CMOS Active Pixel at Lfoundry: Brief description - Agenda
Transcription
CMOS Active Pixel at Lfoundry: Brief description - Agenda
CMOS Active Pixel at Lfoundry: Brief description, main quality factors and criticalities from the device concept to the production phase. VI Scuola Nazionale: "Rivelatori ed Elettronica per Fisica delle Alte Energie, Astrofisica, Applicazioni Spaziali e Fisica Medica“ INFN Laboratori Nazionali di Legnaro (PD) 23 - 27 marzo 2015 Presented by Giovanni Margutti Giovanni.margutti@lfoundry.com Principal Process Integration Eng R/D department - Lfoundry confidential © 2013 LFoundry. All rights reserved. Course Outline Introduction to Photo detectors Principle of operation CCD and CIS PPS vs APS Image Sensors Photodiode 4T Main merit functions and their optimization QE, Dark I/hot pixels, Dynamic range Further image quality improvement: Rolling vs GS, 5T 2 confidential © 2013 LFoundry. All rights reserved. Photodetectors Devices that convert light into electrical signal (charge, current, voltage) Photo conductors Photo transistors Photo diodes Photogates Etc.. The basic principle of operation is the same for all the devices and it will briefly discussed in the next pages The definition of the main merit functions is similar for all the photo detectors and will be also discussed 3 confidential © 2013 LFoundry. All rights reserved. Photodetectors - Principle of operation Photons impinging the semiconductor can have enough energy to create electron/hole pairs; Electrons and/or holes are then separated and collected; An output signal due to the collected charges is generated. In case of “ high energy” photons w e may have the follow ing mechanisms of interactions: -Photoelectric effect - Compton scattering -Pair production .. but the main mechanism of photon absorption in silicon, in case of low energy photons (Emin< E< 10 eV) is the photovoltaic effect 4 confidential © 2013 LFoundry. All rights reserved. Principle of operation: photo generation ~1.1 eV (Si) Silicon band gap betw een valence band and conduction band is 1.12 eV (at 300 K). This is an indirect bandgap, so Incident photons w ith energy> 1.12eV (λ= 1100 nm) may be absorbed, causing electrons to jump from valence to the conduction band,. but the transition must be assisted by phonons. A very useful parameter to represent the chance a photon on a given energy is adsorbed into the silicon is the absorption coefficient I(x)= I(0)exp(- α, usually expressed in cm-1 αx); w here I(x) is the intensity of the light penetrating the silicon at a depth X, X= 0 is the silicon surface; thus at a depth of X= 3/ α 95% of the light is absorbed. To give an idea; Blue light is adsorbed in 0.2 µm; Green in about 1 µm; Red in 3 µm; 5 confidential © 2013 LFoundry. All rights reserved. Principle of operation: charge collection Once electrons and holes are generated, they must be separated and collected. This is performed by an electric field, causing drift of the charges. It’s important at this stage taking into account for the recombination mechanism, whereby an electron and a hole recombine and annihilate each other to reach the equilibrium state: np= ni2 Where n, p are the concentration of free electrons and holes, and ni is the intrinsic concentration of carrier in silicon, at a given temperature Recombination may be due to the following mechanisms: •recombination trough recombination centers (traps, impurities etc..) known as SRH (Shockley–Read–Hall) mechanism •direct recombination (unlikely in silicon) •Auger effect (more likely in heavily doped substrate) So we need to use “good” silicon, with low trap density, ( few silicon defects/low metal contaminations) to avoid recombination . This is key for CIS fabrication. Metal contamination as well as silicon damage must be kept as low as possible as they may be responsible for both recombination or generation (causing dark current, hot pixels.. this will be briefly shown later on) 6 confidential © 2013 LFoundry. All rights reserved. Photodiode - Photogate The Photodiode (PD) basically consists of a N/P junction reverse biased; electrons generated into the depletion region (proportional to the flow of the impinging light) are stored in the PD capacitor. Changes in the PD voltage are then amplified and read out. Photodiodes are the basic element of CMOS image sensors The Photo gate it’s a MOS capacitor with polysilicon as the top terminal. It’s biased in order to have a wide depletion region where electron/holes are generated. The charge generated and collected into the photogate must be transferred to a read out unit to be converted in signal. Photogates are the basic elements of CCD 7 confidential © 2013 LFoundry. All rights reserved. Imager sensors: CCD vs CIS Imaging is the process of using arrays of photo detectors to create and store images. The most important devices used in the image sensor market are CCDs (charge coupled devices) and CMOS Image Sensors (CIS). In both cases the a charge proportional to the incident light is generated and collected, for each pixel. CIS CCD • Charge to voltage conversion is made w ithin each pixel of the array • Voltage signal is pre-amplified by a source follow er transistor on each pixel • Charge is carried across each columns by means of coupled devices • Charge to voltage conversion is made out of the array 8 confidential © 2013 LFoundry. All rights reserved. CIS vs CCD at the beginning… …The gap in terms of image quality is not true anymore: highend cameras are now equipped with CIS 9 confidential © 2013 LFoundry. All rights reserved. CIS: Core and SOC CIS allow s the integration in the same chip! 10 confidential © 2013 LFoundry. All rights reserved. PIXELS - PPS Core of the imager sensors is the Pixel; the active element of the pixel is the photodiode The easiest PD is a diode w ith a w ide depletion region (inverse polarization); electrons/holes are generated inside the silicon; the ones generated in the depleted region are driven by the electric field tow ards the ground potential (holes) or tow ards the anode (positive potential) changing the potential difference of the PD; Charges generated outside of the depleted region can diffuse up to hit the depletion region. Of course in this case electrons have higher chance to recombine w ith holes (in the case depicted in figure). Changes in the PD potential (proportional to the flow of the impinging light) can be now sampled and read. A possible simplified scheme is the follow ing one: The PD is inversed biased . During the integration time the row is disabled and the PD is left floating for a time Dt; • During this time the photo-generated charges are collected in the PD capacitance • When the row is enabled the current flow s into the column (charging a capacitor until all the generated charge is transferred). • Once the charge is completely transferred the PD bias is restored to its initial value 11 confidential © 2013 LFoundry. All rights reserved. Pixels array: PPS and APS PPS (one X-tor per pixel) APS (at least three X-tors per pixel) • High fill factor (only one TX): • High noise (charge to signal conversions and signal amplification is performed at the end of each column and all the components of the noise coming from the signal path are amplified) • Slow read out-reset sequence 12 • Low fill factor (no more true w ith decreasing Xtors size) • Low noise (charge to signal conversions and amplification is performed for each PD) w ith respect PPS but higher than CCD at the beginning.. • fast read out sequence confidential © 2013 LFoundry. All rights reserved. 4-T cell (and pinned Photodiode) The 4T arxchitecture allows to dramaticaly reduce the noise and makes the CIS competitive to CCD The main feature are: The introduction of the pinned photodiode, that eliminates two important sources of temporal noise: the componenet of the dark I coming from the silicon surface; the photodiode reset noise; The introduction of the DCS ( double correlated sampling) that eliminates the FD reset noise; the spatial noise coming from the Reset and Source follow ers Vt pixel to pixel variability. 13 confidential © 2013 LFoundry. All rights reserved. Pinned photodiode and 4-T cell 4-T cell architecture 4-T cell architecture TX Positive value potential Reset TX (FD floating diffusion) Source Follow er Row Select 14 confidential © 2013 LFoundry. All rights reserved. Pinned photodiode and 4-T cell Conventional photodiode (p-n junction) Pinned photodiode (p-n-p structure) TX Silicon surface the N- region is completely When depleted the maximum potential accross the PD is hit; increasing the reset voltage the potential doesn’t change. So if the reset voltage Vrst is high enough the PD potential after reset is unique and determined. This is not true for the stardar NP photodiode, for w hich the potential value and depleted region after the reset are affected by the RST noise 15 confidential © 2013 LFoundry. All rights reserved. Double Correlated Samplig The charge collected in the PD is transferred to the FD and converted to voltage, whose value is Q/Cfd. The capacitance of the FD is usually small, so that we can have a good conversion gain and responsivity. If some noise is generated in the FD area, this is amplified and sent to the read out circuitery. The two main contributors to the FD noise are the reset noise and the charges generated and collected into the FD during the integration time (due to thermal generation, unfiltered light, etc..). The charges are eliminated by resetting the FD right before charge from the PD is transferred. The FD reset noise is eliminated by the DCS (double correlated sampling), instead. DCS sequence PD and FD are reset. Photo electrons are generated in the PD during the integration time. FD is reset and signal stored in the SHR. The signal is Vaapix-Vt Reset-Vt Row Select-Vt Source Follower +Reset noise. Then the charge Q is transferred to the FD, added to the previous signal, sent and stored into the the SHS. The output from SHS and SRH is then sent to operation amplifiers who return a signal proportional to SHSSRH=Q/Cfd. So the reset noise is eliminated. The contributions coming from the Vt of the arrah devices (RST TX, RS TS, SF) are also eliminated, and this improves the spatial noise as we are no more sensivite to pixel to pixel differences. confidential © 2013 LFoundry. All rights reserved. 4-T cell readout - DCS 17 confidential © 2013 LFoundry. All rights reserved. CMOS image sensors Section II Image Sensors confidential © 2013 LFoundry. All rights reserved. Outline Main quality factors: definition and improvement Quantum Efficiency QE increase – stack optimization, LG, deep PD Back Side Illumination (BSI) Dark I Hot Pixels Further improvement of image quality -Rolling vs GS; Dynamic range – 5T 19 confidential © 2013 LFoundry. All rights reserved. QE (quantum efficiency) Quantum efficiency (QE) is the measure of the efficiency with which incident photons are detected. The quantum efficiency is the ratio of the number of detected electrons divided by the number of incident photons. It’s usually expressed in % and detailed for wave lenght (see picture) Of course higher QE results in higher signal and lower SNR, so it’s preferred Color filters are realized with colored photo resists; Bayer pattern 20 confidential Optical microscope image © 2013 LFoundry. All rights reserved. How to increase external QE 1) Micro lenses.. In the APS some «active space» is consumed to built in pixel read out electronics. The dead space can be recovered by added lenses (micro-lenses) focusing the light into the PD; 21 confidential © 2013 LFoundry. All rights reserved. Microlenses to enhance fill factor w /o microlens Without ulens Optical simulation Light is focused into the PD area After coat, expose and develop 22 confidential Bleach and reflow © 2013 LFoundry. All rights reserved. How to increase external QE 1) Micro lenses.. In the APS some «active space» is consumed to built in pixel read out electronics. The dead space can be recovered by added lenses (micro-lenses) focusing the light into the PD; 2) Reduce reflected/adsorbed light A certain portion of impinging light is reflected at the interface between adjacent layers forming the optical stack; 23 confidential © 2013 LFoundry. All rights reserved. How to increase external QE 1) Micro lenses.. In the APS some «active space» is consumed to built in pixel read out electronics. The dead space can be recovered by added lenses (micro-lenses) focusing the light into the PD; 2) Reduce reflected/adsorbed light A certain portion of impinging light is reflected at the interface between adjacent layers forming the optical stack; this can be reduced by eliminating the interfaces (when possible); Low k, high n (1.6/1.7) polymer 24 confidential © 2013 LFoundry. All rights reserved. How to increase external QE 1) Micro lenses.. In the APS some «active space» is consumed to built in pixel read out electronics. The dead space can be recovered by added lenses (micro-lenses) focusing the light into the PD; 2) Reduce reflected/adsorbed light A certain portion of impinging light is reflected at the interface between adjacent layers forming the optical stack; this can be reduced by eliminating the interfaces (when possible); Or reducing the refractive index differences between adjacent layers; 25 confidential © 2013 LFoundry. All rights reserved. How to increase external QE 1) Micro lenses.. In the APS some «active space» is consumed to built in pixel read out electronics. The dead space can be recovered by added lenses (micro-lenses) focusing the light into the PD; 2) Reduce reflected/adsorbed light A certain portion of impinging light is reflected at the interface between adjacent layers forming the optical stack; this can be reduced by eliminating the interfaces (when possible); Or reducing the refractive index differences between adjacent layers; • CFA 26 confidential © 2013 LFoundry. All rights reserved. Step index polymeric light guides Microlens and planarization layer Color filter array Nitride passivation M2 M1 Oxide layers High refractive index Polymer (n = 1.6-1.7) 27 confidential © 2013 LFoundry. All rights reserved. QE (how to increase internal quantum efficiency) We need to increase the generation and collection of fotogenerated electrons; Because of the low recombination probability, we can assume all electrons generated inside the depleted regions are collected in to the PD node; To hit 100% of internal QE all the light impinging the silicon must be adsorbed into the PD depletion region So a possibility is increasing the PD depth(indeed in the 4T APS pixel the PD is empty after the being reset). The photo litography mask be so thick to block implanted ions and the dimensions to be printed (CDs) can be ~1/15 time the mask thickness: this may be very challenging, particularly for small pixels Light absorption in silicon Dee PD isolation 28 confidential © 2013 LFoundry. All rights reserved. BSI - CIS Backside illumination (BSI) flips the image sensor upside down so it absorbs light from the backside As an alternative to the more common front-side illumination (FSI) technique, it offers the most direct path for light to strike the pixel In principle, BSI provides a better fill factor, higher QE, lower cross talk, and metal routing flexibility 29 confidential © 2013 LFoundry. All rights reserved. FSI vs BSI pixel Microlenses Bayer filters Metal runners Antireflective coating Dielelectric stack Access transistors Photodiodes BSI FSI 30 confidential © 2013 LFoundry. All rights reserved. Pixel Shrinking - BSI Bonding loop Carrier w afer & Device w afer preparation Device w afer edge trim Carrier w afer & Device w afer direct bond CFA/uLens Thinning loop Antireflective layer Device w afer grinding Device w afer final thinning (< 5um) Thinned silicon Additional modules Device w afer additional optical/electrical modules Critical items Front Side structures Bonding voids; Device w afer distortion; Silicon surface damage. 31 confidential © 2013 LFoundry. All rights reserved. Dark current and its impact on image quality Undesired signal integrated in dark conditions Sources of dark current and hot pixels 1. STI around the photodiode 2. Photodiode surface 3. Substrate 4. Overlap of PD with transfer gate 5. Contamination & defects Medium light 32 Transfer gate p+ implantation STI TX n- implantation epi substrate Low light confidential © 2013 LFoundry. All rights reserved. Hot pixels • Causes – eletrons generated in the depletion region trough SRH mechanims • Mitigation – reduce generation centers by: • Using “ good silicon” • Optimize processes and thermal budget • Reduce metal contamination (gettering, environment, maintenance, etc..) Mid gap state for various impurity, in silicon – after S. M. SZE KWOK K. NGE - Physic of Semiconductor device confidential © 2013 LFoundry. All rights reserved. Further improvement of image quality Distortion induced by the Electronic Rolling Shutter (ERS) Blooming 34 Dynamic range limiting the view of details in darker and brighter areas confidential © 2013 LFoundry. All rights reserved. High Dynamic Range, 5T Aptina DR-pix technology, aimed to increase dynamic Dynamic range= ratio betw een maximum and minimum signal can be detected in linear regime, usually expressed in db. range The maximum signal is determined by the full w ell, the minimum one by the noise (read noise floor) DCG on Saturation Signal DCG off Variable conversion gain 35 confidential © 2013 LFoundry. All rights reserved. Electronic rolling shutter Reset (from aptina.com w ebsite) Readout Row 1 Row 2 … Row n t 36 confidential The integration time of different row s starts and ends in a different time, causing the artifacts © 2013 LFoundry. All rights reserved. Global shutter Introducing the “in pixel memory” all the rows start and end the integration time at the same time… In pixel memory … While the in-pixel memories can be read row by row. No distortion! But the “In pixel memory” needs an effective shield from light! 37 (from aptina.com w ebsite) confidential © 2013 LFoundry. All rights reserved. Conclusion Technological development has enabled us to improve the performances of CMOS imager and make them competitive with CCDs, keeping the advantages of low cost and low power consumption. This is why their diffusion has increased in the last year in several market segments. Nowadays they are also used for high performances digital cameras. Current and future trends •Die stacking (Sony, already in mass production) •Organic photodiode (announced by Fujifilm) •Black Silicon (SiOnyx) 38 confidential © 2013 LFoundry. All rights reserved. Biblio Some reference: Text books from general guideline to deep understanding of imager sensors architectures J. Nakamura Image Sensors and Digital Processing for Digital Still Camera . CRC Press 2005 O. Yadid-Pecht, R. Etienne-Cummings - Cmos imagers from Phototransduction to Image Processing - Kluw er Academic Publisher Introduction to semiconductors devices (including detailed description of CMOS) S. M. SZE KWOK K. NGE - Physic of Semiconductor device– Wiley and sons Articles on CIS at the beginning of their history E. R. Fossum (1993), " Active Pixel Sensors: Are CCD' s Dinosaurs?" Proc. SPIE Vol. 1900, p. 2–14, Charge-Coupled Devices and Solid State Optical Sensors III, Morley M. Blouke; Ed And some time later E.R. Fossum, " CMOS Image Sensors—Electronic Camera on a Chip” , IEEE Micro, vol. 18(3), pp 8-15, May/June (1998). 39 confidential © 2013 LFoundry. All rights reserved. Thank you visit us on Linkedin confidential © 2013 LFoundry. All rights reserved.