Data Reduction Software for the LBT SCIDAR System

Software Requirements and Current Status

A.     Robert Weiß and Stefan Hippler

March 2001

 

1       Software Requirements:

1.1     SCIDAR Data Reduction Box (DRB) Communications:

1.1.1    Communication between DRB, Telescope, and SCIDAR pre-processing box (PPB)

• Position and brightness of binaries used for measurements (either interactively or selected automatically from a position database).
• Desired resolution and SCIDAR mode.

 

1.1.2    Input from instrument and pre-processing box:

• Auto-correlated frames from SCIDAR measurements.
• Cross-correlated frames from SCIDAR measurements with adjustable time lag (e.g. cross-correlations of the first, the third and the sixth consecutive image after a given image).

1.1.3    Output to user:

• Cn2(h)-profile and v(h)-profile together with estimates of the Fried parameter r₀, the isoplanatic angle q₀, and the Greenwood time t₀ (online mode).
• Time series of profiles and the estimated parameters (offline mode).

1.2     Software Concepts and Necessary Development Work:

1.2.1    Profile extraction from auto-correlated pupil image series:

Algorithm description:

• Map auto-correlated pupil image from cartesian to polar coordinates (use brightest pixel as origin – this coincides with the central peak)
• Suppress given region around central peak (depends on selected SCIDAR mode and binary separation)
• Find radial distance with brightest secondary peak and extract profile from origin to this peak
• Estimate noise profile from directions not contributing to the peak’s direction
• Subtract noise profile from signal profile

1.2.2    Reconstruction of the Cn2(h)-profile:

Algorithm description:

• Map pixels to height distribution
• Map zenith angle to zenith position
• Calculate T-Matrix (i.e. the matrix that connects signal-and-distance distributions in the autocorrelation plane to Cn2-and-height distributions in the atmosphere
• Estimate first profile using a least-squares-inversion
• Iterate over first-profile with a deconvolution algorithm

 

Required information:

• Zenith angle of the observed binaries
• Pixel sampling of the pupil plane
• SCIDAR offset (for generalized SCIDAR mode)
• Binary separation and magnitude difference

1.2.3    Estimation of Turbulence Parameters:

In principle, the parameters of the atmosphere could be derived from the Cn2(h)-profile by direct integration. However, the relatively low resolution of SCIDAR measurements leads to large errors (on the order of 10% or above) in the values of these parameters. It is therefore desirable to use statistical inference methods for this purpose; further research into this matter is necessary.

1.2.4    Reconstruction of wind speed profiles:

Problematic. There exist two algorithms that are both equally difficult to automatize. Further research is required.

1.2.5    User Interface:

Online:

• “Extended Weather Station”
• Show current Cn2(h)-profile and v(h)-profile
• Show atmospheric parameters
• If data processing of current auto-correlated images is not possible then show “N/A” in all diagrams

Offline:

• Provide easy access to the database
• Provide data tables of profiles and atmospheric parameters in ASCII format for further processing

1.2.6    Database Environment:

Data tables (preliminary):

• Time against profiles, parameters and correlation images
• Binaries suitable for SCIDAR measurements
• Optional: time against weather data (temperature, humidity, air pressure etc.)

1.3     Development Environments:

• JAVA (with JFC) for user interfaces
• IDL for general data processing
• C for time critical data processing (interfaced to IDL)
• PostgreSQL as database engine (C interface can be wrapped to interface to IDL

2       Progress summary:

2.1     Data processing

Cn2(h)-profile calculation has been completely implemented in IDL (Fig. 1-3). However, some steps still must be done by hand. A typical example takes around 10 minutes from autocorrelation profile extraction to the final turbulence profile result. The inversion algorithms used are definitely the bottleneck, so we plan to implement these in C and aim at an execution time around 2 minutes.

 

Fig. 1: Scavenger main window

 

 

Fig. 2: Scavenger profile extraction window

 

 

Fig. 3: Turbulence profile calculation window (still iterating)

 

Still missing are the v(h)-profile extraction and the parameter estimation (parameter calculation, however, by direct integration are already implemented and show the expected scattering of results).

2.2     User interface:

User interface (GUI) design and implementation will be the last step of development since we place low priority on it. Data processing and Database handling will be completely separated from the GUI. We aim at producing Java code that could be called from anywhere on the Intranet with arbitrarily many instances (at least in online mode).

2.3     Database Design and Implementation:

• PostgreSQL C-to-IDL wrappers are in the implementation phase
• Data table normalization will take place as the next design step
  

Webversion, 10 July 2001