• SONY PREGIUS TECHNOLOGY
• WHAT IS THE EMVA1288 STANDARD?
• EMVA1288 MEASUREMENT EXPLAINED
• APPARENT SENSITIVITY
In 2014 Sony introduced a new global shutter (GS) CMOS imaging device, featuring a technology called PregiusTM, that redefines the GS CMOS imaging category and provides the best of both worlds with fast frame rate and CCD-like imaging performance. Sony’s first Pregius sensor, the IMX174 raises the bar on CMOS imaging quality and in some applications is a viable alternative to CCD sensors. This white paper will explore IMX174’s imaging performance through the use of EMVA1288 measurement comparing the IMX174 against a popular CCD, the ICX274.
Before the IMX174’s imaging performance is discussed, it’s important to understand what EMVA1288 is. The EMVA1288 standard was put together by the European Machine Vision Association to develop a unified and meaningful method of measuring a camera’s imaging performance. It is different from consumer cameras which are often measured in lux.
Lux is a measurement of intensity as perceived by the human eye. It is modelled using the response of the human eye and may not be representative of how a machine would recognize an image. In addition, the lux value of a camera represents the minimum illumination the camera requires to capture an acceptable image. Not only is the definition of an “acceptable image” subjective, it doesn’t provide any information on image noise.
Instead of lux, the EMVA1288 uses metrics such as read noise and full well depth to describe a camera’s performance. Each measurement is characterized using a standardized method defined by the EMVA1288 standard, providing an objective performance comparison between different cameras from different vendors. For this whitepaper, we will compare the 2.3 megapixel CMOS IMX174 with another 2 megapixel CCD, the ICX274.
Quantum efficiency (QE) is a measurement of the sensor’s ability to convert photons to electrons. A sensor with higher quantum efficiency is better for low light applications due to better conversion efficiency. The quantum efficiency for a given sensor is influenced by its photodiode design and will vary across the light spectrum.
When comparing the two QE curve in the above graph, the IMX174 has equal or better quantum efficiency than the ICX274 across most of the spectrum. At a wavelength of 525nm, the IMX174 is 17% more efficient at converting photons to electrons. But what does this mean in terms of real world performance?
If two sensors had the same pixel size and saturation capacity, and sensor A has 17% higher QE at a particular wavelength than sensor B, then sensor A is more sensitive and will require 17% less light to achieve the same image intensity. This means less illumination is required to achieve the same result. However, since IMX174 and ICX274 have different pixel size and saturation capacity, the same conclusion cannot be drawn based on the quantum efficiency alone.
Temporal dark noise (read noise) is noise generated by the sensor and camera circuitry and is influenced by the electrical design. Temporal dark noise can be amplified when the camera gain is increased, degrading image quality as a result. A low temporal dark noise allows for more signal gain without sacrificing image quality. Historically, CCD sensors have much lower temporal dark noise when compared to CMOS sensors. However, IMX174’s Pregius design features more accurate signal measurement technology (see Sony Pregius Technology section below), enabling the sensor to achieve a low temporal dark noise of 6.83 e-.
Sensors with a larger light sensitive area (larger pixel size) will be exposed to more incoming photons and generate more charge, leading to a higher saturation capacity. The IMX174 sensor has a pixel size that is 1.75 times larger than the ICX274; however, it has a saturation capacity that is 4 times larger due to the improved pixel design. But what does this mean in terms of imaging? It means an analog to digital converter will be able to convert the electrons in to more grey levels, resulting in a higher dynamic range in the captured image.
Applications where it’s important to recognize details in both dark and bright areas such as license plate recognition will benefit greatly from the higher dynamic range. Bright vehicle headlights will typically drive the camera to reduce exposure time, resulting in the license plate being too dark to be recognized. A high dynamic range camera will be able to produce enough detail in the darker areas for the license plate to be recognized.
A common misconception is that a higher sensitivity camera should yield a brighter image for the same exposure time when compared to a low sensitivity camera. This method of comparison ignores the difference in saturation capacity and temporal dark noise between two cameras. To understand how this works, we can use a bucket analogy to explain what happens when a pixel is exposed to incoming light.
Sensor pixels can be viewed as buckets catching rainfall (photons). A larger bucket will have a larger volume (saturation capacity). The volume of rain collected will be proportional to the image intensity. An empty bucket represents a black image while a full bucket represents a white image. If the rate of rainfall is constant (constant exposure and illumination), a small bucket will fill up much quicker than a big bucket, resulting in higher image intensity.
When evaluating camera sensitivity based on image brightness, the camera with smaller saturation capacity will typically appear brighter when compared to a camera with larger saturation capacity. This ignores the benefit that large saturation capacity brings, which is higher dynamic range. For applications that are only detecting whether an object is present and do not require a high dynamic range image, evaluating apparent sensitivity alone is a relevant way to compare. To improve apparent sensitivity, cameras with low temporal dark noise are excellent choices, enabling the use of camera gain to increase image brightness without sacrificing image quality.
A technique called Correlated Double Sampling (CDS) is a common method used by CCD and CMOS cameras to reduce temporal dark noise. In CMOS sensors, this can be done on both the analog and digital level to maximize noise reduction. Sony’s Exmor line of rolling shutter CMOS sensors performs both analog and digital CDS. With global shutter CMOS sensors, an additional storage element next to the pixel is required in order to support analog CDS. However, this additional storage element may reduce the surface area of the photodiode, reducing saturation capacity.
Sony has overcome this challenge by introducing a new global shutter CMOS pixel technology called Pregius, first featured in the Sony IMX174 CMOS sensor. The technology integrates an analog memory storage element as part of the pixel design. After integration, charges are shifted from the photodiode to the analog memory and CDS is applied, reducing temporal dark noise. Sony has leveraged their expertise in CCD pixel design to ensure the additional analog memory does not reduce the pixel’s saturation capacity.
As demonstrated by the Sony IMX174, the Pregius technology delivers impressive imaging performance, exceeding other global shutter CMOS sensor in its category. A comparison with other global shutter CMOS sensors can be seen below.
|Resolution||1920 x 1200||2048 x 2048||1280 x 1024||1280 x 1024|
|Pixel Size||5.86 µm||5.5 µm||4.8 µm||5.3 µm|
|Temporal Dark Noise||7 e-||16 e-||26 e-||25 e-|
|Saturation Capacity||32,691 e-||7,620 e-||10,226 e-||7,507 e-|
|Dynamic Range||73 dB||52 dB||51 dB||49 dB|
|Quantum Efficiency @ 525 nm||77 %||53 %||61 %||61 %|
The EMVA1288 standard not only provides users an objective way for customers to compare different sensors and cameras, it also helps them evaluate new sensor technology such as Sony Pregius and understand how the technology can help them.
Understanding what each measurement represents will help with camera selection and proper evaluation of camera and sensor performance.