Академический Документы
Профессиональный Документы
Культура Документы
Abstract: Moving Target Indication (MTI) and Moving Target Detection (MTD) processing were implemented on TI TMS320C6201 Evaluation Module (EVM) digital signal processor as part of a Radar receiver. Simulation was performed and Algorithm tested in MatLab and Borland C++, before the final implementation. Optimization techniques were applied in effort to improve the processing time.
I.
INTRODUCTION
Radar serves to detect objects within the area of observation and to estimate their positional co-ordinates. The character-ristics of the echo provide information such as the range, angular location of the target, its track or trajectory if it is moving, and predict future location [l]. The objective is to design and implement a simple digital radar MTI and MTD. Digital Signal Processor (DSP) provides great flexibility while its speed is the major concern why DSP is preferred. The final realization is built on TMS32OC6201 Evaluation Module (EVM) from Texas Instruments [7]. Different optimization techniques were to be applied to improve the efficiency of the programs. The initial design is analyzed with the simulation of a complete Radar Doppler System. After extracting necessary data from the simulation results, the MTI/MTD subsystem is designed with software in Matlab and Borland C++.
Processing system [6] using Signal Processing Workbench (SPW). Experiments were done on two moving targets with land clutter and white thermal noise taken into consideration. Table 1 shows a two targets' parameters that are simulated.
SPWParameters Sampling Frequency (Hz) Pulse Repetition Fre uenc (Hz Time Delay for Target
Gain for Target (dR) Velocity of target (m/s) F W Frequency (GHz)
I Target 1 I Target I
I
2.0e6
"
2.0e6
2.0e-4
.5e-4
I I
-10 10.0
I I
7.5
10.0
111. DESIGN
The main concept of the MTI/MTD processing system [3] is bought to the design board from the Pulse processing system in the SPW. The flowchart in Fig 1 shows a simplified structure of the MTI and MTD of the radar processing system.
11. SIMULATION
Research and simulation were carried out with different parameters in a Radar Doppler
674
7 7
Direction = arctan
(1)
Range = -metre
CT 2
(m)
The results of both the design in Matlab and Borland C++ were verified with that of the SPW simulation plots and targets parameters. From Fig 2C, positive amplitude detects the target is coming toward the radar while the negative means the opposite.
Figure 1: Flowchart of MTI/MTD Processing
200
I
The simulated signals from SPW were converted as raw data shown in Fig 2A to be used by Matlab and Borland C++, which are the software programs that are used for the design. The data, arranged into matrix, are then processed by the delay canceller where vectors of each sample containing information of the targets and other interference such as clutters are compared with the previous to get the required processed signal. In the delay canceller, as clutter are considered as stationary objects, the cancellation process will cancel them out, leaving only the moving targets signals. The processed signals are then Fast Fourier transformed to convert the data into frequency domain [4]. The transformed matrix was multiplied with its conjugate to extract the absolute of the MTI signal for display as Fig 2B. Beside indication of the targets, detection is the also been carried out, to find out about the range of the targets and also the direction of where the target is heading [ 2 ] .
-200I
0
I
2000
4000
MKX)
2 w
Am Plllr 1w de
8ooo Tme
1 0 6 3 3 3
12000
14000
16000
1
0
-1 w
B00 I
1
2000
4000
M)W
8000
Tme
IOW0
12000
14000
16000
675
x 15,
Id
IV. IMPLEMENTATION Code Composer Studio Version 1.0 is the software that is used and it came bundled with the DSP TI EVM board [ 101. The aim of the program is to make the processing fast enough to meet the speed requirement of signal processing which is crucial on the indication and detection of targets.
C) MTD nrocessed output Figure 2: Matlab and C results
For the range of the targets, the targets can be seen as:
Several optimization methods for C codes [8] were used to improve the efficiency. For data extraction, a Probe Point within Code Composer Studio is used to read data from a file on the host into the input buffers location. Compiler Intrinsics that map directly to inline EVM assembly instructions were also used throughout the whole coding for further improvement in efficiency. The compiler schedules as many instructions as possible in parallel to maximize code efficiency. To eliminate memory dependencies [9], the following techniques are applied: Use keyword const to indicate those objects that are not modified by a function. Use -pm (program-level optimization) opt.ion that gives compiler global accesses to the entire program or module and allows compiler to be more aggressive in eliminating dependencies. Use -mt option that allows compiler use assumptions to eliminate memory dependencies. Use -ms ensure that redundant loops are not generated, thereby reducing code size.
x is the point of samples as from the plot. O h s is the time duration between each samples at sampling frequency of 2MHz. The result is multiplied with the half the speed of light to get the range values of the targets. Table 2 shows the results of the two programs as compared with that of the SPW simulated values. The MatLabs results were obtained with
higher amplitude as compared w i t h t h e latter.
By using Borland C++, further improvement was attempted to enhance the speed of processing with an average of 24 milliseconds.
Parameter Target 1 Range (m) Target 1 Direction Target 2 Range (m) Target 2 Direction
SPW
Matlab
1 B::td
I
30,000
Away
22,500
Toward
676
* **
139,374.3 278.75
** Tota1;ime
With the appropriate change in the declaration type, an improved processing speed was obtained as shown in Table 4.
32 32 32
V Good Good
Poor
Good
i-buf
For representation, the graphs of the various plots of the refined C code are plotted. From the plot, the direction of the target can be seen clearly as having positive swing meaning that the target is flying towards the radar, while the negative represents the opposite. The range and direction definition is as the same as that of the earlier design represented by equation (1) and (3).
The rest of the work carried out was on the observation of the radar signals received with changes and additions of other
677
parameters such as additional targets, rain clutter, sea clutter etc. Our observations verified the accuracy and consistency our coding and implementation.
V. CONCLUSION In this paper, the MTI and MTD processor was implemented with the TI EVM Digital Signal Processor (DSP). The flow begins from research on the principles, to the simulation with SPW, to the design with Matlab and Borland C++, provide the steps necessary to final implementation. With the simulated data used, the performances of all approaches were analyzed. The later stages saw much of the task to make refinement and optimization to the algorithm. The final result shows that the speed of the processor with refined C codes as the most efficient. Through integration of the MTI & MTD processor with the rest of the modules, a DSPimplemented radar receiver module can be built. REFERENCES
Algorithms, and Applications (3rd ed. ). Prentice Hall. [6] Cadence Design Systems Inc.: SPW Jump Start Training. ALTA Group Rev. 5/95. [7] Texas Instruments: TMS320C6201/6701 Evaluation Module Technical Reference . [SI Texas Instruments: 6TMS320C6201/6701 Evaluation Module User Guide, SPRU269D December 1998 [9] Texas Instruments: TMS320C62W67X Programmers Guide, SPRU198B February 1998 [lo] Texas Instruments: TMS32OC62X DSP Design Workshop (Student Guide), DSP6-Notes-2.0 November 1997
Radar design principles: signal processing and the environment, New York: McGraw-Hill, 1991. Radar: principles, [2] Edde, Byron. technology, applications, N.J.: Prentice Hall, c1993. [3] Schleher, D. Curtis, MTZ and pulsed doppler radar, Boston : Artech House, c1991 [4] Winthrop W. Smith, Joanne M. Smith Handbook of Real-Time Fast Fourier Transform, New York : IEEE PRESS, c 1995 [5] Proakis, J.G., & Manolakis, D.G., 1996. Digital Signal Processing: Principles,
[ 13 Nathanson, Fred E.
678