Detectors and selectivity:
Up to
at least 300GHz (probably a lot more these days) receivers can be built like at 1MHz or 1GHz, with preamps, mixers, filters... Which does not mean that they're as sensitive. Active amplifiers, including low-noise, are presently common at 100GHz; well above that, it's more a domain for varactor mixers and harmonic generators, which only lose power, with all the consequences on the noise figure.
So-called
terahertz waves (they use to be the upper GHz... marketing) have no active amplifiers, no continuous-wave transmitters... Nothing accurate is done with them, just a few images and vague spectra at metre range.
Some day better components may emerge from my suggestions there
http://www.physforum.com/index.php?showtopic=15617&st=30but I must first make drawings and understandable explanations. Presently, it's a holy mess, sorry.
Things improve again around 10µm, for which bolometers exist, as well as quantum detectors like HgTe. These are not selective, they just give one electron per photon.
Same in the near-IR, visible, UV : selectivity and spectra are made by other means. Check for
diffraction grating, Fourier transform spectroscopy,
interference filters... plus in some future my evanescent wave filter, there:
http://www.scienceforums.net/topic/74445-evanescent-wave-optical-filter/The usual means is to have very wide-band quantum detectors. Though, optronics also uses
non-linear crystals the same way as electronics takes non-linear components to make
down-mixers, the output at radio frequency being then processed as in a radio receiver, with great selectivity. It relies on big power density at a stable frequency from the local oscillator, hence works only at the available laser frequencies. Typical for
Lidar.
At
X and gamma rays, detectors have little selectivity. They give several electrons per incoming photon; the number of electrons in a pulse tells the approximate energy, provided the photons come isolated enough and the detector is proportional. For some X-ray energies, atoms exist that absorb more strongly just above some energy threshold (=when the 1s ionization energy is exceeded, also the 2s...), so removable filters can be made, to compare the picture with/without the filter(s) and deduce how much energy arrives in that (many % wide) energy band. This allows to know what atoms compose an observed material.