Traditionally the beamsplitting is based upon a partially transmitting mirror, a reflecting mirror namely you have a this black thick line as the mirror which partly reflects the data and partly transmits it. So you have two ways split and you can repeat this splitting process multiple times and get multiple images just as I mentioned earlier.
This is the classical method, however it gets bulky and there are aberrations in it because each split introduces aberrations, like the pointed there. So this method that we have just recently developed does not involve those difficulties.