Dark Current
Introduction
The sensitivity of a scientific camera is vital, with insufficient sensitivity it may not even be possible to acquire clear images of your sample. At Teledyne Photometrics sensitivity is paramount and our approach to highly sensitive cameras is twofold: maximize signal collection, and minimize noise levels. While the former is dependent mostly on QE and pixel size, the latter is a much larger subject.
Noise on scientific cameras is error that has a number of sources, but can largely be categorized into two categories:
- Random noise, such as read noise, dark current, and photon shot noise. These are not constant from frame to frame and can be described by statistical distributions, and reduced by averaging successive frames.
- Pattern noise, caused by small differences in specific sensor pixels resulting in a 'fixed' pattern of brighter or darker pixels.
This article focuses on dark current noise, as seen in Fig.1.
Figure 1: Thermal build-up and dark current noise at the edges of a camera sensor. This noise will spread across the camera as the heat increases over long exposures and will impact image quality.
Dark Current
Dark current arises from thermal energy within the camera sensor. As a camera sensor is exposing an image, the electronics in the camera will heat the sensor, and this accumulation of thermal energy causes thermal electrons to build up on the sensor. These thermal electrons are independent of the photoelectrons generated proportional to the photons (light intensity) falling on the sensor. Unfortunately, the camera sensor doesn't know the difference between these types of electrons, and so any thermal electrons that accumulated in the sensor pixel wells (along with the photoelectrons) are counted as signal upon readout, despite not being part of the signal from the sample. This is known as the dark current.
Every model of scientific camera, whether using a CCD, EMCCD, or CMOS sensor, will have a dark current specification. This indicates how many electrons build up on each pixel for every second of exposure, typically shown as e–/p/s. The higher the dark current, the less able a camera is to perform long-exposure imaging.
However, as dark current is dependent on the temperature its effects can be reduced using cooling. Scientific cameras can use thermoelectric (TE) or Peltier cooling in combination with forced air or liquid cooling in order to reduce the temperature of the sensor during operation, as seen in Fig.2. In general, for every 6-7 °C the sensor can be cooled, the effects of dark current are halved. This is why scientific cameras often feature fans and will state their operating temperature in the datasheet, in order to reduce the effects of noise sources such as dark current. Ideally, the dark current should be reduced to a point where its contribution is negligible over a typical exposure time.
Figure 2: Scientific camera cooling and Citadel Chamber Technology. The left image shows typical temperature levels within a CMOS camera with Citadel chamber design principles (logo right). The sensor is cooled absolutely and evenly, is an optimal distance from the window, and heat is effectively expelled.
But while dark current is temperature dependent, it is also dependent on the exposure time (the 'second' part in e/p/s). For example, the typical dark current of a Teledyne Photometrics Prime BSI sCMOS is 0.5 e–/p/s, meaning that a two second exposure would result in one electron of dark current per pixel. The important part to consider here is: How long are your exposure times? Typical fluorescence imaging, low-light imaging, and high-speed imaging all feature low signals and short exposure times, well below a second and often below 100 ms. When using a highly-sensitive camera, a sufficient signal can be collected in very short exposures (well below 1 second), and the dark current buildup is negligible in these cases.
With modern camera technologies, aggressive and excessive cooling is unnecessary due to the short exposure times. This is why Teledyne Photometrics can make cameras such as the Prime BSI Express (no liquid cooling) and the Moment (uncooled) that can afford to move away from cooling due to their high sensitivity and low exposure times, and not experience issues with dark current noise unless exposures of 2 seconds or longer are used.
Dark Current Noise and Hot Pixels
While scientific cameras indicate a dark current specification on their datasheet, this can vary across the sensor. The stated dark current specification is an ensemble average of the entire sensor, and there is both a variation in the dark current known as dark current noise, and the presence of pixels with greater-than-average dark current known as hot pixels.
Dark current noise is the statistical variation of dark current across the sensor. For instance, a given camera might have a dark current specification of 1.0 e–/p/s. For a four second exposure, a total of four electrons/pixel are generated. Since dark current noise follows Poisson statistics, the root mean square (RMS) dark current noise is the square root of the dark current or, in this case, 2 e–/p.
Hot pixels have a higher-than-average dark current than the rest of the sensor. While making up a very small minority, these hot pixels will repeatedly have higher backgrounds values than other pixels. Since this is an effect that arises from the sensor manufacturing process, each hot pixel location will remain fixed and can therefore be corrected, with Teledyne Photometrics cameras having customisable pixel despeckle filters to avoid any effect from hot pixels.
CCD/EMCCD and CMOS
One factor to note when it comes to dark current is that typical CCD/EMCCD have far lower dark current values than typical CMOS. For example, the Prime 95B sCMOS has a dark current of 0.55 e–/p/s (when air-cooled, if using liquid cooling this value is 0.3 e–/p/s) compared to the Retiga R6 CCD with a dark current of 0.00073 e–/p/s. This much lower dark current has historically made CCDs popular for long-exposure applications, such as imaging luminescence with exposure times in minutes to hours.
However, CMOS technologies have also developed to a point where long-exposure imaging is achievable, thanks to the Retiga E7 CMOS from Teledyne Photometrics. The Retiga E7 has a dark current of less than 0.001 e–/p/s and is capable of exposures of over an hour, along with numerous other benefits of CMOS technology over CCD, such as increased speeds, sensor sizes and ease of use.
Summary
Dark current is a time- and temperature-dependent factor that can cause the build-up of noise on the camera sensor. The longer the exposure the greater the contribution of dark current, and this factor is why cameras often come equipped with cooling options and hot pixel corrections.
While dark currents were historically lower with CCD technologies, the advent of the Retiga E7 CMOS with ultra-low dark current has brought long-exposure imaging into the CMOS era!