Camera Sensitivity Explained - Jai

20 downloads 206 Views 2MB Size Report
May 30, 2007 - CCDs produce image information by converting light (photons) into electrical current (electrons). The sen
Tech Note NO. TN-0701

DATE: 5/30/07

CAMERA SENSITIVITY EXPLAINED One of the questions often asked by users of CCD cameras is “how sensitive is this camera?” Usually, the person asking the question is hoping for an answer similar to that found in photographic film, where a film with a higher ISO number is known to be more light sensitive than a film with a lower rating (e.g., ISO 200 vs. ISO 64). Unfortunately, in CCD cameras, where the “film” is replaced by an electronic sensor, the question of sensitivity is not as straightforward. CCD camera datasheets typically show several specifications related to sensitivity, including minimum illumination and spectral response, yet even these do not directly translate into a simple measure of camera sensitivity. To understand why, requires a short discussion of CCD physics. CCDs produce image information by converting light (photons) into electrical current (electrons). The sensor elements responsible for making this conversion are called pixels, which consist of a photosensitive diode and some surrounding circuitry used to transfer the electrical signal from the pixel to the output registers. The main CCD issues affecting overall camera sensitivity are fill factor, quantum efficiency (QE), and charge conversion. Fill factor is a measure of how much of the light being directed at the sensor actually strikes the photosensitive diodes in each pixel. Frame transfer CCDs have architectures where nearly all of the pixel area is photosensitive. Thus, fill factors for these CCDs can approach 100 percent. On the other hand, interline transfer (IT) CCDs, which are found in most industrial cameras, must divide the pixel area between photodiodes, transfer gates and shift register circuitry (see Fig. 1a). While this enables these CCDs to perform better in typical machine vision applications that require high-speed output, it reduces the photosensitive area of each pixel to less than 30 percent. To compensate for this, interline transfer CCDs typically have small lenslets, or micro lenses, positioned above each pixel to gather and focus as much light as possible on the photodiode portion of the pixel. Through the use of micro lenses, IT CCDs are able to reach effective fill factors of 60 to 70 percent (see Fig. 1b).

Light

Photodiode Vertical Shift Register (shielded)

Transfer Gate (shielded)

PIXEL

Figure 1a: IT CCD pixel with no micro lens results in low fill factor.

Tech Note Light

Microlens

Photodiode Vertical Shift Register (shielded)

Transfer Gate (shielded)

PIXEL

Figure 1b: Micro lens concentrates 60-70% of the light on photo sensitive area.

While fill factor rates how much of the light coming from the scene is available for imaging by the pixels, quantum efficiency rates how well the pixels convert that light into stored electrons. Specifically, QE is the ratio of photons striking the pixel to the number of electrons dislodged to create the signal in the pixel well. So, if an imager has a QE rating of 33% at a particular wavelength, it indicates that for every three photons striking a photodiode, one electron will be created in the pixel well (see Fig. 2).

Photons

- ----- -- -- -- -

Figure 2:

Quantum efficiency (QE) is the ratio of photons striking the pixel to the number of electrons created. A QE rating of 33% means that one electron is created for every three photons striking the pixel.

This QE ratio varies at different wavelengths, and is what is represented by spectral response charts where the Y-axis is labeled “absolute quantum efficiency” or “absolute response” (see Fig. 3). However, not all spectral response graphs are created equally. On some, the Y-axis is labeled “relative response” or “relative sensitivity.” These indicate how the camera/sensor responds to different parts of the spectrum, visible or otherwise, but cannot be used to directly compare the sensitivity of one camera or imager to another. This is because the maximum value (1.0) on one relative response graph, might represent a significantly different QE number than on another

NO. TN-0701 pg 2

Tech Note relative response graph. Furthermore, the relative response measurement might also include the final parameter of our sensitivity trio – charge conversion. 1.0

0.45

0.9

0.40

0.8

Relative Response

Absolute Quantum Efficiency

0.50

0.35 0.30 0.25 0.20 0.15

0.7 0.6 0.5 0.4 0.3

0.10

0.2

0.05

0.1

0.00 300

400

500

600

700

800

900

1000

0 400

500

600

700

800

900

1000

Wavelength [nm]

Wavelength (nm)

Figure 3:

Spectral response charts show how the QE of a sensor or camera varies across different wavelengths. The chart on the left shows “absolute” response, with the Y-axis indicating the actual QE of the sensor. The chart on the right shows “relative” response. Its shape can be directly compared against other sensors, but its amplitude can’t.

Charge conversion (also referred to as pixel sensitivity) is a measure of how much voltage is created from each electron in the pixel well (see Fig. 4). If we combine the charge conversion number with the QE number (and hold the fill factor constant), we should be able to get a good baseline for our sensitivity comparison. Unfortunately, while Kodak states what the charge conversion factor is for their imagers, Sony does not. The newer Kodak imagers have a stated charge conversion rate of approximately 31 µV per electron. Sony doesn’t list this information, but calculations indicate that for the newer Sony imagers, 8-10 µV per electron is typical.

100 Electrons

--- ---------------- ------------ -- ----------- --

Charge Conversion 31 µV per electron

3.1 mV

200 Electrons

--- ---------------- -------------- --- - - ----------- ------------ -- ----------- -Figure 4:

Charge Conversion 10 µV per electron

2.0 mV

Charge conversion (pixel sensitivity) measures the voltage produced by each electron in the pixel well. In this example, a higher charge conversion enables a pixel with a lower QE to produce more signal than a pixel with a higher QE but lower charge conversion. NO. TN-0701 pg 3

Tech Note Ultimately, sensitivity is a measure of what minimum amount of light is required for a camera/ sensor to produce an output level where the image (signal) is readily distinguishable from the noise of the sensor and camera. Sensors with a good combination of fill factor, QE, and charge conversion can reach this output level in relatively low light. Sensors and cameras that are also very “quiet” (high signal-to-noise ratio), can produce a meaningful output level with even less light because more gain can be applied to the signal without producing an unacceptably noisy image. So, if sensitivity is a complex result of fill factor, QE, charge conversion, and system noise, is there an easy way to tell which camera is the most sensitive? The answer is no, which is why it is important not to simply assume that an imager with a higher QE is a more sensitive sensor. Just as a photographer’s choice of film is based on grain, color saturation, contrast, and other factors in addition to ISO film speed, selection of a camera/sensor combination should be based on multiple factors beyond sensitivity. JAI, like many camera manufacturers, provides several measures of sensitivity on its camera datasheets, including spectral response graphs and pixel sensitivity values (charge conversion), where available. JAI also lists a light value, measured in lux, that represents the minimum illumination required to produce a discernable image. But use caution. These minimum illumination figures are measured at the typical frame rate of the camera. Cameras with higher frame rates will, naturally, require more light to produce an acceptable image than a camera running at a slower frame rate. This does not mean that the camera is less sensitive. A 60 fps camera with a minimum illumination rating of 1.0 lux is equally sensitive as a 15 fps camera with a minimum illumination rating of .25 lux. While these numbers provide some assistance in comparing cameras from a single vendor like JAI, different measurement methods from vendor to vendor make it difficult to do cross-vendor comparisons. An effort is currently underway by the European Machine Vision Association to develop a standardized measure of camera sensitivity which can be applied across different camera vendors. Once completed and approved, this “EMVA 1288 Standard” will define a unified method to measure, compute and present specification parameters for cameras and image sensors used for machine vision applications. Until then, it is best to carefully evaluate cameras against the requirements of your application to see which has the best combination of sensitivity, resolution, speed, image fidelity, and other features. For more information about camera sensitivity, or to discuss your particular requirements, please contact JAI.

NO. TN-0701 pg 4