Photons and solid state detection

Transcription

1 Photons and solid state detection Photons represent discrete packets ( quanta ) of optical energy Energy is hc/! (h: Planck s constant, c: speed of light,! : wavelength) For solid state detection, photons are converted to electrons Electrons can be measured as a current (flow through resistor) or as a voltage (accumulation in capacitor) Detection can be done at a single point (point = small area over which all detected photons are summed up) e.g. photodiode, photomultiplier tubes (PMT) Multiple point detectors can be arrayed to detect an image eg. complementary metal oxide semiconductor (CMOS) or charge coupled device (CCD) sensors. Recall shot noise: photon production is inherently noisy with a standard deviation proportional to the square root of intensity

2 Quantum efficiency and spectral sensitivity Photodiode responsivities. Quantum efficiency = number of electrons generated / incident photon Measured as a unit less fraction or in ampere/watt (called responsivity) QE and responsivity vary with the wavelength of light

3 Photoelectric effect Valence band (VB) is the energy band with bound electrons Conduction band (CB) is the energy band with free electrons Difference is the energy band gap Photoelectric effect describes how electrons can be kicked out of the valence band on supplying energy E g will move electrons to the CB. Eventually they will fall back Ionization energy (IE) will eject an electron from the material. This electron is lost

4 Photomultiplier Tube Photoelectric effect dictates ejection of an electron from the photocathode Photoelectrons focused and made to strike an electrode (called dynode) Each dynode is at a +ve potential (10s-100s volts) above the previous one The electric field between dynodes accelerates electrons as they strike subsequent dynodes Secondary emission causes more electrons to be emitted than the number that struck a dynode causing a multiplicative gain (up to 10 8 ) The anode collects all electrons and leads to a current

5 Photodiode pn junction Pure semiconductor is an insulator but can be doped to p type (fixed negative charges, free holes) or n type (fixed positive charges, free electrons) A hole is not a real thing, it is the lack of an electron A depletion region forms with no free charges. There exists an electric field across this region. Region can be expanded by reverse biasing (n region voltage>p region voltage by a few volts)

6 Photodiode operation Photoelectric effect dictates generation of an electron-hole pair (EHP) In the p or the n region the EHP disappears due to recombination In or near the depletion region, the photogenerated free carriers come under the influence of the electric field and causes a current The fact that useful EHP are generated only in a specific region plays a role in determining QE and spectral sensitivity Photocurrent is linearly proportional to the optical power incident

7 Dark current and read noise Dark current is the output of a photodetector is the absence of any light. It is not a constant offset that can be rejected but is also noisy Caused by thermal vibrations leading to EHP creation (PD) or electron ejection (PMT) Limits the lowest light levels that can be detected Read noise is the excess electrical noise (apart from photon shot noise) that appears at the output of a photodetector Independent of optical power, so dominates at low light intensities Dark current (and dark current noise) and read noise can be reduced by cooling the detector

8 Charge coupled device imager Photoelectrons collected in potential wells A well is created if the gate voltage is high By using a cyclic sequence of voltages the collected photoelectrons are moved

9 Charge coupled device imager Photoelectrons of each pixel are moved horizontally (within the same row) into a series of output wells The charge is then transferred vertically A charge-to-voltage amplifier converts the photoelectron charge into a voltage Key ideas The readout is serial While the array is being read, it MUST be protected from light Charge Transfer Efficiency is a critical parameter and measures how much chargeislostpertransfer.itisdefinedas: (charge after transfer)/(charge before transfer)

10 Complementary metal oxide semiconductor imager Column selection Passive pixel array Any pixel can be connected to the charge-to-voltage amplifier using a row and column select mechanism Row selection Switches implemented using MOS transistors Because of the long metal lines between the pixels and the amplifier, this does not work very well Charge to voltage amplifier Output

11 Complementary metal oxide semiconductor imager Column selection Active pixel array Each pixel has its own chargeto-voltage amplifier Row selection Amps and switches implemented by MOS transistors Random access scheme is retained Key feature: processing circuits can be implemented at the pixel or at the column level Key feature: can read one pixel while others are collecting light Output

12 Array imager features Resolution: number of x and y pixels More lead to more detail, but for a given array size more means each pixel is smaller -> smaller photodiode -> poorer sensitivity Pixel pitch: size of one pixel. Down to 1-2 µm using latest fabrication technology Array size (format): size of the array in units of length. Must match image size Fixed pattern noise: spatial noise that is fixed in time. Due to fabrication related differences between individual pixels. Worse in CMOS because each pixel has its own circuitry Fill factor: ratio of area of photodiode to area of pixel Circuits (switches, amps) take up valuable photodiode space Corrective measures Microlenses: per pixel lens increasing light gathering Back-illuminated sensors Frame rate: images per second < 1/(readout time + exposure time) Intensity resolution (how fine a difference) & dynamic range (maximum/minimum)

13 Pixel pitch & sampling requirements Diffraction sets a resolution limit System magnification maps that resolution limit on to an image sensor Pixel pitch needs to be fine enough for adequate sampling Analogous to Nyquist sampling theorem if a signal has a maximum frequency content of B Hz, it must be sampled at > 2B Hz for reconstruction 200 nm apart (object) 20 µm apart (image) 100x magnification Pixel pitch >20 µm >10 µm <10 µm > : just greater than < : just lesser than Pixel pitch at least 2x smaller than optical resolution limit (Rayleigh limit)

14 Digital image representation An image is a two-dimensional array a[x,y] x,y indicate the position of the pixel within an image a[x,y] is the intensity of that pixel. 1 white, 0 black [ 0 1 ] 1 0 For an NxM image x: 1, 2,. N y: 1, 2,. M

15 Digital image representation

16 Spatial resolution Larger pixels lead to lesser detail because a pixel is the smallest unit of an image. It can hold only one intensity.

17 Intensity resolution Light intensity is continuous (red), but a digital image can only store discrete values (blue) Due to the binary system used in computers, the number of distinct levels in an N-bit image is 2 N Almost all computer displays are 8-bit which means they can show 256 distinct shades between black (gray level 0) and white (gray level 255) A good imager will have >8 bits, however most common image formats JPG, BMP, PNG are only 8 bit leading to potential loss of information TIFF image format can store up to 16 bits so is preferred Why are more bits useful? 8 bit 4 bit 8 bit 4 bit

18 Saturation and dynamic range Light intensity Saturation level. Any light intensity higher that this can not be differentiated from this level. level 2 N -1 level 2 N -2 level 2 N -3 level 2 level 1 level 0 Digital pixel levels Dynamic range (DR) is the ratio of the maximum to minimum light that can be detected ie (2 N -1)/1 May be smaller due to noise 2 N -1/(noise floor) Usually expressed in decibel (db) 20 log 10 (2 N -1)

19 Color image representation Split into three color channels red, green and blue Each channel can be thought of as a grayscale image

20 Sensing color Sense 1 color channel at a time with a grayscale/monochrome imager and combine images in software. Limitations? Bayer array: color filter array on top of each individual pixel color filters Why twice as many green? Needs interpolation to form image pixels Foveon sensor: uses 3D photodiodes and exploits the dependence of light penetration on wavelength Sensitivity, particularly of deeper structures, suffers blue diode green diode red diode

21 Intensity scaling b For an N-bit imager 255 a! 0 b " 2 N -1 b a >256 levels in image levels for display a <256 levels in image Note <256 levels does not neccesarily indicate <8-bit imager. Consider a 12-bit imager (light dependent levels from 0 to 4095) imaging in very low light. It is possible that the brightest pixel only goes up to <256

22 Contrast Contrast is the percieved difference between various intensity levels in an image Consider the same scene imaged using two cameras. Both have 65 levels. Neither image is scaled Intensity scaling enhances contrast allowing visualization of fine details Given images c, d can you tell which came from which camera? a: Levels 0-64 b: Levels IMPORTANT to be aware of scaling before comparing images c: a scaled to d: b scaled to 0-255

23 Temporal noise and SNR Observe the intensity of 1 particular pixel over multiple images Low noise High noise Both have same mean Pixel level Frame number Multiple images of the same scene Signal to noise ratio = mean(pixel level) / standard deviation(pixel level) SNR (in db) = 20 log 10 (SNR) What happens to SNR with increasing light intensity?

24 Imaging parameters Light available: number of photons available per unit time Exposure time: duration for which photons are collected for 1 image Quantum efficiency: electrons generated per photon Amplifier gain: electron-to-voltage or electron-to-current conversion factor How do we optimise SNR? Maximize collected photons signal # number(photons) shot noise # [number(photons)] 0.5 Minimize read noise

25 Photon Transfer Curve Noise in imagers read noise, shot noise, fixed pattern noise FPN can be eliminated. How? Read noise independent of light intensity Shot noise dependent on light intensity Output signal of imager dependent on light intensity PTC is a log-log plot of signal vs. noise log(noise) Three regions read noise shot noise saturation log(signal)