How is NEDT measured?

NEDT (noise equivalent differential temperature) is the key figure of merit which is used to qualify midwave (MWIR) and longwave (LWIR) infrared cameras.  It is a signal-to-noise figure which represents the temperature difference which would produce a signal equal to the camera’s temporal noise.  It therefore represents approximately the minimum temperature difference which the camera can resolve.  It is calculated by dividing the temporal noise by the response per degree (responsivity) and is usually expressed in units of milliKelvins.  The value is a function of the camera’s f/number, its integration time, and the temperature at which the measurement is made.

To make the measurement requires a good quality area source blackbody.  A blockbody is set to 25 degrees Celsius.  The camera is placed as close as possible to it (2 – 5 cm).  For an MWIR camera, set the integration time so that the camera’s focal plane array (FPA) detector is approximately halfway in its dynamic range (half well).  No lens is used for the MWIR measurement.  For an uncooled microbolometer LWIR camera, the integration time is set to its maximum.  The standard 25 mm lens is used for an LWIR camera.

A two-point non-uniformity correction is performed to obtain a uniform image, then three sets of data are taken using the camera’s digital data output. 

For the first dataset, set the blackbody temperature to 20 °C and collect 64 sequential frames of data.  Take the average of those frames for each pixel to obtain a new array which represents the response at 20 °C.  The units of this measurement are A/D counts.

For the second dataset, change the blackbody temperature to 25 °C.  Collect 128 sequential frames, take the standard deviation of each pixel’s values over the 128 frames, and place these standard deviations into an array which represents the temporal noise.  The units of measurement are A/D counts.

For the third dataset, set the blackbody to 30 °C and, as was done for the first dataset, collect 64 sequential frames of data.  Take the average of those frames for each pixel to obtain a new array which represents the response at 30 °C.  The units of this measurement are again, A/D counts.

Subtracting the 20 °C response array from the 30 °C response array and then dividing by 10 gives a responsivity array with units of counts per degree.  Dividing the temporal noise array by the responsivity array produces an NEDT array with units of Kelvins.  Taking the average of all the pixels in this array (neglecting bad pixels) and multiplying by 1000 gives the figure of merit in milliKelvins.