Research camera sensitivity level is determined by three main parameters: QE of the detector, the camera pixel size and noise.
We'll do it after two symposia. This period is important to talk about Ding Ding Xiao Bian that, it seems the easiest to understand, but in fact some of the wonders of QE.
QE- quantum efficiency
QE readily appreciated that the detection efficiency of the detector. Since the absorption efficiency and the wavelength of the optical signal related to the semiconductor material, QE curve we see the horizontal axis represents wavelength and the vertical axis as a percentage. (figure 1)
1. Some typical graph of FIG QE CMOS camera (taken from Princeton Instruments, Kuro sCMOS)
And the wavelength of the semiconductor material permeability
And wavelength-dependent light transmissive semiconductor material. Yellow-green light of 500 ~ 600nm, just to penetrate into the intermediate material is detected, the highest detection efficiency. Wavelength becomes shorter, the more photons are absorbed before reaching the detection zone; wavelength becomes longer, the more photons pass through the detection zone, or because of lack of energy, photoelectrons can not be generated. All we see QE curve is high in the middle, reducing both sides.
2. The semiconductor material and the wavelength of light transmission of FIG.
(Image taken from SONY, back-illuminated CMOS image sensor)
Chip surface structure
Normal CCD / CMOS chip, the pixel portion has a surface covered by an opaque metal structure, and photons can not reach the detection zone through, resulting in loss of QE. (image 3)
3. FIG photons can not pass through the metal surface of the pixel structure
In order to improve the QE, modern chips in each pixel added to make the surface of the microlens (MicroLens). Microlens allows some of the originally irradiated light deflection metal structure, the photosensitive region is focused onto the chip. Using this technique, it can increase the nominal CMOS chip to QE example 82% , however, the microlens has its limitations. Perpendicularly incident light effect its best, the effect is deteriorated large angle of incident light. For example, 82% of QE chip manufacturers to provide, in practical applications can be achieved if there is a doubt.
4. FIG improve chip microlens QE.
Back-illuminated chip technology
背照式芯片将前面提到芯片反过来，半导体信号检测区直接面对信号来源，信号不用穿过上述的金属结构层，上述对信号的阻挡也就不存在了。这样做出的背照式芯片，可以获得接近完美的95%峰值量子效率。而且可以实现低至200nm 的深紫外和高至1100nm的近红外探测。
图5. 正照和背照芯片比较
那么，背照式95%的QE和正照式82%的QE, 探测灵敏度真的只差13%吗？
为比较两者真实的灵敏度差别，我们使用了Argolight 公司的标准定量荧光样品。该样品可以长时间发出稳定的荧光。以下是对比Photometrics 公司Prime BSI (95% QE, 6.5um 像元)和一款正照式sCMOS相机(82% QE，6.5um 像元)的成像结果。
图6. Argoligh Argo-HM 荧光样品
成像条件: 60x, 1.35NA 油镜， 450nm LED 激发光。 两台相机分别接在Cairn TwinCam 50/50 双相机分光器成像端。使用相同的曝光时间(80ms), 各拍摄100帧图像，做平均。测量样品上荧光条带的强度。根据已知的相机增益(Gain)和偏置(Bias), 用以下公式，换算为真实的强度值(电子数)。
信号强度(e-) =（信号灰度值 - 偏置）*Gain.
比较结果
理论上，两者灵敏度差别为 95% / 82% = 1.15
实际测量的强度差别为 563/442 = 1.27
即实际实验结果，95% QE的背照式CMOS相机，在相同实验条件下，能检测到的信号是82%QE CMOS相机的1.27倍。
结论
背照式芯片技术大大提高了相机灵敏度，也有效地展宽了半导体芯片的信号检测范围(紫外和近红外). 相对于前照式芯片，在使用显微成像光学系统时，其对灵敏度的提升，大大超过了芯片标称的 QE 值的差别。