Changes

[[Image:HPD_3dist.png|frame|Distributions on a logarithmic plot from runs with different light levels (roughly 64, 530 and 973 primary detected photons.) Units are arbitrary and histograms have been rescaled for comparison.]]

[[Image:HPD_3dist.png|frame|Distributions on a logarithmic plot from runs with different light levels (roughly 64, 530 and 973 primary detected photons.) Units are arbitrary and histograms have been rescaled for comparison.]]

Fits with far better agreement were found when the above-listed parameters were free to vary individually for each distribution. Similar and more drastic problems occurred when the simpler on-Gaussian model was used for the distribution of secondary photo-electrons. Evidently, the inaccuracy in its shape is compounded with more convolutions present in the high-photon count spectrum. Thus, we suspect that the two-Gaussian approximation of this asymmetric distribution is yet not full adequate. However, this already computationally-intensive model (note the three summations) will be even more difficult with more Gaussian and intractable for an arbitrary shape since the convolutions will no longer be straightforward. Disagreements between the fit and data in the right-hand tails of the spectra may also be due to the truncation of the summation in ''n'' corresponding to the number of peaks present due to the convolution with primary and secondary distributions.

Fits with far better agreement were found when the above-listed parameters were free to vary individually for each distribution. Similar and more drastic problems occurred when the simpler one-Gaussian model was used for the distribution of secondary photo-electrons. Evidently, the inaccuracy in its shape is compounded with more convolutions necessary in the high-photon count spectra. Thus, we suspect that the two-Gaussian approximation of this asymmetric distribution is yet not full adequate. However, this already computationally-intensive model (note the three summations) will be even more difficult with more Gaussian and intractable for an arbitrary shape since the convolutions will no longer be straightforward. Disagreements between the fit and data in the right-hand tails of the spectra may also be due to the truncation of the summation in ''n'' corresponding to the number of peaks present due to the convolution with primary and secondary distributions.