Sheila, Georgia, Craig

Tonight we updated our ASC input matrix, for CSOFT and DSOFT pitch. We have reduced cross-coupling, particularly from DHARD to the soft loops, and between the soft loops. The new input matrix has been put in the guardian (prep_asc_for_full_ifo), and we're now using the A channels of the soft blends (though we're not actually blending with anything).

The sensors for the soft loops are the transmon QPDs (TRX_A, TRX_B, TRY_A, and TRY_B), ideally we use a linear combination of the QPDs that is insensitive to the HARD degrees of freedom. To work out what the coefficents we need, we:

• Drove an 8Hz line in the hard degree of freedom. We used the ASC lockin oscillator, and set the output matrix to drive the test masses in DHARD. We notched the 8Hz line out of the ASC arm loops.
• Measured the transfer function from the DHARD excitation to the transmon QPDs.
• Used the ratio of DHARD measured in the A and B QPDs to find a combination of QPDs which cancels DHARD from X- and Y-arm soft bases.
• Put these matrix elements for X-soft and Y-soft in the unused DC5 and DC7 ASC loops, so we can compare to the current soft loops.
• Drove an 8Hz excitation in the X-soft and Y-soft, and scaled the matrix elements so the line is the same amplitude in the arm basis as in the current CSOFT.
• Combined the X-soft and Y-soft input matrices into the common and differential basis, and again used DC5 and DC7 to compare our new-input-matrix CSOFT and DSOFT to the old CSOFT and DSOFT.

The old input matrix for DSOFT was equally sensitive to DSOFT and CSOFT. The new matrix reduced this coupling. The old input matrix for CSOFT was not sensitive to DSOFT. The new input matrix improves cross coupling from DHARD to CSOFT by more than a factor 10, and cross-coupling from DHARD to DSOFT by a factor of 5. The new CSOFT is a factor of 2 less sensitive to CHARD. The DSOFT sensitivity to CHARD is unchanged.

We've yet to remeasure the soft loops with these new input matrix elements.

edit: Since making this change we've lost lock after LOWNOISE_ASC due to a 0.47 Hz instability in CSOFT P. I've not taken the input matrix change out of the guardian since we also had a not-great alignment for this lock and I'm not sure that the input matrix was the cause. It should be easy for future commissioners to revert these changes in prep_asc_for_full_ifo if necessary.