When an instrument builder or a national lab buys cooled cameras in volume, the first unit rarely gets the most scrutiny. The tenth, the fiftieth and the replacement that arrives three years later matter more, because each one has to behave like the camera the system was designed around.
This post is for the engineers who write those acceptance procedures. It covers which numbers to measure, how to measure them with standard methods, where data-sheet values fit, and how to work with the production records that come with each Finger Lakes Instrumentation (FLI) camera. The methods apply to any cooled CCD or CMOS camera.
A camera inside an analyzer, a beamline detector or a telescope network is one part of a calibrated system. Software thresholds, exposure tables and image-processing pipelines are often tuned to a camera's gain and noise. If a new unit differs, the instrument may still work, but its results shift.
A short, repeatable test at receipt catches problems while they're easy to fix. It also gives you a baseline for that serial number, which helps when you're diagnosing a fault years later.
FLI data-sheet values are typical. They describe the model, not a specific unit. Read noise, full well, dark current and gain vary from sensor to sensor, so an acceptance plan should compare each camera with its own records, with the data sheet as a sanity check.
FLI keeps a production test summary for each camera, and our team can provide it by serial number. The serial number is on the back of the camera, and its last two digits give the year of manufacture. Ask for the test summary when you order, and file it with the camera's other paperwork.
Sensor cosmetics work the same way. Defect limits follow the sensor maker's grade specification, so if your instrument can't tolerate certain column or row defects, ask our sales team about grade options before you order.
Gain, in electrons per ADU, converts the counts in your image into detected electrons.
One common trap: the gain setting names shown in FLI Pilot, such as 1.29 or 7.25 on a Kepler camera, come from the sensor maker. They are labels, not e-/ADU values. To get the real conversion, use your camera's production test summary or measure it yourself. FLI has instructions for measuring linearity and gain in FLI Pilot, and our sales team can send them. Keep Pilot up to date, because some older versions displayed gain inverted, as counts per electron instead of electrons per count.
The standard measurement is the photon transfer method, which the EMVA 1288 standard also uses:
Run the test at the gain modes and readout settings your instrument will actually use, since each one has its own conversion.
Read noise comes from the same bias pairs. Take the standard deviation of the difference between two bias frames, divide by the square root of two, and multiply by the gain to get electrons.
Dark current needs a dark sensor and time. Take dark frames at several exposure lengths at your operating temperature, subtract a bias, and fit the signal against exposure time. The slope, converted with the gain, is dark current in electrons per pixel per second. Because dark current rises quickly with temperature, report the sensor temperature with every result.
Kepler cameras support internal shutters, which makes darks convenient. Aurora cameras aren't offered with a mechanical shutter, so use a blocked position on a filter wheel, such as the CFW8-8RND, a third-party external shutter, or a light-tight cap.
Linearity comes from the same flat series. Plot mean signal against exposure time at constant illumination and measure how far the points drift from a straight line. Full well is where the response flattens or the noise variance collapses near saturation.
Know how your camera builds its image before you judge the curve. On Kepler cameras, the 16-bit high-dynamic-range image combines two 12-bit readouts at different gains, and FLI specifies less than 1% non-linearity for it. Test it in the mode you'll ship. Aurora cameras use Sony sensors that don't support on-sensor charge binning, so binning is done digitally. Measure noise unbinned first, then confirm any binned mode your instrument uses separately.
Cooling is specified in degrees below ambient. The Aurora AR455 and Revelation RV455, for example, are TEC-cooled to about 50 °C below ambient. The temperature a camera can hold depends on sensor size and the room it's in, so set a target with some headroom, a little above the maximum cooling, so the temperature stays stable through the test.
Give cameras time to settle. The KL4040's cooling rate is limited to about 3 °C per minute to protect the sensor, so it takes longer to reach setpoint than older units did.
If a camera doesn't reach the temperature you expect, send our team the setpoint, ambient temperature, cooler power and serial number. We can compare them against that camera's production test summary. Our post on air versus liquid cooling for OEM cameras covers the mounting and airflow choices that affect this.
For programs buying many units, the procedure matters more than any single number. A few habits make results comparable from camera to camera and year to year:
Our posts on choosing an OEM camera partner for life-science instruments and large format cameras for national labs cover the selection steps that come before this one.
Every FLI camera is 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. Browse the Kepler, Aurora and Revelation lines to find the right fit.
Send our sales team your acceptance criteria, the modes you'll test and the number of units you're planning, and we'll help you match a camera and its records to your procedure. Request a quote or email sales@flicamera.com.
