[ ESO ]

CCDs at ESO - Performance and Results
HOME INDEX SEARCH HELP NEWS

EEV CCD 44-82

EEV supplied ESO with custom designed 2k x 4k CCD called EEV 44-82. This device has 2 readout ports. The CCD can be operated either in dual readout mode using two ports, or single port readout. Binning along the X and Y direction is possible using any kind of binning factor provided that the scene is not oversatured. A fast wiping of the device can be achieved by using a dump gate close to the serial register.

EEV CCD 44-82 Geometrical Dimensions and Specifications

EEV CCD 44-82
Picture of the EEV CCD 44-82		          This device is 4 side buttable. For each device, the theoretical non-sensitivity side area is 100 microns and the left-right gaps are 450 microns each. The silicon sensitive surface is installed into a custom ESO made invar package allowing four-side buttability.

Noise

The following table compiles noise performances achieved with FIERA at different readout speeds, conversion factors (e- / ADU) and different instruments. It has been measured using the CCD including the rest of the instrument in normal operation, meaning that this figures also includes the noise from the controller.

Speed (kps) Conversion factor (e- / ADU) Time to readout the full array using one port (sec) Noise (e- rms) Instrument
50 0.54 170 1.9 UVES/ VIMOS
225 0.54 38 3.2 UVES/ VIMOS
325 2.0 27 4.0 WFI
625 1.64 14 4.7 VLT TC 2
1000 2.1 9 10.0 -

These noise figures may vary slightly according to the device used, and are better with higher conversion gain. Figures less than 1 e- / ADU as mainly used for spectroscopy purposes, whereas higher conversion factor is used for imaging applications. The ODT employs also a special procedure based upon the changes of VOD, VDR and JD voltages (3D-voltage cube) to get the best noise / linearity figures.


Quantum Efficiency

EEV has optimized the QE to get good QE in the B and U bands. Nevertheless, depending on the device, the QE in this range may vary slightly. The explanation is the following : The quantum efficiency of these devices is strongly related to of the thickness backside anti-reflection coating and this thickness may vary according to the device ( Hafnium Oxide 50 nm ). ESO measured 12 devices and used to achieve a statistic QE plot. This is a criterion also to sort science grade devices : the best QE in the blue ( except for application requiring a higher red sensitivity like "red arms" of spectroscopes ).          Average QE
Average QE of 12 devices, error bars show the QE standard deviation from these 12 devices. Also the best blue sensitive device is plotted.

If the wavelength is higher than 650 nm, QE is extremely stable devices from devices. All the QE curves have been obtained from the ESO CCD Testbench ( 2 ).


Fringing - Photo Response Non Uniformity ( PRNU )

Fringing is an issue for spectroscopic purposes making images harder to flat field out. ODT has measured this effect with 8 devices. The resulting parameter is called PRNU across the chip (Photo-response non-uniformity). The near IR PRNU depends upon the fringing effect. This effect is related to the thinning of the CCD, and also to the aperture of the incoming beam, the more this beam is open, the less the fringing is visible.

        
Flat fields from the same area at different wavelengths. 5 nm bandwidth, left 320 nm, middle 650 nm, right 950 nm, F/2 beam. 300 x 300 pixels positive images.

In the blue part of the spectrum the PRNU degrades also due to the backside p+ implementation laser annealing. The images show qualitatively these effects. The shown plot is a statistical analysis over 8 CCDs.

Average PRNU
Average PRNU from 8 devices, error bars show the PRNU standard deviation from these 8 devices, F/2 beam.		       Between 400 nm and 850 nm, PRNU is very good, and is almost photon-noise limited ( also the 1k x 512 stitching pattern is visible ). PRNU is measured here using a dust-free area and taking 10 % and 90 % of the computed histogram as deviation figures. Below 400 nm, a diamond pattern is visible. This one can be flat field out easily provided that the CCD temperature is stable within 2 - 3 degrees. Measurements have been done about the stability of this diamond pattern according to the temperature, and a variation of 20 degrees ( 170 K - 150 K ) shown a 1.25 % pp PRNU variation ( 3 ).



Cosmetic Defects

The different kinds of defects, which degrade the cosmetic quality of the device, can be split in three parts :

These defects are based upon the grading of the device, typically a good science grade is one with less than 5 defective columns. The amount of defects visible on long dark exposure is strongly related to the CCD temperature. An operating CCD temperature of -120 °C is used to minimize these effects.

Also the CCD device is made of 1024 x 512 blocks, and the area boundaries show sometime a 1 - 2 % QE variation over 1 row/column due to photolithography stepper mismatches. These small defects flat field out perfectly trap defects are a determining factor for assigning devices to certain applications. These kinds of defects cannot be suppressed because they are created during chip fabrication, and are not temperature dependent.

Finally, the amount of the "defects" induced by cosmic rays should not be more than the one provided by natural radioactivity. The CCD package and the head of the detector do not add additional hits. This value is typically measured between 1 and 1.5 event/min/cm2.

       Block Stitching Effect
Effect of block stitching between 1k x 512 areas.

Charge Transfer Efficiency

ESO measurements shows horizontal CTE of 0.9999995 ( six 9s, 5 ) and vertical CTE of 0.9999988 ( almost six 9s ) using EPER method. It means in the worst case : A pixel having 1.000 electrons located on the opposite side to the readout port ( X = 2048, Y = 4100 ) will lose 6 electrons, once the charge packet will reach the readout node ( photon shot noise is about 31 e- at that level ).


Dark Current

EEV 44-82 devices do not have MPP implant. The dark current is temperature dependent. The CCD must be cooled at a temperature less than - 95 °C for science purposes to achieve a dark current less and equal to 10 e-/pixel/hour. ESO-ODT has measured the dark current according to various temperatures.

Dark Current versus Temperature   
Temperature measured on CCD package Dark current
- 102 °C / 171 K 20 e-/pixel/hour
- 120 °C / 153 K 1 or 2 e-/pixel/hour

Below 180 K, the theoretical curve cannot be applied, the measured dark current remains higher. This could be due to the residual image, even thought the CCD is let in the dark during many hours.


Remanence Effects - Residual Image

Remanence Effect
Remanence Effect		       Remanence effects may occur if a flat field ( mean > 10.000 e- ) is taken prior to the dark frame. The resulting effect resembles an increase of dark current, especially at temperatures lower than 180 K. ODT is currently investigating to cure this problem by applying a special wiping sequence ( with different parallel voltages ). On the left you can see a set of 24 dark frames exposed each 1 hour ( CCD operating temperature at 160 K ). This set was taken after the CCD was exposed to the ambient light. It requires four 1 hour dark frames to eliminate the residual image and to reach the actual dark current value ( about 3 to 5 e-/pixel/hour ).

Bright Star / bright spots remnant might be also visible when doing long exposure afterwards :
  
Figure 7a
Figure 7a : Oversaturated Spot
Figure 7b
Figure 7b : One hour dark exposure, just after figure 7a, binned 10 x 10 pixels, plus cosmic hits.
Figure 7c
Figure 7c : One hour dark exposure, just after figure 7b, binned 10 x 10 pixels, plus cosmic hits, only the central spot is visible.
     
Figure 8a
Figure 8a : 2048 x 512 pixels subframe, one hour exposure dark frame, just after a flat field, the dark current is 43 e-/pixel/hour and the "dark current" is related to the "blue diamond" pattern.
Figure 8b
Figure 8b : Same frame as figure 8a, 2nd one hour exposure dark frame, the diamond pattern is fainter, 10 e-/pixel/hour of dark current. Overscan- / prescan area is visible.
Figure 8c
Figure 8c : Same frame as figure 8b, 28th one hour exposure dark frame, dark current has reached its minimum value ( 4 e-/pixel/hour ).

Amplifier Glowing

To avoid amplifier glowing, voltages less than 25 Volts must be applied to the VOD node of the chip ( Channel voltage = 11 V ).


Linearity

Linearity is measured using two ways :
By optimizing the voltages applied to the CCD readout amplifier, a good linearity can be achieved in the range of 0 - 100 ke- without degrading noise performance. The table below summarizes a typical linearity performance from an EEV device at 225 Kps.

Port Speed (kps) Peak - Peak (%) Rms (%) Range (ADU) Saturation (e-)
A 225 + 0.20 / - 0.35 0.15 0 - 62000 96 k
B 95 k

Linearity of EEV devices

Left and right port ( 225 kps ), using a "TDI" method. Best performance is achieved with p - p linearity of ± 0.35 %. A special voltage ( voltage cube ) optimization is performed to get ± 0.5 % at least across the full range.

       Non-linearity residual plot using the TDI method
Non-linearity residual plot using the TDI method


Flatness

CCD Surface
Typical iso-elevation 2D plot of the CCD surface ( units are microns ).		       ODT has developed a special measurement device to acquire full chip surface profile ( 4 ). A typical value is ± 6 microns at 150 K, a maximum value is ± 10 microns p - p.


MTF

According to ODT's latest measurements, the CCD shows no serious degradation from the Nyquist curve. The very sharp hits resulting from the cosmic ray impacts show that 95 % of the energy released is within one pixel.       Cosmic hits
Cosmic rays hits ( 1 hour exposure ), enlargement showing 4 cosmic rays hits - NEGATIVE IMAGE.


Binning Eclipse Effect on Saturated Stars

Eclipse Effect
Eclipse effect on a saturated star with binning 2 x 2 - POSITIVE IMAGE.		       When using binning more than 1 x 1, a saturated star appears with a strange black hole in the middle. The CCD output amplifier is, in this particular state, completely out of range and flips upside-down ( visible at CCD output signal, probed with a scope ). As there is, anyway, no useful information in the center of a saturated star ( no photometry / astrometry feasible ), this effect is more a cosmetic issue ( which could be eliminated by software ) than a scientific problem.


Blooming Effect

To avoid loss of fill-factor, this CCD has not antiblooming implant. This causes vertical trails over bright and saturated stars. The table below gives a rough indication of the order of magnitude for point sources. L1 denotes the number of pixels concerned toward the serial register, L2 the one in the opposite direction. Measurements made with binning 1 x 1.

Direction Overexposure Factor
2 4 5.8 10 25 50 100 1000
L1 ( pixels ) 0 0 1 2 8 19 39 344
L2 ( pixels ) 0 0 1 3 5 8 13 136

Back to   "CCDs at ESO - Performance and Results"
©   2000/Jan/26 by C. Cavadore, R. Dorn, J. Beletic

Send comments to   odt@eso.org
Last update: Mar 25, 2004 - webdesign by   EVI
[ESO] [Index] [Search] [Help] [News]