Method note
CMOS and CCD detectors compared
Read noise, dynamic range, readout speed, and the three cases where the expected choice turns out to be the wrong one.
The detector question has largely been settled by manufacturing volume rather than physics. Modern back-illuminated CMOS performs very well, improves with every generation, and receives most of the industry’s investment. However, “generally better” is not the same as “better for a particular measurement”, and the exceptions are specific enough to be worth listing.
Where the two technologies differ
Readout speed. CMOS reads out per pixel in parallel, while a CCD shifts charge through a serial register. For applications that require thousands of spectra per second, such as OCT, pulsed-source work and anything gated to a laser, CMOS is in a different performance category.
Read noise at speed. CCD read noise increases as the readout rate increases. CMOS read noise is largely independent of rate. When both speed and sensitivity are required, this is usually the deciding factor.
Full-well capacity and dynamic range. Larger CCD pixels still hold more charge. In a single long integration on a stable source, a good CCD can deliver dynamic range that a comparable CMOS device does not.
Uniformity. A CCD shares one output amplifier, so pixel-to-pixel response is inherently consistent. CMOS has an amplifier per pixel, which produces fixed-pattern noise that must be corrected. Modern devices correct it well, but corrected is not the same as absent, and it matters when the measurement is a small difference between adjacent pixels.
Three cases where the expected choice is wrong
-
Long integration on a faint, stable source. The expected choice is the newest CMOS device. For integrations of several seconds on a stationary source, however, full-well capacity and dark current dominate, and a cooled CCD frequently performs better.
-
Absolute photometric accuracy across the array. Where intensity at one wavelength is compared against another in absolute terms, CCD uniformity is the decisive advantage.
-
Deep-UV work. Response below 250 nm depends far more on the specific device, its coating and its window than on the underlying technology. The part matters more than the category, and a UV-enhanced CCD outperforms a general-purpose CMOS in this region.
Selection procedure
Work through the following questions in order and stop at the first that applies:
- How fast is a spectrum required? Above roughly one kilohertz, CMOS is the choice.
- How faint is the light? For integrations longer than a second, a cooled CCD deserves close consideration.
- Which region of the spectrum? Below 250 nm, the specific device matters more than the technology.
- Is the signal a small difference between neighboring pixels? CCD uniformity is the advantage.
If none of these applies, CMOS is the recommended choice. It will be lower in cost, faster and easier to source over the next ten years.