Telescope state integration

The MeerKAT telescope stores information about each observation using the katsdptelstate package. Information about the models used in an observation (and where to retrieve them) is stored in the telescope state so that they may be retrieved later.

The telescope state contains an sdp_model_base_url attribute which holds a base URL for the models. All other URLs in the telescope state are relative to this. For each model used there are two entries in the telescope state. One has the suffix fixed and references the exact model that was current at the time of the observation. This allows any analysis that was done at the time to be reproduced. The second has the suffix config and references an alias for the configuration. If an improved model is computed for this configuration and uploaded to the model store, using this alias will automatically reference this updated model. This allows analysis to use the best current model when it is not necessary to match any online use of the model done when the observation was made.

Because the MeerKAT Science Data Processor (SDP) allows multiple streams of each type to be grouped into a single telescope state, some keys for specifying models are scoped by the name of the antenna-channelised-voltage input stream to which they apply. This is represented as acv in the names below.

Example

We’ll use primary beam models for our example because it’s a complex case that demonstrates multiple aspects of the design. Consider a primary beam model for the UHF receiver on antenna m012. For the sake of this example, we’ll use the receiver serial number and a year as the configuration name. Then the initial setup as aliases may be as seen in the left side of the figure below. After an observation has been run, the telescope state (as serialized to the .rdb file) appears on the right, where the brackets indicate that this is an element of an indexed key (individual indicates that this is a model suitable for consumers that prefer per-antenna models over array-average models).

Now suppose further holography allowed the model for this configuration to be improved, and a new model was uploaded. This would lead to the following situation:

The telescope state has not been altered, but by using the model_primary_beam_individual_config key, one can access the improved model, without losing track of the model that was used online.

Next, suppose the receiver on m012 was swapped out for a different one. Then the situation might change to look like this:

The existing .rdb file continues to reference the old configuration, but if a new observation were started now, it would follow the alias chain down the left side to reference the new configuration and model.

Model keys by type

Each of these keys contains a relative URL, as described above. The type is either fixed or config.

RFI mask

model_rfi_mask_type

Band mask

acv_model_band_mask_type

Primary beam

acv_model_primary_beam_group_type, where group is one of individual or cohort. This is an indexed telescope state key, with the antenna name as the index.

API

In future katdal may be updated to hide these details and allow models to be fetched directly from a katdal data set. Until then, one can obtain the underlying telescope state object from a dataset as dataset.source.telstate. Pass it to the constructor of katsdpmodels.fetch.aiohttp.TelescopeStateFetcher (asynchronous) or katsdpmodels.fetch.requests.TelescopeStateFetcher (synchronous), along with an (optional) underlying fetcher. Then use get() to retrieve models. Instead of passing an URL (as for the underlying fetcher classes), pass the name of the telescope state key holding the relative URL.

In some cases one may wish to look up the key within a telescope state view. This can be done by passing the view as a telstate keyword argument. Here is an example of fetching a band mask model from a view called telstate_cbf which refers to the antenna-channelised-voltage stream:

with katsdpmodels.fetch.aiohttp.TelescopeStateFetcher(telstate) as fetcher:
    band_mask_model_key = telstate_cbf.join('model', 'band_mask', 'fixed')
    try:
        band_mask_model = await fetcher.get(band_mask_model_key,
                                            katsdpmodels.band_mask.BandMask,
                                            telstate=telstate_cbf)
        return band_mask_model
    except (aiohttp.ClientError, katsdpmodels.models.ModelError) as exc:
        logger.warning('Failed to load band_mask model: %s', exc)
        return None