Optimize QMix

Optimization

• Added more detailed benchmarking and profiling scripts (found in QMix/profile/).
• qmix.qtcurrent.qtcurrent is now ~20% faster for 1 tone.
• qmix.harmonic_balance.harmonic_balance is ~20% faster for 1 or 2 tones, and ~10% faster for 3 tones.

Functionality Changes

• Renamed cct.vph to cct.freq (in qmix.circuit.EmbeddingCircuit). The normalized photon voltage is equivalent to the normalized frequency. I think frequency is easier to understand so I changed it.
• Deleted the command line scripts (bin/). There was only one script and it was pretty obscure so I descided to delete it.
• Moved qtcurrent_all_freq() function from qmix.qtcurrent to qmix.harmonic_balance. Again, it's a pretty obscure function and it's only used in the harmonic_balance module, so I moved it.
• Deleted qtcurrent_std from qmix.qtcurrent. This function calculated the "standard" output values, but this was pretty arbitrary. It also didn't really simplify the package at all so I deleted it.

Documentation

• Travis CI now automatically rebuilds the online documentation on each new commit.
• Added information on how to contribute to QMix to README.
• Added docstrings to many private functions. This should allow new users to understand the inner workings a bit better.

Miscellaneous Minor Changes

• QMix/examples/ renamed QMix/notebooks/
• Removed codemeta.json and setup.cfg.
• Fixed several bugs with Travis CI.

Fix parameters for experimental data

Experimental data

• Process experimental data in a consistent order (i.e., I-V and IF data should be processed in the same way).
• Add warning if normal resistance is too low or too high. This can help to detect if you are using the wrong units for the current.
• Make parameters in qmix.exp.parameters more consistent. Changes include:
  • voffset_range, which defines where the voltage offset is potentially found, is now defined as a list. E.g., if the experimental voltage offset could be found anywhere between -0.5mV and 0.1mV, you now use voffset_range=(-0.5e-3, 0.1e-3). This change is backwards compatible, so you can still define it as a float if you like. For example, if you set voffset_range=1e-3, this is equivalent to voffset_range=(-1e-3, 1e-3).
  • rn_vmin and rn_vmax, which previously defined where the normal resistance was calculated, are now combined into vrn. Previously, this was defined using rn_vmin and rn_vmax. Now, it is defined using vrn as a list. For example, if you previously used rn_vmin=4e-3 and rn_vmax=5e-3, you would now use vrn=(4e-3, 5e-3). This change is backwards compatible, so you can still use rn_vmin and rn_vmax.
  • vshot, which controls where the shot noise slope is calculated, is now a list of lists, so that multiple voltage ranges can be defined. For example, if the shot noise slope is smooth between from 4-5mV and 5.5-6mV, you can define vshot=((4e-3,5e-3),(5.5e-3,6e-3)). This change is backwards compatible, so you can still use a normal list.
  • cut_low and cut_high, which previously defined the region of the first photon step that would be used for impedance recovery, are now combined into fit_range. For example, if you previously used cut_low=0.25 and cut_high=0.2, you should now use fit_interval=(0.25, 0.8). (NB: The way that the upper limit is defined has been changed!). In this example, the script will ignore the first 25% of the photon step and the last 20% during the impdance recovery process. This is backwards compatible, so you can still use cut_low and cut_high if you like.
  • vgap_guess and igap_guess have been removed. They actually weren't needed all along.
  • ifdata_vmax, which previously defined the maximum IF voltage to import, has been removed. QMix now uses the value from vmax instead.
  • ifdata_sigma, which defines the width of the filter for the IF data, is now defined in units [V]. Previously, it was defined by the number of data points. This is also backwards compatible (if ifdata_sigma is >0.5, it will assume that you are defining it by the number steps).

Optimization

• Add basic timing scripts for qmix.harmonic_balance. See speed/ directory.
• Use a better initial guess for the junction voltage in the qmix.harmonic_balance.harmonic_balance function.

Testing

• Improve code coverage. Now 99.0% covered!

Command line

• Add command line script to plot IF response from experimental data (bin/plot-if-response.py).

Reduce simulation time

Optimization

• Optimize qmix.qtcurrent using Numba and JIT. 4 tone simulations are now 1.5 times faster!
• Add basic timing scripts for qmix.qtcurrent. See speed/ directory.

Changes

• Move the code that is used for analyzing the IF response into a new module: qmix.exp.if_response.

New Features

• Add ability to force embedding impedance in qmix.exp.exp_data.RawData. The embedding voltage will then be calculated using this value.

Testing

• Improve test coverage for qmix.circuit, qmix.respfn, and qmix.exp.exp_data.
• Add tests for new module (qmix.exp.if_response).
• Add notebook to recreate plots from Kittara's thesis. (For the purpose of validation.)

Minor Changes

• Improved progress messages from qmix.harmonic_balance.harmonic_balance.
• Add ability to plot response function (qmix.respfn.RespFn).
• Always use default parameters from qmix.exp.parameters (i.e., reduce the number of magic numbers).
• Add ability to call qmix.respfn.RespFn by adding __call__ method. This will return the interpolated response function (a complex value).
• Fix error in qmix.harmonic_balance.harmonic_balance (when Thevenin voltage is checked).
• Other minor changes to documentation.

Update documentation and metadata

• This release was created after the JOSS review.
• There are minor changes to the documentation, but no changes to the package itself.

Changes to plotting functions and experimental data imports

New Features

• All plotting functions now accept and return Matplotlib axis objects.
• When analyzing experimental data, you can now pass either CSV files or Numpy arrays to RawData and RawData0.

Changes

• In the qmix.respfn.RespFn class, the f_idc, f_ikk, f_didc, and f_dikk attributes were all made private by adding a leading underscore. They are replaced by the idc, ikk, didc and dikk methods, respectively.
• In the experimental data module, the parameter iv_multiplier was renamed i_multiplier. RawData0 and RawData use this parameter to correct imported current values. (All imported current data is multiplied by i_multiplier.) I also added v_multiplier, which does the same things, except for imported voltage data.

Testing

• Dropped Travis-CI testing with Python 3.5-dev and 3.6-dev. QMix is now tested with 3.5, 3.6 and 3.7-dev.
• Added automatic coverage testing through coveralls.io.

Bug Fixes

• Fixed potential issue with file paths. Previously, file paths were built assuming a mac operating system (e.g., "some-dir/this-figure.png"). The package was updated to now use os.path.join("some-dir", "this-figure.png"), which is machine independent.

Documentation

• Added package API to web page.
• Added more information about how experimental is stored and imported into QMix.
• Added information on getting support and contributing to QMix.
• Added descriptions of different response function types to docstring in qmix.respfn.py.
• Added more comments on keyword arguments and attributes for classes in qmix.circuit.py and qmix.respfn.py. (This should help to explain some of the SIS-related jargon.)
• Added use cases for many class methods.
• Fixed documentation for qmix.exp.parameters.py.
• Updated workflow examples.
• Other (minor) changes.

Fix description

Bug Fixes

• Fixed description in setup.py. Previously, there were some non-ASCII characters, which were causing issues on Windows installations.

First release of QMix

This is the first release of the QMix software package. This software can be used to simulate the quasiparticle tunneling currents in SIS junctions.