Discrete charge-trapping flash memories are more promising than their floating-gate counterparts due to their physically discrete-trapping and coupling-free nature. Si3N4 is conventional material as charge-trapping layer (CTL) for charge storage. The shortcomings of Si3N4 are its low dielectric constant and small barrier height at its interface with SiO2 tunneling layer. Therefore, this research aims to investigate new materials as CTL for improving the performance of the memory devices. The charge-trapping characteristics of La2O3 with and without nitrogen incorporation were investigated. Compared with the memory device with La2O3 as CTL, the one with nitrided La2O3 (LaON) showed larger memory window, higher program/erase (P/E) speeds and smaller charge loss, due to the nitrided La2O3 film exhibiting less crystallized structure, higher trap density induced by nitrogen incorporation, and suppressed leakage by nitrogen passivation. In order to further improve the performance of the memory device with LaON CTL, a device with band-engineered LaTiON/LaON structure as CTL was also explored, and demonstrated to have better performance than the one with LaON CTL. This was ascribed to the variable tunneling path of charge carriers under P/E and retention modes (realized by the band-engineered charge-trapping layer), high trap density of LaTiON, and large barrier height at the LaTiON/SiO interface. SrTiO 3and BaTiO3 ,both ofwhich are typical perovskite-type dielectrics, also possess distinguished characteristics as CTL, including remarkably high dielectric constant and large conduction-band offset relative to SiO2. The charge-trapping properties of SrTiO3 with and without fluorine incorporation were studied. The device with fluorinated SrTiO3 film showed promising performance in terms of higher P/E speeds at a low gate voltage, better endurance and data retention compared with that without fluorine treatment. These advantages were associated with generated deep-level traps, reduced leakage path, and enhanced strength of the film due to the highest electro-negativity of the fluorine atoms incorporated in the film. The charge-trapping properties of BaTiO3 with and without Zr incorporation were also investigated, where Zr incorporated in BaTiO3 could strengthen the dielectric film and improve its thermodynamic stability. The device with Zr incorporation exhibited similar memory window as the one without Zr incorporation, but higher program speed at low gate voltage, better endurance and data retention, due to the Zr-doped BaTiO3 exhibiting higher charge-trapping efficiency and higher density of traps with deeper energy levels. Besides nitride-based memories, nanocrystal-based memories are another type of charge-trapping memories, where nanocrystals (NCs) embedded into a dielectric are used for charge storage. Memory devices with Ga2O3 NCs as CTL were investigated, which are compatible with the CMOS process. The Ga2O3 NCs displayed higher trap density than the Ga2O3 dielectric film. Moreover, compared with the device with Ga2O 3NCs as CTL, the one with nitrided Ga2O3 NCs showed larger memory window, higher operating speed and better data retention, mainly due to higher charge-trapping efficiency of the nitrided Ga2O3 NCs and nitrogen-induced suppressed formation of interlayer at the Ga2O/SiO interface. / published_or_final_version / Electrical and Electronic Engineering / Doctoral / Doctor of Philosophy

Discrete charge-trapping flash memory is being developed for the next-generation commercial flash-memory applications due to its advantages over the traditional floating-gate counterpart. Currently, Si3N4 is widely used as charge-trapping layer (CTL). However, Si3N4 has low dielectric constant and small conduction-band offset with respect to the SiO2 tunneling layer, imposing limitation on further applications. Therefore, this research emphasized on investigating new dielectrics with appropriate fabrication methods to replace Si3N4 as CTL for achieving improved memory performance. Firstly, GeON CTL annealed at different temperatures was investigated. The memory device with post-deposition annealing at 600 0C exhibited the largest memory window, the best charge retention performance, and the highest reliability. These good results are due to the fact that optimal annealing temperature could suppress shallow traps and also produce new traps with desirable energy levels in the CTL. Since ZnON has a negative conduction-band offset (NCBO) with respect to Si, the traps located in the bandgap of ZnON should have deep energy levels. The memory performances of ZrON film with and without Zn doping were studied. Experimental results showed that ZrZnON film had higher program speed and better charge retention performance due to many deeper trap levels induced by the Zn doping, as well as higher erase speed due to the direct recombination of electrons at these deeper trap levels with incoming holes and the intermediary role of these deeper trap levels under erase mode. MoO3 is another NCBO dielectric with a high K value and many oxygen vacancies. La2O3, a rare-earth metal oxide, is a promising dielectric as CTL. To combine the advantages of both La2O3 and MoO3, Mo-doped La2O3 was proposed as a new CTL. Compared to the device with pure La2O3, the one with LaMoO film as CTL had significantly larger C-V hysteresis window, much higher P/E speeds, and better charge retention due to the deeper-level traps and deeper quantum wells created by the LaMoO film. Nitrogen incorporation is a popular approach to increase the trap density in the bulk of CTL. In this research, the memory performances of GdTiO films with and without nitrogen incorporation were compared. Since the nitrogen incorporation induced smaller equivalent oxide thickness, produced nitride-related traps with desirable energy level and larger cross-section for charge capture, the GdTiON film possessed better memory performance than the GdTiO film. Finally, fluorine plasma was employed to improve the quality of blocking layer. The memory device with AlOF blocking layer obtained higher program speed, better reliability and better charge retention than that based on AlO blocking layer. The improved performance was due to the fact that the fluorine incorporation passivated the defects and removed the excess oxygen in the bulk of the blocking layer. In summary, dielectric plays important roles in the performance of charge-trapping flash memory. Memory devices with GeON, ZrZnON, LaMoO, or GdTiON as charge trapping layer and AlOF as blocking layer can produce large memory window, high program/erase speed and good charge retention. / published_or_final_version / Electrical and Electronic Engineering / Doctoral / Doctor of Philosophy

Not available / text

Computer benchmarking involves running a set of benchmark programs to measure performance of a computer system. Modern benchmarks are developed from real applications. Applications are becoming complex and hence modern benchmarks run for a very long time. These benchmarks are also used for performance evaluation in the early design phase of microprocessors. Due to the size of benchmarks and increase in complexity of microprocessor design, the effort required for performance evaluation has increased significantly. This dissertation proposes methodologies to reduce the effort of benchmarking and performance evaluation of computer systems. Identifying a set of programs that can be used in the process of benchmarking can be very challenging. A solution to this problem can start by identifying similarity between programs to capture the diversity in their behavior before they can be considered for benchmarking. The aim of this methodology is to identify redundancy in the set of benchmarks and find a subset of representative benchmarks with the least possible loss of information. This dissertation proposes the use of program characteristics which capture the performance behavior of programs and identifies representative benchmarks applicable over a wide range of system configurations. The use of benchmark subsetting has not been restricted to academic research. Recently, the SPEC CPU subcommittee used the information derived from measuring similarity based on program behavior characteristics between different benchmark candidates as one of the criteria for selecting the SPEC CPU2006 benchmarks. The information of similarity between programs can also be used to predict performance of an application when it is difficult to port the application on different platforms. This is a common problem when a customer wants to buy the best computer system for his application. Performance of a customer's application on a particular system can be predicted using the performance scores of the standard benchmarks on that system and the similarity information between the application and the benchmarks. Similarity between programs is quantified by the distance between them in the space of the measured characteristics, and is appropriately used to predict performance of a new application using the performance scores of its neighbors in the workload space. / text

The ubiquity of mobile phones provides an opportunity to use them for learning programming beyond the classroom. This would be particularly useful for novice learners of programming in resource-constrained environments. However, limitations of mobile phones, such as small screens and small keypads, impede their use as typical programming environments. This study proposed that mobile programming environments could include scaffolding techniques specifically designed for mobile phones, and designed based on learners’ needs. A six-level theoretic framework was used to design scaffolding techniques to support construction of Java programs on a mobile phone. The scaffolding techniques were implemented on an Android platform. Using the prototype, three experiments were conducted with 182 learners of programming from four universities in South Africa and Kenya. Evaluation was conducted to investigate: (i) which scaffolding techniques could support the construction of Java programs on a mobile phone; and (ii) the effect on learners of using these scaffolding techniques to construct Java programs on a mobile phone. Data was collected using computer logs, questionnaires, and image and video recordings. It was found that static scaffolding, such as a program overview and constructing a program one part at a time, supported the construction of programs on a mobile phone. It was also found that automatic scaffolding, such as error prompts and statement dialogs, and user-initiated scaffolding, such as viewing of the full program while creating parts of a program, supported learners to construct programs on the mobile phone. The study also found that the scaffolding techniques enabled learners to attempt and complete more tasks than a non-scaffolded environment. Further, the scaffolding techniques enabled learners to complete programs efficiently, and captured syntactical errors early during program creation. The results also indicated that after the initial familiarization with the scaffolded environment, the scaffolding techniques could enable faster completion of programs. Learners’ feedback indicated that they found the scaffolding techniques useful in supporting programming on a mobile phone and in meeting learners’ needs. This study provides empirical evidence that scaffolding techniques specifically designed for mobile phones and designed based on learners’ needs could support the construction of programs on a mobile phone.

The present Internet is used to support a diversity of application types, each with different performance expectations. Latency is a fundamental component that affects end-user's Internet experience. Many applications can benefit from reduced latency. The thesis examines the startup latency of the Standard Transmission Control Protocol (TCP Reno) congestion control algorithm when used for short-lived interactive applications. TCP plays a key role in preventing Internet failure by decreasing network utilization when it detects packet drops or receives explicit congestion signals. TCP Reno senders react to congestion by reducing their rate by approximately half. When a flow has recently started, TCP Reno has no way to determine the available capacity, and uses a conservative startup method that can result in high latency for short-lived interactive applications with only a small amount of data to send. This is especially so when there is loss of the connection setup packet before a flow starts. This thesis therefore proposes a new algorithm called TCP SYN Loss (TSL) Startup Algorithm that uses a halving congestion response function during startup, which is less conservative than the current method. After connection setup is completed, standard TCP congestion control is applied for the data transfer phase. The new method is shown to significantly reduce web response time by increasing TCP Reno robustness against startup loss. Startup latency also affects users who rely on broadband satellite for Internet access. Application performance is impacted by the delay of bandwidth-on-demand (BoD) multiple access control (MAC) protocol at the start of a TCP flow when there are no prior capacity assignments. This thesis proposes a new method that combines Random Access (RA) and BoD to reduce web response time and startup delay of media streaming over satellite. Furthermore, a modified algorithm called the Split RADAMA protocol was designed to avoid the negative transport effects when switching between RA and BoD transmissions e.g. packet reordering and spurious retransmission. Together the proposed methods can significantly improve web experience for a wide range of end-users by reducing the response time for short-lived interactive applications.

Thesis (MTech. degree in Educational Technology) / The current information literacy training programme (ILTP) of the Library and Information Services (LIS) of the Tshwane University of Technology (TUT) is set to be reviewed in such a way as to be presented in a multimodal transfer approach. A baseline evaluation serves to examine the existing programme on Module 5: Electronic Databases, after which an online program thereof is proposed. The components of the proposed multimodal transfer include an important role played by a facilitator, as well as the application of a learning management system (LMS). A bounded case study approach is used, applying an action research strategy that comprises two phases that includes a hybrid methodology of qualitative and quantitative research methods. Evaluating the proposed module, it becomes evident that a multimodal transfer approach is indeed suitable to transfer the necessary literacy skills to use electronic databases. The graphical user interface (GUI) appeals to users, and they report on the ease of use, as well as usefulness for personal studies of the module. In conclusion, the TUT programme is in line with similar programs regarding the development of critical thinking and research skills of students. The proposed transfer mode is successful and should be expanded to other modules of the ILTP. In doing so, it can serve as instrument for capacity building and empowerment of library staff members participating in the training venture.

The thesis examines the loosely framed hypothesls that a computer progr3m can be designed to aid the gaining of new insights into practical real problems in a way which is 3n3logous to well established creative problem solving techniques. Two programs were designed and tested by the writer for this purpose. Evidence is provided to show that the programs do in fact assist in the galning of such new insights. The research comprises 3n analysis of the relev3nt literature and field studies taking the form of controlled experiments. The field research design. The first stage comprised the development of 3n early prototype program BRAIN. From this e::ercise ideas for further development of the progr3m were e::tracted together with a methodology for establishing how to record user interaction with the progr3m. The second stage of the field research involved the systematic testing of an enhanced version of the initi3l progr3m. The writer sought to establish how users interacted with the program. There was observed to be strong evidence that users did in fact interact with the program. It was noted that some users found the progr3m somewhat too bizarre for their own liking and had difficulty in making good use of the structure provided by the program. Further developments of the initial program, BRAIN, and the theoretical justification for the design of a second program, ORACLE, were made. ORACLE adopts the role of a process facilitator operating in the mode of a Rogerian type therapist. The computer program is developed from ideas associated with the ELIZA program developed by Weizenbaum and experience with the BRAIN program. The third stage of the research concerned itself with ascertaining whether the programs appeared to help users working with real problems - ie; ones over which they exercised personal ownership. At the same time an attempt was made to evaluate the effectiveness of the improvements made to the BRAIN program. The results obtained indicated that there was evidence to support the view that both programs assisted in the gaining of new insights into real, owned problems.

published_or_final_version / Computer Science / Master / Master of Philosophy

No description available.