Overcoming the 887 barrier for the SNR in an 8-bit A/D
Rasika Rajapakshe, Shlomo Shalev · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1994
We have investigated the number of bits necessary for digitizing fluoroscopic images. An A/D with sufficient bits must be used to prevent degradation of the Signal-to-Noise Ratio (SNR) by quantization noise. The maximum achievable SNR with a 8-bit A/D is 887 assuming that the input signal probability density distribution is constant within a quantization bin width. In special cases of low light level video imaging where images are quasi-stationary (for example in electronic portal imaging), one can increase the inherent image SNR either by integrating the image on the camera target for extended periods or by averaging a number of real time (30 frames per second) images into a frame buffer. An A/D with higher number of bits will be required in the first case since the SNR of the analog signal is increased due to target integration. The same SNR can be achieved with a relatively low resolution A/D by averaging an equivalent number of frames if one neglects readout noise. We have used theoretical reasoning and experimental measurement with real time portal imaging to show that one can achieve an SNR higher than 887 with an 8 bit A/D with frame averaging if the input analog signal has certain amount of random noise.