# How is spectral sensitivity of a camera calculated?

I understand that to recreate the spectral response of a camera's sensor, you need to use a monochromator/spectrometer to get calibrated images. But I'm not sure where you go from there. When looking at graphs, generally I just see spectral response (y) and wavelength (x). There aren't any units for spectral response so I find this quite confusing. Am I just sampling the pixel values of each image taken? Is there some sort of conversion/function/software that let's me calculate this for each of the camera's channels?

Okay, I've seen some of your responses but I'm still not sure I understand. Perhaps there is a more comprehensive resource I could be pointed to? My ultimate goal is to take the spectral response of a camera and use it to generate a model of animal vision by pairing with the results of photoreceptor responses in the retina.

• Have you checked Wikipedia? "In the specific case of a photodetector, responsivity measures the electrical output per optical input." - so as far as I understand it, it is just a relative number (e.g. 80%) Commented May 27, 2018 at 17:12
• Also, these graphs are normally done relative to a fixed light power, not photon density, per nm.
– doug
Commented May 27, 2018 at 22:37
• What units are your animal photoresponse functions expressed in? Commented Jun 16, 2018 at 3:41

There are two ways to express the wavelength dependent sensor responsivity in the context of a system analysis.

First, it can be expressed as a percentage. This can be a peak normalization, in which case the highest value will be 1.0. More commonly, the percentage is expressed in units of quantum efficiency. For most consumer cameras this will average about 30-40% but specialized BS-CMOS sensors could hit 85% without a bayer filter.

Secondly, the value used to express responsivity is the sensor response value. This is either digital count or output voltage at some point in the sensor circuit before the ADC. The value is the numerator in the responsivity ratio where the denominator is watts per meter^2. It is common practice, however, to set the demoninator equal to 1 E.G. one would say 500 digital count per wm^2 rather than 2000 digital count per every 4 w/m^2.

The advantage of the latter value is that the digital count response can be directly calculated from the irradiance. One need only account for loss of light to the lens assembly and, if present, the microlens array. This can generally be done as "mental math." Converting relative or absolute quantum efficiency to digital counts, on the other hand, requires that the digital response be calculated via plank's constant and the sensor quantization formula.