next up previous
Next: Scientific Results in Press Up: IRAM Newsletter 48 (May 2001) Previous: News from GILDAS


An ALMA Simulator

The new release of GILDAS, dated MAY2001, includes a full simulator of the ALMA instrument. Please read the note concerning this release, and especially the use of Fortran-90, in this Newsletter.


The ALMA Simulator was developed to carry out studies of the usefulness of the ALMA Compact Array (ACA), an array of smaller antennas that could be used to recover part of the missing short-spacings and thus complement the main ALMA instrument. To properly assess the impact of ACA on the imaging capabilities of the combined array, it is necessary to perform realistic simulations, including noise and pointing errors.

The simulator includes the following steps:

A deconvolution algorithm has been developed to process the heterogeneous array ALMA+ACA.

Using the simulator

The simulator is using the MAPPING software. It is available through a window interface (Fig. 2), that can be obtained by clicking on the ``ALMA Simulator'' button in the main menu. In this window, the user can enter several input parameters, to define the input image, the configurations to be used, the pointing errors models, the deconvolution parameters, etc. Most of these input parameters have reasonably robust default values. The simulator then runs as a pipeline (COMPUTE button): all simulations are performed up to the clean map stage. The DISPLAY procedure can then be used to plot the results, including a comparison between the simulation and the input model (re-sampled on the same grid, and smoothed to the same angular resolution). Different fidelity estimators are also computed and displayed, in the image and/or in the uv plane. An example is given in Fig. 3.

An expert mode is also available, in which each step (Single-Dish simulation, ALMA simulation, ACA simulation, data merging, deconvolution) can be performed separately. The intermediate results can thus be checked.

In all modes, please check the HELP associated to each button to get a list of the various possibilities and parameters of the simulator.


The simulator typically needs 20 minutes to perform a "normal" simulation run on a Linux PC (AMD Athlon 800 MHz processor with 756 MB of RAM). Take care that a typical simulation requires 100 to 300 MBytes of disk space.

Current limitations and Warnings

The simulator has currently a few limitations: only continuum observations can be simulated, multi-configuration is not implemented, the phase noise is currently baseline-independent.

Although the joint deconvolution of ALMA, ACA and Single-Dish is offered as a possibility in the current version of the ALMA Simulator, we caution the users that this is a highly experimental part. It may require some fine tuning of the parameters to reach convergence, and should not be considered as a final product. We are currently working to make the algorithm more robust.

On the other hand, ALMA only or ALMA + Single-Dish deconvolution is a well tested procedure, and should be quite representative of what ALMA may offer.

Future versions

We expect that the package will evolve quickly in the near future to match the needs of ALMA studies and include a number of new features. On the top of the priority list is the improvement of the ALMA + ACA + Single-Dish tool, as well as a better treatment of the phase and amplitude errors. Developments in the latter area have been performed at DEMIRM by F.Viallefond, J.F.Lestrade and J.R.Pardo and we intend to incorporate these in the ALMA Simulator.

The more recent version of the Simulator can be obtained upon request. Please contact us at for any comments, questions, or suggestions.

Jérome PETY, Frédéric GUETH and Stéphane GUILLOTEAU

Figure 2: Window interface of the simulator. The COMPUTE button allows to run the simulation pipeline. The DISPLAY button is used to display the results (comparison in the image plane, in the uv plane, uv-coverage, beam, etc.). The LOAD button allows to reload the results of a previously computed simulation. The EXPERT button changes the interface to allow tests (mostly for use at IRAM). The input parameters are ordered in different sections that pop up when the user pushes the corresponding PARAMETERS button.
\mbox{\psfig{} }
\end{center} \end{figure*}

Figure: Simulation of an observation made with the ALMA array alone. The input image is an H$\alpha$ map of the M51 galaxy. The simulated observation is a snapshot (0.3 hour) of a 7-fields mosaic (hexagonal pattern). The two top images show the model (right) and the simulated observation (left). The model is smoothed at the resolution of the observation. The two bottom images show complementary information: the difference (model-simulation, left) and the fidelity (model/(model-simulation), right). To quantify the visual impression, several numbers are provided: integrated fluxes (and the corresponding percentage as compared to the model); rms of the difference image; fidelity range. Finally, the histogram of cumulative fidelity is shown, with median fidelities computed for the image pixels whose intensity is larger than 0.3/1/3.3/10% of the peak value.
\mbox{\psfig{,width=14cm,angle=270} }
\end{center} \end{figure*}

next up previous
Next: Scientific Results in Press Up: IRAM Newsletter 48 (May 2001) Previous: News from GILDAS