All ranges are calculated automatically if the inspiral_range package
is provided, and the resultant metrics are printed to stdout.  Just
the "range" is added as plot subtitle.

Kevin Kuns authored 4 years ago and Jameson Rollins committed 4 years ago

Adds suspension thermal noise sub-budgets. This addresses 1/3 of #51. Suspension thermal noise can now be broken up into horizontal and vertical noise of each of the suspension stages. With one exception, which does not change the total noise, the calculation is the same as the current calculation but the contributions of each stage and direction are tracked separately.

In the current calculation the horizontal and vertical spring constants of each stage are broken up into an upper and lower joint and are stored separately. The total suspension thermal noise is then calculated by adding the contributions of each of these joints using the appropriate temperature of each joint. Instead of adding the contributions of each joint all at once, this MR makes it possible to track each stage with separate `Noise` classes and add them all to a sub-budget.

The noise for each stage is the sum of the noise from that stage's lower joint and the noise from the upper joint of the stage below. In the case of the top mass, the contribution from the upper joint is also included. In the current calculation, for the upper stages, the imaginary part of the horizontal spring constant is equally distributed between the upper and lower joints while it is only attributed to the upper joint for the test mass. The one change this MR makes to the calculation itself is to distribute this equally between the top and bottom joints as is done for the upper stages.

The imaginary part of the vertical spring constant is only attributed to the upper joint of each stage. The blade springs are the only contribution to the vertical spring constant for the upper stages while the lower stage also includes the continuum fibers or ribbons. It's unclear how to break this up into an upper and lower stage as is done in figure 14 of https://arxiv.org/abs/2001.11173, so this MR does not have vertical noise of the test mass. (The noise is included, it's just attributed entirely to the PUM.)

* All quantum calculations use the final stage suspension mass instead of calculating it from the test mass radius and thickness
* shotradSignalRecycled uses the suspension susceptibility instead of the free mass approximation

- Break seismic noise up into contributions from horizontal and vertical motion.
- Moved code to calculate the platform motion from noises.py into seismic.py.
- Removed duplicate dhdl function from noises.py.

just calls plot.plot_trace

closes #75

closes #77

that don't support the fallback_version option

This is needed for the test script, where setuptools_scm doesn't work

using setuptools_scm

Kevin Kuns authored 4 years ago and Jameson Rollins committed 4 years ago

Quantum noise now broken up into:
* AS Port Vacuum
* Arm Loss
* SRC Loss
* Filter Cavity Loss
* Injection Loss
* Readout Loss

Brownian amplitude noise is generally quite negligible, so the
CoatingBrownian budget item no longer considers it.  The calculation
remains available in coating_brownian().

Kevin Kuns authored 4 years ago and Jameson Rollins committed 4 years ago

- Updated the filter cavity optimization to work with the recent precomp changes
- Removed a redundant argument in computeFCParams

This specifies a default frequency array for load_budget and the CLI, and makes it easier to provide alternative arrays with a frequency "specification string" (e.g. '5:1000:3000').

- merge coating_thickness into mirror_struct
- merge precomp_mirror into mirror_struct
- merge precomp_power into ifo_power
- merge precomp_cavity into arm_cavity

precomp is no longer used for these helpers, but they're sufficiently
trivial that it doesn't matter whether they are cached.

closes #56

fixes bug where overall budget name was used instead.

This fixes a bug whereby references traces were being included in the sum
total.

This prevents buggy behavior if an ifo is not provided.

This function is useful, so no reason to hide it.

closes #72

those variables will be considered internal.

This prevents the precomp data from being re-calculated if the ifo
hasn't changed.

This struct is created by the precomp_mirror function, and is essentially
the union of the Material and Optic structs, holding all relevant
properties for a given optic for thermal noise calculations.  It can be
used in place of either the Material or Optic structs.

Coating thermal noise functions are modified to accept this struct.

This is some cleanup and simplification of Lee's mapped precomp
series.

cache the load status

Add a Struct.hash() method to return a unique hash over attributes.

Modify Noise.run() method to check for changes to the ifo and only run
the update method if the ifo has changed or other kwargs are supplied.

nb.precomp decorator can be used to wrap BudgetItem.calc methods with
functions that only need to be called once per budget run.

    nb.precomp(precomp_foo)
    def calc(self):
        ...

Precomp functions are all executed by the update() method (after attribute
update), and execution state is cached at the Budget level, to prevent
re-calculating the same functions multiple times per update.

The BudgetItem.update() method is no longer expected to be overloaded by
the user.

This provides a speed-up of about 50% for the Aplus model, but much less
for the CE models which are limited by other noise calculations.

This new object holds the PSD and frequency arrays, as well as the
list of all budget sub-noises, accessible via a dictionary interface
and attribute access.

All interfaces and tests are updated to process the new BudgetTrace
object.

A new HDF5 schemata (version 2) is also introduced that stores trace
data in a way that mirrors the BudgetTrace.  The io functions
(save_hdf5/load_hdf5) are updated to write/read this format.  The
load_hdf5 functions continues to support the version 1 SCHEMA.

Closes #59