# CRUSH: GISMO

Author: Attila Kovacs <attila[AT]sigmyne.com>

Last updated: 2 April 2014



#### Table of Contents

1. __Introduction__

2. __Locating Scans__
    - 2.1. By file names
    - 2.2. By scan numbers 

3. __Automatic Pointing Corrections__
    - 3.1. Setting an IRAM Pointing Model
    - 3.2. Differential Pointing Model
    - 3.3. Residual Pointing Corrections

4. __Extinction Correction__

5. __GISMO-specific pixel divisions__

6. __Glossary of GISMO-specific options__

7. __GISMO-specific log quantities__






## 1. Introduction


This document contains information specific to using CRUSH-2 to reduce GISMO 
data. It is assumed that you are already familiar with the contents of the main 
CRUSH-2 README (inside the distribution directory), especially its Section 1 
(Getting Started). 

If you run into difficulty understanding the concepts in this document, you
will probably find it useful to study the main README document a little longer.





## 2. Locating Scans


GISMO scans are typically stored in a filing hierarchy, organized by source 
name, then an IRAM scan ID (itself composed of a UT calendar date and a scan
number).

For example, files for scan 79 taken on 2010-04-12 (of Mars) are stored under 

    <datapath>/Mars/2010-04-12.79/ 

relative to the main GISMO data directory <datapath>. Inside the scan 
directory, you will find several files. The ones ending with

    [...]-GISMO-IRAM-condensed.fits

contain the merged data, which can be reduced by CRUSH. 

You can instruct CRUSH several ways how to locate your data. Your main options 
are either by their file names, or via a combination of `object` name, `date` 
and scan number. The latter approach is especially useful for specifying 
multiple scans taken on the same object, and over a handful of days, since it 
allows you to list scan numbers and ranges for convenience.

You can find a more detailed explanation below.



### 2.1 By file names

 For single scans, it may be the simplest to specify the file name itself
 (especially if your OS or shell supports autocompletion). Simply give the file
 name, either as a fully qualified path name, or relative to the `datapath` 
 setting. Thus,

    > crush gismo /data/Mars/2010-04-12.79/[...]-GISMO-IRAM-condensed.fits
       
 and

    > crush gismo -datapath=/data Mars/2010-04-12.79/[...]-GISMO-IRAM-condensed\
                  .fits
   
 are equivalent.



### 2.2 By scan numbers

 When reducing several scans together (which is the recommended way to deal
 with datasets), it may be more convenient to use a combination of `datapath`, 
 `object`, `date` and the scan numbers, and let crush find the desired data 
 files based on these. E.g.:
   
    > crush gismo -datapath=/data -object=NGC3627 -date=2010-11-08 32-39 43 \
                  -date=2011-02-13 15 19-22 33

 Of course, for each scan, or set of scans, you may add additional options such
 as pointing corrections (via `pointing`) or in-band opacities (via `tau`), or 
 calibration corrections (via `scale`) etc.




## 3. Automatic Pointing Corrections

A major improvement offered since crush 2.04 is the ability to apply automatic
pointing corrections, (a) via supplying a differential pointing model, and (b)
calculating residual pointing corrections based on the pointing log. The
combination of the two methods allows to reach pointing accuracies down to a 
few arcseconds. Below is a detailed explanation on how to use these methods.


## 3.1 Setting an IRAM Pointing Model

 You can use the `pointing.model` option to specify a file containing the
 effective IRAM pointing constants. These are the P1 through P9 as decribed 
 by Greve, A., et al. A&A, 115, 379, (1996), and the two Nasmyth offsets P10 
 and P11. Additionally P12-P16 contain terms with 2AZ dependence, and 
 refraction correction. The argument of the `pointing.model` option should be
 a file containing all relevant constants, one per line. E.g.:

    P1 = -3.15
    P2 = -8.41
    P3 = -0.18, -2.4, -4.2
    P4 = 0.1, 17.5, 0.4
    [...]
    P11 = 1.0

 As shown in the example above, each constant can have a single value to
 specify a static term, or a comma-separated list of 3 values for a time-
 variant dynamical model: the static term plus two dynamic terms, Pn0, PnC, and
 PnS, from which the pointing constant Pn is calculated for a given UT phase 
 (in radians) as:

    Pn = Pn0 + PnC * cos(UT) + PnS * sin(UT)

 Such dynamical constants can be a powerful way to cyclical variations of
 the pointing model due to a daily cycle (e.g. temperature cycle or heating
 cycle by the Sun). However, if you rather stick to just the static values
 you can set `pointing.model.static` to ignore the dynamical terms even if
 those were defined.

 The actual pointing correction calculated from the supplied pointing model
 are reported on the console output during the processing of each scan.

#### UPDATE: 31.01.2012

  The new GISMO FITS format (ver 1.7) now includes all IRAM coordinate columns.
  As a result, crush uses the scanning offsets rather than absolute AZ/EL 
  positions for mapping. This seems to solve the pointing problems experieced 
  earlier, and provides excellent (<3" rms) pointing in both directions with 
 static pointing constants only. 


### 3.2. Differential Pointing Model
   
 Instead of specifying the full pointing model, you can also use the 
 `pointing.model` option to set incremental corrections relative to the model
 used by IRAM during the observations, if you set the 
 `pointing.model.incremental` option as well. Then, for adjusting P4, P5, P10, 
 and P11, the file could contain the lines:

    P4 = 2.38
    P5 = -6.12
    P10 = -0.68
    P11 = 4.30

 Now, these values will be added to whatever pointing model was used during the
 observations to calculate pointing corrections. Otherwise, the behaviour is 
 the same as described above.

 In addition to the standard IRAM pointing model above, `TX` and `TY` define
 two linear temperature coefficients for the _X_ and _Y_ pointing respectively,
 such that TX = dX(arcsec) / dT(C), and TY = dY(arcsec) / dT(C).


### 3.3 Residual Pointing Corrections

 You can also apply residual pointing corrections based on observations of 
 nearby pointing sources. It is preferred that you observe pointing sources 
 frequently, and preferably at least once withing 30 minutes of all your 
 science scans, and withing 15 degrees distance on the sky. The more pointings 
 you have and the closer they are in distance and time to your source, the more
 reliable will be the derived corrections, which are weighted by their relative
 proximity on sky and in time.

 To apply residual corrections, use the `pointing.table` option whose argument 
 should specify the pointing log file. Such log files, containing pointing 
 information are automatically generated during GISMO runs. However, you should
 use an edited copy of this log, making sure to remove, or comment out, all 
 entries with unsuitale pointings. E.g. remove all bogus pointing on very faint
 sources, where the automatic routine may hit a noise peak or spike rather than
 a real source. Remove also any pointing data on extended source, objects with 
 multiple peaks or otherwise complex structure, etc.

 For the option to work, the pointing log must contain certain mandatory 
 columns, such as `id`, `UTh`, `AZd`, `Eld`, `pnt.X`, `pnt.Y`, `src.peak`,
 `src.dpeak`, and `src.FWHM`. Refer to the main README for details.

 The residual corrections calculated from nearby pointing data are printed on 
 the console during the processing of each scan, together with a quality
 indicator (qualities near or above 1 are considered good). 

 The residual corrections can be used together with a pointing model (which is 
 the preferred way to use this option), in which case increments are calculated
 on top of the model predictions, or without a pointing model for an 'absolute'
 pointing correction. In either case, you can still specify an additional 
 correction manually, if desired, via the `pointing` option, as usual. The 
 total combined correction (pointing model, incremental correction based on 
 pointing logs, and manual setting via `pointing`) are displayed during the 
 scan processing for verification.



## 4. Extinction Correction

GISMO, being a 2-mm camera, typically operates with low atmospheric extinction.
Nonetheless, properly accounting for the extinction is important for getting 
the calibration right. By default, CRUSH will reduce the data with a 
guesstimate of the in-band tau value, based on the last IRAM 225GHz radiometer 
measurement before the scan. The scaling of the 225GHz value to 2-mm passband 
is approximate, and is based on the CSO atmospheric transmission model of Juan 
Pardo. (Addition: during the 2011 April run, we confirmed that the opacity
correction based on the radiometer and the atmospheric model works well even
up to 10mm of precipitable water vapor, yielding ~7% rms blind calibration!!!)

You can also set an in-band zenith tau value based on your estimate directly, 
using the 'tau' option (see the GLOSSARY).




## 5. GISMO-specific pixel divisions

For a general overview of channel divisions, please consult Section 3.
(Correlated Signals) in the main README document.

    mux         A grouping of pixels based on their SQUID multiplexing scheme
                Each 4x8 pixel quadrant of the array is read out through the
                same SQUID amplifier. Therefore, it is not suprizing that 
                correlated signals are present on these quadrants. The
                decorrelating on 'mux' groups is default in GISMO reductions

    pins        The GISMO multiplexing scheme is implemented in the time-domain
                Thus, the first channels of each SQUID are read out at the same
                time, followed by the second channel in each group etc. Thus,
                if there is any pickup of high-frequency signals in the
                multiplexing scheme, one could expect some correlated signals
                to be present on these virtual readout pin groups. There
                is little evidence for these, but the reduction of very faint
                compact sources may benefit from the additional decorrelation
                on these groups.
                You can group neighboring 'pins' together for decorrelation
                via the 'correlated.pins.group' option, or the equivalent
                'pins.group' shorthand.

    rows        Geometric rows of the array (it has 16 rows and 8 columns).
                You can also group n rows together for the decorrelation via
                the 'correlated.rows.group=n' option, or the equivalent
                'rows.group=n' shorthand.

    cols        Geometric columns of the array (it has 16 rows and 8 columns).
                You can also group n columns together for the decorrelation via
                the 'correlated.cols.group=n' option, or the equivalent
                'cols.group=n' shorthand.





## 6. Glossary of GISMO-specific options

    cols                @Alias: 'correlated.cols'
                        @Expert
                        Specifies the grouping of channels by their geometric
                        location of the array divided into columns.
                        @See: 'correlated.<?>', 'rows'

    date=YYYY-MM-DD     Specify the observing date (UT) in the format
                        YYYY-MM-DD. This is used for constructing the IRAM 
                        scan IDs by combining with scan numbers.
                        You also need to specify the 'object' and 'datapath' 
                        before CRUSH will be able to find scans by number.
                        @See: 'datapath', 'object'
  
    mux                 @Alias: 'correlated.mux'
                        @Expert
                        Specifies the grouping of channels by the SQUID 
                        multiplexing.
                        @See: 'correlated.<?>'

    object=<name>       Give the IRAM catalog name (case sensitive) of the 
                        observed object. Use together with 'datapath' and 
                        'date' to allow locating scans by serial numbers.
                        @See: 'datapath', 'date'

    pcenter=<row>,<col>  @Expert
                        Specify the pointing center as a row,col combination 
                        on the array. There are 16 rows and 8 columns on the 
                        GISMO array, and the numbering starts from 1. Thus, 
                        the geometric center of the array corresponds to 
                        8.5,4.5. This option is not used (ignored) if pixel
                        positions are defined by the 'rcp' option (which is 
                        default). In those cases, the pointing center can be 
                        adjusted via the 'rcp.center' option.
                        @See: 'rcp', 'rcp.center', 'pixelsize', 'rotation'


    pixelsize=<a>[,<b>]  @Expert
                        Specify the size of the pixels for calculating pixel 
                        positions based on a regular grid. A better way of 
                        setting the pixel positions is through the 'rcp' 
                        option. 
                        The argument can be either a lateral size (in arcsec) 
                        for square pixels, or two comma separated sizes for 
                        rectangular pixels.
                        @See: 'rcp', 'rotation', 'pcenter'

    pointing.table=<file>       @Advanced
                                @Since: 2.04
                        Specify the reduction log file (produced via the 'log'
                        option) to use for applying incremental pointing
                        corrections base on the nearby pointing observations. 
                        You can use this option together with 'pointing.model' 
                        or separately. For more information see Section 3.2 
                        (above).
                        @See: 'pointing.model', 'pointing'

    pointing.model=<file>       @Expert
                                @Since: 2.04
                        Specify a file containing differential
                        pointing constants (i.e. relative to those used by IRAM
                        during the observations). See Section 3.1 (above) for
                        details on the file format.
                        @See: 'pointing.table', 'pointing'

    rotation=<deg>      @Expert
                        Specify the array rotation (in degrees), when pixel 
                        positions are calculated from a regular grid. A better 
                        way is to specify pixel positions via the 'rcp' option 
                        their rotation via 'rcp.rotate'.
                        @See: 'rcp', 'rcp.rotate', 'pixelsize'

    pins                @Alias: 'correlated.pins'
                        @Expert
                        Specifies the grouping of channels by their virtual 
                        pin numbers in the multiplexing time domain.
                        @See: 'correlated.<?>'

    pins.group=<n>      @Alias: 'correlated.pins.group'
                        @Expert
                        Because there are only four SQUID multiplexers in 
                        GISMO, there is little redundancy among the pixels, 
                        which are read out at exactly the same instant. This 
                        option allows to group together <n> consecutive 
                        samples, to combine pixels read in some time-interval.
                        @See: 'pins'

    rows                @Alias: 'correlated.rows'
                        @Expert
                        Specifies the grouping of channels by their geometric
                        location of the array divided into rows.
                        @See: 'correlated.<?>', 'cols'

    skipFWFix           @Expert
                        @Since: 2.15-2
                        During 2013 Oct/Nov we had lost some data due to 
                        firmware problems. Steve has attempted to reconstruct
                        as much of the affected frames as possible. At the same
                        the the reconstructed astrometry is flagged 
                        accordingly. Therefore, it is possile for CRUSH to
                        ignore the reconstructed data if the user wishes to do
                        so.

    SZ                  @Since: 2.15-2
                        Optimize reduction for the detection of Sunyaev-
                        Zel'dovich signals (decrements). The SZ option is 
                        really just a trigger condition for activating other
                        settings. Check the configuration files under
                        the gismo/ crush folder (esp. 'default.cfg) to see
                        what this does exactly.

    tau.timezone=AAA    @Advanced
                        @Since: 2.15-2
                        Specify the timezone in which skydip lookup tables
                        were compiled. For example the table compiled by G.
                        Bruni for 2012 Nov were in 'CET' whereas the table
                        by Israel Hermelo for 2013 Oct/Nov is in 'UTC'.

    ver=X               @Since: 2.11
                        @Expert
                        Force setting the merge version to X, ignoring the
                        value stored in the FITS.

    ver.[<range>]       @Since: 2.11
                        @Expert
                        Set options for specific FITS merge versions. E.g.
			
                          ver.[<1.7] pointing.model=old.model

                        @See: 'ver'


## 7. GISMO-specific log quantities


    bias                The detector bias value (for column 0).

    foc.dX              Focus X offset (mm).
       
    foc.dY              Focus Y offset (mm).
        
    foc.dZ              Focus Z offset (mm).

    modelX              Static pointing model correction in azimuth (arcsec).

    modelY              Static pointing model correction in elevation (arcsec). 

    nasX                Nasmyth X position of instrument.

    nasY                Nasmyth Y position of instrument.

    obstype             Type of observation.

    rot                 Rotation of the instrument (deg).

    stage2.biases       Second stage biases (4 integers)

    stage2.feedbacks    Second stage feedbacks (4 integers)
	
    stage3.biases       Third stage biases (4 integers)

    stage4.feedbacks    Third stage feedbacks (4 integers)

    tiltX               Azymuth pointing correction from inclinometer (arcsec).

    tiltY               Elevation pointing correction from inclinometer 
                        (arcsec).


------------------------------------------------------------------------------
Copyright (C)2016 -- Attila Kovacs <attila[AT]sigmyne.com>