240 likes | 518 Views
Introduction to Calibration. Brian O’Reilly SciMon Camp 2006. Frequency Domain Calibration. We model the DARM loop in MATLAB Compare this model to measurements of the open-loop gain, electronics in the Actuation and Sensing chains and DC value of the Actuation.
E N D
Introduction to Calibration Brian O’Reilly SciMon Camp 2006 Gxxxxxx
Frequency Domain Calibration • We model the DARM loop in MATLAB • Compare this model to measurements of the open-loop gain, electronics in the Actuation and Sensing chains and DC value of the Actuation. • Optical and loop gain are tracked by time-dependent coefficients which are generated on minute or second time scales. • These coefficients are used to propagate measurements at t0 to other times. Gxxxxxx
+ DARM_ERR DARM_CTRL_EXC DARM_CTRL DD(f) + A(f) (t)CD0(f) sres=(Lx-Ly)/L s=h(t)+n(t) Gxxxxxx
ky kx EXCx(t) + + Ay(f) Ax(f) + Actuation • DARM feeds back to the ETMs. • Measuring the actuation has typically been the least accurate and most angst-ridden part of the calibration. Gxxxxxx
Actuation Function • Calibrate the ASQ signal for a simple Michelson • This establishes the length scale in AS_Q counts. • Use it to calibrate ITMs: • Use single arms to calibrate ETMs with ITMs Gxxxxxx
Actuation Function • Treat mass as a simple pendulum. • Knowing the DC value we can set the scale for the transfer function. • Methods for measuring DC value explicitly have also been tried: • sneaky poles Gxxxxxx
Actuation Function ? Gxxxxxx
Compensate the Electronics Gxxxxxx
The Payoff… Small Errors Gxxxxxx
Digital Filters Know them perfectly? Gxxxxxx
The Input Matrix Gxxxxxx
Sensing Function • Model as a cavity pole • Have to understand the sensing electronics chain • Photodiode, Whitening, Demodulation, Anti-Aliasing etc. • How well do we know the cavity pole? • How well do we know C(f)? Not directly measured. Gxxxxxx
L1 H1 Open Loop Gain Discrepancy 5-10% error on response at 2 kHz Gxxxxxx
Model Inputs Gxxxxxx
Frequency Domain Calibration • Measure Open-Loop Gain at a reference time t0 • G0(f) = A(f)CD0(f)DD(f) • h(f,t) = RDERR(f,t)DERR(f,t) • Similar equations forAS_Q Gxxxxxx
Propagate • This value stays within ~5% of unity, barring any problems with the code. Gxxxxxx
Errors on the Response • By breaking the error down into these components we identify problem areas. Gxxxxxx
Finally • After diligent work we feel we can control calibration errors to the level of 5-10%. • Doing better than this is hard, but: • 15Mpc/10 = 1.5Mpc = range of L1 during S2!! • Other ways to calibrate: • HEPI or Tidal Actuators • VCO • Photon Calibrator • time-domain h(t) • Calibrating eLIGO or advLIGO will present a new set of challenges. Gxxxxxx