Reference peak [in]visibility

[BOPOHOP]

Reference peaks can be hidden at some zoom levels.
MacOS 15.3
Interspec 1.0.13 build date 20241002
Spectrum with 8192 channels, calibcoeff : a=-3.99e-11 b=1.918e-06 c=0.3526 d=1.6
Reference nuclide / disapperings peak:
cs137 -- 32.19 (Ba137m -> Ba)
K40 -- 511

Test spectrum (csv format) and screenshoots with different zoom levels are attached

tst.csv

[wcjohns]

Hi @BOPOHOP ,

I think this is intended behavior.
For each nuclide the reference lines are showing for, the lines for that nuclide are scaled so the largest amplitude line in the visible range, for that nuclide, goes about to the top of the spectrum, and then all other line heights for that nuclide are scaled relative to that largest amplitude one.

In your first screenshot, the only Cs137 lines that are in the visible range, are the x-rays around 30 keV, so they are scaled to go to the top of the spectrum.
In the second screenshot, the 661.7 keV of Cs137 is visible, and has a much larger B.R. than the x-rays, so it is scaled to go to the top, and now since the x-ray are only 3 or 4% of the 661.7 keV, they are barely visible, since they only go up 3 or 4% of the spectrum; if you put the mouse over the x-ray energies, it should still show you information about those lines at the top of the spectrum.
A similar thing is happening with the two K40 lines, between second and third screenshots.

The reasoning behind this is to help you judge if the peaks are not just at the right energy, but also at the expected relative amplitudes. And also, it makes it so busy nuclides, like Pu239, are much more reasonably displayed.

Although this relative scaling is the best method I've found (at least for me), it's not perfect. For example, the line heights are scaled linearly, even when the y-axis is log; I've found this actually matches perception well (and better than scaling in log), because of the continuum, but is a bit confusing. Another issue is if you have not selected a detector efficiency function, then the line heights don't take into account your detectors response.

Ideas for improvement are greatly appreciated.

[BOPOHOP]

I got it: one line of each nuclide that falls within the displayed energy range is always displayed at maximum height, the rest are scaled accordingly. Thanks for explanation!
As for "always linear" or "follow spectrum", I think it would be nice to be able to toggle this in the settings. Maybe it is visually perceived differently with spectra from high resolution HPGE or low resolution NaI. I only have NaI and CsI detectors and have never had to deal with HPGE.

[wcjohns]

The follow spectrum option was explored, and it wouldn't be hard to enable this by editing SpectrumChartD3.js around here, however, here is showing reference lines linearly (e.g. how InterSpec does it by default) while y-axis is log:

Verses showing them log, with y-axis still log:

To me, the first option still looks about right, even if the second option was tweaked a bit more.

One thing I've been thinking about is adding an option to only display lines above a certain significance threshold, and maybe use line opacity and/or thickness to indicate relative intensity, instead of height. Its a bit simpler, and more what people are used to, even if it loses some info.
Let's keep this issue open until I get to exploring that, or any other better idea that comes up beforehand.

[BOPOHOP]

Your example definitely shows that it need to hide some lines when using non-linear scaling. At first glance - the threshold is about 1/10 of the maximum, but this needs tuning.
I don't think different widths or transparency are enough.
Another idea that would probably require a lot of reworking is to build "pseudo-spectrum" - sum of gaussians representing each reference line (and ignore scattering) but in this case we need valid DRF.