When Sony announced the discontinuation of its CCD sensors for 2026, this information raised many questions. After all, users in the field of industrial image processing have appreciated CCD sensors for many years - primarily because of their high image quality and good global shutter functionality. The fact that Sony, as one of the leading sensor manufacturers, is focusing on CMOS sensor technology could indicate that this technology has now been developed to such an extent that the majority of applications can be operated with CMOS sensors. But can CCD sensors generally be replaced by CMOS sensors in the future without any restrictions?

Silicon quantum detectors

CCD and CMOS sensors are quantum detectors. Both technologies are based on the semiconductor material silicon and are therefore sensitive in the same spectral range from approx. 300 to 1000 nm. They differ primarily in the point at which the charge is converted into voltage on the semiconductor element: On a CCD sensor, vertical and horizontal charge transport takes place first. The serial charge/voltage conversion of all pixels takes place outside the sensor in the camera electronics. All pixel charges are converted into an analog voltage via an output outside the sensor.

With CMOS sensors, on the other hand, the charge/voltage conversion takes place in each pixel of the sensor. Depending on the activated line, the signal is amplified, noise-minimized and digitized via the readout circuit and finally transmitted in parallel via a configurable number of LVDS lines (Low Voltage Differential Signaling). This results in significant differences in image quality, resolution and refresh rate.

Old iron CCD sensors?

In a CCD sensor, all pixel charges are converted into an analogue voltage via an output, amplified and digitized. This results in high pixel homogeneity, a very uniform signal with low spatial noise (fixed pattern noise) and typically low dark current and therefore high image quality.

In addition, CCDs achieve high sensitivity and good signal quality at low light intensities due to the higher fill factor (ratio of the photosensitive area to the total pixel area).

Another advantage is the perfect global shutter, i.e. the simultaneous exposure of all pixels. For this reason, CCDs are very well suited for machine vision applications, particularly for applications that require very short exposure times.

However, one disadvantage of CCD sensors is the limited readout speed of the serial data stream. Modern CCDs with higher resolutions are therefore often manufactured using multi-tap technologies (division of the sensor into several image areas) in order to achieve n times the readout speed of single-tap sensors. However, a signal adjustment of the taps is then necessary, as even very small deviations lead to visible differences at the limits of the taps.

Principle of the interline transfer CCD and CMOS sensor

© Allied Vision Technologies

Another shortcoming of CCD sensors is that charges greater than the full-well capacity of the pixel cell reach neighboring pixels. This is visible in the typical blooming effects. This can only be remedied by reducing the amount of incident light. In addition, during readout, incident photons can generate additional charge carriers during serial charge transport in the vertical shift register - this is known as smearing. This smearing can be prevented by using a mechanical shutter in front of the sensor or by using flash illumination.

With a CMOS sensor, the charge/voltage conversion takes place at each pixel and the image information is already converted into digital information on the CMOS sensor chip. This mode of operation requires increased design complexity. CMOS designs with global shutter and CDS (correlated double sampling to reduce spatial noise) are based on pixel cells with five to eight transistors and thus reduce the light-sensitive area per pixel. Each column or pixel has an amplifier that works independently of the others. Due to technological deviations, there is a lack of uniformity between the pixels of the individual columns, which in turn leads to increased spatial noise.

Objects in motion

For applications with moving objects, a global shutter function is required for the sensors. This requires a memory area on the pixel in the CMOS sensor that is as well shielded from light as possible. In practice, this is not the case, especially with older CMOS designs - the area is more or less sensitive to light and shows a parasitic light sensitivity during the readout of the pixel data. Especially with short exposure times in the microsecond range, this is clearly visible as a vertical gray value gradient.