Using the modeled scattered light response transfer function the measurements can be checked against the model, and the backscattered light measured by scalling the model to match the measurement.

For SDB2: Figure 1 and 2 shows that for the two injections the same back scatter coefficient of 0.0004 PPM fits well the data. Given the expected filtering of 37dB by the Faraday isolator on SDB1 it corresponds to actually 2PPM of perfectly backscattered light.

For SNEB: Figure 3 and 4 show that 0.03PPM of perfectly back scattered light fits well the measurements.

For SWEB: Figure 5 and 6, show that 0.005PPM of perfectly back scattered light fits well the measurements. This means that SWEB back scatters 6 times less light than SNEB. It would be good to check if there is an alignment issue on SNEB which could explain the extra back scattering, or if there are any other reasons why SNEB could be back scattering more light.

For SDB1: Hrec_hoft_20000Hz is misleading, as the scattered light arches must also be visible in other degrees of freedom (MICH?), and h(t) tries to subtract it, which doesn't work well as the injection is changing with time. The reliable comparison is with hrec_hoft_raw_20000Hz (as is already done for SNEB and SWEB as Hrec_hoft_20000Hz already tries to subtract scattered light by using the B7_DC and B8_DC signals). Figure 7, 8 and 9 show 3 different amplitude of injection, all are fitted with 0.002PPM of back scattered light. However there is a peak at the end of the scattered light shelf in the measured sensitivity that is not reproduced by the projection. This is probably due to either problems in the SDB1 local controls or in the filtering of them to do the projection.  Figure 10 shows that by replacing the measured SDB1 motion with a perfect sine wave at 0.1Hz, a nice match can be obtained of that feature.

The code for these projections is located in /users/mwas/SBE/scatterProj_20191001.