CMOS Photo-Detectors - Seminar Report

ABSTRACT

The world might have been reaching new heights in technology, man might have made artificial residential environment in outer space. But all this will end up to one primary objective. The aim of every human research is how to make life more comfortable and decrease the mortality of human beings.

For this reason a crucial visual study of animal tissue and environment is highly critical. The number of CMOS technology used as light sensors in different imaging system has been on rise in recent decade. The use of appropriate CMOS photo-detector structures and imagers can bring a revolution in field of medicine, agriculture and bio-defence. Some of the pronounced fields where photo-detectors are regularly used are protein detection, gene expression, cell migration and evaluation of animal models of human cancer.

Therefore, to develop we need to make the available technology cheap and miniaturise it, so that it can be made available through hand held devices, and can be brought to masses as a low cost technology implementation.

INTRODUCTION

Imaging has been very attractive sphere of interest for many individuals. Some refer it to be hobby, some refer it to quest, some refer it to research, and some refer it to be revelation. The desire of man to represent and record the 3-D world around him and to be capable of retrieving the data at any desired moment has led to development of many new technologies which wouldn’t have existed 2-3 decades ago.

Medicine industry has seen many transitions in technology, because it needs state of art artefacts to detect diseases and different infections. The technology currently used is Fluorescence. Next to it is Photo Multiplier Tubes. These devices are capable of rendering high gain to currents, and require high operating voltages.

The technology which is creeping into use is CMOS based technology.  These include pixels with photo current integration, high dynamic range imaging, and avalanche photodiode system. These systems are briefly discussed in this paper and their results have been compared.

Fluorescence

Cells contain molecules, which become fluorescent when excited by UV/Vis radiation of suitable wavelength. This fluorescence emission, arising from endogenous fluoro-phores, is an intrinsic property of cells and is called auto-fluorescence to be distinguished from fluorescent signals obtained by adding exogenous markers. The majority of cell auto-fluorescence originates from mitochondria and lysosomes. Together with aromatic amino acids and lipo-pigments, the most important endogenous fluorophores are pyridinic (NADPH) and flavin coenzymes. In tissues, the extracellular matrix often contributes to the auto-fluorescence emission more than the cellular component, because collagen and elastin have, among the endogenous fluorophores, a relatively high quantum yield. Changes occurring in the cell and tissue state during physiological and/or pathological processes result in modifications of the amount and distribution of endogenous fluoro-phores and chemical-physical properties of their microenvironment. Therefore, analytical techniques based on auto-fluorescence monitoring can be utilized in order to obtain information about morphological and physiological state of cells and tissues. Moreover, auto-fluorescence analysis can be performed in real time because it does not require any treatment of fixing or staining of the specimens. In the past few years spectroscopic and imaging techniques have been developed for many different applications both in basic research and diagnostics.

Photo Multiplier Tubes

The current state of art light sensitive device is PMT (Photo Multiplier Tube). They have very high gain, approaching 1 million. And for the very same reason they are very bulky and also very costly. They require operating voltages of around hundreds to thousands of volts. So they are out of question to be used in hand held device. Another disadvantage of PMT is that the efficiency of photon detection is near about ~4% which is very low and in-efficient. Their bulky size renders them incapable to be used in dense arrays.

Charge-Coupled Devices

Another alternative is CCD. These are solid state replacement for PMTs. CCDs require special Silicon processing for manufacturing, and hence have higher integration and manufacturing costs. Their integration with other electronics and CMOS technologies is very difficult to implement. Other disadvantages of CCDs compared to CMOS photo detectors are its higher power consumption, higher system costs, lower efficiency, lower speed of operation, i.e. higher dead time.

DESIGN, RESULTS & DISCUSSIONS

Pixels with Photo Current Integration

Passive Pixel Structure

This is one of the simplest pixel structures available in Photo current integration mode. In the given figure 3, we can identify a transistor and a photodiode integrated to select and output lines. They combine and operate to capture light intensity. The transistor used is called row select transistor (S). The internal capacitance of photo diode integrates the current running through it (photo current). When the pixel is addressed the select line turns on the transistor, and the output of pixel is recorded onto the output line. The charges from pixel are read in parallel. As can be seen from the figure the circuit has only one photo diode, and only one transistor. Hence it has very high fill factor. (Fill Factor is referred as the ration of area covered by the photo detector to the area covered by the total pixel structure.) But this system is prone to noise as no noise reduction system is present. Noise results in very harmful effects in this system.

Active Pixel Structure

In the figure 4 we can see there are 3 transistors and a photo diode. R is reset transistor, B is buffer transistor, and S is the source follower which isolates the read out column from sense node.

The APS has better signal to noise ration. The B present buffers the output from the photo diode; as a result, a standard noise won’t be able to break through the system. The individuality in every circuit results in non uniformity of output for surface illuminated with same intensity. This is termed a fixed pattern noise (FPN). The large part of FPN is due to the variation on sense node voltage just after reset. One of the ways to prevent FPN is Correlated Double Sampling (CDS). In this procedure we take two samples, one before the read out, and second just after the reset. And the output will be the difference between the two output levels.

Digital Pixel Sensor

In this pixel structure an ADC is already present in the structure, so the output is Digital in nature, which is easy to be incorporated into any arrayed operation or data bus. The figure 5 shows a reset transistor and an ADC connected to the sense node. It is very good for integrated solutions and high speed digital imaging. However the Fill factor is very low, so high density arrayed structure is not possible with DPS.

Avalanche Photodiode System

This is the most advanced technology available in the field of Image sensors using CMOS technology. The circuit consists of two transistors, Quenching transistor, and reset transistor. It also consists of two control blocks namely Quench activation and Reset control. An avalanche photodiode (APD) is a highly sensitive semiconductor electronic device that exploits the photoelectric effect to convert light to electricity. APDs can be thought of as photo detectors that provide a built-in first stage of gain through avalanche multiplication. From a functional standpoint, they can be regarded as the semiconductor analogous to photomultipliers. By applying a high reverse bias voltage (typically 100-200 V in silicon), APDs show an internal current gain effect (around 100) due to impact ionization (avalanche effect). However, some silicon APDs employ alternative doping and bevelling techniques compared to traditional APDs that allow greater voltage to be applied (> 1500 V) before breakdown is reached and hence a greater operating gain (> 1000). In general, the higher the reverse voltage, the higher is the gain.

Operation:

When no light is falling on APD, the potential across it is

Vs = (Vop + Vdd)

When light falls on APD, even of very low intensity, Impact ionization occurs, and current starts to built up in the APD. Hence the potential at sense node starts falling. The Quench control detecting this fall turns on the quench transistor. The quench transistor takes the sense node voltage down to zero. As soon as the sense node reaches zero, quench transistor is turned off and reset control turns the reset transistor on which takes the sense node to Vdd voltage. After ‘Reset’, the transistor is turned off; the sense node returns back to the potential of (Vdd+Vop).

To explain the operation of Avalanche photo diode system we approximate the Geiger mode i.e. single photon detection. And hence the device is named Single Photon Avalanche Photo-Diode (SPAD). The APD operating in Geiger mode is capable of both high speed and high sensitivity operation. We can also say that APD system is the nearest replacement for photo multiplier tubes in CMOS technology.

In the previous Photo Current integration section the pixel structures had one problem which couldn’t be removed. The problem was that if a light pulse had very high illumination but stayed for only a short duration, the photo diodes would have been unable to register it. Similarly, if the intensity of light was very small, but appeared for longer interval of time, its integration is also not possible. But in APDs this problem was overcome, because it can operate in Geiger mode.

CMOS IMAGERS

High Dynamic Range Imaging

Dynamic range and signal to noise ratio of a single collected image is strongly correlated with the shot noise and the so called read out noise of charge transfer and output amplification. Common CCDs offer a dynamic range of near about 50 db-80 db. Most of the integrating CMOS imagers use diffused photo diodes in combination with in-pixel amplification. Their blooming resistance and non destructive read out allow a number of individually timed read outs with same integrating period. For dynamic range extension this read out is typically combined with a piece wise linear compressed output signal.

While HRDC (High dynamic range CMOS) uses exponentially sub threshold characteristics of a MOS transistor. This transistor is directly connected to the photo diode as a permanent working shunt for photo currents.

HDRI is the result of years of research done to optimize the dynamic range within which a photo detector is capable to work. HDRC is based on n⁺ -p junction in a low doped p-substrate. This arrangement provides an optimum combination of low photo diode leakage current and spectral response.

If we look at the circuit closely we will find that the circuit highly matches to that of Active Pixel Sensor. But the difference is that the sub threshold characteristics of these transistors are optimized to get a high dynamic range. This design is implemented on p⁺⁺ substrate as in the figure. The logarithmic HRDC principle is based on a continuous current to voltage conversion by a transistor operating on sub threshold regime of MOS transistor. But in the silicon wafer we need to isolate the devices in order to make the integrated chip functional. Shallow Trench Isolation being one of the techniques, it is not optimum for usage in HRDC, since it is prone to have parasitic edge transistors. The leakage current of such transistor is generally very high. So we need an edgeless transistor as shown in adjoining graph, where it has minimal leakage current. The layout of HRDC pixel cell is shown below. The large feature at the centre of image is the Photo sensitive element. The edgeless transistor is located on the lower side, working in logarithmic mode. The additional circuitry required is also visible in the figure.

Figure HRDC pixel Array layout

The CMOS image sensor as the heart of an imager system on the one hand defines the optical properties of the imager system, such as sensitivity, spectral response and dynamic range. On the other hand its layout and design determines the resolution, readout speed, random access of regions of the viewed scene and the kind of sampling in time of the image information, which are all related to the circuit design. Basically there are two different sampling mechanisms, i.e. line synchronous and frame synchronous sampling, also described as rolling shutter and global shutter.

Rolling shutter (also known as line scan) is a method of image acquisition in which each frame is recorded not from a snapshot of a single point in time, but rather by scanning across the frame either vertically or horizontally. In other words, not all parts of the image are recorded at exactly the same time, even though the whole frame is displayed at the same time during playback. This produces predictable distortions of fast moving objects or when the sensor captures rapid flashes of light.

A global shutter, unlike the rolling shutter, exposes all pixels at the same time.

Uses:

Today HRDC technology is in high demand. The line synchronous and frame synchronous techniques of sampling the image finds wide spread application in different fields. Fig. Adjoining shows a typical scene captured with a HDRC® sensor developed for a night vision driver assistance system. In this automotive application the challenges for the sensor are the wide spectral sensitivity from visible (VIS) to near infrared (NIR), to recognize the lane and obstacles far away on the dark road without getting blinded by the high beam of the approaching car.

In the field of medicine the growing use of image sensors has opened new avenues for photo diode development. The integration of cameras in endoscopes and inspection elements for minimal invasive surgery, led to miniaturization of sensors and optimization of image qualities.

Fig. adjoining shows an example of a CMOS imager in a standard CMOS 0.18μm technology. The array is based on 256 APS pixels of 20 μm × 30 μm size with a 60% FF. Only the photosensitive area of the photodiodes is exposed in the array, and everything else is covered with the top metal layer. Here, row and column scanners were used instead of decoder circuits in order to reduce the control lines coming into the chip. Also, only one input clock is needed to control the row and column circuitry.

The signals received from individual pixel structures are fed to multiplexers. The output of these multiplexers is then connected to Op-amp for getting analog signal equivalent of the image. This in turn is given as input to Sample and Hold circuit. S/H circuit is embedded with capacitor of high value to prevent any charge injection effects. This is followed by Analog to digital converter, then state and machine control, followed by output latch, which gives parallel port output. If serial output is desired, a parallel to serial port convertor is used after output latch. The analog equivalent output can be extracted from Op-amp.

CONCLUSION

We have briefly reviewed different CMOS photo detectors used till dates and their implications, the new technologies which are available to us. We also discussed the prospects of imaging technologies. A few years ago, Passive pixel sensor was used to capture images and different noise reduction algorithms were developed to remove the noise from it. But as complexity of different requirements raised, a demand for high definition image with lower Dead time and lower Noise increased. And hence Active pixel sensor came into being. But still the dead time was high due to usage of 3 transistors and Correlated Double Sampling. So to improve its performance, Digital pixel sensor was introduced, which was designed particularly to have high speed image acquiring capability. HRDC implemented on real world has set standards for other imagers.

Finally Avalanche Photo diode system was introduced. Combining all the features which we require from a Photo detector, and providing alternative technology which is cheap to implement and suitable for mass production, we get APD. APD has all the features we expect a cheap alternative of PMT to have. Its integration with different Imagers is quite easy; soon we will be able to see APD, in different fields.