Computer Engineering Laboratory

1
Computer Engineering Laboratory
A new digital Architecture and Compilerfor
reliable, ultra-low-power systemsThe SiMS concept
Christos Strydis Georgi N. Gaydadjiev Stamatis
Vassiliadis
2
Introduction

• Current and envisioned application fields
• Elderly care
• Fitness
• Disease management
• Three important trends
• Achievements of biomedical technology over the
  last 50 years andincreasing penetration of
  implants in healthcare.
• Recent advances in microelectronic technology
  (power, area, speed) have opened new
  possibilities.
• Specific shortcomings have been overlooked new
  are created.
• Our approach
• Extensive survey on microelectronic implants.
• Determine what is needed
• Propose a solution . Not THE solution ( this can
  not be done ? ) SiMS !

3
Introduction

• Current and envisioned application fields
• Elderly care
• Fitness
• Disease management
• Three important trends
• Achievements of biomedical technology over the
  last 50 years andincreasing penetration of
  implants in healthcare.
• Recent advances in microelectronic technology
  (power, area, speed) have opened new
  possibilities.
• Specific shortcomings have been overlooked new
  are created.
• Our approach
• Extensive survey on microelectronic implants.
• Determine what is needed
• Propose a solution . Not THE solution ( this can
  not be done ? ) SiMS !

4
Implant evolution
Swallowable-pill endoscope (2000)
External wearable pulse generator (1957)
Fully implantable pacemaker (1958)
5
A typical implantable system (1)
Skin interface
?
commands
data
Outside
Processing/Controlling Core (PCC)
External host
Inside
Peripherals
Battery
Wireless transceiver
6
A typical implantable system (2)
Skin interface
?
data
commands
power
Outside
Processing/Controlling Core (PCC)
External host
RF induction
Inside
Peripherals
Battery
Wireless transceiver
7
What we have done so far Implant Survey

• Review of approx. 130 implantable systems
  developed over the last 15 years.
• Goals
• Determine what is important and what is not.
• Taxonomy
• Surveyed implantable systems 4 distinct topics
• Application scenario (medical usage, medical
  research)
• General overview
• Implanted part (implant structure, internal
  components etc.)
• External part (host structure, functionality
  etc.)
• Communication scheme (coding, modulation, bit
  rates etc.)
• Electromechanical specifications
• (power, chip(set) size, package dimensions,
  implementation technology)
• Miscellaneous issues
• (unique/original features, experimental results,
  future work etc.)

8
A case study (Lerch et al., 1995)
9
A case study (Lerch et al., 1995)
Wireless transceiver
PCC (Processing/Controlling Core)
Peripherals (here sensors)
10
Classification of surveyed data

• 8 major categories are specified of 46 parameters
  (implant-related).
• More detailed data are given on the architecture
  of PCC(s) (if present).
• The 8 categories are
• Application (e.g. disease diagnosis, pain
  therapy, paralyzed-limb restoration)
• Functionality (e.g. stimulation, measurement)
• Electromechanical features (e.g. physical
  dimensions, weight, analog/mixed/digital-signal
  approach, PCC type, fabrication process etc.)
• Power features (e.g. power consumption, power
  source, low-power modes)
• General implant features (e.g. number/type of
  peripherals, sampling rate, ADC/DAC resolution)
• PCC features (e.g. core frequency,
  instruction/data-word size)
• Miscellaneous implant features (e.g. adjustable
  settings, multiple-peripheral support,
  modularity, HW/SW-based error handling)
• Communication features (e.g. incoming-command/out
  going-data flow, payload size, encoding/modulation
  techniques, bit rates, error handling)

11
Classification of surveyed data
12
Some interesting conclusions

• Large number of applications!
• A large subset of simple - in computing
  requirements - applications.
• Two MAJOR categories Measurement, stimulation

• About 2/3 of the times a single PCC is used for
  control and data processing inside the implant.

13
Some interesting conclusions

• Most implants contain a PCC but
• Full-custom design diminishes
• Coreless systems decrease ALSO
• Semi-custom and structured-custom insignificant
• off-the-shelf are becoming the winners

14
Some interesting conclusions

• Most implants implement their PCC as an FSM but
• FSMs and coreless solutions gradually give way to
  µPs/µCs

15
Some interesting conclusions

• For 60 cases, average ISA size 31 instructions,
  but most of them have less than 16 with few with
  more than 50.
• For 16 of the 60 Instruction Sets, we have
  utilization
• about 88 on average.
• Smaller than RISC ISAs (MISC) are an interesting
  option.

16
Some interesting conclusions
technology delay

• The two trend lines never cross.
• Operating voltages are dropping (following CMOS
  trends).
• Power consumption is increasing in an overall.

17
Some interesting conclusions
RF-induction principle
3G-polySiGe thermal-energy scavenger
100 µW
Courtesy HOLST

• Both types of power scheme (battery,
  RF-induction) have been favored, depending on the
  application requirements
• Novel (primary-) battery-less schemes such as
  MEMS are good future candidates, especially for
  ultra-low-power (ULP) systems.(Reduce volume up
  to 70.)

18
Some interesting conclusions

• Schemes
• (sub)block duplication
• self-test/-diagnostic circuitry
• test/interrogation modes (SW HW)
• error-det./corr. instruction decode
• design for structural testability
• humidity detection

• Reliability is seriously absent 67-73!
• Commercial- and µC/µP-based implants score highest

19
Interesting conclusions

• Many applications exist with moderate workload.
• One PCC suffices ?
• Commercial PCCs are favored over full-custom.
• µP/µC PCCs are favored over FSM (or hardwired).
• Smaller ISAs (than commercial RISC) are favored.
• The technology delay can be narrowed by using
  standardized, pretested components, bringing
  technology benefits to market faster and cheaper.
• Power-scavenging, ultra-miniature systems with
  ULP-requirements are possible.
• Reliability is low. Currently, in
  commercial-based (free resources) and µP/µC-based
  (mostly SW-based) implants.? Intrinsic SW but
  mostly HW reliability is needed.

We are Juggling with
Power source parameters
Core parameters
20
The SiMS concept

• Power (-scavenging) source
• Minimalistic µ-architecture Compiler
• Standardized I/Fs
• Peripherals
• Transceiver module (bidirectional comm.)

21
Focus of current project
PCC

• Power (-scavenging) source
• Minimalistic µ-architecture Compiler
• Standardized I/Fs
• Peripherals
• Transceiver module (bidirectional comm.)

22
SiMS design space
area
power
performance
23
SiMS design space
area
modularity (systematic approach)
power
performance
24
SiMS design space
area
dependability (reliability, availability), safety
modularity (systematic approach)
power
performance
25
SiMS design space
area
dependability (reliability, availability), safety
modularity (systematic approach)
power
performance
cost
26
SiMS A design platform for future applications

• Risk involved when moving from ideas to actual
  products

• Risks during
• design
• development
• normal operation

27
SiMS-architecture characteristics

• Dependability (reliability, availability)
  achieved
• through design for fault tolerance.
• Required HW/SW-based techniques for
• online error detection
• fault isolation and correction machine recovery
• Foreseen caveats
• One or more errors to be detected, isolated,
  corrected?
• Design for higher error detection but no fault
  isolation?

• Currently used
• Circuitry redundancy
• Watchdog circuitry
• CRC checksums
• Scan-based testing

Courtesy Nelson V.P., Fault-Tolerant Computing
Fundamental Concepts
28
SiMS-architecture characteristics

• Dependability (reliability, availability)
  achieved
• through design for fault tolerance.
• Required HW/SW-based techniques for
• online error detection
• fault isolation and correction machine recovery
• Foreseen caveats
• One or more errors to be detected, isolated,
  corrected?
• Design for higher error detection but no fault
  isolation?
• reconfigurable HW?

• Currently used
• Circuitry redundancy
• Watchdog circuitry
• CRC checksums
• Scan-based testing

Courtesy Nelson V.P., Fault-Tolerant Computing
Fundamental Concepts
29
SiMS-compiler characteristics

• Abstraction functionality for hiding the
  elementary operations of the tiny
  PCC-architecture.
• Code optimizations for reducing the instruction
  count.
• Dependable by accepting application-specific,
  preverified, constraint files along with the
  source code.

30
Issues NOT addressed (yet)

• SiMS compiler design tools (for facilitating
  thedigital architecture)
• (bio)sensors/actuators collectively
  termedperipherals (for interfacing to the
  living tissue)
• dedicated ADCs/DACs (for interfacing the
  peripherals to the digital architecture)
• wireless radio module (for bidirectional
  communication between the in-body, SiMS device
  an out-of-body, external workstation)
• power module
• SiMS packaging
• information security provisions
• system electro-magnetic compliance (EMC) and heat
  dissipation management

31
What WE want to do

• A digital architecture for implantable
  systemswhich is
• Minimalistic (small ISA, few-bit architecture)
• Ultra-low-power consumption
• Dependability (reliability, availability)
• Modularity

32
Standardization is the key
Intel IBM Cisco Sharp Motorola Samsung Philips Pan
asonic Kaiser Permanente Medtronic
Its a dogs breakfast of standards, interfaces
and software to make these things work now and
thats expensive David Watson, CTO - Kaiser
Permanente