I am confused about the way mode-locking is done in practice. If i understand the problem well, normally, in a laser cavity with a large enough bandwidth, we have many modes that would interact in an incoherent way and give a low power (this would be a continuous wave laser). If we are able to create some phase relation of the modes (for example all of them to have the maximum of one of the peaks at the same point) than we create a mode-lock laser. So the problem we need to solve is to somehow shift the phases of the modes in the cavity to go from some random noise to a nice tall peak. However, I read about an active way of achieving this, using an acustic-optic resonator. If I understand it well, this allows frequency modulation which can shift the present modes by c/2L and in the end you remain only with the central mode, while the others get lost as they are not able to reproduce themselves. So basically (part of) all modes are converted to some frequency and the rest is discarded. But I am really confused now. All this would shift the frequencies of the modes, not the phases. So you end up with "bunches" of light of the same frequencies, but out of phase. How is this any better i.e. different than random noise? I imagine that, given that you have just one frequency now, the noise shape will stay the same (before the shape was also changing in time, due to having different frequencies), but I am not sure I see how do you get such a nice, narrow peak from lots of out of phases "bunches" of light and not just noise. Any explanation would be greatly appreciated! Thank you!
In practice, there is usually a passive element relying on Kerr lensing that stabilizes the coherent modelocked beam over the CW one. The mode locked beam has a much higher peak intensity, and so if there is a material whose index of refraction is dependent on the intensity of the field (which is really all materials, the effect is just weak) then the modelocked beam will become spatially convergent, like it had passed through a lens. This allows the modelocked beam to be spatially filtered from the CW beam, effectively blocking it. Does that help?