VISTA Camera Calibration Meeting 10/3/04 10:30 am
-------------------------------------------------

Present:

Mike Irwin                 MJI Cambridge    
Peter Bunclark             PSB Cambridge 
Jim Lewis                  JRL Cambridge
Simon Hodgkin              STH Cambridge
Richard McMahon            RGM Cambridge (part-time)
Dafydd Wyn Evans           DWE Cambridge
Will Sutherland            WJS Cambridge  
Malcolm Stewart            JMS ATC       (chair)
Steven Beard               SMB ATC    
Nagaraja Bezawada (Naidu)  NNB ATC 
Nirmal Bissonauth          NB  Durham
Jim Emerson                JPE QMUL

Introduction (JMS)

The ESO PDR 13 May idea is input for cal plan:

1st goal Documents have to be finished by 8 April (real date is 13th but
Easter gets in the way). We will deliver PDFs electronically
- JPE will confirm with ESO that this is OK                                <<<<

2nd goal Identify calibration procedures

3rd goal Provide input for Data Reduction Specification document - this needs
to be in outline form by PDR

We don't need to discuss telescope calibration, guider, LOWFS and camera
calibrations that don't impact pipeline.

Some discussion on whether pipeline astrometric calibrations can feedback
into the pointing tests as well as guide star feedback.  Preference to keep
them as separate calibrations.

3 documents are needed for the PDR (V0.5 means draft for PDR)
  DFS User Requirement - not too much detail
  Calibration Plan - this meeting
  Data Reduction Specification (outline)
  
Flames calibration plan has been stated by ESO to be a good model and
contains Science data calibration and Instrument monitoring plans.


ESO PDR Deliverables (PSB)

VLT Calibration Plan - It was noted that FLAMES calibration plan is terse
and well laid out but lacking certain items (eg. templates).  Therefore
should we do what we think is necessary as well as just making it ESO
compliant?

JPE said that we have to go by ESO rules (document VLT-SPE-ESO-19000-1618)
PSB clarified that he meant: we give extra. A number of people asked
what exactly do we deliver to ESO? JPE said all was in 1618.  After
some discussion it was agreed that requirements should only include:
instrument signature removal, DQC information generation, and photometric
and astrometric calibration of pawprints.

Detector Characteristics (NNB)

Three engineering detectors have been tested

#15 Large defect (bad pixels) 10-20% of chip
Bands seen - different to plots shown at FDR  - was this due to
scaling or a different read mode?                                          ????
Non-linearity is 1.9% (over 20-80% range)
QE 59% 57% 49%
    J   H   K
    
#16 No large area defects
QE up 76% 76% 65%
Patches of bad dark current characteristics

#18 Has spots - NNB said that they will flat field out.
QE up again 93% 97% 96%!

Persistence decay is about 1 minute - this was done after flooding the
detector.

Flooded array 200ke-
60s later residual flux 400e-   0.2%
70s                     100e-   0.05%
100s                     50e-   0.025%  (at the level of the read noise)
Cause - electrons trapped in the lattice, effective capacitance of detector
pixels? Resets do not seem to help.

Spot tests will be carried out.
MJI asked if the charge spreads out at all? Not known.
Will use cosmic rays as point sources

The new detectors from Raytheon will have higher gain - the main change is
to get rid of the dark current problems. The detectors with the new structure
are expected soon.


Observing Blocks and Templates (SMB)

An Observing Block is the smallest schedulable unit and is made up of a
number of templates. Normally the first template will be an acquisition one.
When an exposure is taken, in science mode all 16 detectors will be output.

PSB clarification - do we need to supply the technical templates since
"1618" states "all templates" for instrument? SMB suggested that for V0.5
we put TBD on a number of engineering templates. JPE added that we should
only detail things needed for the pipeline and other things could be placed
in as a list with TBD.

It was confirmed by SMB that all input to the pipeline will be full 16 frame
files (even if some of these may be duff).

NNB then pointed out that it takes 1s to readout the detector (from first to
last pixel) and similarly 1s to reset. There then followed some discussion
regarding the accurate timing of observations. The conclusion was that there
would be a timestamp in the FITS header and there would be a need for a set
of timing offset values for all pixels.

STH pointed out that the 1s readout would have implications for for taking
twilight flats - different parts of a frame will be exposed at different
times and the sky level is changing rapidly at that time (~1% effect?).

There then followed a discussion regarding the non-linear correction and its
interaction with the above 1s delay. The current idea is that the data is
generally output via a sequence of reset/read/read with the end result being
the difference between the two reads and any reset pattern is removed.
The camera team want to apply a non-linear correction after this point so
that, if it has been done incorrectly, it can be redone. However, it was
pointed out that since there is a 1s delay between the reset and the first
read will contain 1s worth of sky and therefore it would not be correct to
apply the non-linear correction to the difference between the two reads.

This is the reason why WFCAM has the non-linear correction applied within
the DAS ie. on the reads before they are differenced. VISTA should also
consider following this idea (ie. applying the correction within the IRACE
VXworks crate), or the possibility of keeping both reads
applying the non-lineariy correction at the earliest opportunity in the
DAS and then only keeping the difference.

There was also some discussion regarding the possibility of extrapolating
backwards knowing the differences in the measurements at the times of the
two reads and when these reads occurred, relative to each other and to the
reset.  This is probably OK except in rapidly varying conditions, e.g. when
doing the twilight flats.

The various readout modes were then discussed (reset/read/read and
read/reset/read) and if there are advantages to either mode. Overheads and
the non-linear correction were considered and after much blackboard work,
both modes were considered equal with regards to the two mentioned problems.

For the PDR it should be assumed that the current mode will be used.
NNB was asked to find out what mode ISAAC uses.                            <<<<

(lunch)

There was a continuation of the discussion regarding the non-linearity
correction and whether it was possible to store both reads. It was decided
that WJS should investigate this (with help from MJI and JRL) and write a
short report to see what the difference is between the various approaches. <<<<

The next part of the meeting was to discuss the templates. The approach
decided upon was to go through the calibrations needed and decide what
templates are needed. For a number of calibrations the current templates
were sufficient and there were no comments.

Reset Frames:

Dark Frames:

Dome Flats: WJS said that the illumination level would be controlled by 3
pairs of bulbs. They can either be on or off, so this will yield 8
illumination levels.  It is therefore important that the bulb status be
placed in the FITS headers. Similarly, 3 new keywords are needed in the
templates.

Linearity: The control of this will be done via the exposure time. There was
some discussion as to the need for some low-level templates eg. to switch the
lamps on and off. An alternative suggestion was for acq_domescreen to have
the ability to turn the lamps on and off.

Radial Distortion Effect: It was decided that this should be part of the
astrometric calibrations.

Twilight Flats: Should the template be able to measure the sky level and
feed this information back into the next template? This is because there is
only about a 3 minute window during which the twilight flats can be taken and
it is difficult to predict when to start a twilight flat sequence. It was
decided that the best way would be to get a "continous" measurement of sky
from a small window on one of the science devices and wait until the best
time to start the twilight flat sequence rather than compute by dead reckoning
(due to the inevitable variation in sky conditions).

The twilight templates should also have the capability to do little
offsets in order to avoid co-stacking bright stars.

Night Sky Flats:

Sky Subtraction and Fringe Removal:

(Further clarification was given on the definition of an OB - they will
start with an acquisition template and only contain one such template that
moves the telescope ie. only one slew is allowed within an OB. Any further
movement of the telescope must be carried out with an offset - though a
large offset - and implicit reacquisition of guide stars - is allowed).

Reset Anomaly: This can be corrected from the dark frames and should be
moved to that section.

Image Persistence: The reference to "ghost" should be changed to "residual"
to avoid confusion with ghost images caused by the optics.
   There is a need to devise a procedure to measure this eg. bright star
followed by blank filter - this can be done with the current templates.
   How often should this be carried out?
   
Cross-talk: The text needs to detail that this refers to electrical
cross-talk. No new templates would be needed. This calibration would update
data obtained from the laboratory tests.

Dithering: This is to remove the effects of bad pixels. A variety of offsets
and positions would be required.

Astrometric Calibrations: 2MASS (systematics < 100 mas) would be used for
this.

Photometric Calibrations: WJS pointed out that there was a need to have
standards to the North and South of the zenith in order to minimize the slew
time of the telescope. There was then some discussion whether or not the
observing system could take account of this and nobody knew.
STH said that there was probably no need to defocus the telescope in
order to carry out the photometric calibrations. He would help PSB rewrite
that part of the document (start of section 4.1).                          <<<<
There was some mention of the operator having to make decisions at some
point - one of the templates was an offset_sky which gave the operator a 
chance to look for bright stars. This was thought to be a bad idea and that 
everything should have been decided beforehand and be automatic.


Templates

PSB's plan was to edit SMB's Section 8 into his document.
(the next bit of these notes was grabbed from PSB and SMB's emails).

VIRCAM_ is prefixed to all the following templates

8.2.1
howfs_acq

8.2.2
howfs_cal_dark	Delete, use img_cal_dark

howfs_cal_bias  Delete, use img_cal_dark with zero time, or don't need it
		at all

howfs_cal_flat

8.2.3
howfs_obs_exp

howfs_obs_wfront

8.3
img_acq

img_acq_offset

img_acq_dither	Delete; dithers will be incorporated in img_obs_dither

img_acq_sky	Rename to img_acq_twilight

Also the function of this template will change. Rather than acquiring a patch
of blank sky (which is now done with the normal img_acq template), this
template acquires twilight at a particular altitude and azimuth.

Separate templates for acquiring evening and morning twilight will be
needed, or the difference might be specified in a keyword. The act of
acquiring the field might include a check on the sky brightness level and a
wait until the level reaches the right value.

It was suggested that an evening and morning twilight flat could be combined
to reduce the effect of the changing sky brightness during exposure. If the
flats can be distinguished by the pipeline by the DATE-OBS keyword then one
template would be sufficient.

img_acq_domescreen

Kept. With the addition of three keywords (to be defined in the telescope
dictionary and included in the camera to telescope ICD) to switch the
three lamps on or off. The template can also be used at the end of the
calibration to switch the lamps off. (The ESO restriction of one
acquisition template per OB refers to one telescope "preset" [aka slew]
and one logical target field rather than one actual template).

8.3.2
img_cal_dark

img_cal_bias	Delete, use img_cal_dark with zero time

img_cal_darkcurrent	(Delete, replace with OB) (This was reinstated after
                         the meeting as a template).

img_cal_badpixels	Delete

howfs_cal_badpixels

The above needs to be added in response to the deletion of
img_cal_badpixels. The HOWFS needs a bad pixels calibration independent of
the DFS pipeline's calibration. This doesn't affect the pipeline.

img_cal_domeflat

This template will no longer turn the calibration lamps off when completed.
The img_acq_domeflat template can be used to do this instead.

img_cal_linearity	(Delete, replace with OBs) (This was reinstated after
                         the meeting as a template).

img_cal_sky		Delete, function replaced with use of science data

img_cal_skyflat		Rename img_cal_twiflat; some magic needed to
			cope with practicalities of handling twilight

8.3.3
img_obs_stare		Rename to img_obs_exp

img_obs_snapshot	Delete, use img_obs_exp

img_obs_telcalib	Delete, pointing now done on autoguider

img_obs_dither

img_obs_microstep

These two need to remain as templates so a set of named dithering patterns
can be defined. It would also allow dithered maps to be tiled together just
like img_obs_exp templates. 

Post meeting comment from SMB:

"I need to clarify my understanding of what a template generates. I think I
was mistaken yesterday when I said a template results in one exposure.
That's only true if your instrument is a simple camera. Some more complex
instruments take multiple exposures within one template and combine them
into one dataset, so I think a more accurate definition is that a template
results in one dataset. I'll find out."                                    <<<<

8.4
no comment.

JPE suggested that the img_acq templates also have the next filter as a
parameter. This is because changing a filter takes 1 minute and that this
could be done while the telescope is slewing. JMS said that he would check to
see what the VLT does.                                                     <<<<

MJI asked about the offsets coordinate system. It was agreed that
this should not be restrictive.  The TCS should be capable of dealing
with (or generating) delta RA and Dec and offsets in 
η and ξ (tangent plane coordinates).


Conclusions
-----------

It was decided that PSB needs to complete the Calibration Plan as soon as
possible. The next draft needs to be ready within one week (by 17th) and 
circulated.                                                                <<<<
There would then be a possible telecon on the Friday (19th) to round things
off.                                                                       <<<<

PSB asked if there was anything more missing from the Calibration Plan. No-one
thought thought that this was the case.

End 16:15


Actions (in addition to delivering the deliverables)
-------

JPE   Confirm that delivering PDFs electronically is OK with ESO

NNB   Find out what read mode (eg read/reset/read) ISAAC uses.

WJS   Investigate (with help from MJI and JRL) the non-linearity correction 
      and its interaction with the chosen method of readout. Write a report.

STH   Help PSB rewrite Section 4.1 of the Calibration Plan

SMB   Clarify what a template generates

SMB   Update Section 8 of the Observation Software Design Description by
      12th March.

JMS   Find out if the VLT changes filter while the telescope is slewing

PSB   Finish draft of Calibration Plan by 17th March

ALL   Telecon on 19th March


Dafydd Wyn Evans 15/3/04