NOTE: This is the new implementation of split.  The old implementation is available for a short time as oldsplit.

This new split task uses the MSTransform framework underneath. split is the general purpose program to make a new data set that is a subset or averaged form of an existing data set.

split is often used after the initial calibration of the data to make a smaller MeasurementSet with only the data that will be used in further flagging, imaging and/or self-calibration. split can average over frequency (channels) and time (integrations).

split also supports the Multi-MS (MMS) format as input. For more information about MMS, see the help of partition and mstransform.


Parameter descriptions


Name of input MeasurementSet or Multi-MS.


Name of output MeasurementSet or Multi-MS.

NOTE: If a .flagversions file with the name of the output MS exist, this task will exit with an error. The user needs to rename or remove the existing flagbackup or choose a different output name for the MS.


If the input is a Multi-MS the output will also be a Multi-MS. The default value is set to True. The output Multi-MS will have the same partition axis of the input MMS.

General selection:  field, spw, antenna, timerange, scan

The split task uses the same selection syntax as the solving tasks for these parameters. See here for more information.


Give the name of the input data column(s) to process. The default is set to "corrected_data". Allowed values are:

  • corrected
  • data
  • model
  • data,model,corrected
  • float_data
  • lag_data
  • float_data,data
  • lag_data,data
  • all

NOTE: 'all' = whichever of the above that are present. If the requested column does not exist, the task will exit with an error.


Keep completely flagged rows in the output or drop them. This has no effect on partially flagged rows. All of the channels and correlations of a row must be flagged for it to be droppable, and a row must be well defined to be keepable. The default is True, which will keep completely flagged rows in the output MS.

IMPORTANT: Regardless of this parameter, flagged data is never included in channel averaging. On the other hand, partially flagged rows will always be included in time averaging. The average value of the flagged data for averages containing ONLY flagged data in the relevant output channel will be written to the output with the corresponding flag set to True, while only unflagged data is used on averages where there is some unflagged data with the flag set to False.

Channel averaging parameter: width

This parameter gives the number of input channels to average to create one output channel. If a list is given, each bin will apply to one spw in the selection. The default 1 means that no averaging will be applied.

Time averaging parameters: timebin, combine

The parameter timebin gives the width for time averaging. When timebin is greater than 0s, the task will average data in time. Flagged data will be included  in the average calculation, unless the parameter keepflags is set to False. In this case only partially flagged rows will be used in the average. If present, WEIGHT_SPECTRUM/SIGMA_SPECTRUM are used together with the channelized flags (FLAG), to compute a weighted average  (using WEIGHT_SPECTRUM for CORRECTED_DATA and SIGMA_SPECTRUM for DATA). Otherwise WEIGHT/SIGMA are used instead to average together data from different integrations.  

The combine parameter will let the timebin span across scan, state or both. State is equivalent to sub-scans. One scan may have several state ids.

NOTE: For ALMA MSs, the sub-scans are limited to about 30s duration each. In these cases, the task will automatically add state to the combine parameter.

To see the number of states in an MS, use the msmd tool. The default is set to combine='', which will not cross the scan or state boundaries when averaging in time. Options are: 'scan', 'state', 'state,scan'.