Over this summer I have been working in a physics laboratory as part of the Dark Energy Spectroscopic Instrument (DESI). This project is headquartered in Lawrence Berkeley National Lab under the Department of Energy. The project as a whole is constructing to an instrument to place on the Mayall Telescope at Kitt Peak Arizona. This Instrument consists of 5000 foot-long fiber positioner robots divided among 10 petals (500 positioners per petal) known as the focal plane, a fiber view camera to assist in the positioning of the fiber optic cables and systems for controlling and moving the focal plane. Every part of the project has been hand engineered and programmed by project members, making it a huge task.
The scientific goal of the project is to survey the sky for four years, starting late 2018 or early 2019. Every 15 minutes for several years, each of the 5000 positioners will point a fiber optic cable at a galaxy take in light from it and then move to a new 5000 galaxies. This will give us something of a 3D image of a map of millions of galaxies hopefully allowing us to discover how, if at all, dark energy changes over time. Currently there are teams working on choosing the targets and simulating the survey as a whole.
My specific group is located here at the University of Michigan. We are tasked with developing and manufacturing the Fiber Positioners (shown above). Since the positioners must point fiber optic cables at galaxies not visible by the naked eye, they must be very accurate. We require that the positioners are accurate within 100 microns (1/1000 millimeters) on their first blind move and then in the subsequent moves, must correct to have less than 5 micron RMS error. For reference a thin piece of human hair is 40 microns thick so this is incredible accuracy. Even more incredible is when testing properly functioning positioners we find the have about 1 micron RMS error in just 3 moves! Their whole range of points that can be targeted is a disk about 12 millimeters across resulting from two compounded rotating arms with a length of about 3 millimeters. The theta axis has a range of about 土190 degrees and the phi arm is on the end of the theta arm with a range of about 190 degrees. A result from an accuracy test is shown below.
As part of this team, I operate the initial accuracy test for all positioners we produce, the result of which is shown above. Every couple of positioners also undergoes additional testing such as lifetime testing at École polytechnique fédérale de Lausanne of cold temperature testing at Lawrence Berkeley National Lab. To test the positioners we back-light their fiber optic cables with an LED and take pictures of them using a high resolution camera as they move to specific positions.
Now of course the whole process would be very difficult to do by hand so it is scripted in python like the rest of the positioner’s movement. The positioners are given positions to move to so the code calculates the movements in motor rotations and calls the camera to take a picture (the part of the code I wrote — the operation of the camera from python), the code uses this knowledge to go through 3 correction moves on a single point before moving to the next.
The movement commands in the test and the actual instrument are passed to the “petal controller” which is a small BeagleBone computer (similar to a Raspberry Pi). The petal controller then sends the commands to the positioners through CAN.
However, running the setup does not mean simply pressing go on the script. I support the process by writing new tests to identify problems, by making new plots to show the data, by being the most familiar with the code and plots to be able to easily identify problems and by demoing improvements in the code. The process of ensuring these positioners are good and are supported by good, smart and fast code is ongoing and ever improving. This project is a lot of work and we still have a few exciting and difficult years before first light.