The Laser Interferometer Space Antenna (LISA) will be a future gravitational wave observatory in space. It will consist of three spacecraft forming a triangular constellation using laser links.The angles of the constellation will change due to orbital dynamics and require a compensation mechanism in each spacecraft. For this purpose, they each house two independently movable optical benches that are optically connected with each other via the so-called Backlink to exchange the laser light between the benches. The Backlink’s non-reciprocity is described by the differential phase stability of its counter-propagating beams and requires a noise level below pm/√Hz in the LISAmeasurement band. The experimental study of such a Backlink is the topic of this thesis, by constructing and commissioning an experiment to investigate three different Backlink implementations; the Three-Backlink Experiment. The LISA requirement can only be tested if the experiment is also pm-stable, which is reached by using quasi-monolithic optical benches with optical componentsglued onto a glass-ceramic baseplate. The first part of the thesis describes the construction of the complex optical benches, which house in total eight interferometers and four fiber-couplings with entangled construction requirements. A beam measurement and alignment tool, the Calibrated Quadrant Singleton, is investigated and characterized as an essential tool for the construction process.Alignment strategies were developed for the demanding construction steps and are presented in a conceptual form, so they are applicable in future optical bench constructions. The construction of the two benches is successfully completed. They are characterized for their relevant performance parameters and implemented in their experimental infrastructure.The second part of this thesis focuses on the Three-Backlink Experiment’s commissioning and noise analysis. The three different Backlink implementations enable a distinction of their individual noise couplings and contributions. The parameter of interest in the Backlink’s measurement is the nonreciprocity which reaches the requirement of 1pm/√Hz in the frequencies above 0.3 Hz for two Backlink implementations; the direct fiber implementation and the free-beam Backlink. Technical limitations in the current phase read-out measurement system limit the third Backlink implementation, the frequency-separated fiber Backlink. Backscatter at the direct fiber Backlink implementation enables the coupling of laser frequency noise and temperature to the non-reciprocity measurements. Theobserved couplings probe existing models and agree with their predicted noise coupling via Backlink backscatter. A free-beam connection between the two stationary optical benches is established with a closed piezo-mirror control loop that ensures pm-stability above 0.3 Hz.An upper limit for the performance of all three Backlinks is measured; the non-reciprocity is at 1.7pm/√Hz above 0.3 Hz, below 10pm/√Hz above 0.01 Hz, and at the frequencies below at 3pm/√Hz. At the current state, without motion between the optical benches, the free-beam implementation is operating with the lowest non-reciprocity noise contribution.The Three-Backlink Experiment offers a unique LISA-like test-bed on two optical benches enabling the study of coupling of different noises in the individual Backlink implementations. This thesis provides the key part of the test-bed, the optical benches, verifies the concept of the three Backlink’s disentanglement, and includes an analysis and modeling of its limiting noise sources.