APPLYING THE MARX BANK CIRCUIT CONFIGURATION TO POWER MOSFETs

large linear as well as large nonlinear chirp. If an appropriate fibre (i.e. with proper length and dispersion) is used to fully compensate for this large linear chirp, a relatively shorter compressed pulse will be realised (at 1·5 mA, optimal compression was achieved when eqn. 1 was satisfied). Because the fibre compression is a linear technique and it compensates only for the linear chirp, the deviation from the transform limit at low bias levels is mainly due to the uncompensated nonlinear chirp, which is of a significant amount. As the bias level is increased, both 4t and 4l of the original pulse decrease, and the different rates of decrease make the ratio 4t/4l deviate slightly from the optimal compression condition. This leads to an overcompensation of the linear chirp causing broadening of the pulse. However, because of the smaller change in the carrier density at high bias levels, the proportion of the linear chirp in the narrowed chirped spectrum is enhanced compared to that at lower bias levels. Therefore, at higher bias levels, even though the pulse width is slightly wider due to the overcompensation of the linear chirp, the time bandwidth product drops as a result of the significant reduction of the nonlinear chirp. Htnce, to achieve the shortest pulse, biasing the laser far below its threshold is recommended. However, at slightly higher bias levels, but stiJ) well below threshold, lower time bandwidth products can be achieved. With an increase in the bias current level, there is also a rise in the optical power level which could be advantageous if these pulses are to be used as soliton sources. The average power measured at a bias level of 1· 5 mA (bias level that gives shortest pulses) was around 43p.W which corresponded to a peak power of 120mW. Whereas at a higher bias level of 4 mA the peak power increases to around 200mW. We also observed the effect of RF current on the pulse width by slightly varying the RF drive around 27 dBm. However, the change in pulse width was not as sensitive to RF current as that to the DC current.