A camera's data sheet lists pixel size near the top, and it's tempting to treat smaller as better or bigger as more sensitive. Neither is true on its own. Pixel size only means something next to the optics in front of it: the focal length of a telescope, the magnification of a microscope, or the working distance of an inspection lens.
This post walks through the sampling math that system designers use to match a cooled camera to their optics. It applies to any scientific camera, and we use Finger Lakes Instrumentation (FLI) models as worked examples.
Every optical system has a resolution limit set by its aperture, its wavelength and, for ground telescopes, the atmosphere. The camera's job is to sample that image finely enough to keep the detail, without spreading light over so many pixels that each one sees too little signal.
Get it wrong in one direction and you undersample: fine detail falls between pixels, star images look square and small defects disappear. Get it wrong in the other direction and you oversample: you pay for pixels that add no information, and each pixel collects less light, so noise matters more.
That's why the right pixel size for an instrument comes out of a short calculation, not a preference.
The usual rule of thumb comes from the Nyquist sampling theorem: put at least two pixels across the smallest detail the optics can resolve. Many designers aim for two to three pixels across the resolution element, or across the full width at half maximum (FWHM) of a star image. That keeps detail without wasting much light.
Some applications want more. Precise centroiding and photometry often benefit from a little oversampling. Detection of faint sources, where every photon counts, often prefers sampling near the low end.
For a telescope, the key number is the pixel scale, in arcseconds per pixel:
Pixel scale (arcsec per pixel) = 206.265 × pixel size (µm) ÷ focal length (mm)
A few examples with FLI pixel sizes:
If your site's typical seeing is 2 arcseconds FWHM, two to three pixels across a star means a pixel scale of roughly 0.7 to 1.0 arcsec. That's why small-pixel cameras suit shorter focal lengths and larger pixels suit longer ones. FLI describes its ProLine CCD cameras, such as the PL23042 with 15 µm pixels, as "ideal for extended exposures with long focal length optics."
Field of view follows from the same geometry. The KL6060's 61.4 mm sensor on a 4,000 mm telescope covers about 0.88 degrees on a side. For satellite tracking networks, that tradeoff between pixel scale and sky coverage drives telescope and camera choices together; our post on building an SDA sensor network covers the fleet side.
In a microscope or an imaging instrument, divide the pixel size by the total magnification to get the pixel size at the sample. Then compare it with the optical resolution, which for a well-corrected objective is about 0.61 λ / NA (the Rayleigh limit).
For example, at 520 nm with a 0.75 NA objective, the Rayleigh limit is about 0.42 µm, so Nyquist sampling needs pixels of about 0.21 µm or smaller at the sample. A 3.76 µm pixel at 20x magnification is about 0.19 µm at the sample, which meets that target. An 11 µm pixel at the same magnification is about 0.55 µm, which undersamples; it would need more than 50x total magnification, or a relay lens, to match.
That's one reason small-pixel back-illuminated CMOS sensors work well in compact instruments: they reach good sampling with less magnification and a shorter optical path. The Revelation line puts the same Sony sensors as Aurora into a thin OEM form factor for instrument builders.
Inspection systems usually start from the other end: the size of the part and the smallest defect you need to see.
Divide the field of view by the number of pixels across the sensor. An AR455 has 9,568 pixels across its 36 mm width, so imaging a 100 mm wide area puts about 10.5 µm of the part on each pixel. If you need three pixels across the smallest defect, that setup finds defects down to about 31 µm. A smaller field, more pixels or a second camera changes the answer.
Our post on cooled cameras for display, solar and optics inspection covers when a cooled scientific camera makes sense for inspection in the first place.
Pixel count and pixel size together set the sensor's size, and the sensor has to fit inside the area your optics illuminate and correct. Check the sensor diagonal against the lens or telescope's image circle before you choose.
FLI sensor diagonals range widely. The AR533 is 16 mm, and FLI calls it "an ideal replacement for 1" format CCDs," which matters when an existing instrument's optics were designed around that format. The AR455, AR461 and AR411 are 43.2, 54.8 and 66.7 mm. The KL4040 is 52.1 mm and the KL6060 is 86.9 mm. A large sensor behind optics with a small corrected field will show vignetting and soft corners, so the extra pixels add little.
Our post on back focus, flanges and filter wheels covers the mechanical side of fitting a camera to the optical path, and the CCD replacement post for gel and blot imagers covers a common format change.
Binning combines neighboring pixels into one larger effective pixel. It's a way to trade resolution for signal when an application is oversampled, or to run one camera across optics with different focal lengths.
How binning works depends on the sensor. A CCD can sum charge on the chip before readout, so a 2x2 bin pays the read noise once. The Sony sensors in FLI Aurora cameras don't support on-sensor binning, so binning is done digitally after readout. Digital binning still improves the signal per effective pixel, but each original pixel contributes its own read noise. With low-noise CMOS sensors, that penalty is often small, but it's worth including in your noise budget.
FLI cameras are Made in USA, and our sensor chambers are assembled in our 1,000+ sq ft ISO Class 7 clean room. See how on our in-house manufacturing page.
Send our sales team your focal length or magnification, your wavelength band and the field you need to cover, and we'll help you match a pixel size and sensor format to your optics. Request a quote or email sales@flicamera.com.
