stefan.ballmer@LIGO.ORG - posted 16:09, Thursday 20 September 2018 - last comment - 11:49, Friday 21 September 2018(44075)
Correcting for current effective recycling gain in last CARM handoff step.
Sheila, Jenne, Hang, Danny, Stefan As Sheila and Gabriele previously reported, the changes in recycling gain due to alignment made the CARM handoff spotty. The issue is that we go to a certain arm power, but that arm power does no longer corresponds to the same CARM offset when the effective arm loss changed. In alog 43953 they also hinted at a solution: Also looking at REFL DC (normalized by REFL_DC for large CARM offsets) allows resolving that degeneracy for the last few picometers of ARM offset. We now tried to implement an effective correction along those lines in Guardian. Plot 1 shows the theoretically predicted curves for REFL_DC power vs. transmitted arm power for a number of different effective arm losses (50, 75,100, 200ppm). Also shown are markers corresponding to a CARM offset of 1, 3, 5, 10 and 30pm. One can see that spots of constant CARM offset are roughly lying on the diagonal dotted lines. (This approximation gets worse close to resonant, but is likely good enough for our purpose). The algorithm we put in Guardian does the following: 1) We record the REFL_DC power for large CARM offset (in START_TR_CARM). Call this P0. 2) We continue going up to an arm power of 800 (our reference arm power Tref). 3) There we read the actual REFL_DC value. Call this P. 4) Now we can calculate the actual transmitted arm power Tact from the ideal (at time of script tuning) arm power Tideal via 1/Tact = 1/Tideal + 1/Treff * [Pref/Pref0 - P/P0] Here Pref/Pref0 is the power ratio at the time of script tuning. Note that this math corresponds to moving along the diagonal dotted lines on Plot 1 until intercepting the graph for the current arm loss. 5) Finally, the script asks for Tact (instead of Tideal) when doing the final hand-off to REFL RF. Note that the Guardian actually does the math in terms of CARM_TR_OFFSET which is defined as -sqrt(2*T), i.e. OFFSETact = - sqrt( 1/[ 1/OFFSETideal^2 + 1/OFFSETref^2*( Pref/Pref0 - P/P0) ] ) Steps 2-5 all happen in CARM_5_pm. All code is in Guardian, but due to ALS trouble hasn't been tested yet.
Images attached to this report
Comments related to this report
We now tested the script at the level of not breaking normal operation. The nominal script values are: Pref/Pref0 = 0.57 Treff = 800 OFFSETref = -40 Tideal =1458 OFFSETideal= -54 Also, the script stalls and throws an error if the calculation results in a hand-off arm power of 612.5 < Tact < 1512 -55 < OFFSETact < -35 We will have to observe the behavior over the next couple of weeks - sepecially when we for some reason and up with a bad alignment.