High-level synthesis for travel-time tomography
Permanent address of the item is
Korkean tason synteesin soveltaminen matka-aikatomograﬁaan
Tomography is used to acquire an image of inner contents of an object or material. Typically measurements are based on penetrating waves. Measurements require signal generating and recording hardware, data preprocessing and finally computationally heavy mathematical inversion to recover the unknown parameters. The goal is to decrease the data transfer requirement on, for example, tomography mission to a Near-Earth Asteroid. This could enable small, inexpensive spacecraft to collect tomography data. Travel-time tomography, which uses signal travel times, is one of the methods that can be used to achieve this goal. Tomography algorithms are still under heavy development, for which reason hardware prototyping cycle should be very short. High-level synthesis is used to generate hardware by using high level programming language. It helps the designer to implement hardware design changes quickly, especially when the requirements change. In this work, a setup to collect acoustic tomography data was developed. Data preprocessing hardware for travel-time tomography was implemented with Mentor Graphics Catapult high-level synthesis tool. The calculation of travel-time values was implemented first in Matlab and the scripts were then transformed to C code. Catapult was used to implement hardware on the FPGA from these C codes. Evaluation of the workflow was performed and interfacing options for the module to a PC running Matlab were studied. Travel-time tomography was shown to be a feasible method to recover target objects. Determinining the time period to use in measuring a travel-time is an issue. Simulation of signal noise sensitivity on an asteroid mission was accomplished by reducing the accuracy of preprocessor calculations. A method where signal power is integrated over a time period was evaluated and it proved to be surprisingly stable in recovering targets from the test area even with noisy signals. Tomography algorithms changed over the course of the project, and high-level synthesis enabled to implement the designs.