The SI502AB is scientific, slow scan, 512x512 element, 24 mm pixel back-illuminated imager and was used at room temperature for all of the resolution experiments.

Contrast transfer function was measured on both EBCCD tubes as a function of both acceleration voltage and light level. In addition, the ICCD was measured under identical illumination conditions so that a comparison of CTF could be made.

The improvement in the contrast transfer function of the EBCCD over the ICCD is apparent at all light levels. This was also obvious in the visual display quality of the test pattern images.

A difference frame of two identical flat field frames at a 'high' light level was used for the measurement of variance versus mean in the EBCCD tube. Because the EBCCD has nearly noiseless gain, at very low spatial frequencies the ratio of the variance (in electrons) to the mean (in electrons) is just the EBS gain. However, at higher spatial frequencies, charge spreading results in correlation of noise, reducing variance. Figure 7 shows the results of binning various numbers of pixels in software before calculating the variance to mean ratio. The effect was modeled using a simple gaussian spread of standard deviation 13 mm (pixel size is 24 mm)

Figure 7. Ratio of variance to mean versus number of pixels binned into 'super pixels' showing the dependence of signal to noise ratio on MTF.

The high performance of the experimental EBCCD tubes reinforces the long held belief in the advantages of direct electron imaging CCDs. What remains is the development of a video rate EBCCD coupled to a GaAs photocathode to provide perhaps the ultimate in low light level imaging: single photon detection and high speed readout.