UCLES: EEV 2Kx4K 13.5um Pixel CCD

This page contains information on the AAO's science grade EEV 2Kx4K device - hereafter EEV2. This is a grade 1 (science grade) contract device. It has serial number 7461-16-7 and A6284 marked on the CCD header. It was first commissioned on the AAT in April 2001. The information provided below is based on tests made at the Epping Lab by John Barton (a copy of his report can be found here), and observations made at the AAT with the RGO Spectrograph by R. Stathakis, and UCLES by S. Ryder.

The CCD array has a 2048 (Horizontal) x 4096 (Vertical) pixel format (with an additional 4 unused vertical pixels). Each pixel is 13.5 x 13.5 microns in size, for a total active image area of 27.6 x 55.2 mm. The device has been thinned, and is used back-side illuminated. Its performance has been optimised for blue wavelengths. Binning is possible in the the vertical or horizontal directions.  

Quantum Efficiency

The QE as measured by EEV is shown below (no direct measurements possible by AAO). The figure compares the EEV against the MITLL3 device as well as currently decommissioned detectors (MITLL2 and Tek). The data used for the figure can be found here.

 

Wavelength (nm) 300 350 400 500 650 900
Qauntum Efficiency (%) 25 35.9 73.2 91.7 86.2 30.1

 

Data for the standard star LTT4364 through wide slits observed with the RGO spectrograph centred at 6581Å, 1.7 Å/pix and archive TEK data with the RGO centred at 4778 Å, 3.0 Å/pix are shown below. Unfortunately, although the EEV data extends over the full wavelength range covered by the TEK data, the ends of the EEV data are significantly vignetted by the spectrograph camera. So we only get a decent comparison over the wavelength ranges of the EEV corresponding to similar physical locations in the spectrograph focal plane as the TEK - roughly 4000-6000 Å.

The figure below summarises the photons/unit wavelength/time comparison. From 4000-6000 Å the EEV is 30-40% better than the TEK, as expected from the manufacturers numbers above. 

 

 

Fringing

Fringing is apparent with the EEV2 at wavelengths beyond 6000 Å, and becomes worse as one goes further into the red. At 6000 Å, the fringing is at a level of <2% p-p; it is still <3% p-p at 6500 Å, but rises to 6% p-p at 7000 Å, and 36% at 8000 Å (Note earlier observations with the RGO spectrograph indicate spectroscopic fringing at a level of <0.5% p-p at 6000 Å, 0.5% p-p at 6500 Å, 4% p-p at 7000 Å, 22% at 8000 Å, 58% p-p at 9000 Å, and 60% at 10000 Å). However, careful flat-fielding can mitigate this; for instance, dividing the spectrum of a standard star by a flatfield image taken immediately afterward (see figure below) reduces the fringe amplitude at 7500 Å from 20% p-p, to 4% or less. Thus, unless the wavelengths of primary interest are redder than 6500 Å (in which case the MITLL3 should be favoured due to its superior quantum efficiency), the fringing should not be a serious hindrance to most UCLES programs. 

 

 

 

Fringing with EEV2

 

Cosmetics

The device has a number of hot pixels. The hottest pixel on the device at (1842,1525) generates a faint hot column of about 10 e-/pix as the rows are shifted through the hot pixel during the readout. Thus the hot column appears only on the "far" side of the hot pixel and extends right through to the last row of the readout. No other hot columns have been detected.

Flat field cosmetics appear to be very good. Anecdotal reports from the first spectroscopic commissioning run suggest a virtually featureless flat-field in the blue. Its hard to tell what the flat field performance is like in the red, as the observed flat fields are completely dominated by fringing effects, giving a dominant 'diseased zebra' pattern.

The 'diseased zebra' : the image below shows a sample 914x150 pixel spectroscopic flat-field sample. The left edge is ~8500 Å, and the right edge ~10100 Å at 1.7 Å/pixel. The stretch is such the the features have 60% peak-peak amplitude.

Lab results show regularly spaced lines across flat-fields. These occur every 512 pixels at the boundaries of the stitching process employed in the manufacture of these large CCDs. Each line appears as a pair of rows, one brighter and one fainter than the average for the flat field. The worst pair exhibits a row about 1% brighter and the next row about 4% fainter than the average, i.e. a 5% p-p effect. Hopefully these will flat-field out. This worst pair, in fact, form a limit to the full well for the CCD as this pair had a weaker charge handling capacity than the remainder of the CCD. 

There are no light emitting defects. There are several trapping sites on this CCD but they are relatively small. 

BIAS
 

DARK
Click for larger image

Gzipped FITS BIAS (7Mb) Gzipped FITS DARK  (16Mb)
EEV Full bias vertical cut EEV Dark frame cut
BIAS Frame this is a NORMAL speed bias created as median of 20 frames. The vertical cut shows the sort of bias structure present in the image, and indicates over-scan subtraction is essential for low count images. DARK Frame is a single 1800s dark at NORMAL speed created. It has been overscan subtracted, and had the bias frame to the left subtracted. The horizontal cut through the image above shows the dark count rate in the image to be 0.5+-0.1 adu/1800s. A similar dark taken immediately before this one showed 0.9+-0.1 adu/1800s.

 

Cosmic Rays

A quick visual examination of an 1800s dark count image shows about 400 CR hits per 1800s exposure over the whole detector (where a 'hit' is a several hundred adu peak CR). This is much lower than the MITLL3 (about 1500 hits in the same time, though the detector has a 23% larger surface area).

Dark Current

Dark current is only detectable in binned images and then only if account is taken of the clock induced dark current - the latter dominating in short exposures. The dark current generation rate is about 0.15 e-/pix/2000sec or 0.27 e-/pix/hour. On the telescope the dark current in unbinned mode has been measured as 1.2 +/- 0.1 and 0.65 +/- 0.1 e/pix/1800s  in a pair of consecutive exposures at the start of a run (the decrease is probably due to settling after power on as described from lab tests below). This is negligible for most applications, unless exposures are longer than 500 seconds and a binning of 8 or more pixels is used.  36 hours after being installed, the dark current stabilises to a level of 0.9 e-/pix/hr. 

Although only detectable in binned images (eg 5 x 5 binning), clock induced dark currents are comparatively strong on this CCD. The charge build-up is uniform across the CCD and independent of exposure time. It amounts to about 0.10 to 0.12 e-/pix, so is negligble unless binning more than 20 pixels together. It dominates shorter exposures if the CCD is properly cleaned out, i.e. not recently powered on.

This CCD recovers rapidly after power-on compared to others. After 1 hour the dark current rate is down to about 1 e-/pix/2000sec compared with the MITLL CCDs which require 8 hours to get to this level. The CCD cleanout rate following power off and on, with a cold CCD, as measured in dark frames are:

Time after Power on
Intensity 
(e/pix/2000s) 
30s 
123 
5min 
12 
10 
6.3 
20 
2.9 
30 
1.3 
60 
1.2 
80 
0.7 
100 
0.5 
16 hours 
 0.35  

Residual Images

For the EEV, residuals from non-saturated images or from saturating overexposure should not be a problem. Tests were made for the residual cleanout rate by overexposing a small area of the CCD by a factor of 10 times saturation level, i.e. 2Me-/pix, and then taking a series of 100 sec exposures starting 30 secs after the end of the saturating exposure. This test indicated that the residuals accumulated at a rate less than 1 e-/pix/2500sec after 30 seconds and at 4 minutes were undetectable in a 5 x 5 binned image.

When viewed on the XMEM display, EEV2 images have red (lower number) orders on the left, and wavelength increases going up each order. When saved to disk and viewed with figdisp/Ximtool/Skycat, the images are flipped vertically, so that wavelength increases from top to bottom along each order. This is the exact reverse (i.e. flipped in X and Y) of the MITLL3 orientation.

Vignetting

An important point to note is that the current UCLES camera optics cannot illuminate the entire area of the EEV2 detector (27.6 x 55.2 mm). In fact the unvignetted (<10% vignetting) region which can be observed is more like 18.8 x 38.5 mm. The region covered at 50% vignetting is 60 x 34 mm, which is approximately the entire chip, however unless you are working in the very red, the echellogram will not put any light on much of the chip. You can see the effect of this by looking at the figures on the MITLL3 page, but note that the EEV CCD is physically 10% smaller than the MITLL3, so it sees 10% less of the echellogram than the figures indicate, and is therefore less subject to vignetting.