I have seen that the limits of integrations are chosen by the groups who do PLL research. I find it okay to choose the lower cut off because our viewing time window is like a high pass filter not letting us see jitter of low frequency nature. But logically I find it impossible to choose a higher cut off for the integration limits unless there is some low pass filter(PLL) filtering out the phase noise by its bandwidth. Any high frequency phase movement of 5GHz will appear if we're viewing it long enough to account for say 10KHz(lower limit). Mathematically it's possible to take the PSD and filter it out in matlab. But is there any physical possibility of applying such an integration limit<fs/2 even possible without an LPF followed by it?