Editing Spectroradiometer (section)

==Definitions==
{{main|Stray light}}

=== Stray light ===
Stray light is unwanted wavelength radiation reaching the incorrect detector element. It generates erroneous electronic counts not related to designed spectral signal for the pixel or element of the detector array. It can come from light scatter and reflection of imperfect optical elements as well as higher order diffraction effects. The second order effect can be removed or at least dramatically reduced, by installing order sorting filters before the detector. 

A Si detector's sensitivity to visible and NIR is nearly an order of magnitude larger than that in the UV range. This means that the pixels at the UV spectral position respond to stray light in visible and NIR much more strongly than to their own designed spectral signal. Therefore, the stray light impacts in UV region are much more significant as compared to visible and NIR pixels. This situation gets worse the shorter the wavelength. 

When measuring broad band light with small fraction of UV signals, the stray light impact can sometimes be dominant in the UV range since the detector pixels are already struggling to get enough UV signals from the source. For this reason, calibration using a QTH standard lamp can have huge errors (more than 100%) below 350&nbsp;nm and a deuterium standard lamp is required for more accurate calibration in this region. In fact, absolute light measurement in the UV region can have large errors even with the correct calibration when majority of the electronic counts in these pixels is result of the stray light (longer wavelength strikes instead of the actual UV light).

=== Calibration errors ===
There are numerous companies that offer calibration for spectrometers, but not all are equal. It is important to find a traceable, certified laboratory to perform calibration. The calibration certificate should state the light source used (ex: Halogen, Deuterium, Xenon, LED), and the uncertainty of the calibration for each band (UVC, UVB, VIS..), each wavelength in nm, or for the full spectrum measured. It should also list the confidence level for the calibration uncertainty.

===Incorrect settings===
Like a camera, most spectrometers allow the user to select the exposure time and quantity of samples to be collected. Setting the integration time and the number of scans is an important step. Too long of an integration time can cause saturation. (In a camera photo this could appear as a large white spot, where as in a spectrometer it can appear as a dip, or cut off peak) Too short an integration time can generate noisy results (In a camera photo this would be a dark or blurry area, where as in a spectrometer this may appear are spiky or unstable readings).

The exposure time is the time the light falls on the sensor during a measurement. Adjusting this parameter changes the overall sensitivity of the instrument, as changing the exposure time does for a camera. The minimum integration time varies by instrument with a minimum of .5 msec and a maximum of about 10 minutes per scan. A practical setting is in the range of 3 to 999 ms depending on the light intensity.

The integration time should be adjusted for a signal which does not exceed the maximum counts (16-bit CCD has 65,536, 14-bit CCD has 16,384). Saturation occurs when the integration time is set too high. Typically, a peak signal of about 85% of the maximum is a good target and yields a good S/N ratio. (ex: 60K counts or 16K counts respectively)

The number of scans indicates how many measurements will be averaged. Other things being equal, the Signal-to-Noise Ratio (SNR) of the collected spectra improves by the square root of the number N of scans averaged. For example, if 16 spectral scans are averaged, the SNR is improved by a factor of 4 over that of a single scan.

S/N ratio is measured at the input light level which reaches the full scale of the spectrometer. It is the ratio of signal counts Cs (usually at full scale) to RMS (root mean square) noise at this light level. This noise includes the dark noise Nd, the shot noise Ns related to the counts generated by the input light and read out noise. This is the best S/N ratio one can get from the spectrometer for light measurements.