Chris, I assume you want to track phase changes in the optical path in real time. You can use a tracking beam, as your supervisor suggests but, in my opinion, the tracking beam should trace nearly the same optical path as your signal beam. Then, if you modulate your tracking beam with a sinusoidal phase (e.g., using an acousto-optic modulator) and interferometrically detect it (i.e., by mixing it with a splitted version of itself that did not go through the interferometer) the resulting signal will have the same modulation frequency and a phase shift that is time-varying and proportional to phase changes in the optical path.

Then, you can use this "error signal" to correct the path in real-time, for example using a PID-controlled piezo-electric mount that moves one of the mirrors in your interferometer.

-Paulo. 

Thank you for answering Paulo,

The reason why I did not choose to let the reference beam trace more optics just after the microscope objective is that the final lens in front of the CCD would focus the reference beam on to the CCD. By placing the beam splitters just after the tube lens I will make sure that the reference wave is a plane wave that covers the whole active area of the CCD. The part of your answer about modulating the references arm phase and letting it interfer with a non modulated reference beam I did not fully understand as I am quite new to the field of interferometry as well as digital holography in physics. Could you maybe add a sketch I think that it would clear things up for me (you are allowed to edit the setup image I attached in the original question).

-Chris

If I'm correct you focus a image of the (cleanup)  confocal pinhole on a CCD image seonsor. Which is Hetrodyne amplified by your reference beam.  As mentioned by Paulo you can add two Acousto-Optic Frequency Modulators +1 order and -1 order driven from the same waveform generator. The fase difference of the AOM drive frequency directly controls the phase of the reference beam. Practically you need to pickup the reference before the first beam expander and expand the beam after both AOM. Why you don't use a photodiode when light is focussed on the CCD? Locking the only works if phase-freqency-change of the sample is stabile with you exposure time.

I have recently published an article about a somewhat different approach:

Unbiased (reference-free) phase field imaging for general optical fields including phase discontinuities. https://arxiv.org/abs/1610.08089

If the interferometer is well designed, it has a fixed point, which allows to calibrate length differences between branch 1 / branch 2 easily. Actually this is important in experiment.

As I said, a different approach with simple evaluation:  may be that gives you also a new idea.