Skip to content

harrison-caudill/pylink

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

67 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Note to Anyone Reading This

I just closed down business operations at BStar and switched over to consulting. I have some time on my hands right now, so feel free to file an issue if you have a feature request.

PyPi Naming

Please note that there is already a pylink package on PyPi, so it is currently registered as pylink-satcom. I'll repeat this warning in the Installation section below.

Python Link Budget Calculation/Management and General Modelling

This software package is meant to replace the manual-intensive spreadsheet method. This package is intended to permit the following major changes in common methodology:

  • Use of configuration files on a per (satellite, ground-station, radio) basis.

  • Ability to export consistently-formatted PDF link budgets for communcation with external agencies.

  • Ability to easily produce graphs, such as pfd/4kHz for regulatory compliance.

  • Ability to easily solve for required values within a link budget.

  • Ability to tag components with arbitrary values, such as datasheet links, descriptions, and part numbers.

  • Creation of more generalized models for calculating whatever you want (see the Midlife Crisis Example or the HyperSpectral Imaging SNR Budget)

  • Ease the building of monte-carlo simulations (FIXME: need example).

The nature of this package is best described as three things:

  1. A caching DAG (Directed Acyclic Graph) Solver.

  2. A set of utilities common to link-budgets.

  3. A set of pre-defined computational nodes common to link budgets.

If you're looking for a quick-start and/or just want to go with some boilerplate examples, take a look at the Examples directory.

DAG Solver

Spreadsheets are, for the most part, DAGs. If you define C46 = C87 - $B$34 then you are, essentially, saying that 3 nodes exist: C46, C87, and $B$34. You are also stating that to solve for node C46 you take the value of node C87 and subtract the value of node $B$34.

This system works the same way, except that instead of saying C46, we might instead say link_margin_db. And instead of coupling our presentation and data layers, as in a spreadsheet, we might define it as follows:

def _link_margin_db(model):
    # Note how we're just referencing things like required_ebn0_db as
    # instance vars.  No they aren't instance vars, that's just how
    # you reference nodes in the DAG.
    return model.received_ebn0_db - model.required_ebn0_db

my_example = pylink.DAGModel(received_ebn0_db=8.0,
                             required_ebn0_db=6.0,
                             link_margin_db=_link_margin_db)
print('My Example Link Margin: ', my_example.link_margin_db)

The DAGModel class overrides python's __getattr__ method so that you can reference nodes directly, without the added syntactic sugar of extra parens, brakcets, and tick-marks.

If you're curious what this all looks like in a context other than link budgets, take a look at the Midlife Crisis Example. There we create a DAGModel that has nothing at all to do with RF, satellites, etc. There's really nothing that restricts us to link budgets, or even RF. Feel free to write the nodes and use the framework for whatever you want.

Please note that there are two types of nodes:

  • Static Nodes
  • Calculated Nodes

Simply put, static nodes are just plain old values that you pass in, whereas calculated nodes are functions/methods/... You'll see an enum referenced all over the place. That's because it uses node numbers internally, and an enum is convenient way to reference node numbers without having to use a bunch of single-ticks and brackets. For example:

def _link_margin_db(model):
    return model.received_ebn0_db - model.required_ebn0_db

my_example = pylink.DAGModel(received_ebn0_db=8.0,
                             required_ebn0_db=6.0,
                             link_margin_db=_link_margin_db)
m = my_example # m as in model
e = m.enum     # e as in enum

print(e.link_margin_db)
print(m.node_name(e.link_margin_db))
print(m.node_num('link_margin_db')) # the alternative to using the enum

It also includes a multi-round linear solver for convenience. See the Solver Example.

Utilities

There are some utilities that are handy for working with RF. For example, there's a function that fakes an antenna gain pattern for you: pylink.pattern_generator, and another one that calculates the attenuation of PFD from spreading over a distance: pylink.spreading_loss_db.

Pre-Defined Nodes

As shown above, new nodes can be registered with the DAG Model directly:

my_example = pylink.DAGModel(received_ebn0_db=8.0,
                             required_ebn0_db=6.0,
                             link_margin_db=_link_margin_db)

Here we've added 3 nodes:

  • received_ebn0_db
  • required_ebn0_db
  • link_margin_db

It frequently makes sense to group nodes before registration. That's where Tributaries come into play. If you look in the Basic Example, you see that it uses a whole list of tributaries. Geometry is probably the simplest and most straight-forward tributary if you're looking for a production example, otherwise please see the Examples.

Aside from logical grouping, it also makes sense to reuse code. Antennas, for example, have patterns that can be plotted to PNG files irrespective of whether they're a transmit or receive antenna. Instead of duplicating that code, we simply have a single Antenna class that remembers whether it is meant for tx or rx. When it contributes nodes to the DAG, those nodes (instance methods) will be able to refer to their object and know whether to use the tx or rx path.

Installation

Please note that there is a name collision with another pylink package in PyPi. As such, we have registered this package there under a different name: pylink-satcom.

We recommend using Anaconda with Python 3.7. This package can be installed by executing: pip install pylink-satcom

If you want to install it from source: pip install . works as well.

Legacy Support

Migration instructions from previous versions can be found in the Changelog.

Extending and Understanding

Submodules

  • model.py: Contains the actual DAG Model class that houses the core logic of the calculations.

  • utils.py: Standalone utility functions (such as to_db)

  • report.py: Satellite link budget latex report generator.

  • tagged_attribute.py: The TaggedAttribute class for adding metadata tags to individual components.

  • element.py: RF Element container.

  • tributaries/*.py: These each provide boilerplate inputs and calculators that are common to link budgets. For example, you're likely to need a transmitter and a receiver. There will be a channel to carry the signal, etc.

Creating Tributaries

If you just want to add one more computation, or modify one, you can do so by including it in the model itself -- you don't need to create your own tributary. If, however, you do want to create a new one, use the pre-existing source as a guide (it should be pretty clear). Note that you'll need to define the tribute instance variable. This should be a dict of node-names to values. That works both for constants (like apoapsis_altitude_km or speed_of_light_m_per_s), and for functions that calculate values (like slant_range_km or link_margin_db). The DAG Model will expect this value to exist and raise an exception otherwise.

Tagging Architecture

Individual components include a facility for tagging with metadata such as test report links, datasheets, part numbers, etc. The tagging is key/value based, and not restricted or controlled. There are two primary mechanisms by which tagging occurs:

  • Through pre-defined objects, such as Antenna or Element, whcih permit arbitrary keyword arguments that will be automatically added to their metadata.

  • Throug the use of TaggedAttribute objects, which permit adding arbitrary tags to individual values (such as the rx_antenna_noise_temperature).

You'll find examples of both of these usages in the examples directory.

Cycles

In some circumstances, cycles do exist, breaking the DAG nature of this system. Under very special circumstances, we can deal with those. If one of the items in the loop exists within a finite set, then you can do an O(N) search across all of those options, to determine the most appropriate value. A real-life example can be found in modulation.py:

best_modulation_code
 -> additional_rx_losses_db
 -> excess_noise_bandwidth_loss_db
 -> required_rx_bw_dbhz
 -> required_rx_bw_hz
 -> rx_spectral_efficiency_bps_per_hz
 -> best_modulation_code

The way we get around this issue, is to recognize that best_modulation_code exists within a finite set (specifically all available modulation options). That allows us to, essentially, fake the return value of our own function, observe a figure of merit, and return the appropriate value at the end. To introduce a cycle, you'll need to do the following:

  1. Loop through all possible options

  2. In your loop, start by overriding the value you are attempting to compute to the current option

  3. Compute the value that induces a cycle with clear_stack=True

  4. Revert the value you are attempting to calculate

  5. Select the appropriate option by comparing the figure of merit.

  6. Return the result from your calculator.

For example:

def _cycle_inducement(model):
    e = model.enum

    best_cycle = -1
    best_option = None

    for option in model.cycle_inducement_options: # Step 1
        model.override(e.cycle_inducement, option) # Step 2
        cycle = model.cached_calculate(e.cycle, clear_stack=True) # Step 3
        model.revert(e.cycle_inducement) # Step 4

        # Step 5
        if cycle > best_cycle:
            best_cycle = cycle
            best_option = option

    return best_option # Step 6

You can also find a unit-test of this behavior in model_test.py.

HyperSpectral Imaging

BStar pivoted to HyperSpectral in an attempt to address the disparity between customer/partner/regulator interest in our success and investor interest. HyperSpectral Imaging was the selected target (due to the close association with comms and the simplicity of the business model). For expediency, the HSI SNR budget was computed using pylink, and I'm adding it to the repo to avoid having yet-another-repo. If it gathers enough steam, I'll break it out into a separate repo.