KAGRA is a kilometer-scale underground gravitational wave (GW) observatory. It uses a low-latency (delay ≤4 sec.) software pipeline to reconstruct its primary output, the so-called GW strain signal, before sending it to analysis pipelines and long-term storage cluster.

During my postdoctoral fellowship at Institute of Physics, Academia Sinica, I was in charge of the development of this calibration pipeline Python package and the maintenance of the low-latency calibration server (running ScientificLinux 7 / CentOS 7 at the time). The overall code structure was be based on the low-latency pipeline developed earlier for LIGO (a similarly-sized GW detector in the U.S.). At the same time, our calibration pipeline needed to account for differences in the KAGRA and LIGO observatory designs and their compute infrastructures (storage limitations, etc.).

Upon execution, it dynamically constructed a Gstreamer pipeline that received data streams from the real-time system and produced the reconstructed GW strain output. Gstreamer elements (functions) responsible for data processing were wrappers around Gstreamer components mostly written in C which enabled efficient pipeline execution.

The development progress was reported to the rest of the calibration team members on a semi-regular basis. Below is an example of one such report.

On meetings and conferences with outside members (with LIGO, Virgo, or NIST) the development progress on calibration pipelines was discussed as a part of the overall team’s effort (often covered at the surface level), e.g., on the 2019 National Institute for Standards and Technology (NIST) Gravitational Wave Workshop in Colorado, US.

The observation data from KAGRA’s first joint observing run in 2020 released to the broad scientific community, used this calibration pipeline to reconstruct GW strain output.