In a recently concluded development study, a small team at ASTRON, The Netherlands Institute for Radio Astronomy, investigated the concept of Streaming Visibility Processing (SVP) for ALMA. The key idea behind the SVP concept is to stream the visibilities produced by the ALMA correlator into the (initial) data processing stages without storing them first. SVP performs initial visibility processing in real-time up to the point where the data is reduced to more tractable rates without detrimental effects on the scientific quality of the end products. Further details are in the final report of the streaming visibility processing study and by following the 'read more' link.
Streaming Visibility Processing for ALMA
Published: 16 Mar 2026
Images of AB_Auriga: (Left) an image of the raw data made with CASA; (Right) an image showing the result from our low-latency imager, including a direction-independent phase calibration step.
Although the team concluded that, e.g., ALMA’s WVR correction, which is based on interpolation, is not an ideal candidate for streaming processing, they demonstrated low-latency imaging on a few ALMA calibrators including a direction-dependent phase calibration stage. An example with AB_Auriga is shown in the image above with the left panel showing the CASA image of the raw data and the right image showing the result from our low-latency imager including a direction-independent phase calibration step, i.e., site conditions are calibrated out using the signal from AB_Auriga itself instead of radiometer data. Such a low latency imager can provide ALMA operators with feedback almost immediately after an observation.
Another potential use case of SVP is to compress the visibilities before storing them. This can, e.g., be done using a method originally developed for LOFAR called Dysco compression. An interesting use case for this is on-the-fly compression when a user requests a data set in Measurement Set format from the CalMS service provided by the European ALMA Regional Center. Dysco compression would allow compression of the calibrated Measurement Set by about a factor four (with a demonstrated absence of data quality loss) thereby reducing the need to recalibrate the data to save storage capacity while also reducing the amount of data that needs to be downloaded from the service.
The study was undertaken by S.J. Wijnholds, T.J. Dijkema and S. Yatawatta.
This news item was contributed by Carlos de Breuck