Download Single photon detection based devices and techniques for pulsed

C555etukansi.kesken.fm Page 1 Wednesday, November 11, 2015 9:42 AM C 555 OULU 2015 UNIVERSITY OF OUL U P.O. Box 8000 FI-90014 UNIVERSITY OF OULU FINLA ND U N I V E R S I TAT I S O U L U E N S I S ACTA A C TA C 555 ACTA UN NIIVVEERRSSIITTAT ATIISS O OU ULLU UEEN NSSIISS U Lauri Hallman University Lecturer Santeri Palviainen Postdoctoral research fellow Sanna Taskila Professor Olli Vuolteenaho Lauri Hallman Professor Esa Hohtola SINGLE PHOTON DETECTION BASED DEVICES AND TECHNIQUES FOR PULSED TIME-OF-FLIGHT APPLICATIONS University Lecturer Veli-Matti Ulvinen Director Sinikka Eskelinen Professor Jari Juga University Lecturer Anu Soikkeli Professor Olli Vuolteenaho Publications Editor Kirsti Nurkkala ISBN 978-952-62-1043-8 (Paperback) ISBN 978-952-62-1044-5 (PDF) ISSN 0355-3213 (Print) ISSN 1796-2226 (Online) UNIVERSITY OF OULU GRADUATE SCHOOL; UNIVERSITY OF OULU, FACULTY OF INFORMATION TECHNOLOGY AND ELECTRICAL ENGINEERING, DEPARTMENT OF ELECTRICAL ENGINEERING; INFOTECH OULU C TECHNICA TECHNICA ACTA UNIVERSITATIS OULUENSIS C Te c h n i c a 5 5 5 LAURI HALLMAN SINGLE PHOTON DETECTION BASED DEVICES AND TECHNIQUES FOR PULSED TIME-OF-FLIGHT APPLICATIONS Academic dissertation to be presented with the assent of the Doctoral Training Committee of Technology and Natural Sciences of the University of Oulu for public defence in the OP auditorium (L10), Linnanmaa, on 18 December 2015, at 12 noon U N I VE R S I T Y O F O U L U , O U L U 2 0 1 5 Copyright © 2015 Acta Univ. Oul. C 555, 2015 Supervised by Academy Professor Juha Kostamovaara Reviewed by Professor Horst Zimmermann Professor Ari Paasio Opponent Docent Markku Åberg ISBN 978-952-62-1043-8 (Paperback) ISBN 978-952-62-1044-5 (PDF) ISSN 0355-3213 (Printed) ISSN 1796-2226 (Online) Cover Design Raimo Ahonen JUVENES PRINT TAMPERE 2015 Hallman, Lauri, Single photon detection based devices and techniques for pulsed time-of-flight applications. University of Oulu Graduate School; University of Oulu, Faculty of Information Technology and Electrical Engineering, Department of Electrical Engineering; Infotech Oulu Acta Univ. Oul. C 555, 2015 University of Oulu, P.O. Box 8000, FI-90014 University of Oulu, Finland Abstract In this thesis, a new type of laser diode transmitter using enhanced gain-switching suitable for use with a single photon avalanche diode (SPAD) detector was developed and tested in the pulsed time-of-flight laser range finding (lidar) application. Several laser diode versions were tested and the driving electronics were developed. The driving electronics improvements enabled a pulsing frequency of up to 1 MHz, while the maximum laser output power was about 5–40 W depending on the laser diode dimensions. The large output power is advantageous especially in conditions of strong photon noise emerging from ambient light outdoors. The length of the laser pulse matches the jitter of a typical SPAD detector providing several advantages. The new laser pulser structure enables a compact rangefinder for 50 m distance measurement outdoors in sunny conditions with sub-centimeter precision (σ-value) at a valid distance measurement rate of more than 10 kHz, for example. Single photon range finding techniques were also shown to enable a char bed level measurement of a recovery boiler containing highly attenuating and dispersing flue gas. In addition, gated single photon detector techniques were shown to provide a rejection of fluorescent photons in a Raman spectroscope leading to a greatly improved signal-to-noise ratio. Photonic effects were also studied in the case of a pulsed time-of-flight laser rangefinder utilizing a linear photodetector. It was shown that signal photon noise has an effect on the optimum detector configuration, and that pulse detection jitter can be minimized with an appropriate timing discriminator. Keywords: 3D imaging, laser pulser, pulsed time-of-flight, Raman spectroscopy, range finding, recovery boiler char bed, SPAD, timing discrimination Hallman, Lauri, Yksittäisten fotonien ilmaisuun perustuvia tekniikoita ja laitteita pulssinkulkuaikamittaussovelluksiin. Oulun yliopiston tutkijakoulu; Oulun yliopisto, Tieto- ja sähkötekniikan tiedekunta, Sähkötekniikan osasto; Infotech Oulu Acta Univ. Oul. C 555, 2015 Oulun yliopisto, PL 8000, 90014 Oulun yliopisto Tiivistelmä Tässä työssä kehitettiin uudentyyppinen, tehostettua "gain-switchingiä" hyödyntävä laserdiodilähetin käytettäväksi yksittäisten fotonien avalanche-ilmaisimien (SPAD) kanssa, ja sitä testattiin pulssin lentoaikaan perustuvassa laseretäisyysmittaussovelluksessa. Useita laserdiodiversioita testattiin ja ohjauselektroniikkaa kehitettiin. Ohjauselektroniikan parannukset mahdollistivat jopa 1 MHz pulssitustaajuuden, kun taas laserin maksimiteho oli noin 5–40 W riippuen laserdiodin dimensioista. Suuri lähtöteho on edullinen varsinkin vahvoissa taustafotoniolosuhteissa ulkona. Laserpulssin pituus vastaa tyypillisen SPAD-ilmaisimen jitteriä tarjoten useita etuja. Uusi laserpulssitinrakenne mahdollistaa esimerkiksi kompaktin etäisyysmittarin 50 m mittausetäisyydelle ulkona aurinkoisessa olosuhteessa mm–cm -mittaustarkkuudella (σ-arvo) yli 10 kHz mittaustahdilla. Yksittäisten fotonien lentoaikamittaustekniikan osoitettiin myös mahdollistavan soodakattilan keon korkeuden mittauksen, jossa on voimakkaasti vaimentavaa ja dispersoivaa savukaasua. Lisäksi portitetun yksittäisten fotonien ilmaisutekniikan osoitettiin hylkäävän fluoresenssin synnyttämiä fotoneita Raman-spektroskoopissa, joka johtaa selvästi parempaan signaali-kohinasuhteeseen. Fotoni-ilmiöitä tutkittiin myös lineaarista valoilmaisinta hyödyntävän pulssin kulkuaikamittaukseen perustuvan lasertutkan tapauksessa. Osoitettiin, että signaalin fotonikohina vaikuttaa optimaaliseen ilmaisinkonfiguraatioon, ja että pulssin ilmaisujitteri voidaan minimoida sopivalla ajoitusdiskriminaattorilla. Asiasanat: 3D-kuvantaminen, ajoitusdiskriminaatio, etäisyysmittaus, laserpulssitin, pulssin kulkuaikamittaus, Raman-spektroskopia, soodakattilan keko, SPAD Acknowledgements First I would like to thank Professor Juha Kostamovaara for providing me with this incredible and diverse journey into the world of optoelectronics. Throughout the years he has given thoughtful comments and has been a truly excellent supervisor. Secondly, I would like to thank all the current and former lab staff for the nice work atmosphere. To name a few, thanks to Brigitte Lanz, Jaakko Huikari, Sähba Jahromi and Dr. Ari Kilpelä for pleasant cooperation, University Lecturer Kari Määttä and Dr. Sergei Vainshtein for all the advice, and Matti Polojärvi for technical help. Some of the work was done with people from other institutions. Out of these I would especially like to thank Hannu Lindström, Martin Kögler and Jussi Tenhunen at VTT for the good collaboration. Without the expertise and hard work by Dr. Boris Ryvkin from Ioffe Physico-Technical Institute this project would probably not have been as successful. This work has been financially supported by the the Academy of Finland, the Finnish Funding Agency for Technology and Innovation (TEKES), Infotech Oulu and the Tauno Tönning Foundation, all of which are gratefully acknowledged. Thanks to all of the reviewers of the publications and the pre-examiners of this thesis: Professor Horst Zimmermann from Vienna University of Technology and Professor Ari Paasio from the Technology Research Center of the University of Turku. Gordon Roberts revised the language of this thesis. Outside work, thanks to Makke and Masa for various good sports and gaming sessions in Oulu. On the other hand it is always fun with my friends in Helsinki during my visits there. Thanks for the hospitality to my sister and brother Anne and Anssi in Helsinki. Also the uplifting stopovers by Benjamin the dog should be mentioned. Thanks to my parents Mikko and Marjo-Riitta for support, delicious dinners and relaxing sauna nights. Last but not least, thanks to Sanna for her cheerful encouragement. 16.10.2015 Lauri Hallman 7 8 List of terms, symbols and abbreviations Jitter Accuracy Precision 1D 3D AlAs APD CCD CMOS COG ESL FC/APC FC/PC FET FWHM GaAs HP HR InGaAs InP LD LR MEMS MOSFET NIR NTC OCL ORC PCB PDP the short-term deviations of the significant instants of a signal from their ideal positions in time (IEEE 1996) the degree of correctness with which a measured value agrees with the true value (IEEE 1996) the quality of coherence or repeatability of measurement data, customarily expressed in terms of the standard deviation of the extended set of measurement results (IEEE 1996) 1-dimensional 3-dimensional aluminum arsenide avalanche photodiode charge-coupled device complementary metal oxide semiconductor center of gravity equivalent series inductance fiber channel/angled physical contact fiber channel/physical contact field-effect transistor full-width-at-half-maximum gallium arsenide high-pass high-reflectivity indium gallium arsenide indium phosphide laser diode low-reflectivity microelectromechanical systems metal–oxide–semiconductor field-effect transistor near-infrared negative temperature coefficient optical confinement layer Optoelectronics Research Centre (Tampere, Finland) printed circuit board photon detection probability 9 PIN PMT pn PWL RLC rms SMA SBR SPAD TDC TOF p-type/intrinsic/n-type photomultiplier tube p-type/n-type piecewise linear resistor-inductor-capacitor root mean square SubMiniature version A signal-to-background ratio of a single photon counting experiment (defined here as the ratio of average number of detected signal counts to the standard deviation of background count bin populations, with bin widths equal to the signal width Tsignal) signal-to-noise ratio, defined here as the peak signal amplitude to rms noise value ratio (when using an analog detector) single photon avalanche diode time-to-digital converter time-of-flight BG BW c C Ccharge d Epulse f flens F fraccounts G h i(t) is(t) I0 Iamplitude Ioperation Ithreshold k total noise count rate per second bandwidth speed of light capacitance capacitor in the current pulsing circuit thickness of active layer of laser diode laser pulse energy pulsing frequency focal length of the collimating lens excess noise factor fraction of laser pulse detections per number of sent laser pulses gain Planck constant photodetector output current average of the photodetector output current current coefficient current pulse amplitude operating current level lasing threshold current instantaneous photoelectron number SNR 10 L Lstray Lstray_Rdamp Lstray_total M n N nlaser nphot nphotoel hd Pav Prcvd q R Rbias Rdamp rdet RF Rtot t T T0 tper Tsignal tw v Vbias vmax vnoise_rms Vth v(t) w x Z laser diode chip length stray inductance in the current pulsing circuit stray inductance of Rdamp total stray inductance in the current pulsing circuit avalanche photodiode internal amplification refractive index number of averaged signal counts number of laser pulses sent the average number of photons reflecting back from a diffuse target towards a detector lens number of photoelectrons single electron response function average output power of laser diode the average power illuminating the photodetector charge of an electron intrinsic responsivity of the photodetector at gain M=1 resistor through which Ccharge is charged in the current pulser resistor in the current pulsing circuit detector lens radius feedback resistor total resistance in the current loop time temperature characteristic temperature of a laser diode pulsing period full width of laser pulse width of current pulse through laser diode frequency of radiation drain bias voltage in current pulsing circuit signal waveform peak voltage rms voltage noise detection voltage level signal waveform laser diode oxide stripe width the direction in the laser diode from the wire bonding contact towards the substrate target distance 11 α αi βi Γa ΔV Δt ε λ σ σN τ τHP γ(Z) 12 divergence of collimated beam electron ionization rate hole ionization rate optical confinement factor voltage change over switch S1 time interval between start- and stop-pulse diffuse reflectivity of the target central wavelength of the laser standard deviation of collected count times measurement precision after averaging N signal counts efficiency of the optics high-pass discrimination timing constant overlap of the input and output beams as a function of distance Z of the target List of original publications This thesis is based on the following publications, which are referred to throughout the text by their Roman numerals: I II III IV V VI VII VIII Hallman L W & Kostamovaara J (2008) Effect of signal quantum shot noise on the jitter of leading edge detected and high-pass discriminated laser pulse signals. Measurement Science and Technology 19(8): 085303. Hallman L W, Ryvkin B, Haring K, Ranta S, Leinonen T & Kostamovaara J (2010) Asymmetric waveguide laser diode operated in gain switching mode with high-power optical pulse generation. Electronics Letters 46(1): 65–66. Nissinen I, Nissinen J, Länsman A K, Hallman L, Kilpelä A, Kostamovaara J, Kögler M, Aikio M & Tenhunen J (2011) A sub-ns time-gated CMOS single photon avalanche diode detector for Raman spectroscopy. Proceedings of the European Solid-State Device Research Conference (ESSDERC). Helsinki, Finland: 375–378. Hallman L W, Haring K, Toikkanen L, Leinonen T, Ryvkin B S & Kostamovaara J T (2012) 3 nJ, 100 ps laser pulses generated with an asymmetric waveguide laser diode for a single-photon avalanche diode time-of-flight (SPAD TOF) rangefinder application. Measurement Science and Technology 23(2): 025202. Hallman L W, Haring K, Toikkanen L, Leinonen T, Ryvkin B S & Kostamovaara J T (2012) Compact laser pulser for TOF SPAD rangefinder application. Proceedings of SPIE Photonics West. San Francisco CA. Hallman L W & Kostamovaara J (2014) Note: Detection jitter of pulsed time-of-flight lidar with dual pulse triggering. Review of Scientific Instruments 85(3): 036105. Hallman L W, Huikari J & Kostamovaara J (2014) A high-speed/power laser transmitter for single photon imaging applications. Proceedings of IEEE Sensors. Valencia, Spain: 1157–1160. Kostamovaara J, Huikari J, Hallman L, Nissinen I, Nissinen J, Rapakko H, Avrutin E & Ryvkin B (2015) On laser ranging based on high-speed/energy laser diode pulses and single-photon detection techniques. IEEE Photonics Journal 7(2): 7800215. The author of this thesis set up and conducted the measurements for papers I, II and IV–VI, and participated in the measurements for papers III, VII and VIII. Papers I, II and IV–VII were written by the author with the help of the co-authors. For paper III, the main contributions of the author were the characterization of the SPAD device and the design of the gating circuitry. The new laser diode gain-switching principle, first tested in paper II, was invented by Dr. Boris Ryvkin, Dr. Eugene Avrutin and Prof. Juha Kostamovaara. Some previously unpublished new results are also presented in this thesis. 13 14 Contents Abstract Tiivistelmä Acknowledgements 7 List of terms, symbols and abbreviations 9 List of original publications 13 Contents 15 1 Introduction 17 2 Photonic effects in pulsed time-of-flight range finding 19 2.1 Pulsed TOF range finding using a linear detector ................................... 19 2.2 Pulsed TOF range finding using a single photon detector ...................... 24 2.3 Avalanche multiplication ........................................................................ 28 2.3.1 Single photon avalanche diode ..................................................... 29 2.3.2 Avalanche photodiode .................................................................. 31 3 High speed/power ~100 ps laser pulse transmitter for single photon TOF measurements 33 3.1 Enhanced laser diode structure to optimize gain-switching performance for high-power optical pulse generation ............................ 34 3.2 High frequency current pulsing circuit ................................................... 37 3.3 Power waveform measurements ............................................................. 42 3.4 Far-field measurements ........................................................................... 45 3.5 Temperature effects ................................................................................. 46 4 Single photon TOF measurements 51 4.1 SPAD TOF distance measurements......................................................... 51 4.2 Measurement of the char bed level in a recovery boiler ......................... 57 4.3 Raman spectroscopy using gated SPAD detection .................................. 61 5 Effect of signal photon quantum shot noise on linear detection schemes 63 5.1 Jitter of leading edge triggering receiver................................................. 66 5.2 Jitter of high-pass discriminated signals ................................................. 67 5.3 Jitter of dual-triggering receiver.............................................................. 69 6 Discussion 71 7 Summary 75 References 79 Original publications 85 15 16 1 Introduction This thesis considers pulsed time-of-flight (TOF) range finding from the viewpoint of individual photons. On the one hand, noise effects originating from the discrete photonic nature of light are investigated. On the other hand, TOF measurement based on the flight time of single photons is tested. For this purpose, work is done on developing a new type of laser diode especially suitable for single-photon measurements. An important application of the TOF technique is range finding, where the range of a target can be calculated based on the measured flight time of a laser pulse to and from the target. The amount of photons in an optical pulse sent by a TOF rangefinder to a target is typically in the range from millions to billions (for example a commonly used 10 W, 3 ns, 850 nm laser pulse has about 128·109 photons). However, the fraction of photons reflecting back towards a detector lens can be very small, depending on the target reflectivity and distance. Usually, at least a signal-to-noise ratio of over 10 is required in this type of measurement to avoid false detections due to noise (SNR is defined here as the peak signal amplitude to rms noise value ratio). Therefore the minimum detectable laser pulse is composed of about ~500 photons with typical detection methods and parameters (3..5 ns laser pulse, 100 MHz detector channel, 50% quantum efficiency of the detector, 7 pA / Hz receiver channel input reduced noise and an avalanche diode internal amplification M=100). Recently, single photon TOF measurement has become more attractive in practice due to a discovered method of integrating a single photon avalanche diode (SPAD) and the detector electronics in standard CMOS technology. The integration allows inexpensive and compact detector structures for TOF measurements, including detector matrices. In this thesis, research and development is carried out on a new type of laser diode tailored for use with this type of detector. The ~100 ps width of the laser pulse matches with the detection jitter of a typical SPAD, which leads to benefits in measurement precision. However, the main advantage of the new laser diode compared to other similar competing laser diodes is the high energy of the laser pulse (over 1 nJ), providing a considerable benefit for example in outdoor measurements with high background light levels. Also, the high pulse energy is advantageous for use with 3D detector matrices because the reflected pulse energy is spread over the detectors, and therefore less energy is available for each individual SPAD. The new laser diode structure is introduced in chapter 3. 17 The new laser diode was developed for use with an avalanche transistor current pulsing circuit made from commercially available components. Although this circuit provides large >10 A and short ~1.5 ns current pulses, the pulsing frequency is limited to about 20 kHz due to heating of the avalanche transistor. In this thesis, it is shown that the use of a carefully chosen MOSFET transistor in the same topology provides similar or even shorter current pulses at a higher ~0.1–1 MHz pulsing frequency, leading to faster and more precise measurements with the developed laser diode. The overall laser pulser size is about one cubic centimeter. Demonstration distance measurements were performed with the developed laser diode pulser and SPAD detectors to show the feasibility of a compact device for fast distance measurements up to several tens of meters in bright sunny conditions, for example. Distance measurements were also made in a factory environment. The factory measurements showed the possible feasibility of a 3D scan in the factory to help optimize process efficiency. The single photon time-of-flight method is useful also in other applications besides range finding. A Raman spectroscope utilizing a SPAD matrix was made utilizing pulsed TOF methods. It was shown that the fluorescence background typical of Raman spectra can be suppressed by picking out the first few Raman signal photons with time gating of the SPAD detector. This is an advantage compared with conventional continuous-time Raman spectroscopy devices. The results from the first prototype of such a device are presented in chapter 4.3. The photonic nature of light also plays a role when using a traditional linear sensor. This thesis was started with the research of the effect of the photonic nature of light on the jitter of a pulsed time-of-flight lidar using a linear receiver channel. The photonic nature of light presents a noise component which can be the dominant noise source in some conditions. The effect of this noise source on the distance measurement precision was studied, and ways to decrease its effects were found. 18 2 Photonic effects in pulsed time-of-flight range finding Practical important information can be deduced from the time-of-flight (TOF) of particles such as electrons or photons in a variety of applications, like mass spectrometry to identify chemicals (Wilman et al. 2012), positron emission tomography and diffuse optical tomography for medical imaging (Braga et al. 2013, Yudovsky & Pilon 2009), fluorescence-lifetime imaging to provide additional information regarding biological processes in microscope pictures (Cubeddu et al. 2011, Pacheri et al. 2013) and stimulated emission depletion fluorescence microscopy to achieve super-resolution imaging (Hell & Wichmann 1994). The main use of time-of-flight technology in this thesis is pulsed time-of-flight laser range finding, where the distance to a target is calculated based on the measured flight time of photons back and forth to a target. The distance can be scanned in many directions, or measured with a matrix of detectors enabling 3D imaging. The applications include augmented reality devices, man-machine interfaces, robotics, scanning of objects for making replicas with a 3D printer, forest scanning from drones, map creation, autonomous vehicles etc. Furthermore, a Raman spectroscope was constructed in this thesis, which also utilizes time-of-flight technology. With this instrument, molecules can be identified for applications in medicine, agriculture and security. Raman scattering is a weak phenomenon where the energy of a photon interacting with a molecule is shifted. The very small fraction of Raman scattered photons are shown to be distinguishable with time-of-flight technology. 2.1 Pulsed TOF range finding using a linear detector A lidar (light detection and ranging) device uses the same principle as the classic radar (radio detection and ranging), where a distance map of targets and obstacles can be obtained from echoes of sent electromagnetic waves. Shorter wavelength electromagnetic waves are used in lidar (of the order of ~1 µm) than in classical radar (~1 cm), enabling better vertical and horizontal resolution since optical laser beams can be much more easily collimated. Another difference between radar and lidar is that the discrete photonic quantum nature of light emerges to be of practical importance at optical wavelengths. According to insights by Max Planck and Albert Einstein in the early 1900s, a photon can be described as a discrete quantum of energy of magnitude hν 19 which can be gained or lost by the electromagnetic field, where h is the Planck constant and ν is the frequency of radiation (Isaacson 2007). Because radio frequencies are about 10000 times lower than frequencies of visible light, the energy of an optical wavelength photon is about 10000 times larger than the energy of a radio frequency photon. Due to the high energy of optical photons, optical detectors respond directly to individual photons, whereas the quantum nature of radio waves is buried in thermal and other noise, and therefore cannot be measured with current techniques (Heinzen & Wineland 1990). The rather large energy of optical photons enables simple single photon detectors, which are also used in this work. The classic method to perform radar scanning is to record the full detected echo waveform and to calculate the distance to the target based on the delay of the echo and constant speed of the electromagnetic wave (Texas Instruments 2011). However, unless complex signal processing methods such as interpolation and deconvolution are used, the measurement resolution is limited by the sampling period of the detected waveform (Neilsen 2011). For example, at a rate of 1 gigasamples per second, the range estimates would be binned at 15 cm resolution. This means that if a large measurement distance range and high accuracy are desired, the memory requirement and computing power requirement set by the signal processing might rise too high with this approach. A more robust method that avoids memory and power consumptive transient waveform analysis is to set a sufficiently high signal threshold to avoid false triggering from noise, and to determine the target distance from the time instant of threshold crossing. With this method, the signal threshold needs to be about 10 times the sigma value of noise to avoid false triggering due to noise (Ruotsalainen et al. 2001). A block diagram of this approach is shown in Fig. 1. As calculated in the introduction chapter, with typical parameters usually about 500 photogenerated electrons (photoelectrons) are needed in the receiver per pulse to be able to distinguish the optical pulse from noise using this method with a linear detector. 20 start-pulse target laser diode start-pulses photodetector amplification and discrimination counts time-to-digital converter (TDC) minimum signal threshold Δt=tstop-tstart stop-pulses signals from photodetector Fig. 1. Pulsed time-of-flight distance measurement device block diagram. The optical echo from the measurement target is composed of individual photons spread in time with a density envelope corresponding with a convolution of the laser pulse shape and target response function (Neilsen 2011). That is, the possible target roughness or dust in the measurement path modifies the density envelope of photons. The average number of photons nphot reflecting back from a diffuse target towards a detector lens of radius rdet depends mainly on the energy of the laser pulse Epulse and target distance Z, and can be calculated with the radar equation (Collis & Russell 1976, Wang & Kostamovaara 1994) nphot  2 Epulse    rdet      ( Z ) hcZ2 (1) , where λ is the wavelength of the laser, τ is the efficiency of the optics, ε is the diffuse reflectivity of the target, γ(Z) is the overlap of the input and output beams, h is Planck’s constant and c is the speed of light. Nevertheless, even a constant optical power from most sources does not equal a constant photon rate. In Mandel (1965), it was shown that the photon number fluctuation from a single mode laser has a random Poisson distribution. That is, if the average number of photons arriving within a certain period is nphot, then the probability to receive exactly k photons within this time interval is P( nphot_instantaneous  k )  k nphot e  nphot k! . (2) This means that the average number nphot of photons reflecting back towards a detector lens also has a variance equal to nphot (Neville & Kennedy 1964). When a 21 linear photodetector is used, this inherently choppy and random nature of light means that the signal is noisy itself, even in an otherwise noiseless environment. Each absorbed photon gives off a free electron in the detector circuit leading to a current impulse with a shape depending on the frequency response and gain of the amplifier channel. The resulting variance var{i(t)} of the detector output current due to the Poissonian distribution of photons (also called signal photon noise or quantum shot noise) is (Alexander 1997) var{i (t )}  qRF M t 2  Prcvd ( ) hd' 2 (t   )d , (3)  where q is the charge of an electron, R is the intrinsic responsivity of the photodetector at gain M=1 (detector gain explained in chapter 2.3.2), F is the excess noise factor of the detector (also explained in chapter 2.3.2), Prcvd(ξ) is the average power illuminating the photodetector and hd' (t ) is the normalized single electron response of the detector channel that has an area of unity. On the other hand, the average of the normalized detector output signal waveform is(t) is (Alexander 1997) t is (t )  i (t )  RM  Prcvd ( )hd' (t   )d . (4)  Fig. 2 illustrates how signal photon noise forms in a linear analog detector. Example detector impulse current pulses corresponding with the very weak (~1 nWpeak) optical pulse of Fig. 2 a), and their sum is shown in Fig. 2 b). A single electron response function corresponding with a 140 MHz low-pass filter followed by a 400 MHz low-pass filter was used here. Fig. 2 c) shows the average output current and variance according to (3) and (4). 22 photons / ns impinging photodetector P(t) 8 6 4 2 0 0 1 2 current [pA] 300 a) 4 5 6 t [ns] 7 example total photocurrent 200 six example single electron responses 100 0 3 0 1 2 3 current [pA] 600 4 5 t [ns] 7 i (t ) var{i (t )} 400 6 b) 200 example total photocurrent 0 -200 0 1 2 3 4 c) 5 6 7 t [ns] Fig. 2. a) An average photon rate impinging a detector with 0.5 A/W responsivity and an internal amplification of M=1. b) A corresponding example photodetector output current pulse waveform. c) The average of the output current waveform, the standard deviation of current fluctuations due to signal photon noise (square root of variance) and the example current waveform. A very weak optical and corresponding electrical pulse was shown in Fig. 2 to illustrate the nature of photonic noise more clearly. Also, the electronics noise is not shown since it is typically much larger than the above small current pulse amplitude and variance. In practice, the electronics noise of a pulsed TOF rangefinder receiver channel is typically about 70 nArms (100 MHz, 7 pA/sqrtHz). Therefore if a PIN diode is used (internal amplification M=1), the minimum current pulse amplitude needed for an SNR of 10 is 700 nA, corresponding with an optical pulse with about 30000 photons / 3 ns. This photon flux corresponds with a signal noise of about 5 nArms calculated with (3). Therefore when using a PIN diode 23 receiver, the 5 nArms signal shot noise is usually negligible compared to the 70 nArms input reduced rms noise of the receiver. But when an avalanche photodiode (APD) receiver is used, the internal amplification of the avalanche transistor (typically M≈100) amplifies the signal shot noise by M F , where F is the excess noise factor (explained in chapter 2.3.2, typically F≈3), so correspondingly with an APD, the signal shot noise is about 60 nArms, whereas the noise of the receiver channel is almost unchanged in relation to the PIN diode receiver channel case. Therefore signal shot noise is usually non-negligible when using an APD receiver. The effects of this noise source on the timing properties of the detected pulse are investigated in chapter 5. 2.2 Pulsed TOF range finding using a single photon detector Distance measurement based on the flight time of a single photon has been possible for a long time using a photomultiplier tube detector (PMT), for example (Poultney 1972). Instead, in this thesis, single photon avalanche diode (SPAD) detectors are used, which are more compact, and now also available for manufacture with the inexpensive and versatile CMOS technology. One of the advantages of SPAD detectors compared to APD detectors is that typically shorter laser pulses can be used with the SPAD detector. This is because the timing jitter of a SPAD detector is usually shorter than the single electron response function of an APD, which correspondingly determine the shortest sensible laser pulse lengths when using a SPAD or an APD receiver (Becker 2012). In general, a lidar benefits from the use of a shorter laser pulse – if comparing two cases with equal laser pulse energy, the use of a shorter laser pulse leads to greater SNR (Abend 1966). Typically, laser pulse lengths of about 3 ns have been used in the lidar application (Kurtti & Kostamovaara 2012, Nissinen et al. 2012). This is due to the difficulty of creating short and energetic laser pulses. This difficulty stems from laser driver restrictions, since at the required high current pulse amplitude levels (>10 A) the current pulse width is very difficult to reduce below a few nanoseconds due to current pulser inductances and switching speed properties of avalanche transistors (Duan et al. 2014). Since a standard laser diode output follows the input current pulse shape, the resulting laser pulse width is also within a few nanoseconds. Probably due to the aforementioned difficulty of generating short and energetic laser pulses, one of the leading TOF imager groups have lately made 3D imaging studies using a SPAD detector and a long ~4 ns laser pulse (Niclass et al. 2013). 24 Shorter laser pulses can be obtained for example by using gain-switching, but standard gain-switching leads to quite modest optical pulse energy (Ryvkin et al. 2009) and specialized driving current pulse requirements. The best combination for lidar measurements would be a short, energetic laser pulse and a sufficiently high resolution receiver. The enhanced gain-switching principle presented in this thesis together with a single photon detector achieves this set of goals. Another advantage of using a single photon detector in the lidar application is that simpler detection schemes can be used with a SPAD detector compared to a linear detector. For example, if using a 100 ps laser pulse, detecting any one of the sent photons leads to centimeter precision (if not considering noise counts) since light travels one centimeter back and forth to a target in 67 ps. This is a sufficient distance measurement precision for many applications. In contrast, when using a linear detector with, for example, a 3 ns laser pulse, a certain timing spot needs to be picked from the pulse to get sufficient distance measurement accuracy because a 3 ns period corresponds with about 0.5 m back and forth flight time. This linear pulse discrimination task is complicated by the very high dynamic range of optical pulse echo amplitude, as described in chapter 5. As a result, much simpler receiver structures are possible when using a single photon detector, providing other advantages as well. Pulsed time-of-flight range finding using a single photon detector is pronouncedly statistical in nature. Strictly speaking, a multi-mode laser diode is not necessarily Poisson distributed due to interaction of laser modes (Mandel 1965), but in this work the signal and background photons are considered to be Poisson distributed, as was the case, for example, with Abshire (1978). Therefore the output counts from a SPAD are a random sequence of pulses with the pulse density corresponding with the optical echo shape and background light (smeared slightly by the jitter, diffusion tail and afterpulsing of the SPAD detector as explained in chapter 2.3.1). An example case of laser pulse detection with a single photon detector is shown in Fig. 3. Typically, many detection periods do not contain any counts, other periods contain one or a few counts at most, as shown in Fig. 3 b). 25 power average echo shape of laser pulse from target time a) measurement period 1 2 3 4 counts from target dark count / background induced count nlaser b) histogram counts c) time histogram counts d) time Fig. 3. Principle of time-correlated single photon counting. a) Average shape of the laser pulse reflection, b) an example case of random individual SPAD counts, c) the histogram after a short measurement time, and d) the histogram after a longer measurement time. Due to the statistical nature of SPAD counts, single-shot measurement with a single photon detector based on just one SPAD count is not reliable since it is difficult or impossible to tell if a single SPAD count is caused by a valid signal photon or by a background light or detector noise induced electron-hole pair. However, obtaining several counts within a short period increases the probability of valid laser pulse detection. For example Fig. 3 c) shows three counts within a period corresponding with the laser pulse response width. Usually, the position of the laser pulse is not known in advance, so the group of three counts could also be noise or background induced counts. But since the background noise counts are spread out randomly over time, and signal counts are localized within a short period, the probability of 26 closely grouped counts to be signal photon induced is large typically. Therefore the false detection rate can be made low if a sufficient number of counts is required within a narrow time window Tsignal. This can be implemented for example by collecting a count histogram or alternatively, in the case of 3D measurements (for example when using a SPAD matrix), this idea can be extended by utilizing signal counts from neighbouring pixels which measure the distance to nearby directions. Since neighbouring pixels often collect closely timed signal counts from the same object surface, the information provided by adjacent pixels can be used to decrease the false detection rate (Niclass et al. 2013, Zhang et al. 2013, Bao et al. 2014, Shin et al. 2014). The signal-to-background (SBR) -ratio of a SPAD distance measurement to one direction is defined here as the ratio of the average number of detected signal counts to the standard deviation of background count bin populations, with bin widths equal to the signal width Tsignal. Each of these background bins collects on average BG·Tsignal·nlaser background counts, where BG is the total noise count rate per second and nlaser is the number of laser pulses sent. Therefore SBR  fraccounts  nlaser , BG  Tsignal  nlaser (5) where fraccounts is the fraction of laser pulse detections per number nlaser of sent laser pulses. fraccounts has a linear dependence with the energy of the laser pulse (if working in true single photon detection mode, explained in the next chapter), and therefore it has a nearly linear advantage on the SBR according to (5). This speaks in favor of the laser diode developed in this thesis, which has an especially high pulse energy compared to non-enhanched laser diodes. An SBR of over ~10 can be considered reliable, based on Poisson distribution statistics (Neville & Kennedy 1964). Histogram signal processing can also be done to improve the sensitivity and precision, but this increases the system complexity (Hernandez-Marin et al. 2007, McCarthy et al. 2013). At least two signal counts are usually required in (5) because a single SPAD count is often not sufficient for a reliable measurement, unless signal counts from adjacent measurement pixels are used to validate measurement result. In such a case, even one signal detection per SPAD might be sufficient for reliable detection (Shin et al. 2014). Another practical significance of the Poissonian spread of photons is that most single-photon detectors give a distorted output signal if more than one photon is absorbed to take part in the creation of the detection pulse. As can be seen in Fig. 27 4, if on average one photon is absorbed in the detector, then there is almost a 30 percent chance for non-single-photon detection. But when the input to this type of detector is attenuated to on average less than 0.1 detections / pulse, the probability for non-single-photon detection becomes negligible. The practical effect of this and solutions are presented in later chapters. P(k photoelectrons involved in a single detection) 1 0.8 average photoelectron number = 0.1 0.6 average photoelectron number = 1 0.4 average photoelectron number = 5 0.2 0 0 2 4 6 8 10 instantaneous photoelectron number k Fig. 4. Distribution of instantaneous photoelectron numbers (how many photons partake in a photon detection process) for three average values of photoelectron numbers based on Poisson statistics. 2.3 Avalanche multiplication The optoelectronic detectors in this research utilize impact ionization, where an electric field is used to increase the kinetic energy of an electron so high that, upon a collision with an atom, a new free electron-hole pair is created (Fig. 5). Therefore an avalanche of free electrons initiated by just one photoelectron can be created (McIntyre 1999, Johnson 1965). This leads to an amplification of the electric signal inside the detector. Depending on the chosen electric field strength, a single photoelectron can be multiplied leading to tens or hundreds of electrons, or more. In the next chapter, the operation of a single photon avalanche diode (SPAD) is described, where a single absorbed photon can trigger a macroscopic avalanche current pulse. Chapter 2.3.2 describes the operation of an avalanche photo diode (APD), which has a smaller, linear amplification for a continuous stream of photoelectrons. 28 2.3.1 Single photon avalanche diode Despite its name, avalanche breakdown is usually a favourable phenomenon, where the electric field across the depletion region of a pn-junction is raised so high that a substantial current flow forms from an avalanche of electrons initiated by just one photogenerated electron. A single photon detector can be created based on this phenomenon, usually called a single photon avalanche diode (SPAD). A way to manufacture a SPAD detector in the standard CMOS process was discovered by Rochas et al. (2001). The problem to solve was to prevent excessively high electric fields at the edges of the active region due to the strong doping profile differential there, leading to premature breakdown at the wrong location. The avalanche breakdown should happen uniformly over the active area and not only at the edges. Earlier others, using dedicated manufacturing processes, have usually solved this problem by placing a guard ring around the active area (Haitz 1965). In the conventional CMOS process, layers were not available for directly creating such a guard ring. Rochas et al. (2001) solved this using two ntub layers separated by a small gap. Due to lateral diffusion between these n-tubs an efficient n+ guard-ring forms between the n-tubs, which can be made to encircle the SPAD without any additional costly manufacturing steps in the CMOS process. Since then, other, more direct ways to create the guard ring have been found (Rochas et al. 2003, Gersbach et al. 2008). One such structure, using a p-well guard ring, was used in papers III, VII and VIII and is shown in Fig. 5. 29 collision with atom + new electronhole pair photon-induced electron p-well n+ anode photon p+ RPTU p-well B cathode n+ 0.2 µm 1.55 µm 6.5 µm deep n-well p-substrate Fig. 5. Cross section of a typical SPAD structure implemented in a high-voltage CMOS technology with a p-well guard ring, with the high field pn-junction multiplication region approximately marked with red color, and one of many impact ionizations depicted. The fact that such a single photon detector can be manufactured in CMOS technology is beneficial because CMOS is an inexpensive, scaling and high-quality technology leading to well performing, customizable SPADs, with even a possibility for detector matrices. Due to the macroscopic avalanche current formed in a SPAD, additional amplifiers are not needed after this detector, which simplifies the overall detector structure compared to linear analog photodetectors. But due to the self-sustaining avalanche current formed at each detection, a current quenching circuit is needed to enable detection of successive photons. This current quenching circuit can fortunately also be integrated in CMOS technology (Zappa et al. 2002). Also any following signal processing circuits can be integrated in CMOS, leading to an overall cost-effective, small and high-performance optoelectronic receiver. The noise of a SPAD detector is quantified by the so-called dark count rate, i.e. non-photon induced noise count rate, which is typically less than 10 Hz / µm2 in regular CMOS technology (Pancheri & Stoppa 2007). This dark count rate mainly originates from fabrication defects and is often negligible especially compared to solar background light induced counts, as will be shown later. Smaller dark count rates are available in special manufacturing technologies (Hsu et al. 2009, Bronzi et al. 2012). 30 Another imperfection of a practical SPAD leading from fabrication defects is afterpulsing, where a small part of the avalanche pulse charge gets trapped in impurities and defects of the semiconductor, and gets released after some time. This electron trapping and the resulting afterpulsing is a minor harmful effect in the range finding application. The main downside of this type of detector is the low photon detection probability (PDP) due to the planar fabrication process with junctions relatively close to the surface. Therefore long wavelength light will mainly get absorbed after the effective high-field multiplication region. The typical PDP for the GaAs/GaAlAs laser created in this thesis is about 3–8% depending on the wavelength (the laser emission wavelength is either about 860 nm or 810 nm depending on whether a bulk or quantum well laser diode is used) (Pancheri & Stoppa 2007, Guerrieri et al. 2010). The incidence time of individual photons can be determined very precisely with a SPAD. The statistical fluctuation (jitter) of time between photon incidence and the output detection pulse due to transit time spread is typically of the order of 100 ps with a CMOS SPAD (Pancheri & Stoppa 2007), thanks to the strong electric field in the shallow depletion region. A minor non-ideality regarding the pulse timing is that if a photon gets absorbed close to, but outside the depletion region, the induced free electron has a certain probability to drift into the depletion region to cause an avalanche slightly later than if it were absorbed directly in the depletion region. This gives a somewhat harmful exponential diffusion tail in the detection histogram, as seen later in the measurement results of chapter 4. The reverse bias voltage needed to achieve single photon detection is small in the planar CMOS process, typically in the region of 20 to 40 V. The lowest reverse bias voltage at which a self-sustaining avalanche current is accomplished is called the breakdown voltage. For time-gating purposes, the reverse bias voltage can be biased above the breakdown voltage for a short period (i.e. gated operation) for picking out the signal, as explained later. 2.3.2 Avalanche photodiode The avalanche photodiode (APD) is a detector also based on impact ionization, but operated below the breakdown voltage, so that the average free electron amplification is a finite number M. Usually, APDs are manufactured in a special technology to allow a much deeper absorption region compared to a SPAD in 31 CMOS technology, to maximize photon absorption and minimize the capacitance of the detector. The value of amplification M that can be reached depends mainly on the ratio of the electron ionization rate αi to the hole ionization rate βi inside the APD. The amplification of individual free electrons is a random process, the variance of which is also minimized when the ratio of αi to βi is maximized. The variance of amplification is quantified by the excess noise factor F (McIntyre 1999). Even after the amplification of electrons by the APD, the flow of electrons is still usually so small that amplification is needed for accurate detection of the optical pulse. Usually a transimpedance amplifier offers a good compromise between low noise and high bandwidth (Alexander 1997). 32 3 High speed/power ~100 ps laser pulse transmitter for single photon TOF measurements The total performance level available from a laser diode transmitter depends on the laser diode structure and the current pulser driving it. In this study, both of these parts were developed to achieve a high frequency laser pulser emitting powerful laser pulses with a width matching the SPAD jitter, providing benefits in the pulsed TOF range finding / 3D imaging application, as was explained in chapter 2.2. Table 1 lists publications regarding the development of the new laser pulser and summarizes advancements in each paper. Table 1. Publications regarding the new laser pulser and comparative improvements. Paper Achievements II The first report of ~150 ps, ~5 W single pulse Notes IV ~110 ps, ~20 W single pulse, zero-order transverse Current spreading over the laser stripe was emission from the new laser diode structure. emission was shown from the laser diode. TOF improved and the current pulse was distance measurements were demonstrated with shortened. the developed laser diode and a SPAD detector. V VII ~100 ps, ~40 W single pulse emission was shown, Laser diode samples had high variance of and another SPAD TOF distance measurement performance, a “good sample” was found demonstration was done. here. >10 x higher pulsing rate was achieved with a FET A quantum well version of the laser diode pulser. Outdoor distance measurements were showed better tolerance to temperature demonstrated. The laser diode samples had more variation compared to bulk laser diodes due consistent performance due to a commercial to a higher Ioperation / Ithreshold ratio. manufacturer. An in-lab designed CMOS SPAD detector was used. VIII The same laser pulser as in paper VII The paper focused mainly on discussion of range finding with a SPAD detector and the developed laser pulser. The new laser diode structure and the current pulser are presented in chapters 3.1 and 3.2, correspondingly. The following subchapters show conjoint measurement results of the laser diode and current pulser. After these, in chapter 4.1 the laser transmitter is utilized in demonstration measurements in its target application. 33 3.1 Enhanced laser diode structure to optimize gain-switching performance for high-power optical pulse generation Gain-switching is a phenomenon where the laser optical output overshoots for fast edges of input current. This is due the interplay between output optical power and carrier density inside the laser cavity according to the the rate equations which describe the dynamic operation of a laser diode. With a properly chosen current pulse width and amplitude, this overshoot can be utilized to obtain a short, single optical output oscillation. In Ryvkin et al. (2009) a waveguide structure was suggested to significantly increase the energy of short (~100 ps) optical pulse emission from a gain-switched laser diode while keeping the current pulse requirements modest. Simulations of the new laser diode structure predicted the emission of optical pulses with energies of over an order of magnitude higher than with regular gain-switched laser diodes. The new laser diode structure was achieved and optimized through an understanding of the laser diode gain-switching dynamics and simulation of the suggested structure based on rate equations and a traveling-wave model. Several ideas were suggested by Ryvkin et al. (2009) to increase the amplitude of the gain-switching optical peak, the main points being presented here briefly. The main idea was to aim for an extremely high ratio of d / Γa (> 2.5 µm), where d is the active layer thickness and Γa is the optical confinement factor. This was shown to be advantageous from the dynamic point-of-view, since the small confinement factor leads to stronger decoupling of electrons and photons in the thick cavity, delaying the optical pulse emission and therefore allowing a greater carrier build-up in the large active layer for bigger optical pulse emission. The structure was optimized for ~1.5 ns, >10 A injection current pulses which can easily be created with a simple avalanche transistor pulsing circuit. The suggested way of realizing the small optical confinement factor Γa was to detune the active layer from the optical mode by misaligning the active waveguide in relation to the optical mode, as in Fig. 6, i.e., by using an asymmetric waveguide structure. The degree of overlap between the active layer and optical mode is quantified by the confinement factor Γa (=light intensity within active area / total light intensity). The active layer can also be composed of several quantum well active layers, as suggested in Ryvkin et al. (2011), which provides several practical significant benefits as described later. 34 modal intensity distribution refractive index n 3.7 0.8 3.6 3.5 0.4 OCL 0.2 p-cladding n-cladding 0.6 3.4 3.3 3.2 3.1 0.0 -3 -2 -1 0 1 distance x [m] 2 3.0 Fig. 6. A proposed asymmetric waveguide structure and the corresponding modal intensity distribution (Ryvkin et al. 2009, published by permission of IEEE). Fig. 7 shows the main versions and phases of development of the laser diode. All of the versions are based on the gain-switching enhancement principle, briefly described above. The first main improvement was the achievement of a zero-order transverse mode due to a small modification of the internal structure of the laser diode. Several low-reflectivity (LR) and high-reflectivity (HR) coatings were tested. Next, the internal efficiency of the laser diode was improved, also by small modifications to the internal structure of the laser diode, which led to smaller threshold currents. Finally, in papers VII and VIII, a quantum well version of the laser diode was tested, in addition to bulk laser diode versions. A commercial laser diode fabrication facility was used to manufacture these laser diodes. The gain-switching enhancement principle was used in both versions of the laser diode (Ryvkin et al. 2011). The quantum well version of the laser diode provided a lower central emission wavelength of ~810 nm compared to the ~860 nm central wavelength of the bulk laser diode, which is advantageous in terms of the PDP of the SPAD detector. The quantum well laser diode also provides more immunity of pulse emission to temperature changes due to a lower lasing threshold current, as explained in chapter 3.5. The details of the internal developments of the laser diode are beyond the scope of this thesis, which instead focuses on the development of the transmitter as a whole and on the applications. 35 Manufactured by Optoelectronics Research Centre (ORC), Tampere: Laser diode version K2371GSL: -L=1.4 mm chips, w=120 µm oxide stripe width -HR 94% / LR 4% (paper II) zero‐order transversal mode Laser diode version RO633GSL: -L=1.4 mm chips, w=137 µm oxide stripe width -HR 96% / LR 7%, as-cleaved, HR 96% / as-cleaved (three variations) (paper IV & V) internal efficiency ηi improvement Manufactured by Innolume, Dortmund: -Bulk: L=1.5 mm, w=90 µm; L=3 mm, w=30 µm; L=1.5 mm, w=30 µm (three variations) -Quantum well: L=1.5 mm, w=30 µm -HR 99.9% / LR 5% (paper VII & VIII) Fig. 7. The main phases of laser diode development. Fig. 8 shows close-up pictures of an early version of the laser diode. The characterization measurement results of the laser diode are shown in chapter 3.3 after an introduction of the current pulser circuit used to drive the laser diode. laser diode L (1.4 mm in this case) x-axis w (120 μm in this case) submount a) b) Fig. 8. a) Close-up picture of laser diode K2371GSL. b) Combined microscope picture of emitting edge of laser diode and a false color picture of infrared near field laser emission. 36 3.2 High frequency current pulsing circuit A short, large amplitude current pulse is the foundation to emitting a short, high amplitude laser pulse from a laser diode. Simulations based on rate equations show that current pulses larger than ~5 A with a width of 1 to 1.5 ns (FHWM) are needed to drive the designed laser diode. Here also compactness and simplicity of the current pulsing structure were requirements. While the current amplitude needed can be quite straight-forwardly generated, the short pulse width requirement poses a challenge, since high speed ON/OFF current switches at the required current pulse amplitude are not readily available. One typical solution to this is to use the current pulsing circuit shown in Fig. 9, the basic structure of which was introduced in Kilpelä & Kostamovaara (1997). This current pulser is able to generate relatively short and large current pulses, only requiring a switch with a sufficiently fast ONswitching rise time, while the current pulse length is determined by the passive component values as described in the following. This current pulser, which can be built with commercial components, was the starting point to the development of the laser diode described in the previous chapter. The circuit operates by first charging the voltage over capacitance Ccharge to Vbias through the schottky diode D1 during switch disconduction. When the switch S1 is closed, the circuit loop between S1, Ccharge, resistance Rdamp, total loop inductance Lstray_total and laser diode LD closes. The resulting RLC-circuit can be made slightly underdamped to produce a large, narrow current pulse with negligible current oscillations by choosing the value of Rdamp according to Rdamp  Lstray_total / C charge . (6) In this case, the width tw of the forming current pulse through the laser diode becomes (7) t w  2 Lstray_total C charge , 37 Vbias Rbias Ccharge Rdamp Lstray D1 S1 LD Fig. 9. Current pulsing circuit schematic. and the resulting current pulse amplitude is approximately I a m p litu d e   V R tot  L stra y_ tota l / C c h a rg e , (8) where ΔV is the voltage step amplitude and Rtot is the total resistance in the current loop. If the circuit inductances are carefully minimized, large and narrow ~15 A, 1.5 ns current pulses are available from this current pulser using an FMMT415 avalanche transistor switch. Much larger current pulses are also available from an avalanche transistor current pulser at longer pulse widths. Although good performance is available using an avalanche transistor in this circuit, the residual voltage that forms across a Si avalanche transistor from collector to emitter when switched on is harmful, as it decreases the overall voltage change ΔV across the device, limiting the achievable current pulse amplitude. The residual voltage also increases the power consumption of the avalanche transistor, leading to heating and therefore limiting the pulsing frequency to about 20 kHz. The residual voltage is mostly a consequence of the limited difference between electron and hole velocities in the quasi-neutral region, as analyzed in Vainshtein et al. (2003). The achievable current pulse shortness has also been shown to depend on the geometry of the semiconductor avalanche transistor chip (Duan et al. 2014). High-voltage GaAs junction bipolar transistors with a higher difference between electron and hole velocities, and therefore lower residual voltage, have been suggested by Vainshtein et al. (2004) for short and large current pulse generation, but are not commercially available. A faster laser pulsing rate would enable faster distance measurements, and better measurement precision, due to a higher SPAD signal averaging rate. To enable faster pulsing rates, the plan presented in paper IV was to relax the current pulsing requirements by reducing the width of the emitting laser diode stripe, since the current pulse amplitude requirement scales down in proportion to the laser 38 diode stripe width. With this relaxed current pulse requirement, an even simpler FET current pulsing electronics structure was predicted to be feasible for use with the laser diode. Since FET transistors do not have a significant residual voltage, a faster pulsing frequency can be achieved. In practice, with careful FET transistor selection, a current pulse amplitude of ~12 A with a width of 1.5 nsFWHM at a pulsing rate of ~100 kHz, and a current pulse amplitude of ~5 A with a width of ~1.1 nsFWHM at a pulsing rate of ~1 MHz were achieved. In selection of the FET transistor, the most important factor was to find a transistor working at the edge of its performance level in as many aspects as possible. Aiming at about 1.5 ns current pulse width (because the laser diode was designed for this current pulse width), the search for a suitable FET transistor converged on the Fairchild FDMC86244 with a maximum allowed current pulse amplitude of ~12 A. This current pulse amplitude was achieved at the FET’s maximum allowed drain voltage of 150 V, with a Ccharge value of 95 pF and Rdamp value of 5 Ω. This current amplitude – drain voltage pair also matches according to (8). Not matching these values would mean the transistor is overspecified in one way leading to subpar performance. Other important criteria for the FET transistor were to have a sufficiently high saturation current, a low drain to source onresistance and a clearly shorter turn-on time than the current pulse rise time, which is approximately equal to (7). The gate capacitance of large current FET transistors tends to be high, in the range of hundreds of picofarads. To drive the gate, an Intersil EL7158 high current gate driver was chosen. Fig. 10 shows the resulting measured current amplitudes and widths with two values of Ccharge as a function of Vbias using the chosen FET. With a smaller capacitor value of 47 pF narrower, but smaller current pulses were achieved, as equations (7) and (8) predict. Fig. 10 also shows for comparison current pulses achieved using an avalanche transistor and a Ccharge value of 82 pF from paper V. Large and narrow current pulses are available using an avalanche transistor, but at a smaller frequency of about 20 kHz, limited by the heat dissipation mainly due to the high residual voltage from collector to emitter during conduction. The FET pulsers achieve relatively narrow and high current pulse amplitudes at small bias voltages, and high pulsing frequency due to much smaller residual voltages. Also the current pulse width stays more constant as a function of Vbias with a FET transistor due to the smaller residual voltage (due to its Vbias-dependence). All of the current pulse values in Fig. 10 are deduced from the measured voltages over Rdamp, taking the stray inductance of resistor Rdamp into account, as described in the following chapter. The 39 maximum pulsing frequencies at which the laser pulser starts to heat up significantly are also shown. In the FET pulser case, somewhat faster pulsing rates are possible if the laser diode temperature requirements (discussed later) are not dissatisfied. The layout of the current pulser PCB also has an effect on how much the current pulser heats the laser diode (an example thermal image is shown in Fig. 12 b)). Current pulse peak amplitude [A] 20 Avalanche transistor pulser, Ccharge=82 pF, Vbias=309 V max f ≈ 20 kHz (paper V) 18 16 FET pulser, Ccharge=95 pF (paper VII) 14 Vbias=150 V, max f ≈ 100 kHz 12 FET pulser, Ccharge=47 pF, (paper VII) 10 Vbias=240 V Vbias=150 V, 8 max f ≈ 100 kHz 6 4 Vbias=75 V, Vbias=90 V, max f ≈ 1 MHz 1 1.2 max f ≈ 1 MHz 1.4 1.6 1.8 FWHM current pulse width [ns] Fig. 10. Measured achieved current pulse widths and amplitudes as a function of Vbias. The pulsing frequencies at which the current pulser starts to heat up without active cooling (<10 °C heating of laser diode) are shown in the figure. The current pulse shape through the laser diode was deduced from the measured voltage over Rdamp with the principle shown in Fig. 11. The voltage was measured with a 6 GHz, 500 Ω input resistance Lecroy PP066 passive probe. Due to the stray equivalent series inductance (ESL) of Rdamp, the current waveform through Rdamp is not linearly proportional with the voltage over Rdamp. The current waveform was found by a simulation procedure where the current waveform was iteratively changed until the simulated voltage response of the current pulse to a resistor and ESL matched with the measured voltage response. The surface mount 0603 resistor 40 Rdamp was found by measurements to have a stray equivalent inductance of about 0.75 nH (in some cases 0402 resistors were used). Due to this stray inductance of Rdamp, the current waveform lags the voltage waveform by an interval approximately equal to the Lstray_Rdamp/Rdamp-time constant, where Lstray_Rdamp is the ESL of Rdamp. 10 measured voltage over 3.6 Ω & ~0.75 nH divided by 3.6 iterated PWL current source [A] Rdamp 3.6 Ω 8 simulated voltage response to the iteratively acquired current pulse divided by 3.6 Lstray_Rdamp 0.75 nH 6 4 2 0 12 13 14 15 16 17 18 19 20 21 22 time [ns] Fig. 11. The simulation setup to acquire the current pulse shape from the measured voltage over the resistor Rdamp with stray inductance Lstray_Rdamp. The measurement was done with a Vbias value of 290 V, a Ccharge value of 68 pF and an FMMT415 transistor. Fig. 12 shows a photograph of the FET current pulser and a thermal image of the pulser taken with a Trotec EC 060 V infrared camera at room temperature. Without the SMA connectors, the pulser fits into a ~1 cm3 package. To achieve high laser pulsing rates, the laser diode should be able to withstand the heating due to the FET transistor (which depends on the pulsing frequency, see Fig. 10) and other components, as well as ambient temperature changes. These are discussed in chapter 3.5. 41 Rbias 80 ˚C (needs to be distributed) MOSFET laser diode gate driver gate driver 37 ˚C MOSFET 54 ˚C laser diode 43 ˚C b) a) Fig. 12. a) Photograph of FET laser pulser (coin for scale). b) Thermal image of the pulser after pulsing 8 A, 1.5 nsFWHM pulses at 1 MHz for a period of 10 s (poster of paper VII, published by permission of IEEE). 3.3 Power waveform measurements The laser output was measured with a New Focus 1434 (25 GHz, 20 µm) photodetector connected to a 30 GHz Lecroy 830Zi-A oscilloscope. The timing between the laser and current pulse was aligned based on measured delays from the laser output to the oscilloscope input. Also the delay from the current measuring probe tip to the oscilloscope input, and the delay between voltage and current waveforms due to the ESL of Rdamp shown in Fig. 11 were taken into account. Optics with a magnification of unity were used, so the ~100 µm laser stripe did not completely fit into the detector with a diameter of 20 µm, causing some uncertainty in the power shape measurement due to the non-homogeneity of the emitting stripe, especially with early laser diode versions. Streak camera measurements enabled the measurement of the whole laser stripe emission, shown later in this chapter. The peak powers of the laser pulses were deduced from the average output powers Pav of the laser diode. Pav was measured with a 1 cm2 calibrated PIN diode placed close to the laser diode output. Assuming homogeneous output from the laser diode output stripe, the peak optical power is Pm ax  v m ax t p er  Pav  t p er , (9) v (t ) d t 0 where v(t) and vmax are the signal waveform and peak amplitude measured by the photodetector and tper is the pulsing period. Fig. 13 shows the highest amplitude 42 single isolated 4.6 nJ output pulse achieved, and the corresponding driving current pulse measured with the methods described above. A clean single isolated optical output pulse is favourable for the TOF measurements, and in this case the unwanted optical pulse tail amplitude is fairly small at ~1 W. Laser pulse shape dependencies on the injection current pulse amplitude are shown in papers II, IV, VII and VIII. 45 40 35 optical output pulse [W] 87 psFHWM 30 25 20 injection current pulse [A] 16.1 A, 1.5 nsFHWM 15 10 5 0 0 0.5 1 1.5 2 2.5 3 3.5 4 time [ns] Fig. 13. Laser diode output power of RO633GSL measured with a New Focus 1434 photodetector at 24 °C from paper V. The outputs of the laser diodes were also measured with a Hamamatsu C5680 streak camera equipped with a spectrograph. Fig. 14 shows a streak camera measurement result corresponding with the measurement shown in Fig. 13. The pulse measurement was done with a streak camera input slit width of 150 µm, and the optics before the streak camera input were aligned for -1.2 magnification of laser diode stripe. Therefore most of the 137 µm laser stripe output fitted into the streak camera input. Measurements with the streak camera showed slightly lower peak power and wider pulse width compared to PIN diode measurements. This can be attributed to small inhomogeneities in the wide laser stripe output and the averaging of this output power by the streak camera equipped with a monochromator. The streak camera measurement was done in a room with a slightly smaller temperature compared to the New Focus 1434 measurement, which also has a small effect on the shape of the optical pulse. 43 wavelength [nm] 880 output power [W] 35 875 875 30 870 870 25 865 20 860 865 FWHM=100 ps 860 855 15 850 10 850 845 5 845 840 0 1 time [ns] 2 0 0 a) wavelength [nm] 880 855 1 b) time [ns] 2 840 FWHM=6.4 nm 0 1 c) Fig. 14. Streak camera measurement result for the same RO633GSL laser diode sample with the same injection current pulse input as in Fig. 13 at 23 °C. a) Output from the streak camera equipped with a spectrograph, with 150 µm monochromator slit width, b) extracted power as a function of time, and c) extracted line width information from a separate streak camera measurement with a 50 µm monochromator slit width to avoid distortion of the monochromator. Table 2 shows an overview of the best performing laser pulsers achieved in papers II, IV, VII and VIII, the first of which is the highest peak power laser pulser achieved. In this case, an avalanche transistor was needed to accommodate the large ~16 A current pulse amplitude requirement due to the wide 137 µm laser stripe. The maximum pulse repetition frequency was limited to about 20 kHz by the maximum specified power dissipation of the avalanche transistor of 500 mW. The next two commercially grown laser diodes shown in Table 2 were produced with a narrower laser stripe, and small changes to the internal structure of the laser diode led to a better internal efficiency. Therefore FET current pulsers could be used due to smaller current pulse amplitude requirements. The use of FET pulsers enabled higher pulsing frequencies, although the pulsing frequencies were still limited to about 100 kHz in ~25 °C room temperature due to temperature sensitivity of the bulk laser diodes and about 10 °C heating of the current pulser at 100 kHz. Out of these two laser diodes, the shorter 1.5 mm laser pulser emits slightly shorter laser pulses due to a shorter cavity length which can be useful to increase distance measurement precision. 44 A quantum well version of the laser was shown to be more immune to heat, as explained in chapter 3.5. 1 MHz, ~140 ps, ~4 W pulsing was achieved with a quantum well laser in room temperature without active cooling, which was mentioned in the poster of paper VII and thoroughly reported by Huikari et al. (2015). Table 2. Comparison of achieved laser pulsers (measured with New Focus 1434 detector). Laser diode stripe version width w length L chip single optical maximum current pulse amplitude and pulse pulse pulsing width energy FWHM frequency f RO633GSL 137 µm 1.4 mm 4 nJ 87 ps 20 kHz 16 A, 1.5 nsFWHM Bulk commercial 90 µm 1.5 mm 1.5 nJ 79 ps 100 kHz 8.5 A, 1.5 nsFWHM Bulk commercial 30 µm 3 mm 1.2 nJ 125 ps 100 kHz 7 A, 1.1 nsFWHM QW commercial 30 µm 3 mm 0.5 nJ 140 ps 1 MHz 6 A, 1 nsFWHM Another factor that speaks in favor of the narrow (small w) laser diodes is that a narrow laser has smaller divergence of collimated beam α (O’shea 1985)  w [rad], f lens (10) where flens is the focal length of the used collimating lens. When the laser beam is collimated with small divergence, the background photon count rate can also be minimized, since a smaller divergence of the detector beam is also required to encompass the transmitter beam (paper VIII). 3.4 Far-field measurements A narrow far-field output beam from a laser diode is highly desirable to minimize the coupling loss to a collimating lens and to enable a smaller collimating lens. In Ryvkin et al. (2013) numerical simulations showed the effect the thickness of the active layer on the far-field profile. The far-field profile of RO633GSL was measured by rotating a 1 cm2 PIN diode 50 cm from the laser diode facet, the results of which are shown in Fig. 15. Due to the smaller slow axis divergence, the collimating lens can be cut narrower in this direction to minimize the size of the optics. 45 far-field profile, a.u. 1 fast axis FWHM ≈ 18° slow axis FWHM ≈ 2° 0.5 0 -30 -20 -10 0 10 20 30 far-field angle, degrees Fig. 15. Far-field measurement results for laser diode RO633GSL (paper V, published by permission of IOP). 3.5 Temperature effects Simplifying a little, the main idea of using a large d / Γa -ratio is that it leads to a beneficial delay in lasing, leading to larger accumulation of carriers for larger optical output pulse energy (Ryvkin et al. 2009). But this also increases the dynamic threshold current, which is potentially harmful in terms of temperature sensitivity of the laser. However, ways of decreasing the temperature dependence of the optical pulse emission were found, and these are described next. First, to research the temperature dependence of the laser diode, the threshold current was measured in a Heraus Vötsch VM04/100 temperature-stabilized oven. The average output power was monitored with a large area PIN diode. Fig. 16 a) and b) show the average output power measured at different temperatures with two driving current pulse widths. While both measurements demonstrate high dynamic threshold currents, note the even higher absolute current threshold for the narrower injection current pulse due to the lasing dynamics. 46 average power @ 1 kHz [µW] average power @ 1 kHz [µW] 30 °C 25 °C 20 °C 15 °C 10 °C 5 °C 0 °C -5 °C -10 °C 3 2.5 2 1.5 1 2.5 2 1 0.5 0.5 0 5 °C 0 °C -5 °C -10 °C -15 °C -20 °C 1.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 current pulse amplitude [A] a) 0 3 3.5 4 4.5 5 5.5 6 6.5 current pulse amplitude [A] b) Fig. 16. Average output power of a K2371GSL laser diode (grown at ORC) driven with a) ~1.7 nsFWHM current pulses and with b) ~4.1 nsFWHM current pulses at different temperatures (different laser diode samples were used in measurements a) and b)). The threshold current Ithreshold of a double heterostructure laser diode varies according to (Agrawal & Dutta 1993) Ithreshold (T )  I0 eT /T0 , (11) where I0 is a constant and T0 is a characteristic temperature of a laser diode that expresses how sensitive Ithreshold is to the temperature T, where a higher value of T0 means lower sensitivity to temperature. Agrawal & Dutta (1993) state that for typical AlGaAs lasers T0 lies at values above 120 K. Extracted from the information in Fig. 16 a) and b), T0 is about 100 K irrespective of the driving current pulse width for the tested laser diodes. The more important performance characteristic for this laser diode is how immune the single pulse emission is to temperature changes. To measure the pulse shape dependence on temperature, a Peltier element was used to cool and warm the laser diode while measuring the output with a ~15 GHz photodetector setup and monitoring the laser diode temperature with a negative temperature coefficient resistor (NTC) placed next to the laser diode submount. The result in Fig. 17 shows that for this early laser diode version (an intermediate version between RO633GSL and the commercially grown lasers, not shown in Fig. 7) the maximum single isolated pulse emission power is about 2.3 W at a temperature of 22.95 ˚C, and the peak power falls down to about 0.75 W after heating by only about 4 ˚C if the 47 driving current pulse amplitude is kept the same. The reason for this is that for this early laser diode version the maximum operating current pulse value for single pulse emission is only about 1.1 times larger than the threshold current value, almost as much as the threshold current increase due to the very moderate heating in Fig. 17. optical pulse peak power [W] 3.5 3 2.5 2 1.5 3.1 Wpeak, 88 psFWHM not isolated optical pulses single, isolated optical pulses 2.3 Wpeak, 101 psFWHM 22.95 °C 24.09 °C 27.30 °C 1 0.75 Wpeak, 164 psFWHM 0.5 0 2.95 (Ithreshold grows as a function of temperature) 3 3.05 3.1 3.15 3.2 3.25 3.3 3.35 current pulse amplitude [A] Fig. 17. Peak power output of a K3058GSL (w=16 µm) laser diode measured at three temperatures as a function of current pulse amplitude. Three example optical pulse waveforms are shown. The injection current pulse width is ~1.65 nsFWHM in all cases. The key to increasing the temperature immunity of single pulse emission is to aim for a high relative difference between the operating current level and the lasing threshold current level. In other words, the effect of threshold current change on the optical pulse emission is smaller if the operating point is further away from the threshold current level. This relationship is better, for example, with the laser diode presented in paper V. In this case, an acceptable isolated pulse shape is obtained at an operating current amplitude almost twice as large as the threshold current value. Fig. 18 shows its single pulse energy measured as a function of temperature. Quantum well versions of the laser diode (paper VII, Huikari et al. (2015)) showed even greater immunity to temperature variations due to significantly lower 48 threshold current values. Furthermore, the ambient temperature dependence of the output pulse was shown by Huikari et al. (2015) to be largely compensable via Vbias control in a temperature range of 22 °C to 60 °C leading to a very small change in output pulse power and shape. energy per pulse [nJ] 5 4 3 20 25 30 35 40 temperature [˚C] Fig. 18. Temperature dependence of single pulse emission of RO633GSL with a 16.9 A, 1.5 nsFWHM injection current pulse (trailing pulse amplitude <10% of peak amplitude at all measured temperatures). Small laser pulse amplitude variation due to temperature change is not very harmful in pulsed time-of-flight laser range finding, since the attenuation of sent laser pulses varies greatly anyway due to varying target distances and different target reflectivities according to (1). But pulse shape variations might have a certain effect on the distance measurement accuracy as shown later. But if the pulse width stays close to 100 ps, the inaccuracy due to pulse shape variations stays within a few centimeters, which is sufficient in traffic applications, for example. 49 50 4 Single photon TOF measurements 4.1 SPAD TOF distance measurements The main reason for developing the new laser pulser was to study for possibilities of simplifying 1D rangefinders and 3D imagers by using SPAD detectors. To work towards this goal, the feasibility of making 1D distance measurements using the developed laser pulser and a SPAD detector were demonstrated in papers IV, V, VII and VIII. Mainly long range (~10–50 m) measurement performance was examined. The results are shown in papers IV, V, VII and VIII and summarized briefly in this chapter. The modifications needed to make a 2D or 3D measurement device are discussed later. Also a measurement was made to find the highest distance measurement precision available. Previously unpublished results about this measurement are presented at the end of chapter 4.1. In all of these papers (IV, V, VII and VIII), the basic measurement setup shown in Fig. 1 was used with different emitter and detector size combinations, with the exception that an amplifier is not needed for SPAD detector output signals. In paper V, the match between the detector (100 µm diameter commercial circular SPAD) and the laser stripe width (120 µm) was better than in paper IV (40 µm x 40 µm square SPAD and 120 µm stripe laser), optical losses were smaller, and an optimized laser pulser was used (40 W, 90 ps in paper V versus 25 W, 100 ps in paper IV). In paper VII, the optical losses were still smaller, and an in-lab designed CMOS SPAD detector was used to prove that the detector can be customized to a matrix setup for 3D measurements etc. Also in paper VII, outdoor measurements were demonstrated to a distance of 50 m. In Paper VIII, more results were given regarding this outdoor measurement, and a new gating scheme to reduce SPAD blocking due to background light was proposed. The high energy optical pulses from the proposed laser pulser enable a long measurement range with only a few measurement pulses to a passive target even in conditions of strong background light. Fig. 19 shows such a measurement result from paper VII which was made outside on a sunny day at air mass 1.5 conditions with roughly 70 klux measured illuminance at the target. A non-cooperative black biplanar target with a low reflectivity of about 4% was used at a distance of 50 m, i.e., the laser beam was focused halfway between two black planes separated by 5.1 cm, and the 20 µm SPAD was gated on at a time corresponding with a distance of 48.5 m. The laser pulser listed third in Table 2 was used (100 kHz, 1.2 nJ, 125 51 psFWHM), and a 50 nm optical bandpass filter was used before a detector lens with a radius of 9 mm and a focal length of 2 cm. The input and output beams were focused to fully overlap at the measurement distance. In these conditions, a background count rate of 13.7 MHz and a 14 percent signal detection rate were counted (with an exponential background and signal count decay after the gate-on time due to SPAD detector blocking, as described later). According to (5), about 20 laser pulses are needed for a reliable SBR of 10 in this case, corresponding with a valid distance measurement rate of ~5 kHz. With the quantum well version of the laser diode listed fourth in Table 2, the valid distance measurement rate could in these conditions be increased to about 34 kHz according to (5), since the achieved pulse energy is about 60% smaller, but the pulsing frequency is ten times higher, and the PDP of CMOS SPADs is beneficially about twice as high at the emission wavelength of the quantum well GaAs/AlAs laser diode (~810 nm), compared to the ~3% PDP at the ~860 nm wavelength of the bulk GaAs/AlAs laser diode. Note that a very low reflectivity target was used in these measurements, so with typical target reflectivities the measurement rate can be even higher. Another result that can be concluded from Fig. 19 is that different targets at different ranges within the field of view of the detector can be distinguished even with a single SPAD detector element, if the targets are separated by ~5 cm. An even higher range resolution of nearby targets can be achieved with post-processing of the waveform at the cost of increased system complexity (Hernandez-Marin et al. 2007, McCarthy et al. 2013). The diffusion tail adds ambiguity to the interpretation of the average pulse flight time, but fortunately only a minority of counts are diffusion tail induced in Fig. 19, and with advanced SPAD structures the exponential tail can be made smaller (Hsu et al. 2009, Bronzi et al. 2012). 52 detection probability / pulse / 61 ps bin 0.1 5 bin difference (61 ps / bin) ~51 mm 0.01 gate on 0.001 12% trigger rate from background ~14% trigger rate photons before signal from signal photons 48 48.5 322 324 49 326 49.5 328 330 50 332 50.5 distance [m] 51 334 336 338 340 pulse time-of-flight [ns] Fig. 19. Distribution of SPAD counts from a biplanar black target at 50 m in 70 klux background illumination. Measurement time 120 s, pulsing frequency 100 kHz, TDC bin width 61 ps. In the case of Fig. 19, a SPAD gate was applied 10 ns before the signal. That is, the bias voltage of the SPAD was decreased below the breakdown voltage only shortly before the signal arrival time. The advantage of such a gate is that fewer background photons have had the chance to block the SPAD operation before the arrival of signal photons (the SPAD is not operational for an interval called the dead time after each detection due to the required current quenching). Usually, the target position is not known in advance, so such a gate cannot be applied before the signal. However, in paper VIII a gating scheme was suggested to decrease the blocking of SPAD due to background photons, even if the target position is not known in advance. The fraction of time that a SPAD is operational can also be increased using a multi-element SPAD detector. Dividing the input power between several detector elements decreases the background count rate per detector and the resulting blocking of SPAD operation. The resulting decrease of the signal photon rate per detector is also often beneficial in terms of accuracy, as explained in the following chapters. Moreover, 3D-measurements are possible with multi-element detectors. A multi-element detector unfortunately also increases optical losses due to the 53 decreasing fill factor. For example, the fill factor (the fraction of SPAD areas sensitive to light within the total area of the SPAD matrix) of the 128 x 128 SPAD imager by Niclass et al. (2008) was only about 5 percent. Later, however, using a microlens array aligned on top of the SPAD matrix, a concentration factor of over 10 was achieved for the same SPAD imager matrix by Pavia et al. (2014). Niclass et al. (2013) achieved a fill factor of 70% for a 64 x 6 SPAD matrix without microlenses. The background light illuminance is typically much smaller indoors: about 200 lux versus 70 klux in the above case. According to (5) this improves the SBR by more than an order of magnitude, enabling much faster and more precise lidar measurements indoors. One of the advantages of using a SPAD detector in distance measurements is the high, in theory perfect, accuracy of the measurements, as long as each detection is caused by an individual photon. In practice, to achieve this condition, the average photoelectron number per SPAD should be less than 0.1 per laser pulse according to Poisson statistics, as was depicted by Fig. 4. The “walk-error”, i.e. the timing shift for non-single-photoelectron signals was measured in paper IV. The result in Fig. 20 shows that the measurement is indeed walk-error free at average photoelectron numbers smaller than about 0.1, and a distance measurement inaccuracy (i.e. walk-error) of about 5 cm is obtained if the average photoelectron number per SPAD is 100. Splitting the sensitive detector area into a SPAD matrix guarantees walk-error free detection for larger reflection amplitudes. The detector matrix could even have various detector sizes, the smallest of which would be more inclined to operate in true single photoelectron detection mode to give accurate results for highly reflecting targets. 54 center of gravity of count histogram 172 ns 171.9 ns 171.8 ns 171.7 ns 171.6 ns 10 -2 10 0 10 2 10 4 average number of photoelectrons in SPAD per laser pulse Fig. 20. Measured walk-error of a SPAD as a function of photon number (paper IV, published by permission of IOP). The distance measurement precision on the other hand increases as a function of the number of measurement results. The optimum detection scheme for a single photon detector in terms of precision was studied by Bar-David (1975) and Abshire (1978) for several pulse shapes. If the diffusion tail and background noise is filtered out by thresholding, the signal count histogram is close to Gaussian shape. In this case, according to the maximum likelihood method, a simple center of gravity (COG) of timing marks (i.e. average) was found to be the optimum timing estimate by Bar-David (1975) and Abshire (1978). In this case, the measurement precision σN increases as a function of the averaged signal counts N (Neville & Kennedy 1964) N   N , (12) where σ is the standard deviation of the collected count times. σ depends on the laser pulse shape and SPAD jitter. The practical extent of distance measurement precision improvement was tested by averaging a large number of measurement signals. In the test setup, a motorized target was used in a dark room, where the target moved back and forth 2 mm between every 1000 gated SPAD counts. The start signal to the TDC was taken from the optical signal with an optical fiber splitter connected to a fast photodetector utilizing constant fraction timing detection. Fig. 21 a) reveals a drift problem of distance measurement results due to temperature variations in the room. 55 An early laser diode version was used here, the shape of which had a strong temperature dependence. Also a low 1 kHz pulsing rate was used, and therefore the room temperature varied by a few degrees Celsius during this measurement, which was performed over 22 hours. The measurement results however, show that micrometer precision can be obtained especially if the pulsing frequency is increased and the laser pulse shape can be made less temperature dependent, both of which were achieved in the results of chapter 3. time interval measurement result with a SPAD detector [ns] 10.72 COG of thresholded 103 histogram population COG of thresholded 104 histogram population COG of thresholded 105 histogram population 10.71 10.7 10.69 10.68 10.67 10.66 0 106 2·106 3·106 a) 4·106 5·106 6·106 7·106 total number N of SPAD signal counts collected measurement result difference from actual distance of 2.0 mm [mm] 1 σ N standard deviation of measurement results 0.5 0 -0.5 individual measurement results -1 103 104 b) 105 number N of SPAD signal counts collected Fig. 21. a) TOF measurement results from a target alternating between positions ~2 m ± 1 mm. A histogram of 1000 signal counts was collected alternatingly at each distance, and centers of gravities were calculated for thresholded histograms (threshold at 4% of histogram population). b) Distance measurement precision as a function of the signal count number. 56 4.2 Measurement of the char bed level in a recovery boiler A pulsed TOF range finding measurement was also carried out in the recovery boiler of a wood pulp plant to solve a previously unsolved practical problem. The recovery boiler is an important part in the conversion of wood to wood pulp. Black liquor (a waste product from the pulp making process) is burned in the recovery boiler, leading to the recovery of necessary products for pulp production, and as a useful by-product electricity is created at a MW-level. The black liquor is sprayed into the recovery boiler from spraying holes (Fig. 22). The measurement of the level of the resulting char bed at the bottom of the boiler was the target of this measurement. The measurement was made from one of the spraying holes which was cleared for making the measurement (photograph in Fig. 22 b)). This was a single direction proof-of-concept measurement to check the feasibility of making TOF measurements using a SPAD detector. The final goal is to make a real-time multi-point 3D-measurement device for char bed profiling, which would enable the optimization of the recovery boiler process. Using a linear detector is not possible in this measurement situation due to the high backscattering from the flue gas inside the chamber. laser probe spraying holes 45˚ 6.12 m char bed 0–1m a) b) Fig. 22. a) Setup for measuring char bed level fluctuations. b) Photograph of the laser probe head in the measuring position at the spraying hole. The first version of the measurement setup in 2013 used a 532 nm pulsed laser which failed to measure the char bed level variations due to high scattering from 57 flue gas particles in the chamber. In the second measurement setup of 2014 (Figs. 22 and 23), a Manlight MLT-PL-R-OEM-0,5-100-1,5 pulsed fiber laser with a higher wavelength of 1550 nm was used since scattering due to small particles reduces as a function of laser wavelength. On the detection side, a Princeton Lightwave PNA-308-1 InGaAs/InP SPAD detector was used. The results of the successful measurement from 2014 are presented in the following. Agilent E3631A +12 V 5A +5 V fan heat sink 2x Peltier TEC1-12706 1550 nm Manlight laser 10 m RS232 pulse FGA01FC PINinverter diode, Ø0.12 mm start-pulse 0.5 ns, 1.5 µJ, @ 100 kHz 352220 FC/APC collimator 10 m, 105 µm fiber Becker & Hickl SPC-130 counts Thorlabs single-mode fixed optical attenuator, FC/PC 10 m, 105 µm fiber Δt single-mode pigtail, mode field diameter @ 1550 nm: 10.4 µm lens + NIR-filter (9.5 nm pass-band) stop-pulse Agilent E3631A: 15 V Stanford PS310: 76 V Melles Griot MG 06 DTC 101 TEC controller PNA-308-1 Coax Fiber Pigtailed Discrete SPAD in TO8 with three stage cooler Fig. 23. The electronics and optics setup for measurement of the char bed level. Fig. 24 a) shows one of the obtained distance signals from the char bed as a function of time. The dip in measurement result during 600–900 s is most likely due an individual char bed pile momentarily growing in height to the level of the laser beam. Fig. 24 b) shows the number of obtained signal counts from the target, compared with the number of signal counts obtained in a clear space with a white 58 target at an equal range. The average flue gas attenuation was measured to be much higher than this, and the success of the measurement was attributed to a high variance of flue gas attenuation or ballistic pathways opening up for individual photons. Fig. 24 c) shows the SBR as a function of time, which depends on the instantaneous flue gas attenuation and background radiation values. 9 distance measurement result [m] (based on histogram peak position) measurement break 8.5 8 7.5 1/300 500 1000 1500 2000 2500 3000 3500 4000 a) time [s] fraction of signal detections versus measurement without flue gas 500 1000 1500 500 1000 1500 1/200 1/100 30 2000 2500 3000 3500 4000 time [s] b) SBR (ratio of signal counts to the variance of background count bins) 20 10 0 2000 c) 2500 3000 3500 4000 time [s] Fig. 24. a) A successful single direction measurement result of char bed level from Wednesday 7th May 2014, 9.52 – 10.57 with the recovery boiler operating (char bed level changing). The results are based on obtained count histograms from 30 s measurement periods smoothed with a 6-point moving average. b) The fraction of successful distance measurements as a function of time. c) The SBR of the measured histograms as a function of time. 59 Fig. 25 shows a typical count histogram obtained with a measurement period of 30 s. The background count rate was high (about 260 kHz typically) due to strong radiation of the ~1500 ˚C target, even though the signal was strongly attenuated with the large diameter mismatch between the SPAD input fiber and the receiver fiber (about 10 µm and 105 µm, correspondingly). In addition, an input optical attenuator was used, as shown in Fig. 23, to avoid saturation of the detector. If longer measurement periods were to be used, it would need to be taken into account that the target position typically changes significantly in periods above ~30 s, as can be seen from Fig. 24 a). x 10-5 3.5 detection probability / pulse / 58 ps bin signal counts from target 3 2.5 2 1.5 1 0.5 0 0 2 10 4 20 30 6 8 10 distance from probe tip [m] 40 50 60 pulse time-of-flight [ns] Fig. 25. a) An example count distribution histogram from the measurement of Fig. 24, for a 30 s measurement period from 2130 s to 2160 s. A six-point moving average was used to smooth the measurement result. Out of eight measurement attempts, a clear backscatter signal (SBR > 10) was detected from the char bed on three occasions, a less clear signal (SBR < 10) was detected on three occasions, and two measurement attempts had an insufficient SBR to be able to distinguish a signal. Table 3 shows an overview of the results. 60 Table 3. Overview of char bed level measurement results. Measurement time Result Tuesday 14.55–15.43 no signal Tuesday 16.00–16.40 no signal Gas attenuation Notes Bad measurement due to probe shaking too much, no support yet. Wednesday 9.52–10.57 clear signal from Wednesday 11.04–12.15 clear signal from ~100 Result in Figs. 24 and 25 ~75–150 Clear signal for 30 minutes before >150 Partly no signal from target due to small ~200 Partly no signal from target due to small 7.75–8.5 m 7.8–8.4 m Wednesday 12.50–13.13 weak signal from Wednesday 13.15–14.02 weak signal from Wednesday 17.23–17.45 weak signal from spraying hole clogging 8.0–8.3 m SBR 8.3–8.8 m SBR ~200 - - Receiver attenuation was decreased, gas 8.2–8.4 m Thursday 10.08–10.23 clear signal from 8.0–8.2 m 4.3 attenuation not calculable due to this Raman spectroscopy using gated SPAD detection In the above pulsed time-of-flight experiments in chapter 4.1 and 4.2, the diffuse optical reflections from targets were assumed to originate mostly from elastic scattering, and other minor effects were ignored, as is reasonable. However, to be exact, a small portion of the optical pulse is inelastically scattered, meaning that the energy of scattered photons sometimes changes during the interaction with the molecular vibrations, phonons or other excitations in the target system. The shifts in energy of the photons are dependent on the chemical composition of the target. Therefore if these shifts, i.e. the Raman spectra can be measured, the chemical composition of the target can also be identified. A factor that makes the identification more difficult is that many materials emit also fluorescent light at different energies than the energies of the input photons. Typically the fluorescent emission decays exponentially within a period of some nanoseconds or more after the laser stimulus, whereas the Raman scattered photons are scattered almost instantly after laser stimulus (Fig. 26). This phenomenon was utilized first by Martyshkin et al. (2004) by picking the Raman scattered photons with a time gated CCD camera. In paper III, the time gating idea was utilized using a gated SPAD detector. In this first device version, a single mechanically scanned SPAD detector was used to 61 scan the spectrograph output. A commercial pulsed Nd:YAG laser with thermoelectric cooling (150 nJ pulse energy, 532 nm center wavelength, 0.2 nm linewidth, ~350 psFWHM pulse width, 100 kHz pulsing rate) was used in these measurements, and an olive oil sample was used due to its high fluorescence to demonstrate the excellent fluorescence rejection capability. Fig. 26. a) The measurement setup for Raman spectrometry. b) Principle of operation and fluorescence rejection (paper III, published by permission of IEEE). The measurement results from paper III in Fig. 27 show that even though olive oil is highly fluorescent, the Raman scattered photons can be detected clearly. In later versions of the device, a SPAD detector matrix was used and other improvements to the system were made (Nissinen et al. 2015). A spin-off company, TimeGate Instruments Oy, is now commercializing the developed time-gated Raman spectrometer. Fig. 27. Raman spectrum of an olive oil sample a) without time gating b) with a time gate of 300 ps (paper III, published by permission of IEEE). The Raman shift values above are corrected from the erroneous values given in paper III. 62 5 Effect of signal photon quantum shot noise on linear detection schemes Instead of using single-photon detectors as in the previous chapter, a more conventional way to perform pulsed TOF range finding is to use a linear optical detector, such as an APD detector with a finite value of avalanche multiplication. The main advantage of linear detection compared to the single-photon technique used in chapter 4 is the true single-shot sub-centimeter precision and accuracy of distance measurement possible in a wide dynamic range of optical echo amplitude even with a comparatively long (~3 ns) and therefore energetic laser pulse (Kurtti & Kostamovaara 2012, Nissinen et al. 2012). The main downside to linear detection compared to use of SPAD detectors is the need for a complex linear amplifier channel for each linear detector. Whereas the Poisson-distributed nature of photons can be seen concretely in the SPAD detector case, the signal photon noise is usually hidden under electronics detection noise when a linear detector is used. But in some cases the signal photon shot noise can be a significant noise source, also when a linear detector is used. The subtle effects of this noise source on the pulse timing measurement precision in linear detection schemes were researched in papers I and and VI. To research the effect of this noise source on the jitter of a pulsed TOF rangefinder, the measurement setup shown in Fig. 28 was used. An APD detector and amplifiers with a typical amount of gain and bandwidth were used. The oscilloscope settings (20 mV/div) were chosen so that the oscilloscope noise was insignificant compared to other noise sources. The rms detection jitter at the detection level Vth can be calculated by dividing the instantaneous total rms noise voltage by the instantaneous signal derivative (slew rate) (Bertolini 1968). laser diode attenuating plate 20 GHz oscilloscope trigger channel beam splitter New Focus 25 GHz PIN diode RF ≈ 7 kΩ, BW ≈ 140 MHz G = +10 V/V, BW ≈ 350 MHz 200 MHz oscilloscope measurement channel Vth S2382 APD SA5212A AD8009 jitter Fig. 28. Measurement setup for jitter research. 63 Fig. 29 shows a time-domain noise measurement result made using the above measurement configuration. Two values of APD amplification were used, and the optical signal attenuation was chosen so that the average detector channel output signal amplitude was the same in both cases. According to this measurement result, the signal photon shot noise is more prominent with the higher value of M and therefore its effects on the jitter can be conveniently researched with this setup. multiplied (M≈15) photodetector current i(t) [µA] multiplied (M≈51) photodetector current i(t) [µA] 1.6 1.6 1.2 1.2 0.8 0.8 0.4 0.4 0 0 0 4 8 a) 12 0 16 20 time [ns] 100 100 80 80 60 signal quantum shot noise according to (3) 40 20 0 0 4 8 12 c) Fig. 29. a) Detection 16 20 time [ns] voltage 16 20 time [ns] measured total noise as a function of time 120 measured total noise as a function of time 12 input referred current noise calculated based on above measurements [nArms] 140 120 8 b) input referred current noise calculated based on above measurements [nArms] 140 4 waveforms 60 signal quantum shot noise according to (3) 40 20 0 0 4 8 12 16 (SNR=20) divided 20 time [ns] d) by the channel transimpedance at an APD internal gain of M≈15 and b) M≈51 (one hundred waveforms shown, measured with a Lecroy 830Zi-A real-time 25 ps sampling period oscilloscope). c) Solid line: Total input referred current noise as a function of time for the M≈15, SNR=20 signals, calculated from 1000 waveforms. Dashed lines: signal quantum shot noise calculated with (3) based on the measured laser pulse shape and an iteratively obtained channel single electron response function. d) Corresponding calculations for the M≈51, SNR=20 signals. 64 When using a linear detector, a certain timing moment needs to be picked from the optical input pulse, because the typical optical pulse length of several nanoseconds corresponds with tens of centimeters of flight distance. This timing discrimination task is complicated by the large variation of reflected pulse power towards the detector according to (1). Pulse discriminators (also called time pick-off circuits) have been developed for this task of precise and accurate pulse timing for a wide dynamic range of input pulse amplitudes. The three researched discrimination schemes are illustrated in Fig. 30. The simplest timing discrimination method is leading edge detection (paper I), which in principle works with an infinite dynamic range of input amplitudes but needs compensation of the timing shift, i.e. walk error due to varying input echo amplitudes. High-pass discrimination (paper I) on the other hand has no walk error as long as the pulse does not get distorted in the receiver channel. In practice some dynamic range extension scheme is often needed in this case (Palojärvi et al. 1999, Pennala et al. 2000). The third discrimination scheme investigated here is to use two or more time domain parameters (for example leading and trailing edge crossing times) for the timing detection (paper VI). This idea has been shown to work well even when the input signal is clipped, therefore enabling measurements in a very large dynamic area of input power (Kurtti & Kostamovaara 2011 and Kurtti & Kostamovaara 2012). The effects of signal photon shot noise on all three of these discrimination schemes were researched here. vin trailing edge crossing times can be used for compensation of walk error, or high-pass discrimination can be used vout(high-pass) C Vth=10·vnoise_rms typ. leading edge detection timing shift, i.e. walk error t R t high-accuracy timing point but only for non-distorted input signals Fig. 30. Principles of pulse timing discrimination for input pulses with highly varying amplitude. In the following chapters, the effects of signal photon noise on the jitter properties of detected pulses are summarized based on the results given in papers I and VI. First, in chapter 5.1, the signal level dependence of photon noise is shown to have 65 an effect on the leading edge jitter. In chapter 5.2, the signal photon noise transient is shown to have an effect on the optimum high-pass discrimination scheme. And finally in chapter 5.3, the limited bandwidth of the receiver channel is shown to have an effect on the detection jitter when a dual-triggering timing discriminator is used. 5.1 Jitter of leading edge triggering receiver In paper I, the effect of signal shot noise on the leading edge jitter was studied first by simulating the receiver channel in Matlab. A simulation result of the dependence of the optimum detection level as a function of signal amplitude is shown in Fig. 31. When electronics noise dominates (for example at small signal levels using a PIN-diode with no internal amplification), the leading edge pulse jitter is simply minimal at the level of maximum signal slew rate (in this case ~59 percent level). On the other hand, when an APD receiver is used, the internal amplification of the APD usually amplifies the signal photon shot noise to a significant level already at the minimum SNR of 10. Due to the signal level dependence of signal photon shot noise, the optimum detection level in terms of jitter is lower than the point of maximum signal slew rate in this case. Another observation made in paper I concerning the leading edge jitter was that for a certain input pulse amplitude, an optimum value of M exists to minimize detection jitter. 60% optimum detection fraction in terms of jitter 55% 50% 45% AP-diode receiver PIN-diode receiver 40% 35% 10 1 µA 102 10 µA 103 100 µA 104 1 mA 105 10 mA 106 100 mA SNR amplitude of photodetector output current is(t) Fig. 31. The simulated optimum detection level for low jitter in the case of a PIN diode receiver (M=1) and an APD receiver (M=100) for a Gaussian shaped optical pulse with 3.5 ns rise time and a receiver bandwidth of 100 MHz, and an input referred current noise of 100 nArms. 66 5.2 Jitter of high-pass discriminated signals Fig. 30 illustrated the principle of high-pass discrimination. The high-pass discrimination timing constant τHP (=RC) is usually chosen to minimize the jitter at zero-cross, and to provide sufficient over- and underdrive. In paper I, it was shown that the optimum τHP also depends on whether signal photon shot noise is prominent. In practice, this means that the optimum value of τHP depends on whether a PIN diode or APD receiver is used. Small SNR≈10 signals were examined in the simulation of Fig. 32 because the jitter decreases anyway with increasing pulse amplitude. For the used optical pulse shape and single electron response (same parameters as in Fig. 31), the signal slewrate at the zero-cross can be maximized with a 4 ns high-pass time constant, regardless of whether a PIN diode or APD receiver is used (Fig. 32 c) and d)). Also the electronics noise gets filtered more with a smaller high-pass time constant, regardless of whether a PIN diode or an APD receiver is used, as seen in Fig. 32 e) and f), and the signal shot noise at the zero-cross is greater with a smaller time constant due to the noise transfer function in both cases. However, since the signal shot noise is insignificant with the PIN diode receiver, the optimum τHP is smaller for the PIN diode (~3 ns) compared to the APD receiver (~5 ns) in this case. This simulation result was confirmed with measurements in paper I. 67 1 A photocurrent is(t) 500 nA τHP increased HP-discriminated signals 0 -500 nA 200 0 10 20 a) t [ns] slew-rate at zero-cross [A/s] 500 nA 100 1 ns 10 ns c) 100 ns τHP noise at zero-cross [nArms] electronics noise = total noise -500 nA 200 0 10 20 t [ns] b) slew-rate at zero-cross [A/s] 100 1 ns 150 50 0 1 ns shot noise at zero-cross 10 ns 100 ns τHP e) jitter at zero-cross [psrms] 600 400 1 ns PIN: optimum τHP in terms of jitter ~ 3 ns 10 ns g) 10 ns d) 100 ns τHP noise at zero-cross [nArms] total noise 100 50 800 τHP increased HP-discriminated signals 0 150 150 100 multiplied photocurrent is(t) 1A electronics noise shot noise at zero-cross 0 1 ns 900 10 ns 100 ns τHP f) jitter at zero-cross [psrms] 800 APD: optimum τHP in terms of jitter ~ 5 ns 700 100 ns τHP 600 1 ns 10 ns h) 100 ns τHP Fig. 32. Output signal is(t) and HP-discriminated signals with various high-pass time constants for a) a PIN-diode (M=1) and b) an APD receiver (M=100). Slew rate of the HPdiscriminated signals at the zero-cross as a function of τHP for c) a PIN diode and d) an APD receiver. Noise components at the zero-cross of HP-discriminated signals as a function of the HP time constant for e) a PIN diode and d) an APD receiver. Jitter at the zero-cross of HP-discriminated signals as a function of HP time constant for g) a PIN diode and h) an APD receiver (paper I, published by permission of IOP). 68 5.3 Jitter of dual-triggering receiver In Kurtti & Kostamovaara (2011) and (2012), it was noticed that the use of a multitriggering receiver improves the detection precision at low SNR. The mechanism by which the precision improves was researched more closely in paper VI. For simplicity, a dual-triggering receiver was used, where the average of the leading and trailing edge was used as a pulse timing estimate. The jitter was researched by analyzing the waveforms of Fig. 29 a) and b) in Matlab. First, the jitter on the leading and trailing edge was calculated from the measured pulses level-by-level (red and blue trace in Fig. 33 a) and b)). The leading edge jitter is smaller due to a larger edge speed, mainly because of the optical signal shape. As a side note, the red curves in Fig. 33 a) and b) also confirm the simulation results of chapter 5.1, according to which the optimum leading edge detection level is lower when the signal photon noise is prominent. The jitter of the dual-triggering receiver (average of leading and trailing edge) was calculated similarly level-by-level. The main result here is that the jitter of the dual-triggering receiver, plotted with the black curves in Fig. 33 a) and b), is advantageously smaller than either the leading or trailing edge solely. Furthermore, even though in the case of a larger M value the total noise is clearly larger according to Fig. 29, the jitter of the dual-triggering receiver is still about 110 psrms at the ~0.8 µA level with both values of M in Fig. 33 a) and b). In paper VI this was attributed to a negative correlation of signal photon shot noise induced timing shift on the leading and trailing edge. current level relating to the pulses in Fig. 29 a) [µA] current level relating to the pulses in Fig. 29 b) [µA] 1.6 1.6 1.2 1.2 0.8 leading edge jitter trailing edge jitter 0.4 0 jitter of measured average of leading and trailing edge 100 140 180 a) 220 260 jitter [psrm s] 0.8 leading edge jitter trailing edge jitter 0.4 0 jitter of measured average of leading and trailing edge 100 140 180 b) 220 260 jitter [psrm s] Fig. 33. Jitter of the pulses in Fig. 29 a) and b) on the leading and trailing edge, and their average. 69 70 6 Discussion The main result of this research is the development of a new type of laser diode and its driver for high energy and short ~100 ps laser pulse emission at fast pulsing rate. This type of laser pulser is needed for TOF measurement applications such as rangefinders and 3D imagers, and it is especially suitable for use with a SPAD detector due to matching between the laser pulse width and the jitter of a typical SPAD detector. Several TOF studies utilizing a SPAD detector and this laser transmitter, and other <1 ns laser pulse transmitters, were conducted in this work. Other research groups have also carried out TOF measurement studies, such as 3D imaging, using a CMOS SPAD detector. An optimum laser pulser for this application would emit short and energetic laser pulses at high pulsing frequency. Compactness of the laser pulser is also a requirement in many applications. Typically a trade-off exists between the laser pulsing performance and compactness of the device. The forerunners in 3D imaging, Niclass et al. (2005) started their 3D imager studies using a regular gain-switched laser diode. This achieved pulser compactness and laser pulse shortness (150 ps), but suffered from a low pulse energy of 15 pJ (compared to the ~1 nJ pulse energy from the laser pulser presented in this thesis). Later Niclass et al. (2008) used a more expensive and largedimensioned solid-state laser emitting 80 ps, 24 pJ pulses at 40 MHz, together with a 128 x 128 SPAD detector matrix achieving sub-centimeter 3D measurement precision to a maximum measurement distance of about 3.5 m. More recently, Niclass et al. (2013) used a regular compact semiconductor laser diode pulser emitting much wider ~4 ns laser pulses, providing significantly larger energy (~200 nJ) for 340 x 96 pixel 3D scanning up to a distance of 100 m at 10 frames/s. However, due to the long laser pulse length, more complicated processing of the results was needed to reach a good enough accuracy. In this case, Niclass et al. (2013) achieved a sigma value for distance measurement precision of about 2.5–10 cm, which equates to about 15–60 cm peak to peak distance measurement error variation. This is quite large, but might be sufficient for some applications. Elsewhere, Massa et al. (2002) and Buller & Wallace (2007) used a passively Qswitched AlGaAs laser diode emitting 10–20 ps pulses with 7–10 pJ energy at 25 MHz pulsing frequency. Even though the laser pulse energy was quite modest in these cases, narrow ~600 ps and large amplitude current pulses were needed to drive the laser diode, which necessitated a large laboratory instrument. In a more recent work, Villa et al. (2014) used a Picoquant LDH-P-FA-765 as a laser pulse source in their 3D SPAD TOF imager. The 5 nJ, <100 ps laser pulses from this 71 device are quite close in amplitude to those achieved from the laser diode pulser presented in this thesis, but the Picoquant device is a large and expensive laboratory device based on a master oscillator fiber amplifier requiring cooling. McCarthy et al. (2013) did 3D imaging to further ranges of a few kilometers using a modelocked fiber laser with dimensions of 9x12x3.2 cm emitting <100 pJ pulses. Even comparing to the large, expensive laboratory equipment used by many of the groups described above, the ~100 ps laser pulser presented in this work offers rather high ~1 nJ optical pulse energy from a compact ~1 cm3 device. The benefit of using a more energetic optical pulse in a pulsed TOF SPAD rangefinder or 3D imager application can be examined from several viewpoints. If true single-photon sensing is used, a tenfold increase in optical pulse energy leads to approximately ten times better SBR, which translates to ten times faster distance measurement or about three times longer measurement range, or three times better ranging precision in a certain measurement time. On the other hand, the optical pulse can be spread on a larger number of detectors enabling better vertical and horizontal resolution for 2D or 3D measurements, or alternatively the higher pulse energy can provide a faster measurement rate for a fixed vertical and horizontal resolution. The laser pulsing frequency was also improved in this work. This is mainly an issue of laser driver design, as the output from the laser diode is the usually the same for a certain injection current pulse shape, regardless of pulsing frequency (if the laser diode temperature is the same). A faster laser pulsing frequency directly translates to a faster valid distance measurement rate, although to avoid distance measurement ambiguity, only one laser pulse should be in flight at a time (unless pseudo-random pulsing is used (Hiskett et al. 2008)). For example, the maximum unambiguous measurement range with the achieved pulsing rate of 1 MHz is about 150 m. This measurement speed improvement gained from higher pulsing frequency can also be exchanged for better 1D or 3D measurement capability in a similar manner as described above. But note that a faster pulsing rate does not directly compensate for small SBR due to a small laser pulse energy, since increasing the pulsing frequency also increases background photon counts. According to (5), if all else is equal, the SBR of the measurement improves only proportionally to the square root of the pulsing frequency. Therefore the high laser pulse energy achieved can be considered a greater merit than the high pulsing frequency, although both are useful to achieve fast measurement results. One important potential application for this laser pulser is 3D imaging for vehicles. In this application, at least 25 Hz real-time 3D measurement to a distance of about 50…100 m would be preferred. Deducing from the achieved 1D 72 measurement results summarized in chapter 4.1, a valid 1D distance measurement rate of over ~8 kHz to a black (~4% reflectivity) target at 100 m can be expected in sunny conditions with the developed 1 MHz quantum well laser diode pulser and a SPAD detector. The 1D measurement speed can be traded off for 2D or 3D measurement capability using a detector matrix or a MEMS scanner. Therefore, roughly calculating, assuming 50 percent signal losses compared to a 1D measurement, a 160 pixel 3D measurement with centimeter precision could be performed in sunny conditions at a measurement rate of 25 Hz up to a distance of 100 m, and a 640 pixel 3D measurement could be carried out similarly up to a distance of 50 m with this laser pulser. This is quite good considering the ~1 cm3 size of the proposed laser pulser and the very low reflection coefficient of the target used in these calculations, but in practice better vertical and horizontal resolution is needed in this application. This could be achieved, for example, by using an array of laser diodes as is done in Velodyne laser scanners (Schwarz 2010), or with an even more energetic laser pulse. Recent simulations show that the optical pulse energy from the designed laser diode can still be increased by using a saturable absorber implemented with an unpumped laser diode section as was done by Lanz et al. (2013), except creating the unpumped section already at the molecular beam epitaxy fabrication stage should work better than using focused ion beam etching. Another option for increasing the laser pulse energy is to abandon the gainswitching regime and use longer laser pulses, as was done by Niclass et al. (2013). That is, a more energetic optical pulse is easier to create if a longer optical pulse length is allowed. However, using a longer laser pulse increases the SBR at the cost of measurement precision. For example, if using a 4 ns laser pulse in pulsed TOF SPAD distance measurements, centimeter distance measurement precision would need averaging of about 1000 signal counts according to (12), whereas only a few signal counts from a 100 ps laser pulse are sufficient for cm-precision. Yet another way to improve the performance of 1D–3D scanning while keeping the short ~100 ps laser pulse for precise distance measurement would be to create a ~100 ps laser pulser for a higher wavelength using the proposed gain-switching enhancement principles. This would decrease the solar background radiation falling on the signal bandwidth due to the spectrum shape of solar radiation. Also quantum efficiencies can be much higher with NIR InGaAs SPADs (~20–40% at ~1000– 1550 nm wavelength range (Acerbi et al. 2013) compared to ~3–8% with Si CMOS SPADs at 810–860 nm (Guerrieri et al. 2010). Moreover, higher average powers can be used at NIR wavelengths in terms of eye safety (Henderson & Schulmeister 73 2004). The downside to this approach is that integration of SPADs and TDCs is not currently possible in InGaAs processes. As another suggestion for further work, the laser diode could be commercialized by packaging the laser diode together with an integrated laser driver and possible temperature compensation circuit via Vbias manipulation. The main competition in this size category is regular gain-switched laser diodes emitting about ~10 times smaller pulse energies than is available with the gainswitching enhancement proposed here. The Raman and char bed measurements made in this work imply that a wide array of applications are possible with single photon techniques and devices (although special fiber lasers were required in these two particular demanding applications). New applications for the created laser pulser might also emerge due to the improving resolution and functionality of scaling CMOS SPAD image sensors. The presented laser pulser emitting short ~100 ps laser pulses could also be used in conjunction with a linear TOF receiver channel. The use of <100 ps laser pulses with a linear pulsed TOF rangefinder receiver channel was suggested already by Pennala et al. (2000), leading to even millimeter single-shot measurement precision compared to the centimeter precision obtainable with nanosecond optical pulses. However, a very high receiver bandwidth with high current consumption is needed in this case, whereas a low-current consumption SPAD detector inherently responds well to short ~100 ps optical pulses. A natural question that arises is whether single photon detectors or linear detectors are better for use in pulsed TOF range finding and 3D imaging applications. The main advantage of using a linear receiver instead of a SPAD receiver is that single-shot distance measurement precision of a few centimeters is possible even using a long (several nanoseconds) and therefore energetic laser pulse (for example a single-shot precision of ~20–144 ps was achieved in Kurtti & Kostamovaara (2011)). The main advantage of SPAD detectors on the other hand is the much simpler receiver structure as no amplifiers are needed after the SPAD detectors. 74 7 Summary The main interest behind this research work was to study for possibilities to utilize single photon avalanche diode detectors in compact laser distance meters and 3D imagers for automotive applications, computer interfaces, industry etc. For this purpose, a high energy laser pulser was needed. The key achievement of this research was the development of a new small ~1 cm3 laser pulser emitting energetic, short ~100 ps laser pulses at high pulsing frequency. The core idea in the new laser pulser is the gain-switching enhancement principle suggested by Ryvkin et al. (2009), according to which a large ratio (>>1 µm) of active layer thickness (da) to optical confinement factor (Гa) should be used to improve gain-switching dynamics for short and energetic laser output pulse emissions. One suggested way of achieving the low confinement factor is to use an asymmetric waveguide structure. Measurement results showed that the gainswitching enhancement principle works, and provides high laser pulse energies of about ~0.5–4 nJ, depending on the laser diode stripe width. Without the proposed gain-switching enhancement the pulse energies are about an order of magnitude smaller at the ~100 ps laser pulse width regime. To drive the laser diodes, initially an avalanche transistor current pulser circuit was used, which provides large ~15 A and narrow ~1.5 ns current pulses. With these current pulses, a wide 137 µm laser diode could be driven providing large 4 nJ, ~100 ps optical pulses. However, the pulsing frequency was limited to about 20 kHz because of high power dissipation of the avalanche transistor due to its high residual voltage. To fight the pulsing frequency limitation, a modified FET current pulser structure was created. An about fifty times higher pulsing frequency of ~1 MHz compared to previously used avalanche transistor current pulsers was achieved at a decreased current pulse amplitude of ~5 A, but an advantageously narrower current pulse width of ~1.1 nsFHWM. At a lower 100 kHz pulsing rate, ~12 A, ~1.5 nsFWHM current pulsing was available from the FET pulser. One goal of this research was to enable high performance of laser pulsing in a wide range of ambient temperatures (for automotive applications etc.). The main conclusion was that the laser diode output can be made less susceptible to ambient temperature changes by operating the laser diode at a current level as far above the lasing threshold current as possible. This way, changes in the lasing threshold current (temperature dependent) have a smaller relative effect on the output pulse power. Fortunately, the presented gain-switching enhancement principle also works 75 with a quantum well laser structure, which has a lower lasing threshold current than a bulk laser diode. 1 MHz, ~0.5 nJ, ~100 psFWHM laser pulsing was achieved at room temperature with a 30 µm wide, 3 mm long quantum well laser diode, and Huikari et al. (2015) showed that the pulse power halving due to an ambient temperature change from 22 °C to 60 °C can be almost fully compensated with a temperature - current pulse amplitude feedback circuit. This might be more effective with the developed FET pulser compared to an avalanche transistor circuit, since the current pulse length stays much more constant as a function of current pulse amplitude with the FET pulser. The main application for the created compact laser pulser is precise pulsed TOF range finding and 3D scanning with a SPAD detector. The short ~100 ps laser pulse inherently leads to centimeter precision distance measurement, even with the simplest of pulse timing estimates, and the high pulse energy and high pulsing frequency enable long measurement range and high measurement speed. For example, in paper VII, a reliable SBR>10 signal enabling centimeter distance measurement precision was obtained from a black target at 50 m at a frequency of ~5 kHz in sunny 70 klux conditions. With the achieved quantum well laser diode pulser exhibiting a higher pulsing frequency of 1 MHz, and a more favorable emission wavelength (~810 nm) in terms of SPAD PDP, a sixfold increase in the valid distance measurement rate can be expected. The 1D measurement speed can be exchanged for 3D measurement capability with a SPAD detector matrix or scanning mechanics. The effectiveness of single photon measurement techniques was also demonstrated in a demanding factory environment. A real-time char bed level measurement was carried out in the recovery boiler of a pulp mill. Linear pulse detection schemes are not possible in this application due to the high backscattering and attenuation by the flue gas inside the chamber. Single photon techniques proved effective, as rarely, but often enough, a ballistic pathway opened up for signal photons. The char bed level changed within a region of about 1 m as black liquor was sprayed into the chamber. A single direction measurement of the char bed level was made at a 1/30 s measurement rate with a precision of a few centimeters. The results showed that with further system improvements it might be possible to make a real-time 3D scanner of the char bed surface with single-photon techniques. As another single photon measurement demonstration, a Raman spectrometer prototype utilizing a single photon detector was built in this research work. Raman scattered photons were shown to be distinguishable from fluorescent photons with a time-gated single photon avalanche diode. 76 Photonic effects were also researched in the case of TOF measurements with a linear detector. The photonic nature of light leads to an inherent noisiness of the optical detection waveform. In this research, it was shown that its effects are different depending on the pulse discriminator type used. In leading edge detection, the optimum detection level is simply lower than the level of maximum signal slew rate due to the signal level dependence of signal shot noise. In high-pass discrimination, a compromise between the signal slew rate and signal shot noise can be made by choosing an appropriate high-pass time constant. 77 78 References Abend K (1966) Optimum photon detection (Corresp.). IEEE Transactions on Information Theory 12(1): 64–65. Abshire J (1978) A comparative study of optimum and suboptimum direct-detection laser ranging receivers. Greenbelt MD, National Aeronautics and Space Administration, Scientific and Technical Information Office. Acerbi F, Anti M, Tosi A & Zappa F (2013) Design criteria for InGaAs/InP single-photon avalanche diode. IEEE Photonics Journal 5(2): 6800209. Agrawal GP & Dutta NK (1993) Semiconductor lasers. New York NY, Van Nostrand Reinhold. Alexander SB (1997) Optical communication receiver design. Bellingham WA, SPIE Optical Engineering Press. Bao Z, Li Z, Shi Y, Wu E, Wu G & Zeng H (2014) Coincidence photon-counting laser ranging for moving targets with high signal-to-noise ratio. IEEE Photonics Technology Letters 26(15) 1495–1498. Bar-David I (1975) Minimum-mean-square-error estimation of photon pulse delay (corresp.). IEEE Transactions on Information Theory 21(3): 326–330. Braga LHC, Gasparini L, Grant L, Henderson RK, Massari N, Perenzoni M, Stoppa D & Walker R (2013) An 8×16-pixel 92kSPAD time-resolved sensor with on-pixel 64ps 12b TDC and 100MS/s real-time energy histogramming in 0.13µm CIS technology for PET/MRI applications. Digest of Technical Papers IEEE International Solid-State Circuits Conference (ISSCC): 486–487. Becker W (2012) The bh TCSPC Handbook. Berlin, Becker & Hickl. Bertolini G (1968) Pulse shape and time resolution. In: Bertolini G & Coche A (eds.) Semiconductor detectors. Amsterdam, Holland, North-Holland Publishing Co. Bronzi D, Villa F, Bellisai S, Markovic B, Tisa S, Tosi A, Zappa F, Weyers S, Durini D, Brockherde W & Paschen U (2012) Low-noise and large-area CMOS SPADs with timing response free from slow tails. Proceedings of the European Solid-State Device Research Conference (ESSDERC): 230–233. Buller G & Wallace A (2007) Ranging and three-dimensional imaging using time-correlated single-photon counting and point-by-point acquisition. IEEE Journal of Selected Topics in Quantum Electronics 13(4): 1006–1015. Collis RTH & Russell PB (1976) Lidar measurement of particles and gases by elastic backscattering and differential absorption. Chapter in Laser Monitoring of the Atmosphere 14: 71–151. Cubeddu R, Bassi A, Comelli D, Cova S, Farina A, Ghioni M, Rech I, Pifferi A, Spinelli L, Taroni P, Torricelli A, Tosi A, Valentini G & Zappa F (2011) Photonics for life. IEEE Pulse 2(3): 16–23. Duan G, Vainshtein SN, Kostamovaara JT, Zemlyakov VE & Egorkin VI (2014) 3-D properties of the switching transient in a high-speed avalanche transistor require optimal chip design. IEEE Transactions on Electron Devices 61(3): 716–721. 79 Gersbach M, Niclass C, Charbon E, Richardson J, Henderson R & Grant L (2008) A single photon detector implemented in a 130nm CMOS imaging process. Proceedings of the 38th European Solid-State Device Research Conference (EESDERC): 270–273. Guerrieri F, Tisa S, Tosi A & Zappa F (2010) Two-dimensional SPAD imaging camera for photon counting. IEEE Photonics Journal 2(5): 759–774. Haitz R H (1965) Mechanisms contributing to the noise pulse rate of avalanche diodes. Journal of Applied Physics 36(10): 3123–3131. Heinzen DJ & Wineland DJ (1990) Quantum-limited cooling and detection of radiofrequency oscillations by laser-cooled ions. Physical Review A 42(5): 2977–2994. Hell SW & Wichmann J (1994) Breaking the diffraction resolution limit by stimulated emission: stimulated-emission-depletion fluorescence microscopy. Optics Letters 19(11): 780–782. Henderson R & Schulmeister R (2004) Laser safety. Bristol UK, Institute of Physics Publishing. Heinzen DJ & Wineland DJ (1990) Quantum-limited cooling and detection of radiofrequency oscillations by laser-cooled ions. Physical Review A 42(5): 2977–2994. Hernandez-Marin, S, Wallace AM & Gibson GJ (2007) Bayesian analysis of lidar signals with multiple returns. IEEE Transactions on Pattern Analysis and Machine Intelligence 29(12): 2170–2180. Hiskett PA, Parry CS McCarthy A & Buller GS (2008) A photon-counting time-of-flight ranging technique developed for the avoidance of range ambiguity at gigahertz clock rates. Optics Express 16(18): 13685–13698. Hsu, MJ, Esener SC & Finkelstein H (2009) A CMOS STI-bound single-photon avalanche diode with 27-ps timing resolution and a reduced diffusion tail. IEEE Electron Device Letters 30(6): 641–643. Huikari JMT, Avrutin, EA, Ryvkin BS, Nissinen JJ & Kostamovaara JT (2015) High-energy picosecond pulse generation by gain switching in asymmetric waveguide structure multiple quantum well lasers. IEEE Journal of Selected Topics in Quantum Electronics 21(6): 1501206. IEEE (1996) The IEEE standard dictionary of electrical and electronics terms. IEEE Standard 100-1996. IEEE Standards Office, Piscataway, New Jersey, USA. Isaacson W (2007) Einstein: his life and universe. New York NY, Simon & Schuster. Johnson KM (1965) High-speed photodiode signal enhancement at avalanche breakdown voltage. IEEE Transactions on Electron Devices 12(2): 55–63. Kasap SO (2013) Optoelectronics and photonics: principles and practices. Boston MA, Pearson. Kilpelä A & Kostamovaara J (1997) Laser pulser for a time-of-flight laser radar. Review of Scientific Instruments 68: 2253–2258. Kurtti S & Kostamovaara J (2011) An integrated laser radar receiver channel utilizing a time-domain walk error compensation scheme. IEEE Transactions on Instrumentation and Measurement 60(1): 146–157. 80 Kurtti S & Kostamovaara J (2012) An integrated receiver channel for a laser scanner. IEEE International Instrumentation and Measurement Technology Conference (I2MTC). Graz, Austria: 1358–1361. Lanz B, Ryvkin BS, Avrutin EA & Kostamovaara JT (2013) Performance improvement by a saturable absorber in gain-switched asymmetric-waveguide laser diodes. Optics Express 21(24): 29780–29791. Mandel L (1965) Phenomenological theory of laser beam fluctuations and beam mixing. Physical Review 138(3B): 753–762. Martyshkin DV, Ahuja RC, Kudriavtsev A & Mirov SB (2004) Effective suppression of fluorescence light in Raman measurements using ultrafast time gated charge coupled device camera. Review of Scientific Instruments 75(3): 630–635. Massa J, Buller G, Walker A, Smith G, Cova S, Umasuthan M & Wallace A (2002) Optical design and evaluation of a three-dimensional imaging and ranging system based on time-correlated single-photon counting. Applied Optics 41(6): 1063–1070. McCarthy A, Krichel NJ, Gemmell NR, Ren X, Tanner MG, Dorenbos SN, Zwiller V, Hadfield RH & Buller GS (2013) Kilometer-range, high resolution depth imaging via 1560 nm wavelength single-photon detection. Optics Express 21(7): 8904–8915. McIntyre RJ (1999) A new look at impact ionization-Part I: A theory of gain, noise, breakdown probability, and frequency response. IEEE Transactions on Electron Devices 46(8): 1623–1631. Neilsen KD (2011) Signal processing on digitized ladar waveforms for enhanced resolution on surface edges. MSc thesis, Utah State University, Department of Electrical and Computer Engineering. Neville AM & Kennedy JB (1964) Basic statistical methods for engineers and scientists. London UK, Intertext Books. Niclass C, Rochas A, Besse P-A & Charbon E (2005) Design and characterization of a CMOS 3-D image sensor based on single photon avalanche diodes. IEEE Journal of Solid-State Circuits 40(9): 1847–1854. Niclass C, Favi C, Kluter T, Gersbach M & Charbon E (2008) A 128 x 128 single-photon image sensor with column-level 10-bit time-to-digital converter array. IEEE Journal of Solid-State Circuits 43(12): 2977–2989. Niclass C, Soga M, Matsubara H, Kato S & Kagami M (2013) A 100-m range 10-frame/s 340 x 96-pixel time-of-flight depth sensor in 0.18-µm CMOS. IEEE Journal of SolidState Circuits 48(2): 559–572. Nissinen J, Nissinen I & Kostamovaara J (2012) An integrated CMOS receiver-TDC chip for mm-accurate pulsed time-of-flight laser radar measurements. Proceedings of the IEEE International Instrumentation and Measurement Technology Conference (I2MTC): 421–424. Nissinen I, Nissinen J, Keränen P, Länsman A-K, Holma J & Kostamovaara J (2015) A 2 x (4) x 128 multitime-gated SPAD line detector for pulsed Raman spectroscopy. IEEE Sensors Journal 15(3): 1358–1365. O’Shea DC (1985) Elements of modern optical design. John Wiley & Sons. 81 Palojärvi P, Ruotsalainen T & Kostamovaara J (1999) A new approach to avoid walk error in pulsed laser rangefinding. Proceedings of the 1999 IEEE International Symposium on Circuits and Systems (ISCAS) 1: 258–261. Pancheri L & Stoppa D (2007) Low-noise CMOS single-photon avalanche diodes with 32 ns dead time. Proceedings of the 37th European Solid State Device Research Conference (EESDERC): 362–365. Pancheri L, Massari N & Stoppa D (2013) SPAD image sensor with analog counting pixel for time-resolved fluorescence detection. IEEE Transactions on Electron Devices 60(10): 3442–3449. Pavia JM, Wolf M & Charbon E (2014) Measurement and modeling of microlenses fabricated on single-photon avalanche diode arrays for fill factor recovery. Optics Express 22(4): 4202–4213. Pennala R, Kilpela A & Kostamovaara J (2000) An integrated receiver channel for picosecond range laser radar pulses. Proceedings of the 43rd IEEE Midwest Symposium on Circuits and Systems 2: 702–705. Poultney SK (1972) Single Photon Detection and Timing in the Lunar Laser Ranging Experiment. IEEE Transactions on Nuclear Science 19(3): 12–17. Rochas A, Besse P-A & Popovic RS (2001) A Geiger mode avalanche photodiode fabricated in a conventional CMOS technology. Proceeding of the 31st European Solid-State Device Research Conference: 483–486. Rochas A, Gani M, Furrer B, Besse PA, Popovic RS, Ribordy G & Gisin N. Single photon detector fabricated in a complementary metal–oxide–semiconductor high-voltage technology. Review of Scientific instruments 74(7): 3263–3270. Ruotsalainen T, Palojarvi P and Kostamovaara J (2001) A wide dynamic range receiver channel for a pulsed time-of-flight laser radar. IEEE Journal of Solid-State Circuits 36(8): 1228–1238. Ryvkin B, Avrutin EA & Kostamovaara JT (2009) Asymmetric-waveguide laser diode for high-power optical pulse generation by gain switching. Journal of Lightwave Technology 27(12): 2125–2131. Ryvkin BS, Avrutin EA & Kostamovaara JT (2011) Quantum well laser with an extremely large active layer width to optical confinement factor ratio for high-energy single picosecond pulse generation by gain switching. Semiconductor Science and Technology 26(4): 045010. Ryvkin BS, Avrutin EA & Kostamovaara JT (2013) Narrow versus broad asymmetric waveguides for single-mode high-power laser diodes. Journal of Applied Physics 114(1): 013104. Schwarz B (2010) Lidar: mapping the world in 3D. Nature Photonics 4(7): 429–430. Shin D, Kirmani A, Goyal VK & Shapiro J (2014) Computational 3D and reflectivity imaging with high photon efficiency. IEEE International Conference on Image Processing (ICIP): 46–50. Texas Instruments (2011) LIDAR system design for automotive/industrial/military applications. Application note SNAA123. 82 Vainshtein SN, Yuferev VS & Kostamovaara JT (2003) Avalanche transistor operation at extreme currents: Physical reasons for low residual voltages. Solid-State Electronics 47(8): 1255–1263. Vainshtein S, Kostamovaara J, Sveshnikov Y, Gurevich S, Kulagina M, Yuferev V, Shestak L & Sverdlov M (2004) Superfast high-current switching of GaAs avalanche transistor. Electronics Letters 40(1): 85-86. Villa F, Lussana R, Bronzi D, Tisa S, Tosi A, Zappa F, Dalla Mora A, Contini D, Durini D, Weyers S & Brockherde W (2014) CMOS imager with 1024 SPADs and TDCs for single-photon timing and 3-D time-of-flight. IEEE Journal of Selected Topics in Quantum Electronics 20(6): 364–373. Wang J & Kostamovaara J (1994) Radiometric analysis and simulation of signal power function in a short-range laser radar. Applied Optics 33(18): 4069–4076. Wilman ES, Gardiner SH, Nomerotski A, Turchetta R, Brouard M & Vallance C (2012) A new detector for mass spectro-metry: Direct detection of low energy ions using a multipixel photon counter. Review of Scientific Instruments. 83: 013304. Yudovsky D & Pilon L (2009) Simple and accurate expressions for diffuse reflectance of semi-infinite and two-layer absorbing and scattering media. Applied Optics 48: 6670– 6683. Zappa F, Giudice A, Ghigni M & Cova S (2002) Fully-integrated active-quenching circuit for single-photon detection. Proceedings of European Solid-State Circuit Conference ESSCIRC-02: 355–358. Zhang Z, Yuan Z, Yong Z, Long W & Jianzhong S (2013) A real-time noise filtering strategy for photon counting 3D imaging lidar. Optics Express 21(8): 9247–9254. 83 84 Original publications I II III IV V VI VII VIII Hallman L W & Kostamovaara J (2008) Effect of signal quantum shot noise on the jitter of leading edge detected and high-pass discriminated laser pulse signals. Measurement Science and Technology 19(8): 085303. Hallman L W, Ryvkin B, Haring K, Ranta S, Leinonen T & Kostamovaara J (2010) Asymmetric waveguide laser diode operated in gain switching mode with high-power optical pulse generation. Electronics Letters 46(1): 65–66. Nissinen I, Nissinen J, Länsman A K, Hallman L, Kilpelä A, Kostamovaara J, Kögler M, Aikio M & Tenhunen J (2011) A sub-ns time-gated CMOS single photon avalanche diode detector for Raman spectroscopy. Proceedings of the European Solid-State Device Research Conference (ESSDERC). Helsinki, Finland: 375–378. Hallman L W, Haring K, Toikkanen L, Leinonen T, Ryvkin B S & Kostamovaara J T (2012) 3 nJ, 100 ps laser pulses generated with an asymmetric waveguide laser diode for a single-photon avalanche diode time-of-flight (SPAD TOF) rangefinder application. Measurement Science and Technology 23(2): 025202. Hallman L W, Haring K, Toikkanen L, Leinonen T, Ryvkin B S & Kostamovaara J T (2012) Compact laser pulser for TOF SPAD rangefinder application. Proceedings of SPIE Photonics West. San Francisco CA. Hallman L W & Kostamovaara J (2014) Note: Detection jitter of pulsed time-of-flight lidar with dual pulse triggering. Review of Scientific Instruments 85(3): 036105. Hallman L W, Huikari J & Kostamovaara J (2014) A high-speed/power laser transmitter for single photon imaging applications. Proceedings of IEEE Sensors. Valencia, Spain: 1157–1160. Kostamovaara J, Huikari J, Hallman L, Nissinen I, Nissinen J, Rapakko H, Avrutin E & Ryvkin B (2015) On laser ranging based on high-speed/energy laser diode pulses and single-photon detection techniques. IEEE Photonics Journal 7(2): 7800215. Reprinted with permission from IOP Publishing (I, IV), IET (II), IEEE (III, VII, VIII), SPIE (V) and AIP Publishing (VI). The original publications are not included in the electronic version of the dissertation. 85 86 C555etukansi.kesken.fm Page 2 Wednesday, November 11, 2015 9:42 AM ACTA UNIVERSITATIS OULUENSIS SERIES C TECHNICA 539. Karhu, Mirjam (2015) Treatment and characterisation of oily wastewaters 540. Panula-Perälä, Johanna (2015) Development and application of enzymatic substrate feeding strategies for small-scale microbial cultivations : applied for Escherichia coli, Pichia pastoris, and Lactobacillus salivarius cultivations 541. Pennanen, Harri (2015) Coordinated beamforming in cellular and cognitive radio networks 542. Ferreira, Eija (2015) Model selection in time series machine learning applications 543. Lamminpää, Kaisa (2015) Formic acid catalysed xylose dehydration into furfural 544. Visanko, Miikka (2015) Functionalized nanocelluloses and their use in barrier and membrane thin films 545. Gilman, Ekaterina (2015) Exploring the use of rule-based reasoning in ubiquitous computing applications 546. Kemppainen, Antti (2015) Limiting phenomena related to the use of iron ore pellets in a blast furnace 547. Pääkkönen, Tiina (2015) Improving the energy efficiency of processes : reduction of the crystallization fouling of heat exchangers 548. Ylä-Mella, Jenni (2015) Strengths and challenges in the Finnish waste electrical and electronic equipment recovery system : consumers’ perceptions and participation 549. Skön, Jukka-Pekka (2015) Intelligent information processing in building monitoring systems and applications 550. Irannezhad, Masoud (2015) Spatio-temporal climate variability and snow resource changes in Finland 551. Pekkinen, Leena (2015) Information processing view on collaborative risk management practices in project networks 552. Karjalainen, Mikko (2015) Studies on wheat straw pulp fractionation : Fractionation tendency of cells in pressure screening, hydrocyclone fractionation and flotation 553. Nelo, Mikko (2015) Inks based on inorganic nanomaterials for printed electronics applications 554. Kursu, Olli-Erkki (2015) Micromotion compensation and a neural recording and stimulation system for electrophysiological measurements Book orders: Granum: Virtual book store http://granum.uta.fi/granum/ C555etukansi.kesken.fm Page 1 Wednesday, November 11, 2015 9:42 AM C 555 OULU 2015 UNIVERSITY OF OUL U P.O. Box 8000 FI-90014 UNIVERSITY OF OULU FINLA ND U N I V E R S I TAT I S O U L U E N S I S ACTA A C TA C 555 ACTA UN NIIVVEERRSSIITTAT ATIISS O OU ULLU UEEN NSSIISS U Lauri Hallman University Lecturer Santeri Palviainen Postdoctoral research fellow Sanna Taskila Professor Olli Vuolteenaho Lauri Hallman Professor Esa Hohtola SINGLE PHOTON DETECTION BASED DEVICES AND TECHNIQUES FOR PULSED TIME-OF-FLIGHT APPLICATIONS University Lecturer Veli-Matti Ulvinen Director Sinikka Eskelinen Professor Jari Juga University Lecturer Anu Soikkeli Professor Olli Vuolteenaho Publications Editor Kirsti Nurkkala ISBN 978-952-62-1043-8 (Paperback) ISBN 978-952-62-1044-5 (PDF) ISSN 0355-3213 (Print) ISSN 1796-2226 (Online) UNIVERSITY OF OULU GRADUATE SCHOOL; UNIVERSITY OF OULU, FACULTY OF INFORMATION TECHNOLOGY AND ELECTRICAL ENGINEERING, DEPARTMENT OF ELECTRICAL ENGINEERING; INFOTECH OULU C TECHNICA TECHNICA

Top subcategories

Top subcategories

Top subcategories

Top subcategories

Top subcategories

Top subcategories

Top subcategories

Top subcategories

Download Single photon detection based devices and techniques for pulsed