sensors Article
Single-Photon Avalanche Diode with Enhanced NIR-Sensitivity for Automotive LIDAR Systems Isamu Takai *, Hiroyuki Matsubara, Mineki Soga, Mitsuhiko Ohta, Masaru Ogawa and Tatsuya Yamashita Toyota Central R&D Labs., Inc., 41-1, Yokomichi, Nagakute, Aichi 480-1192, Japan;
[email protected] (H.M.);
[email protected] (M.S.);
[email protected] (M.O.);
[email protected] (M.O.);
[email protected] (T.Y.) * Correspondence:
[email protected]; Tel.: +81-561-71-7795 Academic Editor: Nobukazu Teranishi Received: 20 January 2016; Accepted: 25 March 2016; Published: 30 March 2016
Abstract: A single-photon avalanche diode (SPAD) with enhanced near-infrared (NIR) sensitivity has been developed, based on 0.18 µm CMOS technology, for use in future automotive light detection and ranging (LIDAR) systems. The newly proposed SPAD operating in Geiger mode achieves a high NIR photon detection efficiency (PDE) without compromising the fill factor (FF) and a low breakdown voltage of approximately 20.5 V. These properties are obtained by employing two custom layers that are designed to provide a full-depletion layer with a high electric field profile. Experimental evaluation of the proposed SPAD reveals an FF of 33.1% and a PDE of 19.4% at 870 nm, which is the laser wavelength of our LIDAR system. The dark count rate (DCR) measurements shows that DCR levels of the proposed SPAD have a small effect on the ranging performance, even if the worst DCR (12.7 kcps) SPAD among the test samples is used. Furthermore, with an eye toward vehicle installations, the DCR is measured over a wide temperature range of 25–132 ˝ C. The ranging experiment demonstrates that target distances are successfully measured in the distance range of 50–180 cm. Keywords: avalanche photodiodes; light detection and ranging (LIDAR); single-photon avalanche diode (SPAD); time-of-flight (TOF); depth sensor; rangefinder; 3-D imaging; single-photon detector; advanced driver assistance system (ADAS)
1. Introduction Recently, advanced driver assistance systems (ADASs) have been designed and developed not only to make automobiles more comfortable, but also to reduce the number of traffic accidents [1]. Forward collision warning (FCW), autonomous emergency breaking (AEB), and pedestrian detection, to cite a few examples, rely on various technologically advanced sensors. High-accuracy ranging sensors are particularly important components of ADASs. Among ranging sensor technologies, millimeter-wave RADARs and stereo-vision cameras remain the key sensors of choice. However, these sensors do have limitations, so a number of ADAS implementations rely on the data fusion of two or more sensors. In this context, we have been developing an optical long-range sensor technology based on the time-of-flight (TOF) principle for next-generation ADASs [2–4]. This light detection and ranging (LIDAR) sensor offers a very good balance of overall performance in terms of spatial (image) resolution, field-of-view, precision, and depth range. In our LIDAR sensor, avalanche photodiodes operating in so-called Geiger mode have been employed as photodetectors (i.e., optical receivers). The device, referred to as a single-photon avalanche diode (SPAD), is a highly sensitive photodetector capable of outputting a precise and digital trigger signal upon the detection of ultralow-power signals, down to Sensors 2016, 16, 459; doi:10.3390/s16040459
www.mdpi.com/journal/sensors
Sensors 2016, 16, x
2 of 9
Sensors 2016, 16, 459 ultralow-power signals, down to the single photon level. Moreover, since the SPAD can2 ofbe9 fabricated using general complementary metal-oxide-semiconductor (CMOS) technology, SPADs, their front-end circuits, a complete signal (DSP) havegeneral been successfully and the single photon level. and Moreover, sincedigital the SPAD canprocessor be fabricated using complementary integrally designed as a low-cost system-on-a-chip implementation. Our latest LIDAR system [3] metal-oxide-semiconductor (CMOS) technology, SPADs, their front-end circuits, and a complete digital using CMOS(DSP) SPAD technology has achieved 100-m ranging even thesystem-on-a-chip strong outdoor signal the processor have been successfully and integrally designed as aunder low-cost lighting environment of 70 klux. implementation. Our latest LIDAR system [3] using the CMOS SPAD technology has achieved 100-m Although the SPAD is sensitive to lighting a single environment photon, not all photons ranging even under the strong outdoor of 70 klux. are detected. For example, photons that penetrate the SPAD are not detected. Also, photo-generated notexample, always Although the SPAD is sensitive to a single photon, not all photons areelectrons detected.doFor trigger an avalanche breakdown because it is a probabilistic process. Thus, a higher photon photons that penetrate the SPAD are not detected. Also, photo-generated electrons do not always detection is crucial forit any sensing application. Untila recently, CMOS SPAD trigger an efficiency avalanche (PDE) breakdown because is a probabilistic process. Thus, higher photon detection technologies have shown a high PDE only in the visible light band. However, in the near-infrared efficiency (PDE) is crucial for any sensing application. Until recently, CMOS SPAD technologies (NIR) band between 800 nm 1 μm, CMOS SPAD devices have in exhibited rather low(NIR) PDEs.band For have shown a high PDE onlyand in the visible light band. However, the near-infrared example, at the of interest our application (870 rather nm), the typically at been between 800 nmwavelength and 1 µm, CMOS SPADindevices have exhibited lowPDEs PDEs.have For example, the below 5%. In order to increase the measurement range and/or rate of automotive LIDAR systems wavelength of interest in our application (870 nm), the PDEs have typically been below 5%. In order that generally NIR light sources, much higher PDELIDAR in the NIR band is highly desirable. to increase theemploy measurement range and/orarate of automotive systems that generally employ More generally, however, the ranging performance depends on the overall sensitivity, which, in NIR light sources, a much higher PDE in the NIR band is highly desirable. More generally, however, turn, is characterized by the product of the PDE and the fill factor (FF) of the SPAD. Therefore, key the ranging performance depends on the overall sensitivity, which, in turn, is characterized by the to realizing a high-performance automotive LIDAR is having a high PDE in the NIR product of the PDE and the fill factor (FF) of the SPAD.system Therefore, key toboth realizing a high-performance band and a high FF. automotive LIDAR system is having both a high PDE in the NIR band and a high FF. Significant Significantimprovements improvementshave haverecently recentlybeen been achieved achievedin in the the NIR NIR PDE PDE [5,6]. [5,6]. For For example, example, the the current state-of-the-art PDE of 20% at 870 nm has been obtained with an excess bias (V E) of 12 V [6]. current state-of-the-art PDE of 20% at 870 nm has been obtained with an excess bias (VE ) of 12 V [6]. Moreover, fill factor factorof of21.6% 21.6%has hasbeen beenreported reported [7]. This paper presents a new SPAD structure Moreover, aa fill [7]. This paper presents a new SPAD structure that that achieves a similar PDE performance to that reported in [6], while also improving the sensitivity achieves a similar PDE performance to that reported in [6], while also improving the sensitivity by by optimizing FF. Furthermore, the performance of SPAD this SPAD is described theofbasis of optimizing the the FF. Furthermore, the performance of this is described on the on basis various various experimental results. experimental results.
2. 2. Design Designand andChip ChipImplementation ImplementationofofNIR-Sensitivity-Enhanced NIR-Sensitivity-EnhancedSPADs SPADs Figure Figure 11 shows shows the the cross-sectional cross-sectionalstructure structureof of the the newly newly designed designedSPAD, SPAD, which which achieves achieves high high sensitivity the FF. FF. The The SPAD SPADlies liesininthe thep-epitaxial p-epitaxiallayer layerofofa sensitivity in in the the NIR NIR band band without without compromising compromising the aCMOS CMOS wafer comprises SPAD-specific custom and p-SPAD. wafer andand comprises two two SPAD-specific custom layers,layers, namelynamely n-SPADn-SPAD and p-SPAD. Isolation Isolation SPADs (i.e.,ring) a guard ring) isby achieved by the superposition existing between between adjacent adjacent SPADs (i.e., a guard is achieved the superposition of existing of p-well and p-well and deep p-well layers, which areinavailable in most modern CMOS processes. deep p-well layers, which are available most modern CMOS processes.
Figure 1. 1. Simplified Simplifiedcross-sectional cross-sectionalstructure structureofofthe theproposed proposedSPAD. SPAD.The Thetop topview viewofofthe theSPAD SPAD isis Figure shown in the how thethe SPAD forms an electrical contact with the n+the layer. shown theinset insettotoillustrate illustrate how SPAD forms an electrical contact with n+ Two layer.custom Two layers, layers, n-SPADn-SPAD and p-SPAD, are employed for thisfor SPAD, and the p-SPAD is fullyisdepleted when custom and p-SPAD, are employed this SPAD, and the p-SPAD fully depleted the SPAD is biased at or above breakdown voltage.voltage. Moreover, the p-epitaxial layer haslayer a gradient when the SPAD is biased at or its above its breakdown Moreover, the p-epitaxial has a doping profile. The thick depletion layer and doping profile efficiently collect electrons generated by gradient doping profile. The thick depletion layer and doping profile efficiently collect electrons the NIR light, which in a high NIRin PDE. generated by the NIRresults light, which results a high NIR PDE.
Sensors 2016, 16, x
3 of 9
SensorsSince 2016, 16, 459 the absorption
of 9 coefficient for NIR light is low compared to that for visible light, NIR3light can penetrate deeper before it is completely absorbed in the silicon substrate. By increasing the vertical thickness of a depletion layer, the detection probability of incident NIR light is enhanced. Since the absorption coefficient for NIR light is low compared to that for visible light, NIR light The depletion layer thickness can be increased by increasing the SPAD bias voltage. However, a can penetrate deeper before it is completely absorbed in the silicon substrate. By increasing the vertical high bias voltage potentially leads to edge breakdown, which would interfere with the photon thickness of a depletion layer, the detection probability of incident NIR light is enhanced. The depletion counting operation. To avoid edge breakdown and achieve high PDE in the NIR band, the distance layer thickness can be increased by increasing the SPAD bias voltage. However, a high bias voltage between the SPAD and guard ring was increased, which reduces the FF. Conversely, a high-FF potentially leads to edge breakdown, which would interfere with the photon counting operation. SPAD, achievable by thinning the depletion layer, has been proposed, which results in a lower PDE To avoid edge breakdown and achieve high PDE in the NIR band, the distance between the SPAD in the NIR band. Thus, previous SPAD designs have involved a tradeoff between the PDE and FF and guard ring was increased, which reduces the FF. Conversely, a high-FF SPAD, achievable by for sensing NIR lights, which represents an obstacle to higher sensitivity. thinning the depletion layer, has been proposed, which results in a lower PDE in the NIR band. Thus, The proposed structure simultaneously achieves high PDE and FF. The doping concentrations previous SPAD designs have involved a tradeoff between the PDE and FF for sensing NIR lights, of the two custom layers are optimized to obtain a high electric field for the avalanche which represents an obstacle to higher sensitivity. multiplication of photo-generated electrons. Importantly, they are carefully designed so that the The proposed structure simultaneously achieves high PDE and FF. The doping concentrations of p-SPAD layer becomes fully depleted when the SPAD is biased at or above its breakdown voltage. the two custom layers are optimized to obtain a high electric field for the avalanche multiplication of Both a high electric field and a thick depletion layer are achieved under a low bias voltage, and thus, photo-generated electrons. Importantly, they are carefully designed so that the p-SPAD layer becomes edge breakdown is avoided. As a result, the NIR PDE is enhanced without compromising the FF. fully depleted when the SPAD is biased at or above its breakdown voltage. Both a high electric field Furthermore, this structure also exploits a p-epitaxial layer that features a gradient doping and a thick depletion layer are achieved under a low bias voltage, and thus, edge breakdown is profile; i.e., a profile where the doping concentration increases with depth. Such a gradient doping avoided. As a result, the NIR PDE is enhanced without compromising the FF. profile further improves the PDE by promoting upward migration and efficient collection of Furthermore, this structure also exploits a p-epitaxial layer that features a gradient doping profile; photo-generated electrons toward the avalanche multiplication region. This technique has been i.e., a profile where the doping concentration increases with depth. Such a gradient doping profile used in image sensors to improve the quantum efficiency. It has also been reported in CMOS further improves the PDE by promoting upward migration and efficient collection of photo-generated SPADs [5]. electrons toward the avalanche multiplication region. This technique has been used in image sensors Additionally, the proposed SPAD also has a shallow p+ layer at the surface of the SPAD in to improve the quantum efficiency. It has also been reported in CMOS SPADs [5]. order to collect surface-generated carriers, which reduces its intrinsic dark noise. Thus, the Additionally, the proposed SPAD also has a shallow p+ layer at the surface of the SPAD in order to proposed structure prevents dark noise from triggering false avalanche events. collect surface-generated carriers, which reduces its intrinsic dark noise. Thus, the proposed structure Figure 2a shows a photograph of a test chip that was fabricated to characterize the proposed prevents dark noise from triggering false avalanche events. SPAD using a 0.18-μm CMOS process. The test chip contains an array of 5 × 3 SPADs and three Figure 2a shows a photograph of a test chip that was fabricated to characterize the proposed quenching/readout circuits. Only the innermost three SPADs are connected to these circuits, as SPAD using a 0.18-µm CMOS process. The test chip contains an array of 5 ˆ 3 SPADs and three shown in the figure, while the remaining 12 SPADs are used as dummies because the characteristics quenching/readout circuits. Only the innermost three SPADs are connected to these circuits, as shown of the outermost devices in a device array are generally unreliable. The distance between adjacent in the figure, while the remaining 12 SPADs are used as dummies because the characteristics of the SPADs is 25 μm. Figure 2b shows schematics of quenching and readout circuits containing the outermost devices in a device array are generally unreliable. The distance between adjacent SPADs is SPADs. 25 µm. Figure 2b shows schematics of quenching and readout circuits containing the SPADs.
(a) (b) Figure 2. 2. (a) chip containing an an array of 5of× 53 ˆ SPADs and three Figure (a)Photomicrograph Photomicrographofofa asample sample chip containing array 3 SPADs and quenching/readout circuits. Only the innermost three three SPADs are connected to the The three quenching/readout circuits. Only the innermost SPADs are connected to circuits. the circuits. remaining 12 SPADs are used as “dummy” devices; (b) a schematic of three quenching and readout The remaining 12 SPADs are used as “dummy” devices; (b) a schematic of three quenching and readout circuitscontaining containingSPADs. SPADs. circuits
Sensors 2016, 16, 459
4 of 9
Sensors 2016,2016, 16, x16, x Sensors
4 of 94 of 9
SPAD quenching and recharge is typically performed passively by high-resistivity polysilicon recharge is typically performed passively by high-resistivity polysilicon quenching and is typically performed passively bypulses high-resistivity polysilicon resistorsSPAD ofSPAD RQquenching = 300 kΩ.and When arecharge photon enters the SPAD, negative voltage with an amplitude resistors of R Q = 300 kΩ. When a photon enters the SPAD, negative voltage pulses with an resistors of R Q = 300 kΩ. When a photon enters the SPAD, negative voltage pulses an of approximately VE = VAPD ´ VBD are generated at the SPAD cathode, where VBD is thewith SPAD amplitude of approximately VE =VVAPD − VBD are generated at the SPAD cathode, where VBD VisBDthe amplitude of approximately E = VAPD − V BD are generated at the SPAD cathode, where is the breakdown voltage, VAPD is the positive bias voltage, and VE is the SPAD excess bias. The generated SPAD breakdown voltage, VAPD is the positive biasbias voltage, andand VE is Ethe SPAD excess bias.bias. TheThe SPAD positive voltage, is input the SPAD excess pulse is then breakdown capacitivelyvoltage, coupledVAPD via a 5isfFthe metal fringe capacitor (CC ) toVthe of a CMOS inverter generated pulse is then capacitively coupled via a 5 fF metal fringe capacitor (C C) to the input of a generated is then capacitively coupled a 5 fF metal fringe capacitor C) to the input of a that serves as apulse comparator. The inverter inputvia is biased to VDD_CPLG by a (C thick-oxide PMOS CMOS inverter thatthat serves as aascomparator. TheThe inverter input is biased to VDD_CPLG by by a CMOS inverter serves a comparator. inverter input is biased to VDD_CPLG transistor with a constant gate bias VG. The inverter is designed with thick-oxide transistors and isa thick-oxide PMOS transistor with a constant gate bias VG. The inverter is designed with thick-oxide thick-oxide PMOS transistor with a constant gate bias VG. The inverter is designed with thick-oxide powered by the 1.8 V supply voltage process. Positive pulses at Positive the inverter output are transistors andcore is powered by the corecore 1.8in1.8 Vthis supply voltage in this process. pulses at the transistors and is powered by the V supply voltage in this process. Positive pulses at the transformed into rectangular pulses of approximately 8.5 ns by a D-flip-flop-based monostable circuit. inverter output are are transformed intointo rectangular pulses of of approximately 8.5 8.5 ns ns by by a a inverter output transformed rectangular pulses approximately This readout circuit monostable imposes a minimum deadreadout time of circuit approximately 17 ns, regardless of time the actual D-flip-flop-based circuit. This imposes a minimum dead of D-flip-flop-based monostable circuit. This readout circuit imposes a minimum dead time of SPAD recharge time. approximately 17 ns, of the actual SPAD recharge time. approximately 17 regardless ns, regardless of the actual SPAD recharge time. 3. 3.Experimental Results Experimental Results 3. Experimental Results Figure 3 shows static SPAD current characteristics as a function of theof reverse bias voltage supplied Figure 3 shows static SPAD current characteristics as aasfunction the reverse biasbias voltage Figure 3 shows static SPAD current characteristics a function of the reverse voltage to supplied the supplied SPADto cathode. The current measurements are conductedare under dark and light conditions with the SPAD cathode. TheThe current measurements conducted under dark andand light to the SPAD cathode. current measurements are conducted under dark light anconditions integration time of 320 ms using a semiconductor parameter analyzer. From this result, the V of with an integration time of 320 ms using a semiconductor parameter analyzer. From this BDthis conditions with an integration time of 320 ms using a semiconductor parameter analyzer. From the the VSPAD BD of the proposed SPAD is approximately 20.520.5 V. V. theresult, proposed is the approximately 20.5 V.approximately result, VBD of proposed SPAD is
Figure Static SPAD current characteristics as aa as function of bias voltage supplied to to the Figure 3. 3. Static SPAD current characteristics as function of the the reverse bias voltage supplied Figure 3. Static SPAD current characteristics a function of reverse the reverse bias voltage supplied tothe the SPAD cathode. SPAD cathode. SPAD cathode.
Figure 4a shows the the microscope-based measurement system used to accurately measure the FF FF 4a shows microscope-based measurement system to accurately measure FigureFigure 4a shows the microscope-based measurement system usedused to accurately measure the the FF of of the proposed SPAD. of the proposed the proposed SPAD. SPAD.
(a) (a)
(b) (b)
Figure 4. (a)4.Measurement system for high-resolution XY mapping of the response of a of SPAD Figure (a) Measurement system for high-resolution XY mapping of photon the photon response a SPAD Figure 4. (a) Measurement system for high-resolution XY mapping of the photon response of a SPAD at at any desired optical wavelength. An An objective lenslens withwith an NA of 0.6 ×40 ×40 is used in the at any desired optical wavelength. objective an NA of and 0.6 and is used in the any desired optical wavelength. Anusing objective lens with an camera NA of 0.6 and ˆ40 used the microscope; microscope; (b) Image captured an NIR-sensitive mounted onisone of in the microscope; (b) Image captured using an NIR-sensitive camera mounted on one of microscope the microscope (b)eyepieces. Image captured using an NIR-sensitive camera mounted on one of the microscope eyepieces. eyepieces.
Sensors 2016, 16, x Sensors Sensors 2016, 2016, 16, 16, 459 x
5 of 9 55of of99
This system relies on the fact that all of the light coming out of a point source placed in the This system relies on the isfact that all of the lighton coming out ofsurface. a pointTherefore, source placed in the focalThis plane of therelies camera onto point theout sample it provides system on port the factfocused that all of thealight coming of a point source placed in the focal focal of theofcamera port iscounting focused rate ontoofa apoint on at theany sample surface. Therefore, it provides an XYplane mapping the photon SPAD desired wavelength withana plane of the camera port is focused onto a point on the sample surface.optical Therefore, it provides an XY mapping of the photon counting rate of a SPAD at any desired optical wavelength with4ba high spatial resolution. In the microscope, an objective lens with an NA of 0.6 is used. Figure XY mapping of the photon counting rate of a SPAD at any desired optical wavelength with a high high spatial resolution. In confirmed the microscope, an objective lens with an NA of demonstrates 0.6 is used. Figure 4b shows the experimentally resolution thewith proposed that the spatial resolution. In the microscope, an objectiveof lens an NA system of 0.6 isand used. Figure 4b shows the shows the experimentally confirmed resolution of the proposed system and demonstrates that the spot size of an 870-nm focused beamof is the approximately 1 μmand FWHM (full width at half maximum). experimentally confirmed resolution proposed system demonstrates that the spot size of an spot Figure size of 5anshows 870-nm focused beam is approximately 1 μm FWHM (full width at half maximum). a normalized map of the photon response of a SPAD with a resolution of 0.5 × 870-nm focused beam is approximately 1 µm FWHM (full width at half maximum). Figure 5 shows a normalized map of the photon response of a SPAD with a resolution of 0.5 × 0.5 μm, using the system shown inmap Figure 4, measured with a Vof E of 5 V and 870-nm light. The Figure 5 shows a normalized of the photon response a SPAD with a resolution of 0.5 μm, using the system shown in Figure 4, measured with a V E of 5 V and 870-nm light. The measurement area is × 30 μm, fully the 25with × 25a V μm SPAD area. This result 0.5 ˆ 0.5 µm, using the30system shown in encompassing Figure 4, measured E of 5 V and 870-nm light. measurement area is 30 × 30that μm, fully encompassing the 25a high × 25 FF μm SPAD FWHM. area. This result experimentally demonstrates the proposed SPAD achieves of The measurement area is 30 ˆ 30 µm, fully encompassing the 25 ˆ 25 µm 33.1% SPAD area. This result experimentally demonstrates that the proposed SPAD achieves a high FF of 33.1% FWHM. experimentally demonstrates that the proposed SPAD achieves a high FF of 33.1% FWHM.
Figure 5. Normalized photon counting rate map of the proposed 25 × 25 μm2 SPAD. VE is 5 V and Figure 5. Normalized photon counting rate map of the proposed 25 ˆ 25 µm2 SPAD. V is 5 V and the Figure 5. Normalized counting rate map of the proposed 25 × 25 μm2 SPAD.E VE is 5 V and the incident cone of 870photon nm light is approximately 74°. incident cone of 870 nm light is approximately 74˝ . the incident cone of 870 nm light is approximately 74°.
Figure 6a shows a light source, an integrating sphere, and a reference photodiode based Figure shows light an integrating sphere, and aa reference photodiode based Figure 6a 6asystem showstoaameasure light source, source, an of integrating sphere, and reference photodiode measurement the PDE the proposed SPAD. A light source radiates lightbased with measurement system to measure the PDE of the proposed SPAD. A light source radiates light measurementranging system from to measure the PDE of the SPAD. A light source radiates light with with wavelengths 350 to 1100 nm into theproposed integrating sphere. wavelengths wavelengths ranging ranging from from 350 350to to1100 1100nm nminto intothe theintegrating integratingsphere. sphere.
(a) (a) (b) (b) Figure 6. 6. Measurement This system system Figure Measurement system system (a) (a) used used to to determine determine the the PDE PDE of of the the proposed proposed SPAD. SPAD. This Figure 6.mainly Measurement system (a) used to determine the aPDE of the proposed SPAD. This system consists of a light source, an integrating sphere, and photodetector connected to a radiometer; consists mainly of a light source, an integrating sphere, and a photodetector connected to a radiometer; consists mainly of a light source, an integrating and of a photodetector connected to a radiometer; (b) PDE as as aa function function of light light wavelength at an ansphere, excess bias bias V. (b) PDE of wavelength at excess of 55 V. (b) PDE as a function of light wavelength at an excess bias of 5 V.
Sensors 2016, 16, 459
6 of 9
Sensors 2016, 16, x
6 of 9
Figure 6b shows the measured PDE of the proposed SPAD as a function of light wavelength with Figure 6b shows the measured PDE of the proposed SPAD as a function of light wavelength a VEwith of 5 aV.VThe proposed SPAD has a peak PDE of 62.2% at 610 nm and achieves a high PDE of 19.4% E of 5 V. The proposed SPAD has a peak PDE of 62.2% at 610 nm and achieves a high PDE at 870 which the which laser light of our LIDAR system. the best knowledge, of nm, 19.4% at 870isnm, is thewavelength laser light wavelength of our LIDARTo system. To of theour best of our this knowledge, is the highest PDE yet achieved for a CMOS SPAD under similar V bias conditions. E this is the highest PDE yet achieved for a CMOS SPAD under similar VE bias Figure 7 shows dark count rate (DCR) measurement results. The dark counts are caused by false conditions. output pulses, aredark mainly induced by intrinsic dark noise in The the SPAD, despite absence Figure which 7 shows count rate (DCR) measurement results. dark counts arethe caused by of output A pulses, whichworsens are mainly induced performance, by intrinsic dark noise the SPAD, despite lightfalse incidence. high DCR the ranging since falseinpulses obscure true the pulses absence light incidence. high Figure DCR worsens the DCRs ranging false generated byofthe desired laserAlight. 7a shows asperformance, a function ofsince VE for 18pulses SPADobscure samples in true pulses generated by the desired laser light. Figure 7a shows DCRs as a function of VE for 18 six test chips at room temperature. As seen in this figure, the DCR values vary greatly among samples. SPAD samples in six test chips at room temperature. As seen in this figure, the DCR values vary Such large variances in DCR are common, as evidenced by [5,6]. At a VE of 5 V, DCRs of half of SPAD greatly among samples. Such large variances in DCR are common, as evidenced by [5,6]. At a VE of samples are less than 100 counts per second (cps), and the maximum DCR is 12.672 kcps, i.e., false 5 V, DCRs of half of SPAD samples are less than 100 counts per second (cps), and the maximum pulses are generated in cycles of 78.914 µs. However, these DCR values have little impact on the DCR is 12.672 kcps, i.e., false pulses are generated in cycles of 78.914 μs. However, these DCR ranging performance. For example, since a 150 m ranging operation requires only 1 µs measurement values have little impact on the ranging performance. For example, since a 150 m ranging operation time,requires dark counts false pulses) time, only affect one operation approximately 78 operations, only 1(i.e., μs measurement dark counts (i.e., false in pulses) only affectevery one operation in evenapproximately if the worst SPAD sample is used. Figure 7b shows the DCR as a function of chip temperature every 78 operations, even if the worst SPAD sample is used. Figure 7b shows the ˝ C. As the figure shows, the DCR grows exponentially with chip temperature, overDCR the range of 25–132 as a function of chip temperature over the range of 25–132 °C. As the figure shows, the DCR as dark noise is directly with correlated with temperature. grows exponentially chip temperature, as dark noise is directly correlated with temperature.
(a)
(b) Figure 7. (a) Measurement results of DCR as a function of excess bias VE, using 18 SPAD samples in
Figure 7. (a) Measurement results of DCR as a function of excess bias VE , using 18 SPAD samples in 6 test chips at room temperature; (b) Measurement results of DCR as a function of chip temperature, 6 test chips at room temperature; (b) Measurement results of DCR as a function of chip temperature, using one SPAD sample. using one SPAD sample.
The room-temperature DCR is less than 1 kcps. When the chip temperature increases, however, TheDCR room-temperature is less When theatchip increases, however, the reaches 10 kcps DCR at 65 °C, 100than kcps1atkcps. 85 °C, 1 Mcps 110 temperature °C, and 6 Mcps at 132 °C. As
the DCR reaches 10 kcps at 65 ˝ C, 100 kcps at 85 ˝ C, 1 Mcps at 110 ˝ C, and 6 Mcps at 132 ˝ C. As noted previously, a 150 m ranging operation is completed in 1 µs, so the false pulse affects every ranging
Sensors Sensors 2016, 2016, 16, 16, x459
of 99 77 of
noted previously, a 150 m ranging operation is completed in 1 μs, so the false pulse affects every ˝ C. However, our final chip will implement spatiotemporal correlation technology operation above 110 ranging operation above 110 °C. However, our final chip will implement spatiotemporal correlation based on a macro-pixel structure [3], which[3], will effectively suppresssuppress the impact DCR of at DCR high technology based on a macro-pixel structure which will effectively theof impact temperatures. The present measurement results results offer important insightinsight for thefor application of the at high temperatures. The present measurement offer important the application proposed SPAD to actual automotive systems that must operate at high temperatures. of the proposed SPAD to actual automotive systems that must operate at high temperatures. Figure 8a 8a shows shows the the elementary elementary system system used used for for measurements measurements of of the the TOF TOF and and target target distances distances Figure using the the proposed proposed SPAD. SPAD. Mainly, Mainly, this experiment experiment ensures ensures that that the the proposed proposed SPAD SPAD can can accurately accurately using respond to narrow and high-speed laser pulses. This is a first step in the development of a respond to narrow and high-speed laser pulses. This is a first step in the development of a LIDAR LIDAR The system. Theelementary present elementary systemofconsists of a picosecond laser (λ = 635 nm, system. present system consists a picosecond laser source (λ =source 635 nm, FWHM ≈ FWHM 100 ps), test chip containing V is supplied, an oscilloscope 100 ps), « a test chipa containing SPADs toSPADs whichtoawhich VAPDa VAPD of 25.5ofV25.5 is supplied, an oscilloscope (25 (25 GSPS) to record trigger signals from laser source and outputsignals signalsofofthe theSPAD SPADwith withaa period period GSPS) to record trigger signals from thethe laser source and output of 25 ns, as as a measurement target, and an PC. When pulsed light reflected of ns,a areflector reflector a measurement target, andoffline an offline PC. the When the laser pulsed laser light by the reflector enters the SPAD in the test chip, a signal formed to an 8.5 ns PW is output from the reflected by the reflector enters the SPAD in the test chip, a signal formed to an 8.5 ns PW is output chip. Finally, the PC superimposes approximately 5000 frames of the output signal data recorded by from the chip. Finally, the PC superimposes approximately 5000 frames of the output signal data the oscilloscope produces a and TOF produces image similar to animage eye diagram. recorded by theand oscilloscope a TOF similarIntothis anmeasurement, eye diagram.theIntarget this distance ranges from 50 to 180 cm. Figure 8b shows representative TOF images at 50, 90, 130, and measurement, the target distance ranges from 50 to 180 cm. Figure 8b shows representative TOF 170 cm.atAs in the theAs pulse (td ) the based on delay the TOF at (t 50d) cm increases images 50,seen 90, 130, andfigure, 170 cm. seendelay in thetime figure, pulse time based on the linearly TOF at with target distance. 50 cmincreasing increases linearly with increasing target distance.
(a) (b) Figure Figure 8. 8. (a) (a) Measurement Measurement system system for for the the TOF TOF and and target target distances; distances; (b) (b) TOF TOF images images superimposing superimposing 5000-frame SPAD out signals at 50, 90, 130, and 170 cm. From the TOF images, each 5000-frame SPAD out signals at 50, 90, 130, and 170 cm. From the TOF images, each delay delay time time (t (tdd)) over various distances is measured, based on the result at 50 cm. over various distances is measured, based on the result at 50 cm.
Figure showsmeasured measureddistances distancestoto target reflector at room temperature, the Figure 99 shows thethe target reflector at room temperature, basedbased on theondelay delay d_50–td_180) measured by the system in Figure 8. Figure 9a shows distances calculated times times (td_50 –t(td_180 ) measured by the system in Figure 8. Figure 9a shows distances calculated from from each t d as a function of target distance (the actual distance to the target reflector). As shown in each td as a function of target distance (the actual distance to the target reflector). As shown in the the figure, the measured distances are comparable to the target actualdistances. target distances. Figure shows figure, the measured distances are comparable to the actual Figure 9b shows9bthe error the error as a function of target distance. As seen in Figure 9b, the errors are all less than ±2%. These as a function of target distance. As seen in Figure 9b, the errors are all less than ˘2%. These results results demonstrate the proposed cantorespond to narrow and laser high-speed lasercan pulses demonstrate that the that proposed SPAD canSPAD respond narrow and high-speed pulses and thus and can thus be used for ranging systems. be used for ranging systems.
Sensors 2016, 16, 459 Sensors 2016, 16, x
8 of 9 8 of 9
(a)
(b)
Figure Figure 9. 9. Measurement Measurementresults resultsof of target target distances distances at at room room temperature. temperature. (a) (a) Measured Measured distances distances to to the target reflector as a function of actual target distance, based on the result obtained at 50 cm; (b) the target reflector as a function of actual target distance, based on the result obtained at 50 cm; Measurement errors as aas function of target distance, based on the result obtained at 50 (b) Measurement errors a function of target distance, based on the result obtained at cm. 50 cm.
4. 4. Conclusions Conclusions This This paper paper presents presents aa new new SPAD SPAD structure structure to to achieve achieve aa higher-performance, higher-performance, e.g., e.g., longer-range, longer-range, higher-precision, and higher-measurement-rate, LIDAR system for future ADASs. higher-precision, and higher-measurement-rate, LIDAR system for future ADASs. In In the the proposed proposed SPAD, SPAD, two two custom custom layers, layers, p-SPAD p-SPAD and and n-SPAD, n-SPAD, are are employed employed using using the the existing existing 0.18-μm 0.18-µm CMOS CMOS technology. The doping concentrations of the customized layers are optimized to achieve a technology. The doping concentrations of the customized layers are optimized to achieve a high high electric electric field, field, while while the the p-type p-type layer layer becomes becomes fully fully depleted depleted when when the the device device is is biased biased to to operate operate in in Geiger Geiger mode. mode. This This full-depletion full-depletionlayer layer enables enables the the efficient efficient collection collection of of electrons electrons generated generated by by NIR NIR light, light, without without compromising compromising the the FF. FF. Moreover, Moreover, the the p-epitaxial p-epitaxial layer, layer, which which has has aa gradient gradient doping doping profile, profile, further further enhances enhances the the NIR NIR PDE. PDE. In In this this study, study, various various experiments experiments are are conducted conducted to to characterize characterizethe theproposed proposedSPAD. SPAD. In In SPAD SPAD PDE measurements, PDEs of 62.2% at 610 nm and 19.4% at 870 nm are attained. To the best PDE measurements, PDEs of 62.2% at 610 nm and 19.4% at 870 nm are attained. To the best of of our our knowledge, knowledge, this this is is the the highest highest PDE PDE achieved achieved for for aa CMOS CMOS SPAD SPAD under under similar similar excess excess bias bias conditions, conditions, i.e., the actual actual FF measured by i.e., 55 V. V. Moreover, Moreover, the FF is is measured by the the special special measurement measurement system, system, and and aa high high FF FF of of 33.1% is obtained at 870 nm. Since SPAD sensitivity is characterized by the product of the PDE and 33.1% is obtained at 870 nm. Since SPAD sensitivity is characterized by the product of the PDE and FF, FF, proposed structure imparts the SPAD with higher NIR sensitivity, eliminating the PDE-FF the the proposed structure imparts the SPAD with higher NIR sensitivity, eliminating the PDE-FF tradeoff. tradeoff. The DCR levels of the proposed SPAD have only a small effect on the ranging The DCR levels of the proposed SPAD have only a small effect on the ranging performance, even when performance, even whenthe thetest worst SPAD theDCR test samples is used. The at DCR the worst SPAD among samples is among used. The measurement results highmeasurement temperature results at high temperature reveal installations. new insightThe forranging actual experiment vehicle installations. Thethat ranging reveal new insight for actual vehicle demonstrates target experiment demonstrates that target distances are successfully measured within an error of ±2%. distances are successfully measured within an error of ˘2%. We believe that the proposed SPAD will We that proposed SPAD will pave theADAS way for higher-performance LIDAR and ADAS pavebelieve the way forthe higher-performance LIDAR and systems. systems. Acknowledgments: This paper is based on results obtained from a project commissioned by the New Energy and Industrial Technology Development Organization would like commissioned to express our deepest gratitude to Acknowledgments: This paper is based on results (NEDO). obtained We from a project by the New Energy Cristiano Niclass. Advice and comments by Kota Ito (Toyota Central R&D Labs, Inc.) have been a great help. and Industrial Technology Development Organization (NEDO). We would like to express our deepest Author Contributions: M. Soga and T. Yamashita contributed the realization of theLabs, SPAD with enhanced gratitude to Cristiano Niclass. Advice and comments by Kota Itoto(Toyota Central R&D Inc.) have been a NIR-sensitivity; I. Takai, H. Matsubara, M. Soga, and M. Ogawa conceived and designed the experiments; I. Takai great help. and H. Matsubara performed the experiments; I. Takai, H. Matsubara, M. Ohta, and M. Ogawa contributed to the development of the systems for the authors contributed to analyses andSPAD considerations of the Author Contributions: M. Soga andexperiments; T. YamashitaAll contributed to the realization of the with enhanced experimental results; I. Takai wrote the paper. NIR-sensitivity; I. Takai, H. Matsubara, M. Soga, and M. Ogawa conceived and designed the experiments; I. Takai andofH. Matsubara performed thenoexperiments; I. Takai, H. Matsubara, M. Ohta, and M. Ogawa Conflicts Interest: The authors declare conflict of interest. contributed to the development of the systems for the experiments; All authors contributed to analyses and considerations References of the experimental results; I. Takai wrote the paper.
Conflicts of Interest: The authors conflict interest. 1. Ninomiya, Y. Special feature:declare Active no safety. R&DofRev. Toyota CRDL 2012, 43, 1–6. 2. Matsubara, H.; Soga, M.; Niclass, C.; Ito, K.; Aoyagi, I.; Kagami, M. Development of next generation LIDAR. References R&D Rev. Toyota CRDL 2012, 43, 7–12. 1. 2.
Ninomiya, Y. Special feature: Active safety. R&D Rev. Toyota CRDL 2012, 43, 1–6. Matsubara, H.; Soga, M.; Niclass, C.; Ito, K.; Aoyagi, I.; Kagami, M. Development of next generation LIDAR. R&D Rev. Toyota CRDL 2012, 43, 7–12.
Sensors 2016, 16, 459
3. 4.
5.
6. 7.
9 of 9
Niclass, C.; Soga, M.; Matsubara, H.; Ogawa, M.; Kagami, M. A 0.18-µm CMOS SoC for a 100-m-range 10-frame/s 200 ˆ 96-pixel time-of-flight depth sensor. IEEE J. Solid-State Circuits 2014, 49, 315–330. [CrossRef] Niclass, C.; Matsubara, H.; Soga, M.; Ohta, M.; Ogawa, M.; Yamashita, T. A NIR-sensitivity-enhanced single-photon avalanche diode in 0.18 µm CMOS. In Proceedings of the International Image Sensor Workshop, Vaals, The Netherlands, 8–11 June 2015; pp. 340–343. Webster, E.A.G.; Richardson, J.A.; Grant, L.A.; Renshaw, D.; Henderson, R.K. A Single-Photon Avalanche Diode in 90-nm CMOS Imaging Technology With 44% Photon Detection Efficiency at 690 nm. IEEE Electron Device Lett. 2012, 33, 694–696. [CrossRef] Webster, E.A.G.; Grant, L.A.; Henderson, R.K. A High-Performance Single-Photon Avalanche Diode in 130-nm CMOS Imaging Technology. IEEE Electron Device Lett. 2012, 33, 1589–1591. [CrossRef] Webster, E.A.G.; Walker, R.J.; Henderson, R.K.; Grant, L.A. A silicon photomultiplier with >30% detection efficiency from 450 to 750 nm and 11.6 µm pitch NMOS-only pixel with 21.6% fill factor in 130 nm CMOS. In Proceedings of the European Solid-State Device Research Conference (ESSDERC), Bordeaux, France, 17–21 September 2012; pp. 238–241. © 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons by Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).
