Backside illuminated CMOS-TDI line scan sensor for space applications

Backside illuminated CMOS-TDI line scan sensor for space applications Omer COHEN, Oren OFER, Gil ABRAMOVICH, Nimrod BEN-ARI, Gal GERSHON, Maya BRUMER, Adi SHAY, Yaron SHAMAY SemiConductor Devices (SCD) P.O.B. 2250, Haifa, 3102102, Israel ABSTRACT A multi-spectral backside illuminated Time Delayed Integration Radiation Hardened line scan sensor utilizing CMOS technology was designed for continuous scanning Low Earth Orbit small satellite applications. The sensor comprises a single silicon chip with 4 independent arrays of pixels where each array is arranged in 2600 columns with 64 TDI levels. A multispectral optical filter whose spectral responses per array are adjustable per system requirement is assembled at the package level. A custom 4T Pixel design provides the required readout speed, low-noise, very low dark current, and high conversion gains. A 2-phase internally controlled exposure mechanism improves the sensor's dynamic MTF. The sensor high level of integration includes on-chip 12 bit per pixel analog to digital converters, on-chip controller, and CMOS compatible voltage levels. Thus, the power consumption and the weight of the supporting electronics are reduced, and a simple electrical interface is provided. An adjustable gain provides a Full Well Capacity ranging from 150,000 electrons up to 500,000 electrons per column and an overall readout noise per column of less than 120 electrons. The imager supports line rates ranging from 50 to 10,000 lines/sec, with power consumption of less than 0.5W per array. Thus, the sensor is characterized by a high pixel rate, a high dynamic range and a very low power. To meet a Latch-up free requirement RadHard architecture and design rules were utilized. In this paper recent electrical and electrooptical measurements of the sensor's Flight Models will be presented for the first time. 1. INTRODUCTION An all CMOS technology, Backside Illuminated (BSI) multi-spectral line scanner sensor with up to 64 Time Delay Integration (TDI) levels has been designed for continuous scanning Low Erath Orbit (LEO) space applications, as an upgrade for Charged Coupled Devices (CCD's). The sensor (shown in Figure 1) is Radiation Hardened (RadHard) by design in order to mitigate in-orbit radiation effects. Prototypes have been manufactured, the evaluation phase has been completed and Flight Models manufacturing has been commenced. In this article we briefly describe the sensor design, architecture and recent measurements that were performed including some of the Flight Models results. Figure 1. BSI CMOS TDI line scanner image sensor prototype

2. SENSOR DESIGN The sensor two main parts are: a silicon chip and a ceramic package. The sensor's ceramic package houses the silicon chip and provides a highly accurate mounting surface as well as an electrical interface. An optical filter is accurately mounted on the ceramic package in front of the silicon chip and thus the spectral range of the sensor can be defined within the visible and near infrared spectrum. The optical filter is made by an optical coating of a glass substrate. Different selection and arrangement of filters is possible as required by the application. The package can be sealed with adhesives. The sensor has a single silicon chip containing 4 bands where each of the 4 bands can have a specific optical filter. Each band includes an array of 2,600 pixels by 64 TDI levels, Readout channels, a self-sustained internal controller, a communication controller, a video output port and all the necessary peripheral support blocks as described herein. At the silicon chip level, each band is an autonomous sensor sharing only the silicon bulk with the other 3 bands and differs from the other 3 bands only by a communication address. Single band architecture is described in block diagram shown in Figure 2. The electrical interface is provided by two connectors. In order to reduce the number of pins of the sensor, some signals which are common to all 4 bands such as power supplies and clocks are shared at the package level. Figure 2. Single band block diagram The current CMOS TDI sensor transduces the incoming light signal into an electrical signal by a well-known CMOS Four Transistor (4T) Active Pixel Sensor (APS) architecture [1], [2], and [3]. In order to fit the required sensor's parameters such as conversion gain and readout speed a state of the art custom 4T pixel was designed. The line scanning CMOS TDI sensor outputs a line of 2,600 pixels per image cycle, where the other dimension of the image is acquired by the motion of the satellite which is equivalent to scanning orthogonally to the line of pixels. When a high scanning throughput is required only a short exposure period is available which results in a low signal. Thus, in order to improve the Signal to Noise Ratio (SNR) a TDI arrangement was implemented. The TDI design includes several pixels arranged along the scanning direction where each of these pixels collects photons from the same part of the scene at a different instant in time. The signal is then integrated with the appropriate time delay, allowing higher signal and improved SNR. The TDI format in the sensor allows registration of the same part of the scene up to 64 times. Hence, the sensor has an array of 2,600 pixels by 64 TDI levels. The acquisition of the image is done by an electronically controlled rolling shutter. The shutter is controlled by communication commands which enable the exposure time to be changed from the line time down to zero exposure in steps of 1/1024. The sensor internal configuration is designed for two phase exposure described in [4] where the first and second halves of the exposure are half pixel shifted in a synchronized manner with the scanning motion. As a result of the two phase exposure the dynamic MTF of the image captured by the moving rectangular pixels is improved from 63% to 91% without sacrificing any other performance parameter. The whole process of two phase

exposure, signal collection and TDI operation is synchronized and controlled by an internal self-sustained controller. The photon signal is converted to voltage by the pixels, and it is then digitized by an internal Analog to Digital Converter (ADC). The signal out of a single pixel is converted at low resolution and by integration of the 64 TDI levels the depth of sampling is increased to 12bits. A Read Out Channel (ROC) is defined as the medium connecting a pixel to a video output channel. One ROC contains a unique combination of a 2 phase exposure mechanism, ADC and TDI mechanism. A single band contains approximately 170,000 different ROCs. The internal controller mentioned above is responsible for the control of the whole ROC operation and synchronization. The main features and designed performance are summarized in Table 1. Table 1. Sensor's features and typical performance Parameter Value Detector type VIS-NIR CMOS TDI line scanner Spectral bands 4, customer selectable bands (400nm-900nm) Format 4 independent lines, 2600 pixels each Pitch 26µm TDI depth 8 up to 64 (in steps of 8) Quantum Efficiency >80% at peak MTF (at 1 Nyquist) >50% (400nm-800nm) Pixel capacity 300Ke - (variable by communication command) Floor noise <120e - for all 64 TDI levels Dynamic range >70.5dB Linearity <1.5% (5% up to 85% of full scale) Maximum line rate Tested: 7,500 line/sec. Design: 10,000 line/sec. Power dissipation <1.8W (at maximum line rate) Power supplies 3.5V, 1.8V, reference 2.5V Video output Digital, 12bit Communication Serial port Clocks Main: single LVDS up to 58MHz, line sync Readout direction Bi-directional Environment Radiation hardened for Space applications 3. MEASUREMENTS AND PERFORMANCE 3.1. Full Well Capacity The Full Well Capacity (FWC) defines the saturation level of the sensor in terms of electrical charge. The sensor design has a built in feature that enables adjustment of the Full Well Capacity. The Full Well Capacity adjustment was demonstrated in a range of at least 150,000e - to 500,000e -. The first Flight Models lot uses generic calibration for all sensors which is set to Full Well Capacity of above 300,000e -. The results of FWC measurements on Engineering Models and Flight Models with the generic calibration are shown in Figure 3. Each array is independent of the others and has individual calibration parameters. Hence, it is possible to calibrate each array individually and achieve less dispersion of FWC value if required.

Figure 3. Measured median Full Well Capacity of Engineering Models (in red) and Flight Models (in blue). Each count is the measurement result of a single array (band). 3.2. Readout Noise The sensor has a fixed sampling depth of 12 bits. A larger FWC implies that the Gain coefficient converting signal measured in electrons to signal measured in digital level is larger (more electrons collected per one digital level of signal). The gain coefficient is taken into calculation when the readout noise is expressed with electrons. When the readout noise is expressed in number of digital levels, the readout noise calculation does not include the gain and it is almost independent of the FWC. More over the gain measurement using the Photon Transfer Curve method [5] has large tolerance (we measured gains in the range of at least 80 to 110 electrons/digital levels). It is therefore easier to compare readout noise in terms of no. of digital levels. Measurements of both Engineering Models and Flight Models are shown in Figure 4. Each array of the sensor is an independent measurement. A single array spatial distribution of readout noise and its readout noise histogram are shown in Figure 5 and Figure 6 respectively. Figure 4. Distribution of readout noise among the sensor population. Engineering Models (in red) and Flight Models (in blue). Each count is the measurement result of a single array (band).

Figure 5. Spatial distribution of readout noise in a single array showing all the ROC (color map is in DL). X-axis and y-axis are the sensor column and the ROC row counts respectively. 3.3. Dynamic Range Figure 6. Histogram of Readout Noise in a single array (shown in Figure 5.) In order to calculate the sensor dynamic range, we first measure the Electronic Dynamic Range which is defined as the difference between the saturation level and the floor level, as shown in Figure 7. The saturation level is limited by the A/D converter and it is always at 4,096 digital levels. The floor level is dependent on the calibration of each readout channel and additional parameters and it is typically in the range of 100-300 digital levels. The instantaneous Dynamic Range shown in Figure 8 is derived by dividing the Electronic Dynamic Range by the Readout Noise, which is the sensor sensitivity limit. Figure 7. Measurement results of the sensor electronic dynamic range. Engineering Models (in red) and Flight Models (in blue). Each count is the measurement result of a single array (band).

Figure 8. Instantanuous Dynamic Range in db units. Engineering Models (in red) and Flight Models (in blue). Each count is the measurement result of a single array (band). 3.4. Linearity Recent measurements of the sensor's deviation from linearity performed on the first Flight Model are reported here. As described above the sensor has 4 independent arrays. The experimental data from each array is depicted on a separate graph. In order to calculate the deviation from linearity the sensor was measured under illumination in a range of exposure times. By controlling the exposure time the signal corresponding to 0% to 100% of the well fill capacity is varied. A best linear fit was calculated to the signal range of 5% to 85% of the well fill. The deviation from linearity is then calculated as the largest deviation from the best linear fit per each ROC. For each column the ROC with the largest deviation from linear fit is taken. The results of the largest upper deviation and largest lower deviation are shown in Figure 9 (red lines). The specification limits are marked with flat green lines. Also shown in blue line is the median of absolute maximum deviation from linear fit for all ROC per column.

Figure 9. Deviation from lineraity in precentage of Dynamic Range measurement of 4 arrays of a single Flight Model in the range of 5%-85% of the well fill (clockwise direction starting from top left). The blue line is the median of deviations from the best fit line per column. The upper and lower red lines are the worst ROC deviation above and below the linear best fit per each column respectively. The green flat lines are the specification limits.the horizontal axis is the column number out of 2600 active columns. 3.5. Dark Current In [6] we presented the dark current measurement result for this sensor. Further investigation of the dark current yielded much lower results but the results were unstable. Measurement conditions are: 100Hz line rate (close to the lowest possible line rate), maximum integration time available at this line rate and the summing all 64 TDI levels. To minimize the contribution of temporal noise 64 samples are averaged (thus, measurement result in digital levels may be a non-integer value). A more careful observation of the results shows that most of the ROCs output is zero as shown in Figure 10. The minority of ROCs have signal of less than 1 digital level. The Dark Current of the array is an average of all ROCs. It is suggested that the signal contribution of the minority of the ROCs is noise generated by the measurement environment. Thus the dark current is below our measurement resolution of approximately 5,000 e - /sec/pixel (an image pixel signal is the accumulated from all 64 TDI levels). Figure 10. Dark current distribution of a single array measured at 100Hz line rate and maximum integration time. All 64 TDI levels are active and contribute signal. X-axis and y-axis are the sensor column and the ROC row counts respectively.

Row Row 3.6. Intense Light Effects An important quality of a visible spectra Earth observation sensor is the immunity to intense light effects that may occur due to sun reflections often seen from bright objects. First measured is the blooming effect, where an intense light on one pixel causes an abnormal response of the neighbouring pixels. In order to create a sub-pixel intense light source we use a gold mask with pinholes which was evaporated on the sensor directly. The sensor is then illuminated with a uniform light source through the pinholes. The sensor operates in a Test Mode allowing readout of pixels array without the TDI operation. The result is shown in Figure 11. The pinhole mask is not perfectly aligned with the pixels. The image area where the pinholes are better centered is located at the bottom part of the image around column 1,000. A zoom in on the same image is shown in Figure 12. Due to the variations in pinholes size and additional parameters the light intensity is not equal to all pixels. The light intensity in Figure 12 is in the range of 5 to 10 times the saturation level of the pixel which is 2048 Digital levels in the sensor test mode. As can be seen, the first neighbour pixels generate some response to the intense light of the illuminated pixel. The response is related to charge drift in the absorption layer and it is not a result of blooming due to electronic circuitry. The charge drift completely decays at a distance of the second neighbour. Cross sections of the image in vertical and horizontal axes are shown in Figure 13. No blooming effects where observed up to the light intensity levels required for "black sun" effect described hereafter. Column Figure 11. A portion of the sensor illuminated with intense uniform light source through a pinhole mask. The image includes 64 rows (in accordnace with the TDI levels) and a central portion of the sensor from column ~470 to column ~1540. The color scale represent the signal intensity where red is the saturation level and blue is the floor level. Column Figure 12. Zoom in to image shown in Figure 11 near row 61 and columns 740 to 750.

Row Figure 13. Vertical and Horizontal axes cross section of the image shown in Figure 11. The second observed intense light effect is the so called 'Black Sun' [7]. It is a known effect related to 2D CMOS image sensors of some configurations such as 4T pixels performing Correlated Double Sampling (CDS). The CDS operation samples the output at reset and then samples the output of the pixel signal and subtracts the former from the later. This operation eliminates low frequency noise. When the very intense light reaches the pixel it generates charges faster than the ability of the pixel to sink the current. In this case the pixel can't be reset. CDS operation under these conditions samples a saturation level instead of the pixels reset level and when compared to as a reference to the signal sample, which is also in saturation, the result is zero appearing as black pixel in the image. We show here measurements of this effect in TDI scanning operation. The setup for the measurement include a laser light source, a set of Neutral Density (ND) filters to allow wide range of light intensities, rotating mirror with accurately controlled rotation, which can be synchronized with the scanning speed, and the sensor. Figure 14 shows an image of an intense illumination taken with the CMOS TDI sensor in normal TDI scanning mode. The illumination spot is approximately Gaussian with the highest intensity appearing at the center. As observed, the center of the illumination disk is dark instead of bright because of the Black Sun Effect. By varying the ND filters and the Electronic Shutter (ES) exposure time, it is possible to change the spot intensity across a range of 1:63,000. The spot relative intensity defined by the ND filter and the ES is shown in Figure 15. At relative intensity of 1 the spot intensity peak is equal to the saturation photon flux. Images similar to the one shown in Figure 14 were captured with different spot intensities. A vertical cross section of the high intensity spot of each image is shown in Figure 15. Similar performance was demonstrated up to light intensity of approximately 54,000 saturation levels. At high intensities above 1,000 saturation levels our setup was limited due to minute parasitic reflections from objects in the setup causing intense sparkles. Other than the sparkles no other effect was observed. After the removal of the intense light the image was immediately recovered and there was no trace to the intense spot. Column Figure 14. An image taken at normal TDI scanning mode from the CMOS TDI sensor. X-axis and y-axis are the sensor column and the ROC row counts respectively. The disk around column 2,200 is caused by high intensity illumination. The center of the disk illuminated with the highet intensity apears black instead of white and is related to the Black Sun effect. The illumination in the center of the spot is more than 450 times the saturation level.

Row Figure 15. Cross section of high intesity illumination disks with variuos intensities. With highest order N.D. filter and almost closed E.S. the intensity is lower than the saturation as depicted by the purple line (peak intensity is about 85% of the saturation level). When the E.S. is opend to almost maximum value and slightly lower N.D. filter value the intensity is then about 54 time the saturation (see Table 2). No effect is notivced other than the spot is wider due to its Normal distribution. Further increase of intensity up to N.D. filter 1 and 98% open E.S. equivalent to approx. 5400 saturation levels shows Black sun effect where the spot center signal is dropped to the floor level and it goes up to saturation at the disk sorrounding the spot center. Table 2. Spot relative intensity as a function of the ND filter and the ES exposure portion. ND Filter Electronic Shutter Spot relative intensity (intensity of 1 equals the saturation level of the sensor) 0 98% 53720 1 98% 5372 2 98% 537.2 3 98% 53.7 3 4.9% 2.7 3.5 4.9% 0.85

4. CONCLUSIONS A very high performance CMOS TDI line scan sensor was designed providing a high QE, high MTF and high dynamic range sensor. Among the CMOS-TDI sensor advantages are the high level of integration including onchip ADC, on-chip controller, and CMOS compatible voltage levels, thus reducing the power consumption and the weight of the supporting electronics as well as providing a simple interface. In this paper we present the very recent measurement result of Flight models in comparison to the previously presented Engineering models. Flight models performance are slightly improved compared to the Engineering models. The dark current value is under the sensitivity of our test setup of approx. 5,000 e - /sec/pixel (accumulated signal from all 64 TDI levels) and cannot be accurately specified. In addition the effects of intense light spot were measured and reported. No blooming effect was noticed up to the intensity causing Black Sun effect. Black Sun effect was demonstrated at TDI mode of operation with light intensity of up to 54,000 saturation levels with no effects on the sensor beyond the intense light spot and the sensor image was immediately recovered after the removal of the intense light. Further measurements and tests will be performed to complete the Flight models characterization. REFERENCES [1] E. R. Fossum, D. B. Hondongwa, "A Review of the Pinned Photodiode for CCD and CMOS Image Sensors", IEEE Journal of The Electron Devices Society, Vol. 2, No. 3, May 2014, pp. 33-43. [2] R. Coath, J. Crooks, A. Godbeer, M. Wilson, R. Turchetta, "Advanced Pixel Architectures for Scientific Image Sensors", in Proc.Topical Workshop Electronics for Particle Physics, Paris, France, Sep.21 25, 2009, pp. 57-61 [3] R. Guidash, T. Lee, P. Lee, D. Sackett, C. Drowley, M. Swenson, L. Arbaugh, R. Hollstein, F. Shapiro, S. Domer, A 0.6 m CMOSpinned photodiode color imager technology, in Proc. Technical Digest IEEE Electron Device Meeting,Washington, DC, Dec. 7 10, 1997, pp. 927 929. [4] Kodak CCD Primer, #KCP-001, "Charge-Coupled Device (CCD) Image Sensor", Eastman Kodak Co., Microelectronics Technology Div., Rochester, NY. www.kodak.com. [5] D. Gardner, Characterizing Digital Cameras with the Photon Transfer Curve. (2002), https://www.semanticscholar.org/paper/characterizing-digital-cameras-with-the- Photon/859600415df0290714f8d928f40889f4eb6db5a2. [6] O. Cohen, N. Ben-Ari, I. Nevo, N. Shiloah, G. Zohar, E. Kahanov, M. Brumer, G. Gershon, O. Ofer, "Backside illuminated CMOS-TDI line scanner for space applications", Proceedings Volume 10562, International Conference on Space Optics ICSO 2016; [7] E. Fox, "Test & Inspection: CMOS Imaging Technology Advances" Online Quality Magazine, Apr. 5 th, 2012, https://www.qualitymag.com/articles/88478-test---inspection--cmos-imaging-technology-advances