Types Of Camera Sensor --- 相机传感器的类型

https://www.photometrics.com/learn/camera-basics/types-of-camera-sensor

Introduction

Quantitative scientific cameras are vital for sensitive, fast imaging of a variety of samples for a variety of applications. Camera technologies have advanced over time, from the earliest cameras to truly modern camera technologies, which can push the envelope of what is possible in scientific imaging and allow us to see the previously unseen.

The heart of the camera is the sensor, and the steps involved in generating an image from photons to electrons to grey levels. For information on how an image is made, see our article of the same name. This article discusses the different camera sensor types and their specifications, including:

• Charge-coupled device (CCD)电荷耦合器件
• Electron-multiplying charge-coupled device (EMCCD)电子倍增电荷耦合器件
• Complementary metal-oxide-semiconductor (CMOS)互补金属氧化物半导体
• Back-illuminated CMOS背照式CMOS

This order also shows the chronological order of the introduction of these sensor types, we will go through these one at a time, in a journey through the history of scientific imaging.

Sensor Fundamentals

The first step for a sensor is the conversion of photons of light into electrons (known as photoelectrons). The efficiency of this conversion is known as the quantum efficiency (QE) and is shown as a percentage.

传感器的第一步是将光的光子转换为电子（称为光电子）。这种转换的效率称为量子效率 （QE），以百分比表示。

All the sensor types discussed here operate based on the fact that all electrons have a negative charge (the electron symbol being e–). This means that electrons can be attracted using a positive voltage, granting the ability to move electrons around a sensor by applying a voltage to certain areas of the sensor, as seen in Figure 1.

这里讨论的所有传感器类型都基于所有电子都带负电荷的事实来工作（电子符号是 e – ）。这意味着可以使用正电压吸引电子，从而通过向传感器的某些区域施加电压来使电子在传感器周围移动，如图 1 所示。

image-38.png

Figure 1: How electron charge is transferred from pixel to pixel across a sensor. Photons (black arrows) hit a pixel (blue squares) and are converted into electrons (e–) and are stored in a pixel well (yellow). These electrons can be transferred to another pixel using a positive voltage (orange), and moved anywhere on a sensor, pixel by pixel. 图 1：电子电荷如何在传感器上从一个像素传递到另一个像素。光子（黑色箭头）击中像素（蓝色方块）并转化为电子（e – ）并存储在像素井（黄色）中。这些电子可以使用正电压（橙色）转移到另一个像素，并逐个像素地移动到传感器上的任何位置。【正电荷吸引电子？】

In this manner, electrons can be moved anywhere on a sensor, and are typically moved to an area where they can be amplified and converted into a digital signal, in order to be displayed as an image. However, this process occurs differently in each type of camera sensor.

通过这种方式，电子可以移动到传感器上的任何位置，并且通常被移动到可以放大并转换为数字信号的区域，以便显示为图像。但是，此过程在每种类型的相机传感器中发生的情况不同。

CCD

CCDs were the first digital cameras, being available since the 1970s for scientific imaging. CCD have enjoyed active use for a number of decades and were well suited to high-light applications such as cell documentation or imaging fixed samples. However, this technology was lacking in terms of sensitivity and speed, limiting the available samples that could be imaged at acceptable levels. CCD 是第一台数码相机，自 1970 年代以来一直可用于科学成像。CCD几十年来一直被广泛使用，非常适合高光应用，如细胞记录或固定样品成像。然而，该技术在灵敏度和速度方面有所欠缺，限制了可以在可接受水平下成像的可用样品。

CCD Fundamentals

In a CCD, after exposure to light and conversion of photons to photoelectrons, the electrons are moved down the sensor row by row until they reach an area that isn’t exposed to light, the readout register. Once moved into the readout register, photoelectrons are moved off one by one into the output node. In this node they are amplified into a readable voltage, converted into a digital grey level using the analogue to digital converter (ADC) and sent to the computer via the imaging software. 在CCD中，在暴露于光并将光子转换为光电子后，电子在传感器上逐行向下移动，直到它们到达未暴露在光下的区域，即读出寄存器。一旦进入读出寄存器，光电子就会一个接一个地移动到输出节点中。在该节点中，它们被放大为可读电压，使用模数转换器（ADC）转换为数字灰度电平，并通过成像软件发送到计算机。

image-40.png

Figure 2: How a CCD sensor works. Photons hit a pixel and are converted to electrons, which are then shuttled down the sensor to the readout register, and then to the output node, where they are converted to a voltage, then grey levels, and then displayed with a PC. 图 2：CCD 传感器的工作原理。光子撞击一个像素并被转换为电子，然后电子被沿着传感器传送到读出寄存器，然后到输出节点，在那里它们被转换为电压，然后是灰度电平，然后用PC显示。

The number of electrons is linearly proportional to the number of photons, allowing the camera to be quantitative. The design seen in Fig.2 is known as a full-frame CCD sensor, but there are other designs known as frame-transfer CCD and interline-transfer CCD that are shown in Fig.3. 电子数与光子数成线性正比，使相机能够定量。图 2 中的设计称为全画幅 CCD 传感器，但还有其他设计称为帧传输 CCD 和行间传输 CCD，如图 3 所示。

image-41.png

这到底是个俯视图还是侧视图？！根据后面太阳图片，应该是俯视图

Figure 3: Different types of CCD sensor. The full-frame sensor is also displayed in Fig.2. Grey areas are masked and not exposed to light. The frame-transfer sensor has an active image array (white) and a masked storage array (grey), while the interline-transfer sensor has a portion of each pixel masked (grey). 图 3：不同类型的 CCD 传感器。全画幅传感器也如图 2 所示。灰色区域被遮盖，不暴露在光线下。帧传输传感器具有活动图像阵列（白色）和屏蔽存储阵列（灰色），而行间传输传感器具有每个像素的一部分屏蔽（灰色）。

In a frame-transfer CCD the sensor is divided into two: the image array (where light from the sample hits the sensor) and the storage array (where signal is temporarily stored before readout). The storage array is not exposed to light, so when electrons are moved to this array, a second image can be exposed on the image array while the first image is processed from the storage array. The advantage is that a frame-transfer sensor can operate at greater speeds than a full-frame sensor, but the sensor design is more complex and requires a larger sensor (to accommodate the storage array), or the sensor is smaller as a portion is made into a storage array. 在帧传输CCD中，传感器分为两个：图像阵列（来自样品的光照射到传感器）和存储阵列（在读出前临时存储信号）。存储阵列不会暴露在光线下，因此当电子移动到该阵列时，可以在图像阵列上曝光第二个图像，同时从存储阵列处理第一个图像。优点是帧传输传感器可以比全帧传感器以更高的速度运行，但传感器设计更复杂，需要更大的传感器（以容纳存储阵列），或者传感器更小，因为一部分被制成存储阵列。

For the interline-transfer CCD, a portion of each pixel is masked and not exposed to light. Upon exposure, the electron signal is shifted into this masked portion, and then sent to the readout register as normal. Similarly to the frame-transfer sensor, this helps increase the speed, as the exposed area can generate a new image while the original image is processed. However, each pixel in this sensor is smaller (as a portion is masked), and this decreases the sensitivity as fewer photons can be detected by smaller pixels. These sensors often come paired with microlenses to better direct light and improve the QE. 对于行间传输CCD，每个像素的一部分被遮蔽，不暴露在光线下。曝光后，电子信号被转移到这个掩蔽部分，然后像往常一样发送到读出寄存器。与帧传输传感器类似，这有助于提高速度，因为在处理原始图像时，曝光区域可以生成新图像。然而，该传感器中的每个像素都较小（因为一部分被遮蔽），这会降低灵敏度，因为较小的像素可以检测到的光子更少。这些传感器通常与微透镜配对，以更好地引导光线并改善 QE。

CCD Limitations

The main issues with CCDs are their lack of speed and sensitivity, making it a challenge to perform low-light imaging or to capture dynamic moving samples.

The lack of speed is due to several factors: CCD的主要问题是缺乏速度和灵敏度，这使得执行低光成像或捕获动态移动样品成为一项挑战。

• There is only one output node per sensor. This means that millions of pixels of signal have to be shuttled through one node, creating a bottleneck and slowing the camera. 每个传感器只有一个输出节点。这意味着数百万像素的信号必须通过一个节点传输，从而造成瓶颈并减慢相机的速度。
• If electrons are moved too quickly, it introduces error and read noise, so most CCDs elect to move electrons slower than maximum speed to attempt to reduce noise. 如果电子移动得太快，就会引入误差和读取噪声，因此大多数CCD选择以低于最大速度的速度移动电子，以试图降低噪声。
• The whole sensor needs to be cleared of the electron signal before the next frame can be exposed 在下一帧曝光之前，需要清除整个传感器的电子信号

Essentially, there are very few data readout channels for a CCD, meaning the data processing is slowed. Most CCDs operate at between 1-20 frames per second, as a CCD is a serial device and can only read the electron charge packets one at a time. Imagine a bucket brigade, where electrons can only be passed from area to area one at a time, or a theatre with only one exit but several million seats. 从本质上讲，CCD的数据读出通道很少，这意味着数据处理速度会变慢。大多数 CCD 以每秒 1-20 帧的速度运行，因为 CCD 是一种串行设备，一次只能读取一个电子电荷包。想象一下，一个水桶大队，电子一次只能从一个区域传递到另一个区域，或者一个只有一个出口但有几百万个座位的剧院。

In addition, CCDs have a small full-well capacity, meaning that the number of electrons that can be stored in each pixel is limited. If a pixel can only store 200 electrons, receiving a signal of >200 electrons leads to saturation, where a pixel becomes full and displayed the brightest signal, and blooming, where the pixel overflows and the excess signal is smeared down the sensor as the electrons are moved to the readout register. 此外，CCD的满阱容量很小，这意味着每个像素中可以存储的电子数量是有限的。如果一个像素只能存储 200 个电子，则接收 >200 个电子的信号会导致饱和，即像素变得饱满并显示最亮的信号，以及光晕，即像素溢出，多余的信号在电子移动到读出寄存器时涂抹在传感器上。

In extreme cases (such as daylight illumination of a scientific camera), there is a charge overload in the output node, causing the output amplification chain to collapse, resulting in a zero (completely dark) image. 在极端情况下（例如科学相机的日光照明），输出节点中存在电荷过载，导致输出放大链崩溃，从而产生零（完全黑暗）的图像。

Sat-and-Bloom.png

Figure 4: Examples of blooming caused by saturation of a CCD sensor pixel. Left) Picture of a sunset. The sun is so bright in the image that there is blooming on the sun itself, leaking into the surrounding pixels, and a vertical smear across the whole image. Right) A similar situation with the blooming and smear labeled. 图 4：CCD 传感器像素饱和引起的光晕示例。左）日落的图片。太阳在图像中是如此明亮，以至于太阳本身就会开花，泄漏到周围的像素中，并且整个图像上有垂直的涂抹。右）类似的情况，带有开花和涂抹标记。

CCD pixels are also typically quite small (such as ~4 µm) meaning that while these sensors can achieve a high resolution, they lack sensitivity, as a larger pixel can collect more photons. This limits signal collection and is compounded by the limited QE of front-illuminated CCDs, which often only reaches 75% at maximum. CCD像素通常也非常小（例如~4μm），这意味着虽然这些传感器可以实现高分辨率，但它们缺乏灵敏度，因为较大的像素可以收集更多的光子。这限制了信号收集，并且由于前照式CCD的QE有限而变得更加复杂，通常最大只能达到75%。

Finally, CCD sensors are typically quite small, with an 11-16 mm diagonal, which limits the field of view that can be displayed on the camera and means that not all of the information from the microscope can be captured by the camera.

最后，CCD传感器通常非常小，对角线为11-16毫米，这限制了相机上可以显示的视野，并意味着相机无法捕获显微镜的所有信息。

Overall, while CCDs were the first digital cameras, for scientific imaging purposes in the modern day they are lacking in speed, sensitivity and field of view.

EMCCD

EMCCDs first emerged onto the scientific imaging scene in 2000 with the Cascade 650 from Photometrics. EMCCDs offered faster and more sensitive imaging then CCDs, useful for low-light imaging or even photon counting.

EMCCDs achieved this in a number of ways. The cameras are back-illuminated (increasing the QE to ~90%) and have very large pixels (16-24 µm), both of which greatly increase the sensitivity. The most significant addition, however, is the EM in EMCCD: electron multiplication. EMCCD以多种方式实现了这一点。相机采用背照式（将QE提高到~90%），并具有非常大的像素（16-24μm），这两者都大大提高了灵敏度。然而，最重要的补充是EMCCD中的EM：电子倍增。

EMCCD Fundamentals

EMCCDs work in a very similar way to frame-transfer CCDs, where electrons move from the image array to the masked array, then onto the readout register. At this point the main difference emerges: the EM Gain register. EMCCDs use a process called impact ionisation to force extra electrons out of the silicon sensor, therefore multiplying the signal. This EM process occurs step-by-step, meaning users can choose a value between 1-1000 and have their signal be multiplied that many times in the EM Gain register. If an EMCCD detects a signal of 5 electrons and has an EM Gain set to 200, the final signal that goes into the output node will be 1000 electrons. This allows EMCCDs to detect extremely small signals, as they can be multiplied up above the noise floor as many times as a user desires. EMCCD的工作方式与帧传输CCD非常相似，其中电子从图像阵列移动到掩膜阵列，然后移动到读出寄存器上。此时，主要区别就出现了：EM增益寄存器。EMCCD使用一种称为冲击电离的过程来迫使额外的电子从硅传感器中排出，从而使信号成倍增加。此EM过程是逐步进行的，这意味着用户可以在1-1000之间选择一个值，并在EM增益寄存器中将其信号乘以多次。如果EMCCD检测到5个电子的信号并将EM增益设置为200，则进入输出节点的最终信号将是1000个电子。这使得EMCCD能够检测到极小的信号，因为它们可以根据用户的需要在噪声基底以上多次增加。

image-43.png

Figure 5: How an EMCCD sensor works. Photons hit a pixel and are converted to electrons, which are then shuttled down the sensor to the readout register. From here they are amplified using the EM Gain register, then sent to the output node, where they are converted to a voltage, then grey levels, and then displayed with a PC. 图 5：EMCCD 传感器的工作原理。光子击中一个像素并转化为电子，然后电子沿着传感器向下穿梭到读出寄存器。从这里开始，它们使用EM增益寄存器被放大，然后发送到输出节点，在那里它们被转换为电压，然后是灰度电平，然后用PC显示。

This combination of large pixels, back-illumination and electron multiplication makes EMCCDs extremely sensitive, far more so than CCDs. 这种大像素、背照和电子倍增的结合使EMCCD非常灵敏，远远超过CCD。

EMCCDs are also faster than CCDs. In CCDs, electrons are moved around the sensor at speeds well below the maximum possible speed, because the faster the electrons are shuttled about, the greater the read noise. Read noise is a fixed +/- value on every signal, if a CCD has a read noise of ±5 electrons and detects a signal of 10 electrons, it could be read out at anywhere between 5-15 electrons depending on the read noise. This has a big impact on sensitivity and speed, as CCDs move electrons slower in order to reduce read noise. However, with an EMCCD you can just multiply your signal up until the read noise has a negligible effect. This means that EMCCDs can move signal around at maximum speed, resulting in huge read noise values from 60-80 electrons, but signals are often multiplied by hundreds of times, meaning that the read noise impact is lessened. In this manner, EMCCDs can operate at much higher speeds than CCDs, achieving around 30-100 fps across the full-frame. This is only possible due to the EM Gain aspect of EMCCDs. EMCCD也比CCD更快。在CCD中，电子以远低于最大可能速度的速度在传感器周围移动，因为电子穿梭得越快，读取噪声就越大。读取噪声是每个信号的固定+/-值，如果CCD的读取噪声为±5个电子并检测到10个电子的信号，则可以根据读取噪声在5-15个电子之间的任何位置读出。这对灵敏度和速度有很大影响，因为CCD为了降低读取噪声而移动电子的速度较慢。但是，使用EMCCD，您只需将信号相乘，直到读取噪声的影响可以忽略不计。这意味着EMCCD可以以最大速度移动信号，从而产生来自60-80个电子的巨大读取噪声值，但信号通常会乘以数百倍，这意味着读取噪声的影响会降低。通过这种方式，EMCCD可以以比CCD高得多的速度运行，在全画幅上实现约30-100fps。这只有在EMCCD的EM增益方面才有可能。【提高电子速度导致提高噪声，但是信号倍乘远大于噪音】

EMCCD Limitations

Despite the advantages of electron multiplication, it introduces a lot of complexity to the camera and leads to several major downsides. The main technological issues are EM Gain Decay, EM Gain Stability and Excess Noise Factor. 尽管电子倍增具有优点，但它给相机带来了很多复杂性，并导致了几个主要缺点。主要技术问题是电磁增益衰减、电磁增益稳定性和过量噪声因数。

EM gain decay or ageing is a phenomenon that is not fully understood, but essentially involves charge building up in the silicon sensor between the EM electrode and photodetector. This build-up of charge reduces the effect of EM gain, hence EM gain decay. The greater the initial signal intensity and the higher the EM gain, the faster the EM gain will decay. Using an EM gain of 1000x on a large signal would quickly result in EM gain decay. This results in the EM gain not being the same each time, leading to a lack of reproducibility in experiments, limiting the usefulness of the camera as a quantitative imaging tool. EMCCDs essentially have limited lifespans and require regular calibration, leading to these cameras needing to be used in a certain way, limiting the EM gain that can be used in an experiment without damaging the camera. When a camera has been purchased and will be used daily in a research lab, it can be disappointing to learn that the camera will become less and less reliable over time. EM增益衰减或老化是一种尚不完全了解的现象，但本质上涉及EM电极和光电探测器之间的硅传感器中的电荷积聚。这种电荷的积累降低了EM增益的影响，从而降低了EM增益衰减。初始信号强度越大，EM增益越高，EM增益衰减得越快。在大信号上使用1000倍的EM增益会很快导致EM增益衰减。这导致每次的EM增益都不相同，导致实验中缺乏可重复性，限制了相机作为定量成像工具的实用性。EMCCD基本上具有有限的使用寿命，需要定期校准，导致这些相机需要以某种方式使用，从而限制了可以在实验中使用的EM增益，而不会损坏相机。当购买了相机并将在研究实验室中每天使用时，得知相机会随着时间的推移变得越来越不可靠，这可能会令人失望。

In addition, the EM gain process itself is not stable, different fluctuations can occur. One such example is EM gain being temperature-dependent, in order for EMCCDs to have reliable EM gain they typically operate at temperates from -60 ºC to -80 ºC, meaning they require extensive forced-air or liquid cooling. This all adds to the camera complexity and cost, especially if a liquid cooling rig needs to be installed with the camera. 此外，EM增益过程本身并不稳定，可能会发生不同的波动。一个这样的例子是EM增益与温度有关，为了使EMCCD具有可靠的EM增益，它们通常在-60ºC至-80ºC的温度下工作，这意味着它们需要大量的强制空气或液体冷却。这一切都增加了相机的复杂性和成本，特别是当相机需要安装液体冷却装置时。

While an EMCCD can multiply signal far above the reaches of read noise, these cameras are subject to other sources of noise, unique to EMCCDs. The number of photons a camera detects is not the same every second, as photons typically fall like rain rather than arrive at the sensor in regimented rows. This disparity between measurements is called photon shot noise. Photon shot noise and other sources of noise all exist in the signal as soon as it arrives on the sensor, and these noise sources are all multiplied up along with the signal, resulting in the Excess Noise Factor. The combination of random photon arrival and random EM multiplication leads to extra sources of error and noise, with all sources of noise (predominantly photon shot noise) being multiplied by a factor of 1.4x. While an EMCCD may eliminate read noise, it introduces its own sources of noise, impacting the signal-to-noise ratio and the ability of the camera to be sensitive. 虽然EMCCD可以增加远高于读取噪声范围的信号，但这些相机会受到EMCCD特有的其他噪声源的影响。相机每秒检测到的光子数量并不相同，因为光子通常像雨一样落下，而不是成排到达传感器。测量值之间的这种差异称为光子散粒噪声。光子散粒噪声和其他噪声源一旦到达传感器就存在于信号中，并且这些噪声源都随着信号相乘，从而产生过量噪声因子。随机光子到达和随机电磁倍增的组合会导致额外的误差和噪声源，所有噪声源（主要是光子散粒噪声）都乘以 1.4 倍。虽然EMCCD可以消除读取噪声，但它会引入自己的噪声源，从而影响信噪比和相机的灵敏度。

Finally, the large pixels of an EMCCD lead to these cameras having a lower resolution than CCDs; EMCCDs have a small field of view due to their small sensors; and even today (20 years later) EMCCDs are still the most expensive format of scientific camera. 最后，EMCCD的大像素导致这些相机的分辨率低于CCD;EMCCD由于传感器小，视野小;即使在今天（20 年后），EMCCD 仍然是最昂贵的科学相机格式。

While EMCCDs greatly improved on the speed and sensitivity of CCDs, they brought their own issues and continued to limit the amount of information that could be obtained from the microscope.

CMOS

While MOS and CMOS technology has existed since before CCD (~1950’s), only in 2009 did CMOS cameras become quantitative enough to be sufficient for scientific imaging, hence why CMOS cameras for science can be referred to as scientific CMOS or sCMOS. 虽然MOS和CMOS技术早在CCD之前就已经存在（~1950年代），但直到2009年，CMOS相机才变得足够量化，足以进行科学成像，因此用于科学的CMOS相机可以被称为科学CMOS或sCMOS。

CMOS technology is different to CCD and EMCCD, the main factor being parallelization, CMOS sensors operate in parallel and allow for much higher speeds. CMOS技术与CCD和EMCCD不同，主要因素是并行化，CMOS传感器并行运行并允许更高的速度。

CMOS Fundamentals

In a CMOS sensor there are miniaturized electronics on every single pixel, namely a capacitor and amplifier. This means that a photon is converted to an electron by the pixel, and then the electron is immediately converted to a readable voltage while still on the pixel. In addition, there is an ADC for every single column, meaning that each ADC has far less data to read out than a CCD/EMCCD ADC, which has to read out the entire sensor. This combination allows CMOS sensors to work in parallel, and process data much faster than CCD/EMCCD technologies. By moving electrons much slower than the potential max speed, CMOS sensors also have a much lower read noise than CCD/EMCCD, allowing them to perform low-light imaging and work with weak fluorescence or live cells.

在CMOS传感器中，每个像素上都有微型电子元件，即电容器和放大器。这意味着光子被像素转换为电子，然后电子在像素上立即转换为可读电压。此外，每列都有一个ADC，这意味着每个ADC要读出的数据比CCD/EMCCD ADC少得多，后者必须读出整个传感器。这种组合使CMOS传感器能够并行工作，并且处理数据的速度比CCD/EMCCD技术快得多。CMOS传感器的电子移动速度远低于潜在最大速度，因此读取噪声也比CCD/EMCCD低得多，从而可以进行弱光成像，并与弱荧光或活细胞一起工作。

image-44.png

Figure 6: How a CMOS sensor works. Photons hit a pixel and are converted to electrons, and then converted to voltage on the pixel. Each column is then read out separately by individual ADCs, and then displayed with a PC. 图 6：CMOS 传感器的工作原理。光子撞击像素并转化为电子，然后转换为像素上的电压。然后，每列由单独的ADC单独读出，然后用PC显示。

CMOS sensors have also been adopted by the commercial imaging industry, meaning that nearly every smartphone camera, digital camera, or imaging device uses a CMOS sensor. This makes these sensors easier and cheaper to manufacture, allowing sCMOS cameras to feature large sensors and have much larger fields of view than CCD/EMCCD, to the point where some sCMOS cameras can capture all the information from the microscope. CMOS传感器也被商业成像行业采用，这意味着几乎每个智能手机相机、数码相机或成像设备都使用CMOS传感器。这使得这些传感器的制造更容易、更便宜，使sCMOS相机能够配备大型传感器，并且具有比CCD/EMCCD大得多的视场，以至于一些sCMOS相机可以从显微镜捕获所有信息。

In addition, CMOS sensors had a large full well capacity, meaning they had a large dynamic range and could simultaneously image dark signals and bright signals, not subject to saturation or blooming like with a CCD. 此外，CMOS传感器具有较大的满阱容量，这意味着它们具有较大的动态范围，可以同时成像暗信号和亮信号，而不会像CCD那样受到饱和或光晕的影响。

Early CMOS Limitations

Early sCMOS cameras featured much higher speeds and larger fields of view than CCD/EMCCD, and with a range of pixel sizes, there were CMOS cameras that imaged at very high resolution, especially compared to EMCCD. However, the large pixel and electron multiplication of EMCCDs meant that early sCMOS cameras couldn’t rival EMCCD when it came to sensitivity. When it came to extreme low-light imaging or the need for sensitivity, EMCCD still had the edge. 早期的sCMOS相机比CCD/EMCCD具有更高的速度和更大的视场，并且具有一系列像素尺寸，因此CMOS相机可以以非常高的分辨率成像，尤其是与EMCCD相比。然而，EMCCD的大像素和电子倍增意味着早期的sCMOS相机在灵敏度方面无法与EMCCD相媲美。当涉及到极低光成像或需要灵敏度时，EMCCD仍然具有优势。

image-46.png

Figure 7: Camera sensitivity. While early CMOS was far more sensitive than CCDs due to lower read noise, early CMOS couldn’t compete with EMCCD and the near elimination of read noise. 7：相机灵敏度。虽然早期的CMOS由于读取噪声较低而比CCD更灵敏，但早期的CMOS无法与EMCCD竞争，并且几乎可以消除读取噪声。

image-47.png

Figure 8: Split sensor sCMOS patterns and artifacts. While early CMOS was far more sensitive than CCDs due to lower read noise, early CMOS couldn’t compete with EMCCD and the near elimination of read noise. 图 8：分离式传感器 sCMOS 图案和伪影。虽然早期的CMOS由于读取噪声较低而比CCD更灵敏，但早期的CMOS无法与EMCCD竞争，并且几乎可以消除读取噪声。

In Fig.8 we can see the bias of a split sensor camera, showing a horizontal line separating the two halves of the sensor, along with the other horizontal scrolling lines. This is due to each sensor half never being exactly the same due to noise and fluctuations. This effect is exacerbated when 100 image frames are averaged, as seen in the lower image. Here the sensor split is also clear, as are vertical columns across the image. This is fixed pattern column noise and is again due to the ADC pairs of the sensor. This noise can interfere with signal in low-light conditions. 在图 8 中，我们可以看到分体式传感器相机的偏置，显示了一条水平线将传感器的两半分开，以及其他水平滚动线。这是因为由于噪声和波动，每个传感器的一半永远不会完全相同。当平均 100 个图像帧时，这种效果会加剧，如下图所示。在这里，传感器的分裂也很清晰，图像上的垂直列也是如此。这是固定模式的列噪声，同样是由传感器的ADC对引起的。这种噪声会干扰弱光条件下的信号。

These early sCMOS sensors were front-illuminated and therefore had a limited QE (70-80%), further impacting their sensitivity. 这些早期的sCMOS传感器是前照式的，因此QE有限（70-80%），进一步影响了它们的灵敏度。

Some early sCMOS, in an effort to run at a higher speed, featured a split sensor, where each half of the sCMOS sensor had its own set of ADCs and the camera image at speeds up to 100 fps. However, this split caused patterns and artifacts in the camera bias, which would be clearly visible in low-light conditions and would interfere with the signal, as seen in Figure 8. 为了以更高的速度运行，一些早期的sCMOS采用了分离式传感器，其中sCMOS传感器的每一半都有自己的一组ADC，并且相机图像的速度高达100 fps。然而，这种分裂导致了相机偏置中的图案和伪影，在低光条件下清晰可见，并且会干扰信号，如图8所示。

This combination of front-illumination, split sensors, patterns/artifacts, and smaller pixels all led to early sCMOS lacking in sensitivity. 这种前照式、分离式传感器、图案/伪影和较小像素的组合都导致早期的sCMOS缺乏灵敏度。

Back-Illuminated sCMOS

In 2016 Photometrics released the first back-illuminated sCMOS camera, the Prime 95B. Back-illuminated (BI) sCMOS cameras greatly improve on sensitivity compared to early front-illuminated sCMOS, while retaining all the other CMOS advantages such as high speed, large field of view. The combination of a much higher QE due to back-illuminated (up to 95%, hence the name of the Prime 95B), the single sensor (no split), more varied pixel sizes, and a cleaner background, BI sCMOS is the all-in-one imaging solution. 2016年，Photometrics发布了第一款背照式sCMOS相机Prime 95B。与早期的前照式sCMOS相比，背照式（BI） sCMOS相机的灵敏度大大提高，同时保留了所有其他CMOS优势，如高速、大视场。BI sCMOS具有背照式（高达95%，因此得名Prime 95B）、单传感器（无分割）、更多样化的像素尺寸和更清晰的背景，因此具有更高的QE，是一体化成像解决方案。

BI sCMOS Fundamentals

Back-illumination allows for a large increase in camera QE across wavelengths from UV to IR, due to the way that light can access the camera sensor. Figure 9 highlights the differences between a front-illuminated and back-illuminated camera sensor. 背照允许在从紫外到红外的波长范围内大幅增加相机QE，因为光线可以进入相机传感器。图 9 突出显示了前照式和背照式摄像头传感器之间的差异。

image-48.png

Figure 9: Front-illumination vs back-illumination for camera sensors. Front-illuminated sensors (CCDs and early sCMOS) have the light come in from the front, where it passes through microlenses, wiring, electronics, and more before reaching the photodetector. Back-illuminated sensors (EMCCD and BI sCMOS) have a flipped sensor where the light comes in from the ‘back’, immediately reaching the photodetector. 图 9：相机传感器的前照式与背照式。前照式传感器（CCD和早期的sCMOS）让光线从正面进入，在到达光电探测器之前，通过微透镜、电线、电子设备等。背照式传感器（EMCCD 和 BI sCMOS）具有翻转传感器，其中光线从“背面”射入，立即到达光电探测器。

Every stage that light has to travel through will scatter some light, meaning that the QE of front-illuminated cameras is often limited from 50-80%, even with microlenses specifically to focus light onto each pixel. Due to the additional electronics of CMOS sensors (miniaturized capacitor and amplifier on each pixel), there can be even more scattering. 光线必须经过的每个阶段都会散射一些光，这意味着前照式相机的 QE 通常限制在 50-80% 之间，即使使用专门用于将光线聚焦到每个像素上的微透镜也是如此。由于CMOS传感器的附加电子元件（每个像素上的小型电容器和放大器），可能会有更多的散射。

By rotating the sensor and bringing the photodetector silicon layer to the front (from the ‘back’), light has less distance to travel and there is less scattering, resulting in a much higher QE of >95%. While back-illumination was achieved earlier with some CCDs and most EMCCDs, it took longer for CMOS due to the complex electronics involved, and the specific thickness of silicon required to capture different wavelengths of light. Either way, the result is a good 15-20% QE increase at peak, and a 10-15% QE increase out to >1000 nm, doubling the sensitivity in these regions. The lack of microlenses also unlocked a new QE region from 200-400, great for UV imaging.

BI sCMOS have a much greater signal collection ability than FI sCMOS due to the increase in QE and the elimination of patterns/artifacts with a clean background. Along with the low read noise, BI sCMOS is able to match and outperform EMCCD in sensitivity, as well as already featuring much higher speed, resolution, and larger field of view.

Summary

Scientific imaging technologies have continued to advance from CCD, to EMCCD, sCMOS, and back-illuminated sCMOS, in order to deliver the best speed, sensitivity, resolution, and field of view for your sample on your application. Choosing the most suitable camera technology for your imaging system can improve every aspect of your experiments and allow you to be quantitative in your research. While CCD and EMCCD technologies enjoyed popularity in scientific imaging, over the past few decades sCMOS technology has come to the fore as an ideal solution for imaging in life sciences.