With higher pump power the population inversion is created faster which means that after the first pulse the upper level is depleted, the pump power rapidely generate  inversion population for the next pulse. For one pulse regime  I suggest that you use pulsed pumping regime, with a periode short enough to generate only one pulse

Modulating the pump diode power at f/2 may do the job (an acousto-optic modulator may be required). However this behavior is intrinsic to three-level laser systems...   

I think that besides of increasing of gain also available pulse spectral confinement determine the pulse state. When pump is increased pulse spectrum is broaden, but if it is confined (gain spectral band, spectral filter) the process of spectral width increasing cannot be continued and laser rebuilt its pulsing state and launch two pulses in a cavity so decreasing its spectral width. May be do you know is any work that investigate that phenomena and derive some criteria for pulse stability ?

Hi, first of all what kind of mode-locking do you use in your fiber laser ? Then what is the total cavity dispersion ?

Hi, Its not a real laser. Im simulating a pulse evolution in a fiber cavity in a case of passive mode-lock. The intensity dependent losses is provided by a saturable absorber like a sesam. I launch a some kind of pulse in to the cavity and observe it evaluation. In some cases pulse get some stable state and its evaluation ceased anymore. But if the gain is increased it stable state destroyed and laser found some other equilibrium state by launching a second pulse to previous period. As it simulation I dont use real parameters. I think absolute volumes of cavity parameters doesn't matter but its relation to each other is important. So the my question is how to adjust a cavity parameters to each other to make a stable, highest pulse energy fiber laser