Fundamental Technology on Dependable SoC and SiP

for Embedded Real-Time Systems

Nobuyuki Yamasaki (Keio Univ.)Kikuo Wada (NECAT)

Masayuki Inaba (Tokyo Univ.)

IntroductionTarget: High-end embedded real-time systems

Traditional researches: most researches on softwareTime quantum: 1msec

Requirements for embedded real-time systemsVery small size, various functions, low power, scalability, noise tolerance, heat tolerance, etc.

VLSI (SoC) operates only after it is mounted on a substrate.

Dependability on VLSI system→ Depends on a substrate mounting technology

SoC-SiP-Board co-design→Improvement of dependability

Software can take full advantage of hardware. RTOS-Processor-Communication co-designRTOS-Middleware co-design

Kojiro (Tokyo Univ.)

Control board

SiPSoC

Requirements for a high power motor driverReal-time processing under high speed communication

Motor temperature estimation processing for very high power motor driving such as 20 times overdrive rated at 200W

• Current control cycle: 10msec → 10μsecReliability, availability, and safety on communication and control under high-stress environment

Huge current noise, unusual situation such as cable disconnection, etc

⇒ Prevention of fatal accidentsRequirements for a large scale distributed motor driver

Microminiaturization of the controller (size: 36x46x7mm)Area constraint of the digital control part: 20mm squareReal-time communication and control under the size constraint

• Poor processing power of current MPU (H8S/2215 16MHz)• Limit of control cycle: 1msec → 10μsec• External computation servers (Xeon 3.4GHz x 2) are required.

• High communication traffic 7.2MB/sec• Limit of Inter-device synchronization cycle: 8msec (USB) →

100μsec (Responsive Link)Reliability of communication under the size limitation

• Severe noises under the logic servo systemsPower saving scheme under the large scale distributed control

Static power of a whole logic part: 80W@idle → 1W

Requirements in the Field of High-end Robots

Kojiro (Tokyo Univ.)

•Freedom 82 DOFs•Driven -muscle 109•# of controllers 60

High power leg: HRP3L-JSK (Tokyo Univ.)Continuous current 80[V],100[A],15[sec]Peak current 80[V],200[A],10[msec]

Evaluation indicatorsEvaluation itemsClassification

Dependability Evaluation Indicators

Reliability

Availability

Real-timePower

Noise tolerance

Heat radiation

Safety

Maintainability

Footprint

Plug-and-Play

Heat control

Parts replacement

Board replacement

Failure analysis

Hard/soft time constraint (T/F)Time quantum (sec)Jitter (sec)Dynamic power (W)Static power (W)S/N ratio (%)Thermal resistance (deg C/W)Integration to robots (T/F)Correct operating (T/F)Network reconfig time (sec)

Self-monitoring (T/F)

Repair and replacement (sec)

Repair and replacement (sec)

Failure analysis (sec)

Network failure

①②③④

⑤⑥⑦⑧⑨

Humanoid robot levelD-RMTP SiPs are connected via Responsive LinkAdaptive ECC for Responsive LinkNetwork reconfiguration to avoid faulty linksTask migration from faulty SiPs

SiP levelD-RMTP SoC and memory modules are integrated on a board by FFCSPThermal control using self-monitoringRedundant FFCSP links to improve yield

SoC levelProcessing cores are connected via NoCRedundant processing cores and network links

Multi-Level Dependability Support

DRMTPSiP

DRMTPSiP

DRMTPSiP DRMTP

SiP

DRMTPSiP DRMTP

SiP

ResponsiveLink

Humanoid Robot

DRMTP SoC

MemoryModules

D-RMTP II SoC

D-RMTP I SiP

Robot Level Dependability

DRMTPSiP

DRMTPSiP

DRMTPSiP DRMTP

SiP

DRMTPSiP DRMTP

SiP

ResponsiveLink

Humanoid Robot

ECC code (4Byte)

ECC code(1Byte) Line code

Reed-Solomon(48, 32)

BCH (16, 8)

BS+NRZI (9, 8)

8b10b (10, 8)

4b10b (10, 4)

Hamming(12, 8)

BS+NRZI (9, 8)

8b10b (10, 8)

4b10b (10, 4)

None8b10b (10, 8)

4b10b (10, 4)

NoneBCH

(16, 8)

BS+NRZI (9, 8)

8b10b (10, 8)

4b10b (10, 4) Received waveform w/ noise

D-RMTP SiPs are connected via Responsive LinkPermanent faults (links & boards)

Network reconfiguration to avoid faulty linksTask migration from faulty D-RMTP SiPs

Transient faults (links)Adaptive ECC & line codes for Responsive Link

(1)Permanent faults by link disconnection

(2)Transient faults by motor noise

③

⑦⑧⑤

D-RMTP SoC and DRAM modules are integrated on a board by FFCSPOverheating (DRAM & CPU)

Thermal control using self-monitoring

Permanent faults (links)Redundant vertical links to improve yield

SiP Level Dependability

20μsec

20μsec

0.80V

1.10V

Voltage transition (0.81.1V)

Voltage & thermal control (D-RMTP I)

DRAMDRMTP IIFPGA Substrate

Vertical chip stacking (D-RMTP II)Redundant vertical links

①④

③

Processing cores are connected via Network-on-ChipPermanent faults (cores)

Redundant processing cores and task migration from faulty cores

Fault analysis (cores)Instruction-level trace buffer (much like a flight recorder)

SoC Level Dependability

Redundant cores (D-RMTP II)

⑤

⑥

Trace buffer (D-RMTP I)

Redundant cores

FY20

12FY

2013

FPC D-RMTP Ⅱ DRAMOrganic Interposer

Silicon Interposer

FPC Metal FrameHeat Sink

D-RMTP III

FPC

MetalFrame

DRAMChip-on-ChipConnection

Chip-on-ChipConnection

Multistage bonding by FFCSP technology

Approx 1,100pins of Chip-to-Chip Connection through Organic Interposer

Embedded Sensors had already been completed in FY2011.

Network-on-Chip (NoC) is extended to 3-D NoC that interconnects chips in a package to improve performance and dependability.

Same chips are bonded on both side of the Interposer

Sensor

Approx 100pins of Chip-to-Chip Connection through FPC

Approx 10,000pins of Chip-to-Chip Connection through Silicon

InterposerApprox 100pins of

Chip-to-Chip Connection through

FPC

①

SiP Design Plan for Dependablity

①

②

③

③

④

④

SiPStructure

Chip-to-Chip Connection

Pin Count : Approx 100 pinsLength : 16mm

Pin Count : Approx 500 pinsLength : 0.4mm

Chip-to-Chip Interface

I/O Buffer (SSTL, LVDS, etc.)Driving Power:1.5V/15mA(SSTL15)

LSI Internal Logic LevelDriving Power:1.0V/several μA

Network-on-Chip (NoC) is extended for 3-D NoC that interconnects chips in a package to improve performance and dependability.

D-RMTP IIDRAM

Organic Interposer

Same chips are bonded on both side of the Interposer

FPC

Sensor

Embedded Sensors had been already completed in FY2011.

Low Power ConsumptionHigh Power Consumption

Low Band Width, More Cross Talk High Band Width, Less Cross Talk

NoC

信号

NoC

信号

150μm pitch Vertical Connection

SiP Design for Dependability

Organic Interposer Silicon Interposer

Structure

NoCConnection

Approx 1,100 Connections/sq. 5mm

Approx 10,000 Connections/sq. 5mm

Stress of Chips

More(CTE of Interposer is larger than Chips)

Less(CTE of Interposer is equal to Chips)

D-RMTP II DRAM

OrganicInterposer

D-RMTP II DRAM

SiliconInterposer

150um

NoC

信号

NoC

信号

50um

NoC

信号

NoC

信号

NoC

信号

NoC

信号

150um PitchNoC Connection

Laser Via Through Silicon Via

9 TimesDenser

50um PitchNoC Connection

Dependability is Further Improved by using Silicon Interposer

Additional Idea for Further Dependability

Dependability Verification for RoboticsRobot technology for social practical use

High-power humanoids for home assistive robotsMulti-DOFs human-like robots for friendship and familiarity

1)Under high-power environmentFast/low latency real-time communicationReliability of communication and control(Noise/heat tolerance)Computing power for fast control tasks

2)Under multi-DOFs distributed environmentReal-time communicationAchievement of VLSI tecnology for small motor driverReliability under complicated communication pathLow power based on RT-DVFS control

Extension High-power leg platform (HRP3-JSK-leg)

Multi-Dofs musculoskeletal humanoid (Kojiro)

Basic Experiments of High-power Actuation

Testing environment for high-power dependability verification

D-RMTPI SoC/SiP(30mm^2)

Control board with D-RMTPI SoC/SiP

Water cooling motor driver module with D-RMTPI SoC/SiP(85x60x34mm)

Water cooling box

Motor driver

Motor with low gear ratio

Light/high stiffness link

Tactile sensing by using rotary encoder

Collision object

ＦＰＧＡ

Experiment of motor control on RMTPReal-time control thread: 1msecTransmitter/Receiver communication thread: 100msec

Experiment of communication between multi-RMTPDriver board with RMPT - PCI testing board with RMTPResponsive Link (RL) communication

One axis testing environmentDemonstration of avoiding collision object

Motor

RL receiver thread

Serial transmitter thread

Motor control thread

Serial receiver thread

RMTPSerial receiver thread

RL receiver thread

Serial transmitter thread

Motor command

Monitoring

M-RMTP

High-power Leg Experiment: Push Recovery

Push recovery control against sudden big impact to trunk bodyLarge computing power to search more stable foot steps

Calculation time on Intel E8600 4.0GHz: 1msecHigh power joint performance to achieve calculated joint trajectory

Max instant current 100A for achievement of step width 250mm and step time 250msec

External force while push recovery demonstration

0 0.5 1 1.5 2 2. 5 3-600

-500

-400

-300

-200

-100

0

100

time[s]

force [N]

xyz

Push recovery demonstration

Searching more stable foot steps

Peak force: 590[N]

Pulse width: 140[msec]

Robotics Functional SafetyDifficulty to define explicit functional safety standard

Robotics applications are on developing immature stagecf. Automobile area is enough mature : ISO26262

Toward international standard criterion for roboticsRobot safety center by life support robot commercialization project

Testing functional safety for home assistive robots→Proposal of international safety standard ISO 13482 for service robots

Test environment for collision safety

Test environment for static safety

Testing safety of service robots under various situations

Robot Technology Middleware with SIL3 Capable of IEC 61528

“RTM Safety” by SECCommercially available, May 2012Middleware for service robots

RTM Safety consists of two packagesRTM Safety Package

Framework for development of RT componentsSafety Library Package

Functions for development of safety software

Watchdog APL Components with functional safety standards

Functional safety library

Component framework

OS abstraction layer N/W abstraction layer

Trouble detection function

Internal CLK

(Peripheral circuit)

Power supply/Reset

Function Data reference Data setting ClockPackage framework of RTM Safety

System diagram of RTM Safety

Wheel chair robot with RTM Safety