PLATO Main Electronics Units (MEU) based on inputs from P. Plasson and the system team J. Miguel Mas-Hesse Centro de Astrobiología (CSIC-INTA)
PLATO Kick Off November 2010J. Miguel Mas-Hesse2 On-board data processing: Introduction Stellar fieldPLATO image DPS goal: automatically process and extract accurate stellar light curves from the raw, messy CCD images.
PLATO Kick Off November 2010J. Miguel Mas-Hesse3 On-board data processing: baseline concept The normal-telescopes onboard data processing system will be formed by: –4 CCD + 1 FEE per focal plane (telescope) 32 telescopes –1 Data Processing Unit (DPU) per 2 focal planes 16 DPU boards –4 DPUs per 1 Main Electronics Unit (MEU) 4 MEU boxes –1 Instrument Control Unit (ICU) per 4 MEUs 1 ICU (+ 1 fully redundant ICU)
PLATO Kick Off November 2010J. Miguel Mas-Hesse4 On-board data processing: architecture 1 N-DPU for 2 N-cameras F-cameras
PLATO Kick Off November 2010J. Miguel Mas-Hesse5 On-board data processing: architecture Present MEU baseline concept: + 1 DPU per 2 focal planes + 1 single DPU PSU box, external to the MEU’s To ICU-MTo ICU-R SpW
PLATO Kick Off November 2010J. Miguel Mas-Hesse6 PLATO MEU: FEE – DPU SpW link FEE – DPU link baseline: –2 SpaceWire links per FEE (100 Mbps) Alternative: –1 SpaceWire link per FEE (200 Mbps) Feasibility has to be demonstrated. SpW 100 Mbps
PLATO Kick Off November 2010J. Miguel Mas-Hesse7 On-board data processing: architecture
PLATO Kick Off November 2010J. Miguel Mas-Hesse8 On-board data processing: architecture
PLATO Kick Off November 2010J. Miguel Mas-Hesse9 PLATO MEU: operation Window Extraction up to stars / 25 sec Fullimage Acquisition x 8 2 images / 6.25 sec Data correction and LC extraction Masks & window positions updating 6.25 sec 6.25 sec 6.25 sec 5400 sec Target count for 2 x 4 CCD per DPU (to be processed every 25 sec.): Nominal sample 2x( stars/telescope) = stars/DPU Extended sample (+50% margin) = stars/DPU Background win.2 x 400 Imagettes up to 2 x Offset windows2 x 2 x 4 half-row (2.255 pixels) Smearing rows 2 x 4 x 10 overscan rows Total memory requirement, including 50% margin on star count: 250 Mbytes per DPU
PLATO Kick Off November 2010J. Miguel Mas-Hesse10 PLATO MEU: CPU load The CPU load has been estimated for various processors, based on current baseline for number of stars + extraction algorithms. At this phase the %CPU load should be ≤ 50%. The present concept seems feasible, but: –with 1 DPU/2 telescopes, no margin on star counts acceptable. –with 1 DPU/telescope an extended sample with +50% margin in star counts could be considered, if required.. Weighted mask photometry assumed
PLATO Kick Off November 2010J. Miguel Mas-Hesse11 PLATO MEU: open points Backup concept: + 1 DPU board per telescope + 4 DPUs per MEU + 8 MEU More powerful, but too heavy
PLATO Kick Off November 2010J. Miguel Mas-Hesse12 PLATO MEU: open points BaselineAlternative Number of telescopes per DPU 21 MEU-PSU1 MEU-PSU box for 4 MEU boxes 4 MEU-PSU inside each MEU box FEE / DPU number of SpaceWire links 2 SpaceWire links per FEE (100 Mbps) 1 SpaceWire links per FEE (200 Mbps) FEE DPU image transfer mechanism and protocol RMAPDMA transfer tables DPU Boot mechanismBoot Software in PROMPure RMAP hardware boot
PLATO Kick Off November 2010J. Miguel Mas-Hesse13 PLATO: next steps The industrial MEU definition study is starting now. Goals: –Definition and dimensioning of the N-DPUs Global architecture, choice of processors, number of SpaceWire data links, location of PSUs, mass, power comsumption,… Trade off on algorithms potential increase of complexity – Development of simulators/breadboards for critical interfaces Spacewire link FEE – DPU Final design within mass and power budgets, but compliant with scientific requirements!