For this application you want as sensitive as you can get, since you want to detect minute differences in background intensity. That points to a big ole photodiode.
Hence why I suggested that they use an off axis sensor to do online dark subtraction. Locally if there was no star, the diodes should give you the same intensity. If a star enters the field of view of one, but not the other, then you have a detection. It's more complicated than that apparently, but the idea is sound.
From the patent it actually looks like they use some kind of optical chopper and a single diode. A chopper is a disc with slits cut out, in the fold of view. If there's a star that doesn't fill the field of view, the chopper will give you a periodic signal as it rotates. If the sky is empty then you'll get roughly the same signal all the way round. (from a quick skim)
Referencing the link from my other comment, the nutator and optical modulator referenced in the patent combine to remove the residual sky background and apply FM modulation to the star aligned with the telescope OA.
By measuring the phase and depth of the FM modulation, you get the angle and distance of offset between the star and OA.