Validation script modifications

[jaydeshpande]

https://core.ac.uk/download/pdf/76989828.pdf validating against fig 3.21

from pvtrace import *
import numpy as np
import matplotlib.pyplot as plt

def create_waveguide():
    # Simulation wavelength range
    x = np.arange(400, 801, dtype=np.float)

    # LSC plate size
    size = (l, w, d) = (4.8, 1.8, 0.260)  # cm

    # Make LSC model
    lsc = LSC(size, wavelength_range=x, n1=1.49)
    #lsc.add_solar_cell({'left', 'right', 'near', 'far'})

    # Use Fluro Red dye with peak absorption coefficient 11.39 cm-1
    lsc.add_luminophore(
        'Fluro Red',
        np.column_stack((x, fluro_red.absorption(x) * 11.387815)),  # cm-1
        np.column_stack((x, fluro_red.emission(x))),
        quantum_yield=0.95
    )

    # Include constant background absorption coefficient of 0.02 cm-1
    lsc.add_absorber(
        'PMMA',
        0.02 # cm-1
    )

    # This function returns an approximation of the lamp spectrum used in the experiment
    def lamp_spectrum(x):
        """ Fit to an experimentally measured lamp spectrum with long wavelength filter.
        """
        def g(x, a, p, w):
            return a * np.exp(-(((p - x) / w)**2 ))
        a1 = 0.53025700136646192
        p1 = 512.91400020614333
        w1 = 93.491838802960473
        a2 = 0.63578999789955015
        p2 = 577.63100003089369
        w2 = 66.031706473985736
        return g(x, a1, p1, w1) + g(x, a2, p2, w2)

    # Add a custon light
    lamp_dist = Distribution(x, lamp_spectrum(x))
    wavelength_callable = lambda : lamp_dist.sample(np.random.uniform())
    position_callable = lambda : rectangular_mask(l/2, w/2)
    lsc.add_light(
        "Oriel Lamp + Filter",
        (0.0, 0.0, 0.5 * d + 0.01),  # put close to top surface
        rotation=(np.radians(180), (1, 0, 0)),  # normal and into the top surface
        wavelength=wavelength_callable,  # wavelength delegate callable
        position=position_callable  # uniform surface illumination
    )
    return lsc

runs = {}
for i in range(10):
    lsc = create_waveguide()
    lsc.simulate(800, emit_method='full')
    counts = lsc._make_counts(lsc._df)
    opdt = 100*counts[['Solar Out', 'Luminescent Out']].sum(axis=1)/800
    runs[i] = opdt.to_list()

import pandas as pd 
combinedDat = pd.DataFrame.from_dict(runs, orient='index')
combinedDat.rename({0:'left', 1:'right', 2:'near', 3:'far', 4:'top', 5:'bottom'}, axis=1, inplace=True)

[danieljfarrell]

What was the change you made?

[jaydeshpande]

I suspect ... I changed too many things when I tried it the last time. For this script above:

• started with https://github.com/danieljfarrell/pvtrace/blob/dev/lsc-device/examples/lsc_custom_illumination.py

• set parameters to match that from the thesis (thickness, refractive index)

• set emit_method to 'full'. Partly because, 'kT' creates wavelengths out of bounds. I looked at modifying pvtrace/material/distribution.py. I was trying to re-write this part to "pass" when the nm is out of bounds. But quickly realized that this pass would destroy the photon accounting elsewhere. So instead, just decided to go with full.

            eV = 1240.0 / nm
            eV = eV + 3 / 2 * kB * T  # Assumes 3 dimensional degrees of freedom
            nm = 1240.0 / eV
            p1 = dist.lookup(nm)

Comparison below:

Surface	Model	Fraction (%)
Bottom	Raytrace ICL	49.2537313
Bottom	Raytrace ECL	49.8507463
Bottom	Thermodynamic	50
Bottom	PVtrace	50
Top	Raytrace ICL	13.5820896
Top	Raytrace ECL	13.5820896
Top	Thermodynamic	13.7313433
Top	PVtrace	13.95
Long	Raytrace ICL	7.1641791
Long	Raytrace ECL	7.01492537
Long	Thermodynamic	7.01492537
Long	PVtrace	7.03125
Short	Raytrace ICL	6.56716418
Short	Raytrace ECL	6.26865672
Short	Thermodynamic	5.67164179
Short	PVtrace	5.7875

[jaydeshpande]

By the way - I truly believe in the capability of PVTrace. I might do some pull requests with new additions if you are open to it!

I like how clean and easy to understand it is.

[danieljfarrell]

What features are you thinking of adding? I would certainly consider any pull requests. You want to make a proposal first so that it's definitely something I want to include in the project.

[danieljfarrell]

Thanks for preparing the script above and the table! Makes it very clear. In fact I was half way along doing this myself. This is why I isolated this on its own branch.

[jaydeshpande]

Sure! I will think about it a bit more - I will run the proposal by you. So far, I feel that the ability to import meshes would be the most welcome addition. I see that you have already created an issue about that. Would be great if we can import STL meshes.

btw, feel free to use the script above as a demo script in the next version if you'd like.

[danieljfarrell]

Yes, I will include the script above in the next release.

There is an example using an icosahedron mesh here. pvtrace can already import STL meshes using the Mesh geometry and trimesh.

[jaydeshpande]

Thanks! Those examples went under my radar. I will take a look.

[danieljfarrell]

Validation notebook has been added!

https://github.com/danieljfarrell/pvtrace/blob/dev/lsc-device/examples/Validation.ipynb

Thanks for your input in getting this completed.

[danieljfarrell]

See #37