Make noise configuration more flexible
We currently generate all noise time series inside LISA Instrument, which does not allow for custom noise functions and correlations. The parameterization is fixed and tied to the predefined noise shape (e.g., an amplitude and a knee frequency), and the user passes these arguments on a MOSA or SC-basis on Instrument's initialization.
We propose a more general approach, where the user can pass either
- callables, stored until the simulation is run -- then executed to generate the noise time series,
- or time series in the form of Numpy arrays, already generated by the user.
Of course, callables and arrays are casted to MOSA or SC containers, so that the user can differentiate the callables (noise generation function parameters) or the time series on a MOSA or SC basis.
Callables
The default value for noise arguments in Instrument
's init function would be noise generation functions found in lisainstrument.noises
.