Stefan, TVo, Hang

We were able to power up to 28 W and stay there stably for about 15 mins or so. However, after that we still had two kinds of locklosses.

=======================================

In the first time we lost lock when POP18 dropped to 50 and then a fast lock loss. Since we had re-distributed the gains in the common mode board (as powering up, in stead of dropping gains at the very end using IMC-REFL_SERVO_FASTGAIN, we decreased the gain upstream in LSC-REFL_SUM_A_IN2GAIN), this thus ruled out that saturation inside the common mode board was the cause for the fast lock losses. More investigation needed.

Also we noticed that at 20 W the LSC-REFL_A_RF9_Q was very close to zero, meaning a good balance of the upper and lower sidebands, while as we powering up, the REFL_9_Q increased by a lot.

=======================================

In the second time, still after about 15 mins at 28 W, an oscillation at 0.4 Hz grew in CHARD YAW ASC. This might be related to the CHARD blending (for yaw, DC part from REFL 9 and AC part from TR QPD combo) which was designed to happen at 0.4 Hz. As we kept losing optical response in the REFL9 signal, this might then lead to an instability in the signal blending. This hypothesis is also consistent with the fact that we did not see similar behavior in CHARD PITCH, as we changed the input matrix for CHARD PIT DC path to REFL45 (LHO:43633) which decayed much less then the RF9 signal and therefor the pitch stayed stable. The solution might thus be either changing the CHARD YAW to using REFL 45 like we did for pitch, or monitoring the signal strength in both REFL9 and QPD combo to keep their ratio constant, or (for a single lock) servo CHARD YAW only to QPD combos as a quick test.