Project Status by yaosaigeng

VIEWS: 3 PAGES: 56

									Digitizing the Analog World:
Challenges and Opportunities

February 18, 2010
Boris Murmann
murmann@stanford.edu



Murmann
          Mixed-Signal Group
Outline
   A (superficial) snapshot of the semiconductor industry
   The need for analog interfaces
   Research examples
       Digitally assisted A/D Converters
       Medical ultrasound systems
       Structural health monitoring
       Brain/human machine interfaces
       Biomolecule detection
       Large area electronics


                                                             2
The Beginning of an Industry
           Transistor               Integrated Circuit
Bardeen, Brattain, Shockley, 1948      Kilby, 1958




                                                         3
    Moore’s Law
    In 1965, Gordon Moore predicted exponential growth in the
     number of transistors per integrated circuit




                                                            4
 … And He was Right




[ITRS 2007]




                      5
A Gigantic (Economic) Feedback Loop

  Science




                          Engineering/
    +       Constraints
                           Design Mill
                                         State-of-the-Art



                                                            Time

                           Leverage
                          Mechanism



        $$$
                                                                   6
State-of-the-Art Semiconductor Fab
   E.g. Intel’s “Fab 32” (Chandler, Arizona) ~ $3 Billion




                                                             7
State-of-the-Art Silicon Chips

                              Single-Chip GSM Radio
                             [Staszewski, ISSCC 2008]




       Intel “Tukwila”
    >2 Billion Transistors



                                                        8
45nm Technology (Intel)




                           Steve Cowden
                          THE OREGONIAN
                             July 2007
Transistors Will Continue to Shrink




                                      10
Increasingly Complex Applications




               [Walden Rhines]

                                    11
Outline
   A (superficial) snapshot of the semiconductor industry
   The need for analog interfaces
   Research examples
       Digitally assisted A/D Converters
       Medical ultrasound systems
       Structural health monitoring
       Brain/human machine interfaces
       Biomolecule detection
       Large area electronics


                                                             12
Murmann Mixed-Signal Group




                             13
Sponsors




           14
The Age of “Digital” Stuff




                             15
Wait a Minute…
                  Single-Chip GSM Radio
                 [Staszewski, ISSCC 2008]




                         What‟s this?




                                            16
The Reality of Modern Electronic Systems


                  Signal
                                 A/D
                Conditioning
   Sensors,
 Transducers,              CLK           Signal
  Antennas,                            Processing
     etc.         Signal
                                 D/A
                Conditioning



                        Analog         Digital



                                                    17
Analog Interfaces are Everywhere




                                   18
Analog vs. Digital Speed Performance
                         10000
  Relative Performance




                         1000
                                                     300x


                          100



                           10



                            1
                            1990    2000      2010


                                   20 years


                                                            19
Energy for Digital (NAND) Computation
           10000

                          0.5um
                   1000
                                    0.35um
ENAND2 [fJ/gate]




                    100                        0.25um                               300x
                                                          0.18um
                                                                    0.13um
                    10
                                                                          90nm

                      1
                      1997        1999       2001       2003       2005      2007
                                                Year
                                                                                     20
                Survey Data on A/D Converter Energy
                                     http://www.stanford.edu/~murmann/adcsurvey.html
                     -6
                10

                     -7
                10

                     -8
                10
P/fs [Joules]




                     -9
                10                                                                                             P                   fJ
                                                                                                   FOM                  100
                                                                                                           fs  2ENOB           conv-step
                     -10
                10

                     -11
                10                                                               Flash
                                                                                 Pipeline
                     -12                                                         SAR
                10
                                                                                 Sigma-Delta
                     -13
                                                                                 Other
                10
                          20   30   40      50      60   70            80       90     100   110
                                                   SNDR [dB]
                                                                                                                                 21
A/D Converter Energy Trend
                                                    Slope ~0.5x/2.1years

                       6
     log(P/f s [pJ])




                       4

                       2

        0
     2008
        2006                                            120
           2004                                   100
              2002                         80
                 2000                 60
                    1998         40
              Year          20        SNDR [dB]


   Reduction of “only” ~30x over 10 years
                                                                    22
Digitally Assisted A/D Converters
                                 Additional digital processing for
                                   performance enhancement




                  Signal
                                    A/D
                Conditioning
 Analog Media                                        Signal
                           CLK
     and                                           Processing
 Transducers      Signal
                                    D/A
                Conditioning



                        Analog                     Digital

                                                                23
The Human Ear
   Incredible performance
       P=14mW, DR=120dB, 1GFLOPS (spectral analysis)
   Achieved through clever analog/digital partitioning!




                                                           [Sarpeshkar]



                                                                 24
 Pipeline ADC
                                     Align & Combine Bits
                                                                      Dout

Vin
        SHA        Stage 1                   Stage n-1      Stage n




                                               Vres1
                                      G
                                       G1
                                 -
          ADC      DAC




     Bottleneck: Highly linear, high-speed, low noise amplifier
                                                                             25
Digital Nonlinearity Compensation

    Simplified Amplifier
                                                         Analog
                                                       Nonlinearity


     Vin                                                     Backend
                                            g
                                                               A/D
                                       -
             A/D            D/A
                                           Coefficient
                                           Estimation
                   MOD
      Dout
                                                g-1
             12


                                                          Digital
                                                         Inverse

                                                                       26
High Frequency Linearity of a Typical ADC




                                 National Semiconductor ADC14155
                         14-bit, 155-MS/s, 1.1-GHz Bandwidth CMOS ADC



                       Linearity degradation at high
                        frequencies is problematic for
                        several applications
                         Wireless base stations
                         Instrumentation equipment


                                                                 27
Digital Post-Processing to the Rescue
                       interpolated samples



                                    main sample

                                5
                Interpolation
                    Filter
                                                  15 coefficients

                  X3                X
                  X2                    X                    Dout
                                                      
                                              X
Vin
      A/D




                                                                    28
Experimental Data
   Data from ADS 5474 (Texas Instruments, 14bits, 400MS/s)
   Three training tones spread across frequency range for calibration
                     95
                                                      before calibration
                                                      after calibration
                     90
          SFDR(dB)




                     85


                     80


                     75


                     70
                      220      270            320              370
                            Frequency (MHz)


                                                                           29
Cell Phones vs. Medical Ultrasound
                           Today’s cell phones
                               Leverage highly
                                optimized, power & cost
                                efficient integration

                           Today’s ultrasound
                            machines
                               Lots of progress in
                                miniaturization
                               Insufficient progress in
                                holistic system
                                optimization, SoC
                                integration and cost
                                reduction
Today’s System Architecture

                 [TI]

                               Digitization path
                               (~12b, 50MS/s)
 N = 16…512                     using discrete
Piezo Elements                   components
                              replicated N times
Working on the Next-Gen Front-End
Fly-by-Feel Aircraft
                           Fly-by-Feel Autonomous Vehicle




                                       Intelligent, Self-Repairing Materials

  Multi-Scale Fabrication and
          Integration




                                  Stretchable Network
                                                                      [Fu-Kuo Chang]


                                                                                       33
Stretchable Network




                      34
Interface for Structural Heath Monitoring




                                            35
Microarray Biomolecule Detection




   Expensive ($10k-$100k)
   Bulky (100-200 lbs)
   Available only in a laboratory
Bio-Sensor Chip




                  37
 The Spin-Valve as a Biosensor
                                                1
                                   Spin-Valve


  Spin-Valve                                    2   Sensor resistance
 sensor array                                       changes from 2  3

                                   Spin-Valve       R
                                                                   3
                    Magnetic                    3                  net
                   Nanoparticles                          2        signal
Magnetic biochip   H field
                                                                        t
                                                        DR ~ 10m ~ 5ppm
                                   Spin-Valve


                                                                       38
Test Station




               39
DUTs




       40
Typical Detection Experiment




                               41
Recognition




      IEEE Student Humanitarian
           Supreme Prize
       NanoLab: A Hand-Held
       Diagnostic Laboratory
     Drew Hall and Richard Gaster
  Stanford University, California, USA



                                         42
Neural Prosthetics
   Cortical motor prosthetics
       Neurons in the motor cortical areas of the brain encode
        information about intended movement




         Courtesy K.V. Shenoy                 Courtesy L.R. Hochberg     43
                                              Nature Magazine June „06
Neural Signal Acquisition
     Electrode signals consist of multiple sources
          DC Offset, about 15mV from electrode/tissue interface
          Local field potential (LFP), ≤3mV peak, 10Hz to 100Hz
          Spikes from nearby neurons, 35μV – 1mV peak, 500Hz to 5kHz




    Courtesy M. Sahani                          Courtesy C.L. Klaver


                                                                       44
Tw-Channel Interface Pixel

           SAR ADC




           Frontend


                             45
Chip Layout (96 channels, 5mm x 5mm)




                                       46
Other “Human Interfaces” in Sight



                     “3-D Gesture-Understanding”
                              Samsung


   “Digital Hand”
 Michael Linderman
    Norconnect




                                                   47
Telematics



                                                    Brian Stark
                                                   BOSCH RTC




        Just point the phone at a restaurant and
       download the menu or make a reservation
                                                            48
The Future?




              49
Organic Semiconductors
                   Mechanically flexible
                   Suitable for solution processing
                       Cover large areas at low cost
                       Make disposable devices




                                                   50
Our Driver: Jelly Fish Autonomous Node




      http://muri.mse.vt.edu/


                                     51
Organic Circuit Process
                                                                                          [Klauk, Nature 2007]
                    OFET
                                                                                                          Capacitor
     Source (Au)                        Drain (Au)
                                                                                        Via
              Organic Semiconductor                                                                              Metal 2

                      SAM (OTS)
                   Dielectric (Al2O3)

                      Gate (Al)                                                                                  Metal 1
                      Gate (Al)




                                                     Substrate (glass, plastic, etc.)




   Al2O3/SAM dielectric is only 5.7 nm thick to minimize
    charge trapping
   Channel length ~20mm
                                                                                                                           52
The “First” Organic A/D Converter

                    W. Xiong, U. Zschieschang, H. Klauk, and B.
                    Murmann, “A 3V, 6b Successive
                    Approximation ADC using Complementary
                    Organic Thin-Film Transistors on Glass.”

                    Abstract
                    A successive approximation ADC is implemented
                    in an organic thin film process. The design uses
                    an auto-zeroed, inverter-based comparator and
                    on-chip calibration to achieve 6b precision in
                    presence of large component variations. The ADC
                    is fabricated on a glass substrate and uses 53 p-
                    and n-type transistors and 19 capacitors.




                                                                53
Summary
   More than 50 years of R&D in integrated circuit
    technology & design have created an incredibly powerful
    platform for electronic information processing
       We are only beginning to understand what we can do with the
        billions of transistors available today
   Virtually all of today’s systems rely on highly
    sophisticated analog-digital interfaces
       This makes my job exciting
   No apparent “technological” end of innovation in sight


                                                                  54
Limits to Growth?

                               ~2050




                       Source: European
                    Nanotechnology Roadmap




                                             55
Limits to Growth?


                          ~100x




                       Source: European
                    Nanotechnology Roadmap




                                             56

								
To top