Mixedsignal Circuits For The Datadriven World

• doi: 10.1109/EDSSC.2016.7785197
• title: Mixed-signal circuits for the data-driven world
• publisher: IEEE
• isbn: 978-1-5090-1831-4
• issn:
• rank: 3016
• access_type: LOCKED
• content_type: Conferences
• abstract: In today's world, we have come to depend upon highly sophisticated and tightly interwoven electronic systems that facilitate sensing, communication and computing tasks. However, as we move forward, many grand challenges exist in raising the performance of such systems to the next level. For instance, the energy consumption of most sensors and data transceivers is still far too high to operate autonomously and for extended periods of time. On the other end of the spectrum, the power levels required to run sophisticated machine learning algorithms still appears too large for exponential growth and/or massive deployment. In the past, the desired progress was often fueled by engineering better components and by leveraging the immense improvements delivered by CMOS feature size scaling. For example, mixed-signal designers have invested an immense amount of time to port and optimize data converters from node-to-node and have thereby enabled wider bandwidths at steadily lower conversion energies [1], Similarly, RF designers have leveraged the high transit frequencies of modern devices to enable low-noise RF signal processing up to mm-wave frequencies. As we enter the next stage of innovation, it is becoming increasingly clear that future progress will come from a “top-down,” system-driven approach. The sheer complexity of today's systems provides us with a new degree of freedom wherein the system architecture and their underlying components (and possibly transistors) can be co-designed, thus providing larger gains than those enabled within the usual scaling and porting framework. This talk will provide an overview of my group's research on system-driven mixed-signal design, which is guided by the parallel consideration of theory, systems, algorithms and circuits. For instance, we are researching a number of sensing approaches aimed at “fooling Nyquist” and extracting desired analog-domain information using low-rate and low-bandwidth observations. In a second line of work, we investigate the application of mixed-signal co-processors in machine learning algorithms [2], tailored for example toward low-power image classification.
• article_number: 7785197
• pdf_url: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7785197
• html_url: https://ieeexplore.ieee.org/document/7785197/
• abstract_url: https://ieeexplore.ieee.org/document/7785197/
• publication_title: 2016 IEEE International Conference on Electron Devices and Solid-State Circuits (EDSSC)
• conference_location: Hong Kong, China
• conference_dates: 3-5 Aug. 2016
• publication_number: 7764658
• is_number: 7784488
• publication_year: 2016
• publication_date: 3-5 Aug. 2016
• start_page: 4
• end_page: 4
• citing_paper_count: 0
• citing_patent_count: 0
• download_count: 353
• insert_date: 20161219
• index_terms:
  • ieee_terms:
    • Decision support systems
  • author_terms:
    • analog circuits
    • CMOS
    • low-power
    • machine learning
    • mixed-signal circuits
    • technology scaling
  • dynamic_index_terms:
    • Mixed-signal Circuits
    • Mixed-signal Integrated Circuits
    • Learning Algorithms
    • Machine Learning Algorithms
    • Stanford
• isbn_formats:
  • format: Print on Demand(PoD) ISBN, value: 978-1-5090-1831-4, isbnType: New-2005
  • format: Electronic ISBN, value: 978-1-5090-1830-7, isbnType: New-2005
• authors:
  • Author Name: Boris Murmann
    Affiliation: Stanford University, Stanford, CA, US
    Author URL: https://ieeexplore.ieee.org/author/37300093100
    ID: 37300093100
    Order: 1
    Author Affiliations:
    • Stanford University, Stanford, CA, US

Image Sensor

• sensor_type: CMOS
• resolution: 1920x1080
• dynamic_range: 70 dB
• pixel_size: 1.2 µm
• dark_current: 5 e-/s

Optical Data

• focal_length: 25 mm
• aperture: f/2.8
• field_of_view: 45°
• distortion: 2%

Performance Metrics

• frame_rate: 60 fps
• signal_to_noise_ratio: 40 dB
• sensitivity: ISO 3200
• shutter_speed: 1/125s
• power_consumption: 300 mW
• noise: 2 e-

Applications & Benefits

• cell_imaging: Used for high-resolution imaging in medical diagnostics.
• benefits: High sensitivity and low power consumption, suitable for mobile and portable devices.

Supporting Organizations

• supported_by: IEEE, SEMI, and other industry consortiums.

Manuscript Details

• publication_date: 3-5 Aug. 2016

Relevancy Score

• score: 9
• missing_fields:
  • specific design techniques
  • case studies of performance improvements

Processed JSON Filename

request_fef61c24-3c42-4349-ae44-c0cafd76de9d-mixedsignal_circuits_for_the_datadriven_world.json