AMPHI

AMPHI corrects the visibilities in the input merge-file by comparison with the predictions of an input model. Both MERGE-formats 1 and 2 are accepted. It uses the method of Cornwell & Wilkinson (MNRAS 1981, 196, 1067) to minimize the differences between observed and model visibilities. The observed closure quantities are preserved by allowing only corrections which vary the complex gain at each station. A map may be made from the output merge-file by running INVERT. Lists of error estimates (GAINERR and PHSERR, see below) control how much the observed visibilities may be changed.

AMPHI cannot process a file containing data from more than 20 stations, or more than 2000 sections of data, and it cannot use a model comprising more than 10000 components.

WARNING: AMPHI uses the phases in the input data file, rather than the closure phases. If the data have been averaged, the phases will be inconsistent with the closure phases; to correct the phases, run program PHASE (any input model will do).

The amplitudes and phases output by AMPHI can be examined using the program LIST (specify "AMP" and "PHASE" in the control parameters). AMPHI output does not require closure phases in the input file, but they are retained if present. The running time of AMPHI is proportional to the number of baselines, the number of records in input MERGE file and the number of components in the model; the time can be reduced with the NUMBER option, or by running AVERAGE (and PHASE) on the data before AMPHI.

Example

% amphi
INPUT = "81FEB345.CAL"  
OUTPUT = "HYB345.AMP"  
MODFILE = "3C345.MOD" 
LISTFILE = "AMPHI.LIS"
LONG = 1
NCPRINT = 5   
NUMBER = 200    
CGNUM = 5  XCG = 0.0  YCG = -4.0
CLAMP  
TIMESCALE = 360,30,240,0,120
GAINERR = 0.02,0.03,0.02,0.05,0.02   
/

Parameters

INPUT name of input merge-file (default "FROM")

OUTPUT name of output merge file for INVERT (default "TO")

MODFILE name of input model-file (default "MODEL")

LISTFILE name of output listing file (default "amphi.lis")

These file names are up to 40 characters.

MAXS

MAXSECS

These two parameters are obsolete and are ignored by AMPHI (Version 8.3 or later).

CGNUM, XCG and YCG (default no shifting)

If CGNUM is specified, the centroid of the first CGNUM components of the model is found, and the model is shifted so that the centroid lies at (XCG,YCG) milliarcsec. Note that x increases to the East (left) and y increases to the North (up), so increasing XCG moves the source to the East and increasing YCG moves it to the North relative to the map-center.

NCPRINT = m

Maximum number of model components to list (default is to list the first 10).

NUMBER = n

Select only the first n model components for calculating model visibilities (default is to use the entire input model, up to a maximum of 10000 components).

LONG (no value)

List changes made to the station gains for each UT section in the data. If omitted, the program does not list any corrections. The listed station-gain factors multiply the amplitudes on every baseline to the given station. Where no data involving a particular station are found, the listing shows an asterisk. LONG=1 lists changes to station phases (in degrees) as well; for each UT section, the listing indicates which station phase the program has (arbitrarily) set to zero.

CLAMP and SCALE (no values)

One of these options should be selected if changes in the observed amplitudes are required. CLAMP changes the station sensitivities for each UT section independently (see TIMESCALE option below). SCALE allows only a constant correction to the sensitivity of each station, derived using all the input data. The magnitudes of the corrections are controlled by the GAINERR list (see below). CLAMP is ignored if SCALE is selected.

GAINERR = e1,e2,e3,...

A list of estimated fractional errors in the sensitivity of each station in the data. Station gains can only be changed if either the CLAMP or SCALE option is set. If the GAINERR list is omitted the default (large) values assume that the amplitudes are uncalibrated, and thus make full use of the closure amplitudes.

TIMESCALE = t1,t2,t3,... (default 0,0,0...)

CUTOFF = x (default 0.1)

A list of timescales (in minutes) over which station gains may vary. This list implies the CLAMP option, but is ignored if SCALE is selected. The default TIMESCALE list is zero, giving corrections which are independent for each UT section. A truncated Gaussian is used as the smoothing function, with FWMH = TIMESCALE and truncated at a level CUTOFF of the peak. TIMESCALE applies to amplitudes only, not phases.

PHSERR = p1,p2,p3,...

A list of estimated errors (in degrees) in the visibility phase at each station. The default list (large values) corresponds to the usual VLBI case of uncalibrated phases, and thus PHSERR is normally omitted.

FLOATGAIN (no value)

This option suppresses the application of an overall scaling factor to corrections derived by the SCALE or CLAMP options. Normally, FLOATGAIN is omitted, and AMPHI computes an adjustment to the overall flux scale of the data, such that the mean correction factor (weighted by GAINERR**-2) is unity. If CLAMP is selected, the corrections are time-averaged before calculating the scale factor. Thus maps made with and without FLOATGAIN will differ only by a constant factor. Omission of FLOATGAIN is recommended to prevent drifts in the flux scale, although this normally results in a larger agreement factor.

NOTE: in program versions earlier than V 7.2, the operation of FLOATGAIN with the CLAMP option was different.

NOLOOKUP (no value)

Use standard subroutines for sin and cos in the model visibility calculation, instead of a lookup table. By default, a 4096-point table is used, saving about 50% in cpu time (for large model files, this step dominates the total running time). The rms error in sin/cos from the table is 0.00044, and it has a negligible effect on the accuracy of the gain solution. If the model contains Gaussian or other component types (disk, sphere, ring or line), the visibility is correctly calculated, but the lookup table is not used.

UCUT, UVRATIO (defaults 0.0, 1.0)

These parameters select data inside an ellipse in the u,v plane with semi-axes U = UCUT, V = UCUT*UVRATIO (in units of millions of wavelengths) to be excluded from the gain and phase solutions. No actual editing of the input or output MERGE file is performed. UVRATIO may be omitted, in which case the selected area is circular. Specifying a negative UCUT excludes data from outside, rather than inside, the given area. UCUT and UVRATIO may be used in the case where the input model is a poor fit on the short baselines (as might happen with a limited number of CLEAN components); excluding these baselines should avoid biassing the solutions derived from longer baselines.

FLAGFAIL

Occasionally, AMPHI finds the data ill-conditioned for a reliable gain solution, in which case its default action (FLAGFAIL omitted) is to issue a warning and apply no corrections. Specifying FLAGFAIL makes AMPHI edit out the bad section, so that INVERT and other programs will ignore it; this editing can only be reversed using the RESTORE option in VLBEDIT.

History