Keita, Jenne, TVo, Danny, Georgia, Hang
We developed a script to decompose a camera image into Hermite-Gaussian (HG) modes. Please see the attached images for an example of the code's usage.
The first image showed the raw camera image of AS AIR when the IFO was locked in MICH DARK. We tried to decompose the image into the HG basis up to 4th order (that is, 2nd order in misalignment and 2nd order in mode mismatch). The second image showed the best modal reconstruction, which captured the main features of the raw images reasonably well, and the third image showed the fitting residual (the rms residual power was 0.1% of the total).
From the image we could separate the even-order higher-order-modes (HOMs) due to mode mismatch, from odd-order HOMs due to alignment offsets. For example, in this specific image we estimated that 62% of the power was from mode mismatch and 38% from alignment offset. Unfortunately the image slightly clipped on the camera which distorted some even modes into the odd ones. Nonetheless, after we fixed the camera pointing we should be able to get a better estimation of the contrast defect by subtracting out the odd-order contributions. For example, the last plot showed the even-order HOM-only part of the image, and we could see that the 02 and 20 components were out-of-phase, indicating the heating in the horizontal and vertical directions might not be even.
In addition to helping TCS tuning, the code should also be able to measure (at least estimate) the contrast defect and alignment drift for the full IFO without needing to setting up an OMC scan. Camera images should be enough for this purpose. The code might also be used to estimate the beam quality in other places.
The code is available at:
/ligo/home/hang.yu/Desktop/small projects/modeFit/modeFit.ipynb
Nice. Is there still a problem of engaging WFS in full lock?
In the past sometimes we had to manually pre-align the interferometer after CARM offset reduction before all WFS loops could be engaged. We did it by looking at the AS and REFL cameras plus DC powers (X,Y arm, SASY 90 and sometimes baffle PDs). I guess you could implement this scheme to work automatically.
It might also be possible to go further and do initial alignment by processing images from the cameras with flashing beams. Since it is possible to do it manually, there might be a way to train a script to do this.