New CFH12K Focal Plane Performance
			(CFH12K Revisited)

Here is a summary on the performance of the new CFH12K focal plane. It
includes data from the laboratory tests (X rays and optical) as well
as sky test data from the 3rd of September Engineering night. A lot
of data still need to be reduced (photometric zero points in particular)
but what follows will give a clear picture of the current status. All
running parameters should now be based on the current document (gain,

First of all, we now have a great 12K by 8K mosaic. QE, CTE and cosmetic
look great and the camera is still as fast as it was. Image quality is good
over the whole field. All the available filters have been mounted
and tested on the sky (Z focus value is right in the middle range of
the bonnette for most of them). The various low-noise issues have been 
solved in the past month and now the camera has a really low noise
when mounted on the telescope (as good as in the lab). Dark current
is still very low even though the running temperature has been raised
to -88C to reduce the brick wall pattern/improve red response/improve
the CTE. The new window is perfect: the halo contamination in the blue
has been reduced by a factor of 16 in the B band. Residual images from
saturated objects still remains an issue for some CCDs when using narrow 
band filters. It was a 37 Gb night and a 11 Gb lab data collection. 

OAs and SAs: the following document should be mentionned to the 
observers on top of the other documents placed on the CFH12K page
(the present document is included in that page).

Thank you to all of you who've participated in this new effort to put
an even better camera on the sky: more sensitive (new window), larger
and better (new CCDs), more reliable and controllable (ACE) and more 
efficient for the observations.


Mosaic geometry
After replacing and moving CCDs in the CFH12K focal plane the map of
the outputs has evolved:

Hi Rho parts = HZ

 B---------- B---------  B---------  B---------  B---------  B---------  
|           |           |           |           |           |           |
|   CCD06   |   CCD07   |   CCD08   |   CCD09   |   CCD10   |   CCD11   |
|           |           |           |           |           |           |
------------------------------------------------------------------------  East
|           |           |           |           |           |           |
|   CCD00   |   CCD01   |   CCD02   |   CCD03   |   CCD04   |   CCD05   |
|           |           |           |     HZ    |     HZ    |     HZ    |
 A---------  A---------  ----------B A---------  A---------  A---------  

The CCDs are linear within 1% over the whole ADC dynamic (0-65K). The
pixel time is 7 microseconds (145 Kpixels per second per channel, to
be multiplied by 12 for the global data flow, say 1.8Mpixels per second).

Gain (e-/ADU):          1.8     1.7     1.4     1.5     1.8     1.5
                        1.5     1.5     1.6     2.1     1.3     1.7

Noise (e-):             5.0     5.1     5.9     3.9     3.4	5.2
                        4.2     3.0     5.4     4.6    10.5     3.4

Noise (ADU):            2.8     3.0     4.2     2.6     2.2     3.5
                        2.8     2.0     3.4     2.2     8.1     2.0

The bias level in the image area is uniform and equal to the overscan
level. A simple overscan level correction will correct for the bias
level. However, CCDs 05 and 08 have structures (bright vertical columns)
that are part of the bias but this is still only a small fraction of
the whole surface (1%), hence going for a master bias construction and
then subtract it during data reduction would most probably lead to
a general degradation of the data while slightly correcting for these
bright areas (but a bias can be subtracted only for those two CCDs of
course). It may be better to mask them as bad areas.

The dark current is low and uniform for most of the CCDs, and a constant 
can then be subtracted rather than a 2D map of each CCD when the sky
background dominates. CCDs 00, 06, 10 and 11 exhibit some low level
structures. The values for a 10mn exposures are:
(should be checked regularly by taking a single dark exposure)

	 7(*) 	3	6	6	15(*)	34(*)
	 19(*) 	4	4	2	6	8

(*) = Structure in the dark map (5 to 10 ADUs)

Image quality
The image quality was checked on a 0.6" seeing R-band exposure.
The goal was to look for some low serial CTE effect and non-flatness 
of the focal plane. 

It seems like there is a slight tilt of the focal plane along the
North/South axis even though the Center-South area has good image
quality. So, except for the outer parts of CCDs 06 and 07 where 
a degradation of the image quality can be measured (0.15"), the 
image quality is uniform over the whole field.

No CTE effect can be measured. The stars were selected in each
corner of all CCDs (the brightest available within the linearity range).
The angle between the minor and the major axis of the ellipse fitting
the objects is the same over the whole field. 

Image quality (major/minor axes) over the whole field (average):

	South-West	Center-South	South-East

	Center-West	   Center	Center-East

	North-West	Center-North	North-East

	0.74"/0.71"	0.67"/0.60"	0.74"/0.61"

	0.66"/0.62" 	0.59"/0.54"	0.62"/0.55" 

	0.64"/0.60"	0.62"/0.54"	0.58"/0.52"

CTE was measured with a Fe55 source in the lab: CCDs 06 and 10 both
had 0.99994 and as mentioned in my message, this was high enough for
not causing an image degradation for our application (0.2"/pixel
sampling with typical 0.6" seeing). In is indeed the case. On the
other side of the dynamic, for saturated objects the transfer seems
bad for some chips with excellent measured CTE: CCDs 00, 01, 02 and
07 suffer from this effect. But the level is low enough such that
it will be totally dominated by the sky background noise in broad band
filters applications. Anyway, this does not extend over more than
200 pixels.

Residual image
After storing a large number of electrons at saturation (full well
overflows and the pixel bleeds on its neighbors), pixels sites exhibit 
slow release of charges over time. The extra signal will be totally dominated 
by the sky background noise in broad band filters applications. But it has to
be a concern with narrow band filters. As a typical example, a 5mn exposure will
contain the following extra signal on previously saturated pixels (after 30mn,
the release of charges has reached a non detectable level for most of the CCDs).

    20 (36 e-)     4 (7e-)    30 (42e-)    10 (15e-)    4 (7e-)    100 (150e-)

    60 (90e-)     13 (20e-)   15 (24e-)     4 (8e-)     2 (2e-)     17 (29e-)     

A note about saturation: all CCDs but 04, 05 and 08 have a peculiar way
to saturate: level goes up to 65K and then drop to a lower value (depends
on the CCD) then goes down from 65K as the flux from the object decreases.
Some softwares may complain about about such particularity.

Cosmetic is excellent. Only CCD11 can be qualified as B grade on that matter:

        00: ~6 bright columns
        01: ~16 dark columns
        02: Perfect
        03: ~7 bright columns, 3 horizontal combs
        04: 3 bad column, 1 bright column, pop-corn noise
        05: A lot of hot columns (1/3 of the chip at least)
        06: Perfect - QE gash across device
        07: 5 bright columns
        08: Perfect
        09: Perfect
        10: Perfect
        11: Perfect

Fringing was measured for the I band filter: a fringe frame should be
built from a set of I-band exposures and subtracted after proper scaling.

The amplitude of the fringes (%) is measured as: (Max-Min)/Min. Note that
these numbers come from an exposure taken at an airmass of 1.06 and that
the moon (33%) was up 2 hours above the horizon.
These numbers will be different for different airmasses and darker or
brighter sky level, and also for the Z filter (higher).

		0.8%	3.0%	0.2%	1.0%	0.8%	3.0%

		5.0%	1.0%	1.1%	0.5%	0.5%	0.5%	

Bonnette alignment
The camera was aligned along the West-East axis by taking drift exposures.
The reference CCD is 02 (perfect alignment: 0 pixels over 2048), but since
the other CCDs are not mounted parallel to it the alignment will be different
for all the others (to be measured from the acquired data). Then the 
locking screws were set tight on the side of the bonnette. They should
not be released such that we always get the same geometric configuration on
the sky (current angle read is 89.92 degrees).

New dewar window
The new dewar window is great: from the very first exposure it was
clear that the reflection problem is no more. From comparison with the 
first semester data, the effect of a 9th magnitude star is similar to 
what we had with a 12th magnitude star (improvement by a factor of 16 
at least). And then, even with the low background in the B filter, 
the outer halo (600 pixels diameter) is lost in the sky background 
photon noise and the inner halo is slightly visible. This is no more 
an issue for the CFH12K data quality.

All eight available filters for CFH12K were mounted to check the
Z focus value (and the B,V,R,I and Halpha where calibrated one
chip at a time to get a perfect photometric calibration since
the night was photometric). The focal plane has been pushed back
a bit into the dewar while a spacer between the cryostat and
the mechanical stack has been removed, resulting in a 200 microns
offset of the focal plane towards the filter. This is a good
situation where most filters focus in the middle range of the
bonnette. The Z values are (will depend on temperature and is
meant to be an indication for a focus sequence starting point):

	B	V	R	I	Z	Ha      HaOFF   Tio
	4.3	1.7	1.6	3.1	2.1	1.6     1.8     1.8

Jean-Charles Cuillandre
Canada-France-Hawaii Telescope Corporation, 1999