Matching Items (61)
Filtering by
- Genre: Masters Thesis
Description
In thesis, a test time reduction (a low cost test) methodology for digitally-calibrated pipeline analog-to-digital converters (ADCs) is presented. A long calibration time is required in the final test to validate performance of these designs. To reduce total test time, optimized calibration technique and calibrated effective number of bits (ENOB) prediction from calibration coefficient will be presented. With the prediction technique, failed devices can be identified only without actual calibration. This technique reduces significant amount of time for the total test time.
ContributorsKim, Kibeom (Author) / Ozev, Sule (Thesis advisor) / Kitchen, Jennifer (Committee member) / Barnaby, Hugh (Committee member) / Arizona State University (Publisher)
Created2013
Description
With the advent of parallel processing, primarily the time-interleaved pipeline ADCs, high speed and high resolution ADCs became a possibility. When these speeds touch giga samples per second and resolutions go beyond 12-bits, the parallelization becomes more extensive leading to repeated presence of several identical blocks in the architecture. This thesis discusses one such block, the sub-ADC (Flash ADC), of the pipeline and sharing it with more than two of the parallel processing channels thereby reducing area and power and input load capacitance to each stage. This work presents a design of 'sub-ADC shared in a time-interleaved pipeline ADC' in the IBM 8HP process. It has been implemented with an offset-compensated, kickback-compensated, fast decision making (large input bandwidth) and low power comparator that forms the core part of the design.
ContributorsBikkina, Phaneendra Kumar (Author) / Barnaby, Hugh (Thesis advisor) / Mikkola, Esko (Committee member) / Kitchen, Jennifer (Committee member) / Arizona State University (Publisher)
Created2013
Description
In this thesis two methodologies have been proposed for evaluating the fault response of analog/RF circuits. These proposed approaches are used to evaluate the response of the faulty circuit in terms of specifications/measurements. Faulty response can be used to evaluate important test metrics like fail probability, fault coverage and yield coverage of given measurements under process variations. Once the models for faulty and fault free circuit are generated, one needs to perform Monte Carlo sampling (as opposed to Monte Carlo simulations) to compute these statistical parameters with high accuracy. The first method is based on adaptively determining the order of the model based on the error budget in terms of computing the statistical metrics and position of the threshold(s) to decide how precisely necessary models need to be extracted. In the second method, using hierarchy in process variations a hybrid of heuristics and localized linear models have been proposed. Experiments on LNA and Mixer using the adaptive model order selection procedure can reduce the number of necessary simulations by 7.54x and 7.03x respectively in the computation of fail probability for an error budget of 2%. Experiments on LNA using the hybrid approach can reduce the number of necessary simulations by 21.9x and 17x for four and six output parameters cases for improved accuracy in test statistics estimation.
ContributorsSubrahmaniyan Radhakrishnan, Gurusubrahmaniyan (Author) / Ozev, Sule (Thesis advisor) / Blain Christen, Jennifer (Committee member) / Cao, Yu (Committee member) / Arizona State University (Publisher)
Created2010
Description
An investigation of phase noise in amplifier and voltage-controller oscillator (VCO) circuits was conducted to show that active direct-current (DC) bias techniques exhibit lower phase noise performance than traditional resistive DC bias techniques. Low-frequency high-gain amplifiers like those found in audio applications exhibit much better 1/f phase noise performance and can be used to bias amplifier or VCO circuits that work at much higher frequencies to reduce the phase modulation caused by higher frequency devices. An improvement in single-side-band (SSB) phase noise of 15 dB at offset frequencies less than 50 KHz was simulated and measured. Residual phase noise of an actively biased amplifier also exhibited significant noise improvements when compared to an equivalent resistive biased amplifier.
ContributorsBaldwin, Jeremy Bart (Author) / Aberle, James T., 1961- (Thesis advisor) / Bakkaloglu, Bertan (Committee member) / Ozev, Sule (Committee member) / Arizona State University (Publisher)
Created2010
Description
Portable health diagnostic systems seek to perform medical grade diagnostics in non-ideal environments. This work details a robust fault tolerant portable health diagnostic design implemented in hardware, firmware and software for the detectionof HPV in low-income countries. The device under device under test (DUT) is a fluorescence based lateral flow assay (LFA) point-of-care (POC) device. This work’s contributions are: firmware and software development, calibration routine implementation, device performance characterization and a proposed method of in-software fault detection. Firmware was refactored from the original implementation of the POC fluorescence reader to expose an application programming interface (API) via USB. Companion software available for desktop environments (Windows, Mac and Linux) was created to interface with this firmware API and conduct macro level routines to request and receive fluorescence data while presenting a user-friendly interface to clinical technicians. Lastly, an environmental chamber was constructed to conduct sequential diagnostic reads in order to observe sensor drift and other deviations that might present themselves in real-world usage. The results from these evaluations show a standard deviation of less than 1% in fluorescence readings in nominal temperature environments (approx. 25C) suggesting that this system will have a favorable signal-to-noise (SNR) ratio in such a setting. In non-ideal over heated environments (≥38C), the evaluation results showed performance degradation with standard deviations as large as 15%.
ContributorsLue Sang, Christopher David (Author) / Blain Christen, Jennifer M (Thesis advisor) / Ozev, Sule (Committee member) / Goryll, Michael (Committee member) / Raupp, Gregory (Committee member) / Arizona State University (Publisher)
Created2022
Description
Readout Integrated Circuits(ROICs) are important components of infrared(IR) imag
ing systems. Performance of ROICs affect the quality of images obtained from IR
imaging systems. Contemporary infrared imaging applications demand ROICs that
can support large dynamic range, high frame rate, high output data rate, at low
cost, size and power. Some of these applications are military surveillance, remote
sensing in space and earth science missions and medical diagnosis. This work focuses
on developing a ROIC unit cell prototype for National Aeronautics and Space Ad
ministration(NASA), Jet Propulsion Laboratory’s(JPL’s) space applications. These
space applications also demand high sensitivity, longer integration times(large well
capacity), wide operating temperature range, wide input current range and immunity
to radiation events such as Single Event Latchup(SEL).
This work proposes a digital ROIC(DROIC) unit cell prototype of 30ux30u size,
to be used mainly with NASA JPL’s High Operating Temperature Barrier Infrared
Detectors(HOT BIRDs). Current state of the art DROICs achieve a dynamic range
of 16 bits using advanced 65-90nm CMOS processes which adds a lot of cost overhead.
The DROIC pixel proposed in this work uses a low cost 180nm CMOS process and
supports a dynamic range of 20 bits operating at a low frame rate of 100 frames per
second(fps), and a dynamic range of 12 bits operating at a high frame rate of 5kfps.
The total electron well capacity of this DROIC pixel is 1.27 billion electrons, enabling
integration times as long as 10ms, to achieve better dynamic range. The DROIC unit
cell uses an in-pixel 12-bit coarse ADC and an external 8-bit DAC based fine ADC.
The proposed DROIC uses layout techniques that make it immune to radiation up to
300krad(Si) of total ionizing dose(TID) and single event latch-up(SEL). It also has a
wide input current range from 10pA to 1uA and supports detectors operating from
Short-wave infrared (SWIR) to longwave infrared (LWIR) regions.
ing systems. Performance of ROICs affect the quality of images obtained from IR
imaging systems. Contemporary infrared imaging applications demand ROICs that
can support large dynamic range, high frame rate, high output data rate, at low
cost, size and power. Some of these applications are military surveillance, remote
sensing in space and earth science missions and medical diagnosis. This work focuses
on developing a ROIC unit cell prototype for National Aeronautics and Space Ad
ministration(NASA), Jet Propulsion Laboratory’s(JPL’s) space applications. These
space applications also demand high sensitivity, longer integration times(large well
capacity), wide operating temperature range, wide input current range and immunity
to radiation events such as Single Event Latchup(SEL).
This work proposes a digital ROIC(DROIC) unit cell prototype of 30ux30u size,
to be used mainly with NASA JPL’s High Operating Temperature Barrier Infrared
Detectors(HOT BIRDs). Current state of the art DROICs achieve a dynamic range
of 16 bits using advanced 65-90nm CMOS processes which adds a lot of cost overhead.
The DROIC pixel proposed in this work uses a low cost 180nm CMOS process and
supports a dynamic range of 20 bits operating at a low frame rate of 100 frames per
second(fps), and a dynamic range of 12 bits operating at a high frame rate of 5kfps.
The total electron well capacity of this DROIC pixel is 1.27 billion electrons, enabling
integration times as long as 10ms, to achieve better dynamic range. The DROIC unit
cell uses an in-pixel 12-bit coarse ADC and an external 8-bit DAC based fine ADC.
The proposed DROIC uses layout techniques that make it immune to radiation up to
300krad(Si) of total ionizing dose(TID) and single event latch-up(SEL). It also has a
wide input current range from 10pA to 1uA and supports detectors operating from
Short-wave infrared (SWIR) to longwave infrared (LWIR) regions.
ContributorsPraveen, Subramanya Chilukuri (Author) / Bakkaloglu, Bertan (Thesis advisor) / Kitchen, Jennifer (Committee member) / Long, Yu (Committee member) / Arizona State University (Publisher)
Created2019
Description
High-efficiency DC-DC converters make up one of the important blocks of state-of-the-art power supplies. The trend toward high level of transistor integration has caused load current demands to grow significantly. Supplying high output current and minimizing output current ripple has been a driving force behind the evolution of Multi-phase topologies. Ability to supply large output current with improved efficiency, reduction in the size of filter components, improved transient response make multi-phase topologies a preferred choice for low voltage-high current applications.
Current sensing capability inside a system is much sought after for applications which include Peak-current mode control, Current limiting, Overload protection. Current sensing is extremely important for current sharing in Multi-phase topologies. Existing approaches such as Series resistor, SenseFET, inductor DCR based current sensing are simple but their drawbacks such low efficiency, low accuracy, limited bandwidth demand a novel current sensing scheme.
This research presents a systematic design procedure of a 5V - 1.8V, 8A 4-Phase Buck regulator with a novel current sensing scheme based on replication of the inductor current. The proposed solution consists of detailed system modeling in PLECS which includes modification of the peak current mode model to accommodate the new current sensing element, derivation of power-stage and Plant transfer functions, Controller design. The proposed model has been verified through PLECS simulations and compared with a transistor-level implementation of the system. The time-domain parameters such as overshoot and settling-time simulated through transistor-level
implementation is in close agreement with the results obtained from the PLECS model.
Current sensing capability inside a system is much sought after for applications which include Peak-current mode control, Current limiting, Overload protection. Current sensing is extremely important for current sharing in Multi-phase topologies. Existing approaches such as Series resistor, SenseFET, inductor DCR based current sensing are simple but their drawbacks such low efficiency, low accuracy, limited bandwidth demand a novel current sensing scheme.
This research presents a systematic design procedure of a 5V - 1.8V, 8A 4-Phase Buck regulator with a novel current sensing scheme based on replication of the inductor current. The proposed solution consists of detailed system modeling in PLECS which includes modification of the peak current mode model to accommodate the new current sensing element, derivation of power-stage and Plant transfer functions, Controller design. The proposed model has been verified through PLECS simulations and compared with a transistor-level implementation of the system. The time-domain parameters such as overshoot and settling-time simulated through transistor-level
implementation is in close agreement with the results obtained from the PLECS model.
ContributorsBurli, Venkatesh (Author) / Bakkaloglu, Bertan (Thesis advisor) / Garrity, Douglas (Committee member) / Kitchen, Jennifer (Committee member) / Arizona State University (Publisher)
Created2017
Description
In this work, the development of a novel and a truly in-shoe force measurement system is reported. The device consists of a shoe insole with six thin film piezoresistive sensors and the main circuit board. The piezoresistive sensors are used for the measurement of plantar pressure during daily human activities. The motion sensor mounted on the main circuit board captures kinematic data. In addition, the main circuit board is responsible for the wireless transmission of the data from all the sensors in real-time using BLE protocol. It is housed within the midsole of the shoe, under the medial arch of the foot. The real-time quantitative data and its analyses, enables athletic performance evaluation, biomedical ailment detection, and everyday fitness tracking. A test subject walked 20 steps on a flat surface at a comfortable speed wearing a shoe fitted with the insole and the main circuit board. Measurements were captured using a BLE enabled laptop and the test results were validated for accuracy. From the real-time data captured, the number of steps walked, the speed and the plantar pressure applied can be clearly established. Moreover, additional kinematic data from the motion sensor was captured. Further processing of kinematic data using techniques such as machine learning is essential to get meaningful inferences.
ContributorsBadarinath, Abhishek (Author) / Kiaei, Sayfe (Thesis advisor) / Bakkaloglu, Bertan (Committee member) / Kitchen, Jennifer (Committee member) / Arizona State University (Publisher)
Created2018
Description
The photovoltaic systems used to convert solar energy to electricity pose a multitude of design and implementation challenges, including energy conversion efficiency, partial shading effects, and power converter efficiency. Using power converters for Distributed Maximum Power Point Tracking (DMPPT) is a well-known architecture to significantly reduce power loss associated with mismatched panels. Sub-panel-level DMPPT is shown to have up to 14.5% more annual energy yield than panel-level DMPPT, and requires an efficient medium power converter.
This research aims at implementing a highly efficient power management system at sub-panel level with focus on system cost and form-factor. Smaller form-factor motivates increased converter switching frequencies to significantly reduce the size of converter passives and substantially improve transient performance. But, currently available power MOSFETs put a constraint on the highest possible switching frequency due to increased switching losses. The solution is Gallium Nitride based power devices, which deliver figure of merit (FOM) performance at least an order of magnitude higher than existing silicon MOSFETs. Low power loss, high power density, low cost and small die sizes are few of the qualities that make e-GaN superior to its Si counterpart. With careful design, e-GaN can enable a 20-30% improvement in power stage efficiency compared to converters using Si MOSFETs.
The main objective of this research is to develop a highly integrated, high efficiency, 20MHz, hybrid GaN-CMOS DC-DC MPPT converter for a 12V/5A sub-panel. Hard and soft switching boost converter topologies are investigated within this research, and an innovative CMOS gate drive technique for efficiently driving an e-GaN power stage is presented in this work. The converter controller also employs a fast converging analog MPPT control technique.
This research aims at implementing a highly efficient power management system at sub-panel level with focus on system cost and form-factor. Smaller form-factor motivates increased converter switching frequencies to significantly reduce the size of converter passives and substantially improve transient performance. But, currently available power MOSFETs put a constraint on the highest possible switching frequency due to increased switching losses. The solution is Gallium Nitride based power devices, which deliver figure of merit (FOM) performance at least an order of magnitude higher than existing silicon MOSFETs. Low power loss, high power density, low cost and small die sizes are few of the qualities that make e-GaN superior to its Si counterpart. With careful design, e-GaN can enable a 20-30% improvement in power stage efficiency compared to converters using Si MOSFETs.
The main objective of this research is to develop a highly integrated, high efficiency, 20MHz, hybrid GaN-CMOS DC-DC MPPT converter for a 12V/5A sub-panel. Hard and soft switching boost converter topologies are investigated within this research, and an innovative CMOS gate drive technique for efficiently driving an e-GaN power stage is presented in this work. The converter controller also employs a fast converging analog MPPT control technique.
ContributorsKrishnan Achary, Kiran Kumar (Author) / Kitchen, Jennifer (Thesis advisor) / Kiaei, Sayfe (Committee member) / Bakkaloglu, Bertan (Committee member) / Arizona State University (Publisher)
Created2015
Description
This thesis presents efficient implementations of several linear algebra kernels, machine learning kernels and a neural network based recommender systems engine onto a massively parallel reconfigurable architecture, Transformer. The linear algebra kernels include Triangular Matrix Solver (TRSM), LU Decomposition (LUD), QR Decomposition (QRD), and Matrix Inversion. The machine learning kernels include an LSTM (Long Short Term Memory) cell, and a GRU (gated Recurrent Unit) cell used in recurrent neural networks. The neural network based recommender systems engine consists of multiple kernels including fully connected layers, embedding layer, 1-D batchnorm, Adam optimizer, etc.
Transformer is a massively parallel reconfigurable multicore architecture designed at the University of Michigan. The Transformer configuration considered here is 4 tiles and 16 General Processing Elements (GPEs) per tile. It supports a two level cache hierarchy where the L1 and L2 caches can operate in shared (S) or private (P) modes. The architecture was modeled using Gem5 and cycle accurate simulations were done to evaluate the performance in terms of execution times, giga-operations per second per Watt (GOPS/W), and giga-floating-point-operations per second per Watt (GFLOPS/W).
This thesis shows that for linear algebra kernels, each kernel achieves high performance for a certain cache mode and that this cache mode can change when the matrix size changes. For instance, for smaller matrix sizes, L1P, L2P cache mode is best for TRSM, while L1S, L2S is the best cache mode for LUD, and L1P, L2S is the best for QRD. For each kernel, the optimal cache mode changes when the matrix size is increased. For instance, for TRSM, the L1P, L2P cache mode is best for smaller matrix sizes ($N=64, 128, 256, 512$) and it changes to L1S, L2P for larger matrix sizes ($N=1024$). For machine learning kernels, L1P, L2P is the best cache mode for all network parameter sizes.
Gem5 simulations show that the peak performance for TRSM, LUD, QRD and Matrix Inverse in the 14nm node is 97.5, 59.4, 133.0 and 83.05 GFLOPS/W, respectively. For LSTM and GRU, the peak performance is 44.06 and 69.3 GFLOPS/W.
The neural network based recommender system was implemented in L1S, L2S cache mode. It includes a forward pass and a backward pass and is significantly more complex in terms of both computational complexity and data movement. The most computationally intensive block is the fully connected layer followed by Adam optimizer. The overall performance of the recommender systems engine is 54.55 GFLOPS/W and 169.12 GOPS/W.
Transformer is a massively parallel reconfigurable multicore architecture designed at the University of Michigan. The Transformer configuration considered here is 4 tiles and 16 General Processing Elements (GPEs) per tile. It supports a two level cache hierarchy where the L1 and L2 caches can operate in shared (S) or private (P) modes. The architecture was modeled using Gem5 and cycle accurate simulations were done to evaluate the performance in terms of execution times, giga-operations per second per Watt (GOPS/W), and giga-floating-point-operations per second per Watt (GFLOPS/W).
This thesis shows that for linear algebra kernels, each kernel achieves high performance for a certain cache mode and that this cache mode can change when the matrix size changes. For instance, for smaller matrix sizes, L1P, L2P cache mode is best for TRSM, while L1S, L2S is the best cache mode for LUD, and L1P, L2S is the best for QRD. For each kernel, the optimal cache mode changes when the matrix size is increased. For instance, for TRSM, the L1P, L2P cache mode is best for smaller matrix sizes ($N=64, 128, 256, 512$) and it changes to L1S, L2P for larger matrix sizes ($N=1024$). For machine learning kernels, L1P, L2P is the best cache mode for all network parameter sizes.
Gem5 simulations show that the peak performance for TRSM, LUD, QRD and Matrix Inverse in the 14nm node is 97.5, 59.4, 133.0 and 83.05 GFLOPS/W, respectively. For LSTM and GRU, the peak performance is 44.06 and 69.3 GFLOPS/W.
The neural network based recommender system was implemented in L1S, L2S cache mode. It includes a forward pass and a backward pass and is significantly more complex in terms of both computational complexity and data movement. The most computationally intensive block is the fully connected layer followed by Adam optimizer. The overall performance of the recommender systems engine is 54.55 GFLOPS/W and 169.12 GOPS/W.
ContributorsSoorishetty, Anuraag (Author) / Chakrabarti, Chaitali (Thesis advisor) / Kim, Hun Seok (Committee member) / LiKamWa, Robert (Committee member) / Arizona State University (Publisher)
Created2019