The 3T pixel is the minimum configuration for a CMOS pixel. It has the advantage of circuit simplicity and allows the area of the photodiode structure  within each pixel to be maximised. Since transistors emit light, a 3T structure also minimises the dark current of the detector. One disadvantage of 3T is that the bandwidth of the external electronics needs  to extend down to DC. This makes the imaging system vulnerable to electronic drifts, mains interference and "1/f" noise. Additionally, the capacitance of the photodiode contributes to the measurement node capacitance (the parasitic capacitance shown in the above diagrams) and keeps the charge to voltage conversion gain relatively low. In the 4T pixel an additional transistor switch, known as the transfer gate, is introduced between the photodiode and the measurement node. Charge is transferred onto the node only at the end of the exposure. The peripheral elements of the video processing chain can then be  AC-coupled in order to reject the low frequency noise sources to which the 3T pixel is so vulnerable. A further advantage is that the measurement node does not see the photodiode capacitance due the shielding effect of the transfer gate.  High conversion gains are possible and CCD-like read noise levels can be achieved.

Key points :

• A reverse biased photodiode behaves like a light sensitive capacitor. Illumination causes the capacitor to discharge.
• The transistors within the pixel are used as "on/off" switches except for the  buffer transistor which is used as a unity-gain linear amplifier.
• The "measurement node" is a tiny capacitance that transforms a signal charge into a signal voltage. It is not necessarily a discrete structure, rather, it can consist of "stray" capacitances of other structures such as the gate of the linear buffer or the photodiode itself.
• The transfer gate does not just share charge between photodiode and measurement node, it causes total transfer of charge onto the node. For this to work the photodiode must be "pinned".
• The voltage signal output by the pixel is inversely proportional to the node capacitance. Lowest-noise performance can therefore be achieved by minimising the node capacitance. The transfer gate achieves this by effectively shielding the photodiode capacitance from the measurement node.