Light sources like "arc lamp sources" when they age, they put less out in the blue. We needed an AM 1.5 global (sunlight) source and we calibrated based on the spectral response of our house made detectors. When we no longer had enough "blue", it was time for a lamp change. Essentially a combination of lamp hours, change in lamp power, calibrated intensity (normal) and calibrated intensity (somewhat blue). We had some cells calibrated at a government lab and then we calibrated many (about 10) of our own. Calibration always consisted of the last source used + one calibration cell, lamp power for 100 mW/sqcm.
Since our detectors were "naked" with no filters, damage was possible. The recycling kept everything in check. If it didn't work, you had to have three calibration standards that agreed. We only had about 4 of the "blueish" sensitive ones. We had systems that could measure the spectral response of the sensor, but we did not have something to measure the spectral response of the lamp output. The detector can measure the "integrated" response of it's sensitivity.
So, if we had a home made device that had more of blue response, then the other wavelengths could be mechanically filtered out.
There were filters that shaped the arc lamp output to AM1.5G. One of those filters were about 1mm thick and 3-4" in diameter. Close to a hot bulb and a pretty clunky electric shutter that I had to calm down. That lens was about $1000.00 USD.