About katcbfsim

katcbfsim is a correlator and beamformer simulator, which is intended to match the output formats used by the MeerKAT CBF.

Features

Correlator

The correlator simulation currently models the following effects:

  • Unpolarised point sources

  • Statistical noise

  • A simple antenna-independent Airy beam model (all antennas are assumed to point in the same direction, and thus share a perceived sky).

The following are currently implemented internally, but not yet exposed:

  • A per-frequency per-antenna full-Jones gain

  • System noise, via a single system equivalent flux density

  • Polarised sources

The following are not (yet) modelled:

  • Direction-dependent effects other than a primary beam

  • Atmospheric effects

  • Relativistic aberration

  • Relativistic Doppler effect

  • Time and frequency smearing (values are point-sampled)

  • Extended sources

  • Gain drift

  • Pointing errors

  • Frequency- or antenna-dependent system noise

  • RFI

  • Quantisation effects

In addition, a simplified statistical model of noise is used, which does not account for correlations between visibilities.

Beamformer

The beamformer simulation is much more basic. It simulates only the metadata, and the data is meaningless. As such, the antennas, sources, target etc do not need to be specified.

Installation

Installation uses the normal setup.py, and so can be installed with pip or other Python packaging tools.

A CUDA-capable GPU and corresponding drivers must be present. This is currently the case for the beamformer simulation as well, even though it is not used in this case.

Using katcbfsim

There are several ways to run a simulation, which can also be mixed.

  1. Command-line only. In this mode, the target (phase centre) is fixed for the simulation, and it is also the pointing direction.

  2. Using katcp. The sources, antennas, target etc are specified through katcp commands. In this mode, it is possible to simulate multiple streams, and to sent visibilities to HDF5 files instead of the network.

  3. Using katsdptelstate. Static configuration is stored in the config dictionary, as if for command-line arguments. Time-varying configuration is supplied as timestamped sensors.

  4. Mixed katcp and katsdptelstate. In this case, static configuration can be loaded into katsdptelstate later on (as attributes, rather than in the config dictionary), and then latched by a katcp command.

  5. Using katcbfsim as a library.

The configuration is split into information describing the virtual world (antennas, sources and so on) and information about the virtual correlator or beamformer (called a “stream”). Because the katcp interface supports multiple streams, the per-stream commands all take a stream name.

The world information needed is:

  • The antennas. Each is given by a string that can be parsed by katpoint.Antenna.

  • The point sources. Each is given by a string that can be parsed by katpoint.Target, and should include a flux model. The source is ignored for frequencies outside the support of the flux model. If no flux model is given, 1 Jy is assumed at all frequencies. If no sources are given, a point source is simulated at the initial phase centre.

  • The sync time, as a UNIX timestamp. This is the start time of the first dump, and also the time reported as the sync_time in the SPEAD metadata.

  • The target, i.e., phase centre. This is also a string that can be parsed by katpoint.Target. This can be changed over time.

  • The pointing direction, used for the primary beam model. Like the target, this is a katpoint.Target string and can be changed over time. If it is never specified, it defaults to being the same as the target.

  • A gain (scaling factor between flux densities and counts). Generated values in Jansky are converted to output values by scaling by this gain. It is expressed as a scale factor per Hz of channel bandwidth per second of integration time.

The stream information is:

  • A name, which is used in katcp requests and sensor names.

  • An ADC clock rate, bandwidth, number of channels, and centre frequency.

  • A destination, which is a hostname and port for the SPEAD stream, or the name of an HDF5 file.

  • For correlation:

    • An accumulation length for integrations, in seconds. The actual value is rounded in the same way that the MeerKAT correlator would.

  • For beamforming:

    • The number of time samples included in each heap.

    • The number of bits per sample.

Command-line

Run cbfsim.py --help to see the command-line options. Only a few key options are documented here.

--create-fx-stream <NAME>

This creates a correlator stream with the given name. If this option is not specified, then the katcp request stream-create-correlator must be used to create streams.

--create-beamformer-stream <NAME>

This is equivalent to --create-fx-stream but for beamformer streams.

--start

Start the capture for the stream. If this option is not specified, the katcp request capture-init must be used to start the capture.

--cbf-antenna <DESCRIPTION>

Specify a single antenna. Repeat multiple times to specify multiple antennas.

--cbf-antenna-file <FILENAME>

Load antenna descriptions from a file that contains one per line.

--cbf-antenna-mask <LIST>

Comma-separated list of antenna names. They will be given a fake position, which can later be replaced using the configure-subarray-from-telstate katcp request (see below).

--cbf-sim-source <DESCRIPTION>, --cbf-sim-source-file <FILENAME>

These are similar, but for sources rather than antennas.

--cbf-sim-gain <FACTOR>

System-wide gain, as described above

Telescope state

Command-line options can be loaded through katsdptelstate in the standard way. Antennas and sources are slightly different, however. The antennas must be placed in a cbf_antennas key (in the config dictionary), which is a list of dictionaries. Each dictionary has a description key, which is the antenna string. This is to allow for future expansion. The sources are similarly placed in a cbf_sim_sources key.

The target is read from the telescope state sensor cbf_target, using the latest value strictly prior to the start of the dump. Thus, all values for a simulation can be pre-loaded.

The pointing direction is specified by the telescope state sensors ant_pos_actual_scan_azim and ant_pos_actual_scan_elev, where ant is replaced by the name of the first antenna. These provide the azimuth and elevation, in degrees, for the first antenna. In future, other antenna directions might be used, but for now they are ignored.

katcp protocol

Use the ?help command to obtain a full list of commands. The general flow is

  1. Define a stream with ?stream-create-correlator.

  2. Set world and correlator static properties.

  3. Start the data flow with ?capture-start.

  4. Set dynamic properties as the simulation proceeds.

  5. Stop the data flow with ?capture-stop.

Note that static properties cannot be changed while a capture is in progress, but can be modified between captures.

Mixed katcp and telstate

If the subarray static properties are not known at the time the simulator process is started, they can still be loaded from telstate later, using the ?configure-subarray-from-telstate request. This takes an optional parameter, which is a comma-separated list of antenna names, and requires that --telstate was given on the command line. If no antenna names are listed, the names of the antennas configured at startup are used, replacing their positions (this is a good match for --cbf-antenna-names).

For each antenna named name, the attribute name_observer is used to obtain the antenna. It can be specified as either a description string or an antenna object.

This flow can also be used in the opposite direction: antennas configured on the command line with --cbf-antenna or --cbf-antenna-file are stored into telstate as name_observer.