International symposium on the development of detectors Electron multiplying CCDs P.A. Jerram, P. J. Pool, D. J. Burt, R. T. Bell e2v Technologies, Chelmsford, Essex, UK
Brief Introduction to the EMCCD Structure Performance requirements Overview of non-space applications Possible applications in Particle, Astro-particle and synchrotron radiation detection All performance data derives from e2v L3Vision CCDs
L3Vision CCD Iφ1 Iφ2 Iφ3 A standard CCD becomes an EMCCD by the insertion of a multiplication register Rφ1 Rφ2Rφ3 Rφ2HV
Operation
Operation and this is the signal out from the device from single electrons. If you have spent your life looking at CCD waveforms this is really exciting.
Read Noise In a CCD, charge transfer is noiseless, so the only remaining noise source is in the charge detection circuit. In the EMCCD impact ionisation in silicon provides gain in the charge domain before detection. This gain effectively reduces the read noise by the gain factor. However, for high gain values, the gain is subject to statistical fluctuations which appears as noise of N electrons, where N is the input signal (electrons). The combination of this excess noise factor, together with shot noise, results in a noise on the signal of (2N) : an excess noise factor of 2
Read Noise The excess noise factor of 2 has been shown to be present for the gain factors used in most applications. Theory
Noise v signal 16.00 Noise rms electrons 12.00 8.00 4.00 3 5 7 L3 Noise 0.00 0 10 20 30 40 50 60 70 80 90 100 Signal electrons Consequently, EMCCD Noise is low for small signals, but higher for large signals. However, EMCCD noise can be independent of pixel rate
Dark Noise Dark Current of E2V CCDs 1E3 MAXIMUM VALUES 1E2 Electrons/sec/square micron 1E1 1E0 1E-1 1E-2 1E-3 1E-4 Back illuminated Standard MPP 1E-5 1E-6-120 -100-80 -60-40 -20 0 20 40 Temperature, C Having lost read noise, an EMCCD should be operated under conditions in which the dark charge/pixel does not contribute significantly to the noise.
Clock Induced Charge (CIC) If sufficient cooling is not available, multi phase pinned (MPP) operation can be used. This makes the system susceptible to CIC, but this can be minimised by careful choice of operating conditions. In MPP operation, the low clock inverts the silicon surface, saturating it with holes. CIC is caused by impact ionisation of the holes as they move in and out of the Si/SiO 2 interface during clocking. The charge generated is dependent on the number of transfers through the CCD and not the integration time. Dependent on clock amplitude, transfer rate, and clock timing
Clock Induced Charge (CIC) Clock Induced Charge (e/p/f) 6 5 4 3 2 1 0 2Ø FT @ 927kHz 2Ø FT @411.5kHz 2Ø FT @ 246.8KHz 10 11 12 13 V HIS (Volts) Increasing clock speed CIC minimised by low clock swing and high clock speed, in particular minimising the time a clock is held high.
Quantum Efficiency 100% BI QE Curves at -50 C 80% Quantum Efficiency 60% 40% 20% basic process MIDBAND AR enhanced process BLUE AR enhanced process UV AR deep depletion NIR AR Data from the SECCHI programme (US Naval Research Laboratory) 0% 250 350 450 550 650 750 850 950 1050 wavelength (nm) To complement low read noise and minimal dark noise, EMCCDs benefit from the high quantum efficiency (QE) available from back illumination. Appropriate processing promotes useful QE anywhere in the wavelength range 1 Ångstrom to 1 micron
Overview of three current applications Surveillance Ground-based Astronomy LIDAR
24 Hour Surveillance Daylight With variable gain and resistance to overload damage, the EMCCD is ideal for 24 hour surveillance applications. As expected the EMCCD outperforms an intensified CCD in daylight, because gain can be turned down, so it s like a standard CCD.
24 Hour Surveillance Overcast Starlight When it s seriously dark, the gain can be increased, so it still does the job. This image represents ~1 electron/pixel/frame. Some of you may have concluded that this does not show a real golfer. Nobody has a good swing when it s this dark, so he was the only volunteer
Ground-based Astronomy. Road-map for ESO adaptive optics 8 outputs 4 outputs e2v-ccd-39 e2v-ccd-50 e2v-ccd220 8 outputs 4 outputs MAD-WFS CCD 80x80 pixels 4 outputs 500Hz frame rate Noise: 8-6 e - rms NAOS-WFS CCD 128x128 pixels 16 outputs 25-600 Hz frame rate Noise: 2.5-6.5 e - rms Future-WFS CCD220 CCD 240x240 pixels 8 L3 outputs 0.25-1.2 khz frame rate Noise: < 1 e - rms
Single pixel CCD for LIDAR To operate at even higher rates for example to look at the output from a single fibre a single pixel CCD could be used. A structure of the type below is currently being developed by e2v for LIDAR with ESA funding. Supplementary Channels ABD Image Section DD SS RØ2HV RØDC ABD ØC IØ1 IØ2 DD SW TG ID Parallel Summing Well OS Binning Register Gain Register IG DD BØ2 ØDG BØ1 RØ1 RØ2 RØ3 OG ØRRD OD
Particle and Synchrotron radiation detectors L3 devices provide significant advantage when: A high frame rate is required at low noise The signal is very low single photons Frame rate can be >1kHz with an equivalent noise of <1 electron Single pixel frame rate can be > 1MHz But give no advantage or degrade performance when noise is dominated by shot noise
Particle Detection For example for the detection of light from scintillating fibres in a positron detector. The photodiodes could be replaced by an L3 CCD with the fibres bundled onto the CCD. To photodiodes or PMTs Could be replaced with a single L3 CCD
Particle Detection For example for the detection of light from scintillating fibres in a positron detector. The photodiodes could be replaced by an L3 CCD with the fibres bundled onto the CCD or with LIDAR devces. To photodiodes or PMTs OR replace with a monolithic array of a L3 LIDAR chips
Particle detectors any other applications? Thanks for your attention Paul Jerram E2v technologies Waterhouse Lane Chemlsford UK paul.jerram@e2v.com