Noiseless, high frame rate (> kHz), photon counting arrays for use in the optical to extreme UV John Vallerga, Jason McPhate, Anton Tremsin and Oswald Siegmund Space Sciences Laboratory, University of California, Berkeley Bettina Mikulec and Allan Clark University of Geneva
Future WFS detector requirements High optical QE for fainter guide stars Lots of pixels - eventually 512 x 512 More accuators More complex LGS images (parallax, gated, etc) Off null / open loop operation Very low (or zero!) readout noise kHz frame rates
Photon Counting Charge integrating Events Q V ± sv Events ± sEvents Threshold Events Count (x,y,t) Charge integrating Q ADC V ± sv Events ± sEvents
Centroid in presence of noise: 8 x 8 Noiseless 35% QE 10 photons - 100 photons 1000 photons 8 x 8 2.5 e- rms 90% QE 6 x 6 2.5 e- rms 90% QE 4 x 4 2.5 e- rms 90% QE
Centroid error vs. input fluence
Imaging, Photon Counting Detectors Photocathode converts photon to electron MCP(s) amplify electron by 104 to 108 Rear field accelerates electrons to anode Patterned anode measures charge centroid
Bandpass by photocathode selection
MCP Detectors at SSL Berkeley COS FUV for Hubble (200 x 10 mm windowless) 25 mm Optical Tube GALEX 68 mm NUV Tube (in orbit)
Wavefront Sensor Event Rates 5000 centroids Kilohertz feedback rates (atmospheric timescale) 1000 detected events per spot for sub-pixel centroiding 5000 x 1000 x 1000 = 5 Gigahertz counting rate! Requires integrating detector
Our AO detector concept An optical imaging tube using: GaAs photocathode MCPs to amplify to ~104 Medipix2 ASIC readout
Medipix2 ASIC Readout Each pixel has amp, discriminator, gate & counter. 256 x 256 with 55 µm pixels (buttable to 512 x 512). Counts integrated at pixel. No charge transfer! Developed at CERN for Medipix collaboration (xray) ~ 500 transistors/pixel
First test detector Demountable detector Simple lab vacuum, no photocathode Windowless – UV sensitive
UV photon counting movie
Sub-pixel spatial linearity Lamp Pinhole Detector
Imaged pinhole array Pinhole grid mask (0.5 x 0.5 mm) Gain: 20,000 Rear Field: 1600V Threshold: 3 ke- Gap: 500µm
Avg. movement of 700 spots 1 pixel
Position error (550 events/spot) rms = 2.0 µm
Vacuum Tube Design
Vacuum Tube Design
Vacuum Tube Design
Vacuum Tube Design
Medipix on a Header
Summary Noiseless detectors outperform CCDs at low fluence per frame Photocathode choice to fit application Medipix ASIC readout allows for a huge dynamic range, fast frame rate. MCP/Medipix Status First tube in Fall 2005 GaAs tube in 1st half of 2006
Future Possibilities Medipix 3 now being discussed 130 nm CMOS technology Faster front end for less deadtime per pixel Faster readout rate (10 kHz frame rate) Radiation hard Si APDs rather than MCPs as photon converter/amplifier Higher optical QE Near IR response Cooling will be required to reduce dark count rate
Acknowledgements This work was funded by an AODP grant managed by NOAO and funded by NSF Thanks to the Medipix Collaboration: Univ. of Barcelona University of Cagliari CEA CERN University of Freiburg University of Glasgow Czech Academy of Sciences Mid-Sweden University University of Napoli NIKHEF University of Pisa University of Auvergne Medical Research Council Czech Technical University ESRF University of Erlangen-Nurnberg
Hexagonal multifiber boundaries Flat Field MCP deadspots Hexagonal multifiber boundaries 1200 cts/bin - 500Mcps
Histogram of Ratio consistent with counting statistics (2% rms) Flat Field (cont) Histogram of Ratio consistent with counting statistics (2% rms) Ratio Flat1/Flat2
256 bit fast shift register Readout Architecture 3328 bit Pixel Column 0 3328 bit Pixel Column 255 3328 bit Pixel Column 1 256 bit fast shift register 32 bit CMOS output LVDS out Pixel values are digital (13 bit) Bits are shifted into fast shift register Choice of serial or 32 bit parallel output Maximum designed bandwidth is 100MHz Corresponds to 266µs frame readout
256 bit fast shift register 3328 bit Pixel Column 0 3328 bit Pixel Column 255 3328 bit Pixel Column 1 256 bit fast shift register 32 bit CMOS output LVDS out
“Built-in” Electronic Shutter Enables/Disables counter Timing accuracy to 10 ns Uniform across Medipix Multiple cycles per frame No lifetime issues External input - can be phased to laser
EUV and FUV (50 - 200 nm)
GaN UV Photocathodes, 100- 400 nm The sensitivity of the activated photocathode is considerably higher, with a quantum efficiency of as high as 37% at about 450 Å and a sensitivity cut off energy of about 6.2 eV. The observed decrease of sensitivity at 988 and 520 Å (and probably not observed at ~670 Å) is likely to correspond to 2xEgap+Eaff,, 4xEgap+Eaff,, and 3xEgap+Eaff,, respectively. A much better spectral continuity of the illumination source is required in order to study the UV absorption fine structure.
Hayashida et al. Beaune 2005 NIM GaAsP Photocathodes Hayashida et al. Beaune 2005 NIM
Avalanche Photodiodes (APDs, Geiger mode) Single photon causes breakdown in over-voltaged diode QE potential of silicon Arrays in CMOS becoming available But Appreciable deadtime Low filling factor High dark counts, crosstalk and afterpulsing
APD arrays 32 x 32 Edoardo Charbon Ecole Polytechnique Federale de Lausanne
L3CCD (e2V Technologies) Integrates charge Multiplies charge in special readout register Adjust gain such that se < 1e- But Multiplication noise doubles photon noise variance Single readout limiting frame rate
Assumed performance parameters CCD Medipix-MCP Binning 2 x 2 6 x 6 8 x 8 QE (%) 90 35 Readout noise 2.5 e- Seeing width (pxls FWHM) 0.75 2.25 3 Diffract. width (pxls FWHM) 0.5 1.5 2
Gaussian weighted center of gravity algorithm: From Fusco et al SPIE 5490. 1155, 2004
Advantages of multi-pixel sampling of Shack-Hartmann spots Non-linearity of 2 x 2 binning
Advantages of multi-pixel sampling of Shack-Hartmann spots Linear response off-null Insensitive to input width More sensitive to readout noise
Technology advantage High QE CCDs Number of pixels CCDs, Medipix Readout noise APD, Medipix, L3CCD Frame rate Medipix, CCD Gating Medipix
Soft X-Ray Photocathodes