Data Logging Solution for Digital Signal Processors Brian Newberry Nekton Research, Inc. [email protected] James M. Conrad University of North Carolina at Charlotte [email protected] Abstract Digital Signal Processors (DSPs) are some of the most powerful processors in the world. With clock speeds topping 1 GHz and architectures capable of performing multiple instructions per clock cycle, they are ideally suited for working with audio and video signals with high data rates. The signal processing algorithms that are used to analyze such data are complex and can be difficult to debug if results are not as expected. Having the ability to store the data that is being analyzed by the DSP allows for post processing of data so that the algorithm can be perfected outside of actual testing conditions. This capability of "playback" can drastically reduce the time and cost of system testing. For this purpose, an embedded DSP data logger was designed to store audio data that is being collected and analyzed by a DSP. The data logger is designed to download audio data from the DSP at a high rate, store the audio data as a binary file on a static storage medium, and be able to upload the data to a separate PC from the static storage medium. 1. Introduction The Global Positioning System (GPS) has solved many of the problems of navigation that have faced mankind for thousands of years. The ability to find the exact location of any device on the planet has also changed the way that autonomous mobile systems operate. However, GPS systems have their limitations. The signals sent from the GPS satellites cannot penetrate the earth or into the water that covers nearly two thirds of the earth's surface. This means that for underwater vehicles, the problem of navigation is not yet solved. Manned submarines use gyros and inertial navigation systems that are able to provide navigation information, but these systems are prohibitively expensive and large for use on small autonomous underwater vehicles (AUVs). One possible solution to the problem of underwater navigation is to use buoys, placed in known locations that emit specific frequencies into the water. By finding the Doppler shift that an AUV perceives relative to that specific carrier frequency emitted by a source and with the added knowledge of the velocity of the vehicle, the heading relative to the frequency source can be determined. Using multiple buoys each with their own specific frequency would allow the exact location of the AUV to be determined. This concept is currently in the development phase and much testing remains to be done. The algorithm that is being used for finding the Doppler shift and then computing the heading relative to the frequency source is complex and the audio signal that is received by the system may require conditioning because multi-path reflections and background noise may be significant in certain underwater environments. The purpose of this project was to provide a means of data logging for the system that will be analyzing the frequency data collected by the AUV. The ability to log the analog data gathered by the system would allow for post processing using actual test data rather than simulated data. For this particular application, this will represent a significant reduction in the cost of testing. A complete system test would require at least two people an entire day to perform and would necessitate a great deal of setup and travel. If enough data could be collected and recorded in one day of testing, then that data could be archived and used to perform tests on algorithm refinements in a laboratory setting while not sacrificing any of the characteristics of the actual testing environment. The data logging part of the system could eventually be removed completely after the algorithm is perfected and no recording of audio data is desired. Since this system is being designed for an embedded application, there are certain restrictions on the design and the components that can be used. The entire system must be as low power as possible and be able to fit inside a five-inch diameter cylinder. The low power constraint comes from the fact that all of the power the system uses will have to be carried on batteries inside of the AUV. The five-inch size constraint comes from the size of the particular AUV that the system is going to be tested. These two constraints limit the choice of processors that can be used and also what sort of storage medium can be used to log the analog data that is being collected. It is also important to have an easy method of extracting the logged data because the entire design will be enclosed in a watertight compartment that will not be easy to disassemble. The transfer of data out of the system to a separate PC for post processing must be done using a high bit rate because of the large amount of data that will be accumulated while the system is in operation. The data collection and analysis need to take place at very high speeds and thus will need to be run on a very fast processor. A Digital Signal Processor (DSP) is a good fit for this system because of its high performance low powerconsumption compared to conventional processors. A Pentium III processor running at 1 GHz will consume 29 Watts of power while a 6416T Texas Instruments (TI) DSP running at 1 GHz with all supporting hardware installed will consume about 2 Watts [1,2]. Conservatively, a Pentium based system would drain the available power supply fifteen times faster than the same system using a DSP. Although using a DSP makes perfect sense for this system when considering power constraints, this choice introduces several drawbacks. Whereas there is wide operating system support for general-purpose processors like the Pentium, there is little to no operating system support for DSPs. This limitation is important to this system because an operating system simplifies the task of
View Full Document
Unlocking...