Traditional CCDs vs. EM CCDs
Traditional Multi-port CCDs are better than EM (Electron Multiplying) CCDs due to the following issues:
Increased Shot Noise
EM CCDs are excellent for photon counting since the readnoise decreases to zero as the signal decreases to single electrons. However, when electron multiplication is turned on, the random multiplication process in the serial register increases the shot noise by at least sqrt(2).
In EM CCDs, a phenomenon known as spurious charge leads to noise spikes, which manifest as noisy pixels. There can be several to many noisy pixels per frame due to spurious charge.
EM CCDs use high speed clocks in the serial register, one of which uses a very large voltage swing, which determines the gain. Any small variation in the voltage or timing of this clock signal will lead to variations in the gain of the output register, affecting uniformity within and between image frames.
High Unity-Gain Readnoise
EM CCDs tend to have few outputs due to the complexity and size of the electron-multiplying serial register and the great difficulty of matching the output characteristics of these very sensitive circuits. This means that cameras that use EM CCDs must use much higher pixel rates than cameras that use traditional multi-port CCDs. EM CCD cameras can be run in unity gain mode to avoid the shot noise penalty, but the typical readnoise of EM CCD cameras using unity gain mode can be 70 e- or more. Traditional CCDs can have a readnoise of just a few electrons at the same frame rates.
CCDs in the Real World
Almost all real world applications are not in the photon counting regime. There is usually some kind of background light level and the signal of interest must be visible above the background signal. A comparison between traditional and EM CCDs is shown below. This shows a real world example of a Genetically Encoded Voltage Indicator (GEVI) known as ArcLight which was reported in 2013. If the investigator had been using an EM CCD camera instead of a SciMeasure CCD39 camera (RedShirtImaging NeuroCCD-SMQ), ArcLight would not have been discovered as a viable probe.