header-about

EECS Spring 2014 Department Seminar

Presenter: 
Soumyajit Mandal
Title: 
Analog Circuits for Ultra-Broadband Sensing Applications
Affiliation: 
Faculty Candidate
Location: 
White Bldg., Room 411
Time: 
11:30 AM - 12:30 PM
Date: 
April 3, 2014

 In this talk I will describe some difficult sensing problems and how they can be solved using clever circuit design. My first example deals with broadband spectral analysis. Our work was inspired by the biological cochlea, or inner ear, which robustly analyzes incoming sounds over a frequency range of three decades (20 Hz to 20 kHz) on a very limited power budget. We began by mapping the differential equations that model acoustic wave propagation in the cochlea into dynamically equivalent integrated circuits. I shall show that the resulting exponentially tapered transmission-line structure performs constant-fractional-resolution spectrum analysis with O(N) expenditure of both analysis time and hardware, where N is the number of analyzed frequency bins. This is the most efficient scaling relationship of any spectrum-analysis algorithm that we are aware of, including the constant-resolution FFT, which scales as O(N log(N)).  This biologically-inspired algorithm therefore appears promising for performing real-time, on-chip RF spectrum analysis. Our prototype chip, implemented in standard 0.13 μm CMOS technology, decomposes the RF spectrum from 600 MHz to 8 GHz into 50 log-spaced channels, consumes less than 300 mW of power, and has 70 dB of dynamic range. Such real-time spectrum analysis capabilities make it attractive for ultra-broadband cognitive radio systems.
My second example deals with the front-ends of nuclear magnetic resonance (NMR) and nuclear quadrupole resonance (NQR) spectrometers. These systems commonly use resonant circuits for efficient RF transmission and low-noise reception. However, such tuned circuits are narrow-band analog devices that are not suitable for broadband and multi-frequency experiments. We have addressed this issue by developing an ultra-broadband or non-resonant MR front-end that operates in the 0.1-3 MHz frequency range without using conventional resonant circuits for either transmission or reception. The system uses one of the lowest-noise preamplifiers ever reported, with an input-referred noise that is lower than that of a 1.5 Ω resistor over a 3 MHz bandwidth at room temperature. It thus allows robust multi-frequency operation while retaining efficient RF power transmission and low-noise reception. I will demonstrate the usefulness of our approach through a variety of novel NMR and NQR experiments in this frequency range, which is of particular interest for field applications such as oil well logging, rock core analysis, and detection of illegal explosives and drugs.

Biography: 

Soumyajit Mandal received a Ph.D. in Electrical Engineering from MIT in 2009. His doctoral research focused on biologically-inspired integrated circuits and energy-efficient biomedical electronics, and received the MIT Microsystems Technology Laboratories doctoral dissertation award. He has been a scientist at the Schlumberger-Doll research center in Cambridge since early 2010. His recent work has focused on novel hardware and pulse sequences for nuclear magnetic resonance (NMR) in weak and inhomogeneous magnetic fields.