25-26 October 2018
Chimie Paris Tech
Europe/Paris timezone

Extremely Fast Real-Time Computer for the Next Generation of Adaptive Optics Systems

25 Oct 2018, 16:45
Chimie Paris Tech

Chimie Paris Tech


Mr Andrea Guerrieri (EPFL)


In 2016, a rocky exoplanet has been discovered in the habitable zone of Proxima Centauri, the closest star from our solar system. This may represents our best current opportunity to search for life outside the Solar System. As a leading institution in exoplanet research, the Observatory of Geneva has started a feasibility study for an instrument that would allow direct detection of Proxima Cen b in visible reflected light, and characterization of its atmosphere, by spatially resolving the planet to feed a high-resolution spectrograph.
This instrument includes a powerful Adaptive Optics (AO) system to provide wavefront correction at visible wavelengths, capable to run at 4KHz in closed loop. To estimate the required performance of the RTC (Real-Time Computer), we assume some realistic hardware components for the WFS (WaveFront Sensor) optics, WFS camera, and DM (Deformable Mirrors). In particular, we assume a Pyramid WFS using an OCAM2K camera, and a Boston Micromachines MEMS-DM with 500-4000 actuators as reference. The WFS is able to generate up to 1'500 frames/s of 240x240 pixels in size, or more than 3'000 frames/s of 120x120 pixels with 16 bits depth, with a data rate of up to ~4.2 Gbits/s through a CameraLink interface. The RTC should be able to acquire new images every ~250 us, compute the correction factor and send the correction command vector to the DM after an estimated delay of ~50us, in order to control up to 4'000 actuators in real-time. Our RTC prototype is based on the Intel Arria10SX660, a 20nm SoC (System on Chip) which include an HPS (Hard Processing System) and an FPGA fabric in the same die. In fact, the FPGA side includes more than 1'600 hardened floating-point digital signal processing (DSP) blocks capable to run up to 400MHz, allowing for a high parallelism to execute the matrix-vector multiplications, to execute the correction algorithm in real-time. At the same time, by running embedded linux on the dual-core ARM Cortex-A9 CPU of the HPS we can provide access to the RTC from the external world, for the purpose of monitoring, controlling, and changing the configuration parameters of the AO in real-time.

Presentation Materials

Your browser is out of date!

Update your browser to view this website correctly. Update my browser now