered" Equal" Broadened"
!Effective source SET designer must care
Effective SET counts
-
Thank you! ¨ [email protected]
48