Since we recently observed a limited CMRR at the homodyne detector, we investigated the source of the unbalance, which in principle can be a combination of different effects:
1) unequal optical power splitting at the BS
2) different clipping losses at photodiodes
3) different responsivity of the photodiodes
To get rid of clipping losses we align each photodiode with the procedure described here. The residual transverse position error is around 20 um, and the residual axial offset from beam waist is below 400 um. As a result the clipping losses are expected to be lower than 0.1% for both photodiodes.
In principle the responsivity of the two photodiodes should not differ by more than 1%, as they are both quoted with QE>99%. To check whether this is true, we measured the contribution of power splitting. we did repeated measurements of the beam power at BS outputs with a power meter, using as input the BAB and the LO at maximum power (i.e. 2 mW each), in the optical configuration of last CMRR measurement. The resulting splitting ratio is 1.04+-0.01, i.e. the optical power towards PD2 is higher than the optical power towards PD1 by (4+-1)%.
With the BAB well centered on both photodiodes, we measure a residual beat note amplitude resulting from imperfect substraction. To compare this with power unbalance, we measured how much the power ratio on PDs must change in order to recover perfect balancement. To this purpose we artificially lower the detected power on PD2 by movind a bit the photodiode to induce a small clipping. We found that the power on PD2 must be lowered by (3.5+-1)% in order to extinguish the 4 MHz beat and balance the HD, see attached picture. This confirms that the residual unbalance of a few % is dominated by the BS power splitting. In order to remove the uncertainty from the measurement of splitting ratio, we swapped the position of the two photodiodes.