The simplest spectrometer is dark sealed tubes of doped sodium iodide crytals with a counter at the ends.
When an energetic gama burst hits a bit of crystal it flashes in a manner that gives up the energy .. over the course of a second and a few thousand events (normal background radiation) you have a spectrum [1].
You no have an indication of all the radiation all arounf the tube.
All around. With no indication of direction. Literally they could have come from anywhere, up, down, left, right, behind you, etc.
What we have here appears to be a vertical tube (or tubes) with mobile masking templates that rotate about and provide some axis of ingress information coupled with any energetic flahes that appear at the same instant.
There is some topographical statistics that needs to be performed and overall there are reduced signal levels (as some incoming gammas are masked and never make a flash) but you can end up with a positional image of where various radiation sources are in the space about the tube.
The other wrinkle they gloss over ('compressed', 'needs fewer counts') suggests (it's either obfuscated or I haven't read deeply enough) they have an eigen library of responses (ie. mean Australian background, Pure Potassium, Pure Uranium, Pure Thorium, Cs-137) baked in to facilitate faster recognition of spectrum patterns.
It's a pattern matching in N^256 type problem in a space that reduces to about N^11 or so when an SVD is applied .. and there are ways to "cheat" and do this faster on the fly.
When an energetic gama burst hits a bit of crystal it flashes in a manner that gives up the energy .. over the course of a second and a few thousand events (normal background radiation) you have a spectrum [1].
You no have an indication of all the radiation all arounf the tube.
All around. With no indication of direction. Literally they could have come from anywhere, up, down, left, right, behind you, etc.
What we have here appears to be a vertical tube (or tubes) with mobile masking templates that rotate about and provide some axis of ingress information coupled with any energetic flahes that appear at the same instant.
There is some topographical statistics that needs to be performed and overall there are reduced signal levels (as some incoming gammas are masked and never make a flash) but you can end up with a positional image of where various radiation sources are in the space about the tube.
The other wrinkle they gloss over ('compressed', 'needs fewer counts') suggests (it's either obfuscated or I haven't read deeply enough) they have an eigen library of responses (ie. mean Australian background, Pure Potassium, Pure Uranium, Pure Thorium, Cs-137) baked in to facilitate faster recognition of spectrum patterns.
It's a pattern matching in N^256 type problem in a space that reduces to about N^11 or so when an SVD is applied .. and there are ways to "cheat" and do this faster on the fly.
[1] https://en.wikipedia.org/wiki/Gamma_spectroscopy