The modulation transfer function (MTF) is the number most optics buyers use to judge image quality. Lens makers quote it, defense programs write it into acceptance specs for targeting and surveillance optics, and electro-optical (EO) system builders measure it before a payload ships. Every MTF bench ends at a camera, and that camera can either disappear from the result or quietly become part of it.
This post is for optics manufacturers, EO test engineers and test-equipment builders choosing the detector for an MTF bench. It uses Finger Lakes Instrumentation (FLI) Aurora, Revelation and Kepler cameras as worked examples. For general sampling math, see our post on pixel size and sampling.
A typical bench projects a target, such as a slit, pinhole or edge, through a collimator into the lens or system under test. The camera records the image, and software turns the image profile into contrast versus spatial frequency. ISO 9334 defines the optical transfer function, and ISO 9335 describes how to measure it.
Some benches place the sensor directly at the image plane. Others put a relay microscope objective behind the lens under test, so the camera sees a magnified image of the spot or slit. Either way, the camera's own response is in the measurement.
MTFs of a linear imaging chain multiply. Boreman's SPIE tutorial text on MTF treats the detector as one more term in that product, set largely by the pixel's size. For a square pixel of pitch p that fills its cell, the pixel MTF follows a sinc curve that falls to about 0.64 at the Nyquist frequency of 1/(2p).
With the 3.76 µm pixels of Aurora and Revelation cameras, Nyquist is about 133 line pairs per millimeter (lp/mm). With the 9 µm pixels of the KL4040BI, it is about 56 lp/mm. If the lens under test delivers contrast well beyond those numbers, imaging it straight onto the sensor will understate it, unless your software divides out the detector term.
The relay objective solves that problem. With a 10x relay, a lens feature at 100 lp/mm reaches the sensor at 10 lp/mm, well below the Nyquist limit of 3.76 µm pixels. The pixel MTF at that frequency is close to 1, so the measurement reflects the lens, not the camera. These numbers are illustrative arithmetic; your bench's relay, target size and frequency range set the real ones.
Magnification has a cost. A larger image spreads the same light over more pixels, so the signal per pixel drops, and the field you can see shrinks. That trade leads directly to noise.
The slanted-edge method in ISO 12233, described in detail by Burns (IS&T PICS, 2000), tilts an edge a few degrees against the pixel grid. Rows that cross the edge at slightly different phases are combined into an oversampled edge profile. That profile is differentiated into a line spread function, and its Fourier transform gives the MTF.
Differentiation amplifies noise, so the edge image needs a clean signal. Aurora and Revelation list 1 e- rolling read noise and a 50,000 e- full well. Near half of full well, shot noise sets the limit: about 158 to 1 at 25,000 e-, the square root of the signal. Averaging several frames improves it further.
Linearity matters as much as noise. A response that bends near the top or bottom of the range reshapes the edge and changes the computed MTF. ISO 12233 expects the data to be linearized through the camera's opto-electronic conversion function before analysis. Kepler cameras produce a 16-bit high dynamic range image from two 12-bit readouts at different gains, with less than 1% non-linearity. To check gain, noise and linearity on your own unit, see our post on camera acceptance testing. A production test summary is available for each camera by serial number.
Lens specs usually give MTF on axis and at several field heights, often across a range of focus positions. A sensor large enough to see the full field lets you measure several field points without moving the camera. The AR455 is 9568 × 6380 pixels on a 36 × 24 mm sensor (43.2 mm diagonal). The KL6060FI is 6144 × 6144 pixels with 10 µm pixels and an 86.9 mm diagonal, for larger image fields. For compact benches and instrument heads, the RV533 offers a square 3003 × 3003 format, and Revelation's thin body fits tight layouts.
Through-focus scans and multi-field runs mean many frames over a long session. A regulated sensor temperature keeps the response steady from the first frame to the last. Aurora and Revelation cool to 50 °C below ambient with air, with liquid cooling as an option, which helps on a vibration-isolated table. For mounting and focus offsets, see our post on back focus and flanges.
MTF depends on wavelength, so test in the band the system will use. Aurora and Revelation offer window options: the F116 extends farther into the near-IR and the F101 farther into the near-UV, and the window is chosen before the clean-room build. Kepler is offered in front-illuminated and back-illuminated versions.
Stray light and offsets lower measured contrast, so take dark and background frames and subtract them. Aurora is not offered with a mechanical shutter, so block the source for darks. Kepler supports internal shutters.
EO programs test targeting, surveillance and tracking optics against MTF specs at several stages, from incoming lens inspection to final system acceptance. Test equipment for those programs also brings supplier paperwork. FLI operates under AS9100D / ISO 9001:2015 · ITAR compliant · CMMC Level 2 compliant. We machine housings in-house, finish and test boards, and test every camera before it ships. Our defense camera supplier checklist covers qualification in more depth.
For bench automation, the Kepler SDK supports Windows and Linux, and every Revelation camera includes our software development kit (SDK).
Send our sales team your frequency range, relay design, field size and waveband, and we will suggest the standard Aurora, Revelation or Kepler configuration that fits. If a bench needs a modest change to a standard camera, we are glad to discuss it. Request a quote or email sales@flicamera.com.
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.