Introduction:
-------------
Goal:
- NCS delegates the telescope calibration to reduction packages
- need a telescope calibration library, perhaps inside GILDAS
"Desirable" Features
- Re-use what is available
- avoid duplication of efforts
- Standardization
- Easy maintenance
Contraints
- many actors
- short timescales
- Portability
- backward compatibility (must be able to process old data)
- new ATM version
Alternative(s)
- per package solution ? leads to maintenance issues, and serious loss of time...
- make a common library
- requires to define needs and interfaces
- requires clients to adapt to the library constraints
Beware:
- RED lifetime was 20 years...
- Wide distribution of the library through GILDAS
- IRAM has a record of success, which leads to high expectations
- Failure is not an option (success is easily forgotten)
For a library, we need to define:
- Functionalities
- Interfaces with outside world (ATM / Reduction packages / NCS / other libraries)
- Internal data format
- Internal design (backend handling, fitting, ...)
***************************************************************************
Conclusions:
------------
- A somewhat cryptic table of the current situation is attached.
- List of desirable library functionalities:
* Focus
* Pointing
* Chopper
* Skydips
* Sideband rejection
* Array geometry
* Array gains
* Baseline
* Regridding
(* Fitting maybe for specialized algorithms)
- J.Penalver is now in charge of Pointing Models for the 30m. Pointing
models thus are outside the scope of the library at least in its first
iteration.
- Language to be used: F90 with derived types, modules for at least derived
types and definition of function interfaces.
- Goals:
* 1st version of library should be available by end of year
(November 2004 if possible).
* 1st version meaning same functionalities as providing now by red:
focus, pointing, chopper (for spectra), skydips (for spectra),
sideband rejection
* Priorities in library building:
1 Focus
2 Pointing
3 Chopper
4 Skydips
5 Sideband rejection
- Work distribution:
o Related work without timescale:
* A.Sievers to distribute an example of 30M-FITS file with corresponding
documentation for each back-end as soon as they become available
starting with bolometer and then 4MHz.
* H.Wiesemeyer to provide translation from 30M-FITS to CLASS data
format.
* A.Bacmann and S.Guilloteau to make a memo on bandpass calibration
at the 30m and PdBI to take into account bandwith increase and new
ATM possibilities.
o Library work:
* Small description of sideband rejection measurement at the 30m.
=> A.Sievers (ASAP)
* Iteration on interfaces and practical derived types:
=> J.Pety and S.Guilloteau 31.05.2004
* Test on minimization routines (eg can we realiably use a 2-D
minimization routine to fit a 1-D problem?):
=> S.Guilloteau 30.06.2004
* One fully working example (focus):
=> J.Pety and H.Wiesemeyer 30.06.2004
* A point will then be made enabling new work distribution round.
***************************************************************************
Pointing Methods
- least square fit
* Input
- lambda offsets
- beta offsets
- lambda derivatives
- beta derivatives
- intensities
- weights
- fitting shape (e.g. Gaussian,n-Gaussians)
Parameters, Guesses, Fixed/Fitted, Boundaries???
* Output
- parameters
- errors
- quality flags
- evalute fitted function
* Input
- fitting shape, parameters
- input coord (lambda,beta)
* Output
- intensities
- plotting: Fit overplotted over data for two crosses
* Input:
- System, Units
* Output:
- Screen plot
- results for NCS
* Input:
- Output of fit
- Logging info
* Output:
- XML file
- sic procedure (OBS)
Comments:
- dealing with spillover on multi-beam receivers ?
- start with "single-pixel" version
- can deal with multi-pixel in two different ways
- changing the lambda(i), beta(i) for each pixel
- or putting the relative positions of the beams into the fitting function
- is there a gain to be obtained by fitting a broadened beam for
planets ? need to evaluate that issue separately.
---------------------------------------------------------------------------
Focus Methods: Very similar to pointing...
---------------------------------------------------------------------------
Sideband Rejection:
- currently being done using the Martin-Pupplet settings, by coupling one
of the sidebands with the cold load, the others with the hot load.
- is the method accurate enough ?
- can we rely on the engineer tables ?
- measurement seems to be the most accurate method, but the precision is
unknown ?
- should it be frequency dependent ?
First step: transfert what is currently done in RED into the library ..
---------------------------------------------------------------------------
Chopper:
* Input:
- El
- sideband rejection(i)
- RF Frequency nu(i) and Bandwidth dnu(i)
- Signed IF Frequency
- Hot,Cold,Sky,Offset(dark count) temperature T(i)
- Hot and Cold Temperatures T(i) and coupling coefficients f(i)
- Telescope specific: Pamb, Water scale height, altitude, latitude
- Date
Pin = fhot B(Thot) + (1-fhot) B(Tamb) = B(Teff)
* Output:
- Trec(i)
- Water
- Water and Others Zenith Opacities(i)
Tcal:
* Input:
- El
- RF Frequency nu(i)
- Signed IF Frequency
- Sideband rejection
- Set of Tamb (cabine, ground) and coupling coefficients f(i)
- Water
- (Interpolated) Trec(i) for dual load
* Output
- Tcal(i)
- Water and Others Zenith Opacities(i) for DSB analysis
---------------------------------------------------------------------------
Skydip
* Input:
- El(j)
- sideband rejection(i)
- RF Frequency nu(i) and Bandwidth dnu(i)
- Signed IF Frequency
- Hot,Cold,Offset(dark count) temperature T(i) TSky(i,j)
- Hot and Cold Temperatures T(i) and coupling coefficients f(i)
- Telescope specific: Pamb, Water scale height, altitude, latitude
- Date
- Bolometer bandpass(i)
* Output:
- Value and errors of Water vapour(i)
- Value and errors of Feff(i)
- Quality flag
---------------------------------------------------------------------------
Baseline 1-D
- Fitting:
* Input
- Abscissa(i)
- Ordinate(i)
- Weights(i) (may be 0 to exclude a window)
- Fitting shape (eg polynomial,spline)
* Output
- Fitted parameters
- RMS
- Baseline computation:
* Input
- Abscissa(i)
- Fitting shape
- Fitted parameters
* Output
- Baseline(i)
- Baseline application
* Input
- Baseline(i)
- Ordinate(i)
* Output
- Baseline(i)-Ordinate(i)
---------------------------------------------------------------------------
Baseline 1-D: On-Off => To be discussed later.
---------------------------------------------------------------------------
Baseline 2-D: spectral x time
- Fitting:
* Input
- Frequency nu(i)
- Space x(j)
- Ordinate y(i,j)
- Weights(i,j) (may be 0 to exclude a window)
- Fitting shape:
* Factorized polynomials
* Box averaging of independant polynomials
* Cubic spline interpolation of independent polynomials
* Cubic spline interpolation of independent cubic spline
* sinusoidal(nu) with coefficients being polynomials(t)
* Output
- Fitted parameters
- RMS
- Baseline computation:
* Input
- Abscissa(i)
- Fitting shape
- Fitted parameters
* Output
- Baseline(i)
- Baseline application
* Input
- Baseline(i)
- Ordinate(i)
* Output
- Baseline(i)-Ordinate(i)
On-off: simultaneous fitting of background and baseline
---------------------------------------------------------------------------
Baseline 2-D: Spatial data => to be discussed later
---------------------------------------------------------------------------
Regridding 1-D
* Input
- Abscissa(i)
- Ordinate(i)
- Weights(i)
- New reference pixel, value, increment, dimension
- Algorithm
* Output
- New abscissa(i)
- New ordinate(i)
- New weights(i)
- Transformation matrix for signal and weight
Comment: Special case for already regularly sampled data
---------------------------------------------------------------------------
Regridding 2-D
- If it can be factorized (eg space x spectral), then use 1-D
regridding
- Else
* Input
- Abscissa(i)
- Ordinate(i)
- Weights(i)
- New reference pixel, value, increment, dimension
- Algorithm
* Output
- New abscissa(i)
- New ordinate(i)
- New weights(i)
- Transformation matrices for signal and weight
---------------------------------------------------------------------------
Array geometry: No consensus yet. In the first iteration of the library,
this functionality will be delivered by MOPSIC.