#TITLE_ALTERNATIVE#

• Main Author: CLARK HADINATA (NIM 23396007), PHILIP
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/11136
• Institution: Institut Teknologi Bandung

In many application (such as ultrasonic nondestructive testing) the signal under consideration is sometimes so small that it is masked by noise. To reveal the original signal, it is necessary to decrease the noise, thus efforts to increase S/N ratio have to be carried out. One of the efforts is the Split Spectrum Processing (SSP) method. This method makes use of the signal's wide bandwidth. The signal is filtered by a series of frequency-shifted filters, so there are a number of signals, with each signal has its own original signal and noise component. These series of signals are then compounded again by some method, such as Average of Squared Signals, Squared of Averaged Signals, and Minimization techniques. There are 3 paremeters which determine this method's caracteristic and performances. They are the signal's bandwidth, the number of filter used, and the filter's bandwidth.This research is carried out by changing one parameter, while the others are kept constant. In this research 6 simulated signals with different Gaussian frequencies are used to optimized the parameters. Each signal is processed by SSP with quarter power, half power, 34 power and full power bandwidth to obtain the optimal bandwidth, 5-90 filters to obtain the optimum number of filter, and 250 kHz - 5.5 MHz filter bandwidth to obtain the ultimate bandwidth filter. It is obtain from the simulation that the optimum signal bandwidth is half power bandwidth (HPBW). The minimum number of filter is 15 filters, more filters doesn't improve the S/N ratio. The product between the signal's bandwidth and the filter's bandwidth is nearly constant, but the signal has to be wide enough to be carried out with SSP (with HPBW, 30 filters, the product is approximately 5.1012). These results is carried out to some masked signals, and it is proved that SSP can improve the S/N ratio up to 10 times. Next, SSP with above parameters is used to improve S/N from real signals, obtained from ultrasonic hardwares. There are 3 ultrasonic transducers : 10 MHz with 3 MHz bandwidth, 5 MHz with 2.5 MHz bandwidth, and 5 MHz with 1.5 MHz bandwidth. Each transducer is used to obtained 2 signals : with 1.5:1 S/N and 1:1 S/N. If the signal is not totally covered by noise (1.5:1 S/N), the SSP can improve the S/N from all signals up to 2 times. But if the signal is not totally covered by noise (L5:1 SIN), the SSP can improve the S/N only signals with wide bandwidth.