LED-BASED DEVICES AS SENSORS FOR OBJECT DETECTION, COUNTING, AND CLASSIFICATION
LED-BASED DEVICES AS SENSORS FOR OBJECT DETECTION, COUNTING, AND CLASSIFICATION
The objective of this project is to investigate the employment of standard solid-state signaling devices based on light emitting diodes (LEDs) for imperceptible detection, counting, and identification of various objects.
Standard LEDs can be simultaneously used as both light emitting devices and photodetectors. Moreover, the spectral ranges of the emitted light and photosensitivity have a significant overlap that allows for employment of a pair of these devices for both signaling and imperceptible detection, counting, and identification of opaque objects interfering with the shared optical path established within the field of view of both LEDs constituting an opto-pair. The relatively broad range of the LED’s spectral sensitivity in the visible part of the spectrum allows also for imperceptible detection and classification of the objects interfering with the ambient day or night light incident on the LED.
Imperceptibility of the system is based on the unnoticeable to the human eyes pulsing of the LEDs in a kHz range used for simultaneous real-time optical emission/detection.
Various applicable LED-based opto-pair configurations can be first simulated and then an experimental laboratory based electronic control, acquisition and analysis circuitry for the employment of standard signaling LED based devices, such as for example traffic lights, for imperceptible object detection, counting, and identification can be designed, fabricated, and tested. Optimization of the electronic circuitry, adding portability and wireless capability, should be also considered.
Light-emitting diodes (LEDs) are widely used as indicator lights and numeric displays on consumer electronic devices. New LED materials and improved production processes have produced bright LEDs in colors throughout the visible spectrum, including white light. With efficacies greater than incandescent (and approaching that of fluorescent lamps) along with their durability, small size, and light weight, LEDs are finding their way into many new applications within the lighting community. The most recent advancement in light emitting diode (LED) technology is the new aluminum indium gallium phosphide (AlInGaP) Precision Optical Performance lamps designed specifically for use in traffic management applications1. These Precision Optical Performance LED lamps offer light output performance superior to that of other technologies with predictable, superior, and stable long term performance.
Light-emitting diodes can also be used in a wide range of applications as inexpensive, readily available optical detectors2. Typically, an LED detects light at the same or somewhat shorter wavelength than the light it emits. In fact, a standard LED can perform double duty in the same circuit without changing its physical or electrical connections.
Employment of LEDs as photodetectors has been first disclosed in 1986 in a US Patent1 describing a proximity sensor using an LED as a simultaneously working light emitter and detector. More than a decade ago, a publication described a Zn-doped InGaN-based blue LED having an emission peak at 450 nm and an extinction peak at 380 nm and a GaAlAs red LED
1 Hewlett-Packard. Application Brief I-004. 1997
2 R. Stojanovic and D.Karadaglic, “Single LED Takes On Both Light Emitting and Light
Detecting Duties”, Electronic Design, Vol. 55, No. 16, 2007
3 US Patent Number 4,564,756, Jan. 14, 1986
having an emission peak at 660 nm and an extinction peak at 620 nm4. The LEDs had a nanosecond response when a reverse bias was applied to the junctions as for p-i-n photodiodes. Most LEDs are based on p-n-junctions, which allows for their employment as wavelength-selective photodiodes without application of a reverse bias. The spectral range of such photosensitive LEDs is determined by the energy gaps of the p-n juction layers and waveguide materials used for the LED fabrication.
LEDs are widely used in commercial analytical systems and dedicated detectors. LEDs are the most energy-efficient device for producing monochromatic light, and provide a concentrated small cool emitter ideal for miniature analytical devices5.
The dual LED use has been already considered during the device design and fabrication stages. An InGaN/GaN multiple quantum well structure commonly used for light emitting diodes has been employed for dual functions of optoelectronic devices exhibiting photodetector properties in reverse bias, while at the same time preserving the distinct identities of LED in forward bias6. The turn on voltage in forward bias and the breakdown voltage in reverse bias are about 3.2 V and 30 V, respectively. The higher photo- and dark-current densities were detected for larger size of devices. The contrast ratio calculated between photo- and dark-current densities of large-size device decreases more rapidly as compared to that of a small-size device. Thus, one can easily integrate photodetectors with LEDs using the same epi-structure to realize a GaN-based optoelectronic integrated circuit (OEIC).
Projects on the development of instrumentation for bio-medical research, conducted at Integrated Micro Sensors Inc (IMS), resulted in the design, fabrication, and testing of several multi-wavelength multifunctional absorption/scattering/fluorescence sensors based on LEDs7,8,9,10,11
4.Eiichi Miyazaki, Shin Itami, Tsutomu Araki. Using a light emitting diode as a high-speed, wavelength selective photodetector. Review of Scientiffic Instruments, 69(11), pp. 3751-3754 (1998).
5.Purnendu K. Dasgupta, In-Yong Eom , Kavin J. Morris and Jianzhong Li. Light emitting diode-based detectors
Absorbance, fluorescence and spectroelectrochemical measurements in a planar flow-through cell. Analytica Chimica Acta, Volume 500, Issues 1-2 , 19 December 2003, Pages 337-364.
6. Y.D. Jhou, C.H. Chen , R.W. Chuang , S.J. Chang , Y.K. Su , P.C. Chang , P.C. Chen , H. Hung , S.M. Wang , C.L. Yu. Nitride-based light emitting diode and photodetector dual function devices with InGaN/GaN multiple quantum well structures. Solid-State Electronics 49 (2005) 1347–1351.
7. D. Starikov, F. Benkabou, Nasr, Medelci, and A. Bensaoula. “Integrated Multi-Wavelength Fluorescence Sensors”. ISA/IEEE Sensors for Industry (SIcon/02) Conference proceedings,15-18, (2002). http://ieeexplore.ieee.org/document/1159798/?reload=true
8. Boukadoum, M., Tabari, K., Bensaoula, A. & Starikov, D. & Aboulhamid, E.M. (2005) FPGA implementation of a CDMA source coding and modulation subsystem for a multiband fluorometer with pattern recognition capabilities, proc. International Symposium on Circuits and Systems (ISCAS 2005), Kobe (Japan), May 2005, pp. 4767-4770.
9. Starikov, D., Clement, J., Bensaoula, A. & Boukadoum, M. (2005) Chip-based integrated filterless multi-wavelength optoelectronic biochemical sensors, proc. IEEE Sensors for Industry Conference (SICON 2005), Houston, February 2005, Conference Proceedings pp. 129-132.
10. Clement Joseph, Mounir Boukadoum, Joe Charlson, David Starikov and Abdelhak Bensaoula. High-speed front end for LED-Photodiode basedfluorescence lifetime measurement system. Proceedings of IEEE Symposium Circuits and Systems Society (ISCAS 2007), New Orleans, 27 -30 May, 2007. http://ieeexplore.ieee.org/document/4253454/
11. Clement Joseph, Hanae Naoum, Mounir Boukadoum, David Starikov, Earl J. Charlson and Abdelhak Bensaoula, “Integrated Solid-State Optoelectronic Sensor System for Biochemical Detection and Quantification”, Solid State Electronics Journal, 2008 http://ieeexplore.ieee.org/document/4422246/
The sensors were assembled from commercially available LED chips with maximum wavelengths covering a spectral range from near UV to almost near IR. The LEDs were arranged in a circular design (Figure 1) each serving as a light source when a forward bias was applied, and as a wavelength-selective photodetector at no bias. The sensor was trained by a pattern recognition based artificial neural network experiment with data measured manually from 7 different compounds. The testing results indicate that very low analyte concentrations (down to few ppb) and high classification rates up to 98% (Figure 2) can be achieved by using such sensors.
The previous research and our preliminary results indicate that employment of commercial LEDs as photodetectors is feasible. Moreover, a spectral match in the LED emission and detection as well as simultaneous dual function operation can be achieved. The new challenge is to take the advantage of these unique features in order to develop a system for imperceptible object detection, counting, and identification.
Employment of LEDs as Photodetectors.
An experimental model and details on application of a standard 5-mm red LED as a very low cost efficient photodiode has been recently described by Dejan Karadaglic and Radovan Stojanovic in the Electronic Design magazine (published by Texas Instruments)2.
Because LED photodiodes are considerably less sensitive than commercial photodiodes (with a photocurrent about 10 to 100 times smaller), direct measurement of the photocurrent is difficult without amplification. Typically, it requires a picoammeter and expensive operational amplifiers (Figure 3). However, most modern microcontrollers have bidirectional I/O ports with configurable internal pull-ups or tri-state (high-impedance) inputs.
Using a high-impedance input, the circuit can make a very accurate and precise measurement of the photocurrent by employing a simple threshold technique and the microcontroller’s built-in timer-counter. In detector mode, the LED “charges” to +5 V very quickly (100 to 200 ns). This charge is sustained by the diode’s inherent capacitance, typically 10 to 15 pF (Figure 4, Step 1). Then P1 on the microcontroller is switched to the high-impedance mode (approximately 1015 Ω resistance), Step 2.
Under reverse-bias conditions, a simple model for the LED is a capacitor in parallel with a current source, iR(φ), which represents the current induced by light intensity . The model includes leakage current iL through P1, which, at about 0.002 pA typically, is insignificant when compared to a typical photocurrent iR(φ) of 50 pA through the diode in normal ambient lighting. Figure 5a2 shows the experimental results of the LED discharging, VP1(t), for φ1 and φ2, where 2 > φ 1.
A software routine (written for 16-bit timer-counter TCNT1 on the 8-bit microcontroller) continually polls VP1(t) through its digital equivalent, the logic state of P1, until the logic 0 threshold VTR (approximately 2.2 V) is reached. The decay time Td, in microseconds, is proportional to the amount of light detected and, therefore, measures the diode photocurrent, iR(φ).
As the amount of light received increases, the diode discharges more rapidly and Td decreases, and vice versa (Fig. 5a, again). If the decay time is more than a user-specified light-intensity threshold, represented by Tdcr (critical), the microcontroller can switch the LED on and it emits light as an alarm (Fig. 4, again, Step 3). In addition, other pins on the microcontroller can be used as relay outputs or light-controlled, pulse width-modulation outputs. Figure 5b shows the voltage output at P2 during the operating steps.
This very low-cost approach provides an inherently digital measurement of light intensity without amplification. Its signal-to-noise characteristics are excellent, due to the signal integration over the measurement. The technique improves the sensitivity of the photodiode, making it more attractive than a conventional (and more expensive) photodiode. A conventional photodiode discharges the capacitance more quickly, making time based measurement more difficult and expensive.
The C code for this application was written for Atmel’s AT90S2313 AVR microcontroller, using the CodeVision AVR compiler. However, other microcontrollers, including those from Microchip Technology (PIC) and Texas Instruments (MPS430), are also suitable.
2. Simultaneous LED/photodiode operation.
As was mentioned above, the simultaneous LED emission/detection mode was first proposed a long time ago in a proximity detector3. Since LED emits light only when it is forward biased, simultaneous operation of a single LED as both a light emitter and a light detector requires that light detection by the LED is done while the forward bias is applied to the LED.
A conventional preamplifier designed to sense high impedance sources cannot be used directly to detect the impedance variation caused by the forward biased LED response to returning light. A forward biased LED has such a low impedance that a small variation in the voltage across the LED or a small variation in the current flowing through the LED cannot be observed in the presence of noise which is also simultaneously detected unless the preamplifier either has been designed specifically to sense low impedance sources or an impedance matching device such as a transformer.
This can be done in theory either by driving a constant current through the LED and sensing variations in the voltage across the LED or by holding a constant voltage across the LED and sensing variations in the current flowing through the LED.
Figure 6 schematically shows a circuit arrangement in which a voltage signal across a forward biased LED caused by the LED response to returning light is detected by a common emitter amplifier. LED (10) is a TIL40 series diffused GaAs LED, which is forward biased by supply voltage V0 and resistor (12), such that a constant current of about l0 mA flows through the LED, causing the LED to emit about 450 microwatts of infrared light at about 940 nm wavelength.
The emitted light (14) illuminates a region adjacent to LED. If an object is in the illuminated region, reflected light returns to the LED and the presence of an object adjacent to LED can be detected from the impedance change of LED caused by the returned light.
Another approach to using two similar LED-based signals for light emission and detection can be realized by pulsing of both devices at the same frequency, with a 50% duty cycle, and a half period phase shift. The photocurrent generated on either of the devices can be acquired in between the emitting pulses by using a simple frequency lock-in amplifier. Such mode is advantageous as it is insensitive to the continuous and off-frequency background light. In addition, pulsing of the LEDs in a kHz range is undetectable for the human eye.
3. Spectral match of the LED optical emission-detection.
In previous projects on the LED based multi-wavelength optoelectronic sensors we have investigated the spectral properties of some standard LEDs used as both light emitters (Figure 7a) and photodetectors (Figure 7b). The spectral responses of 8 LEDs emitting in the range from 350 to 700 nm are covering the entire visible range. Comparison of the emission and photoresponse spectra for each LED indicates that the shorter the LED wavelength the less matching between the emission and photoresponse spectra can be observed. Among those with the best spectrally matched emission and photoresponse are the yellow and the red LEDs (Figures 8a and 8b, respectively). The spectral responses of these LEDs are substantially broad,
so they can be used for detection of not only the same LED light, but also for cross LED light detection, e.g. the yellow light can be sensed by each of the red LEDs and red light can be sensed by the yellow LED. In addition, both yellow and red LEDs can be used for sensing the visible ambient light, such as daylight and night composed of various natural (solar, moon, star) and artificial (lighting, signals) sources12 (Figure 9).
4. Employment of a traffic signal light as a photodetector array.
LED based traffic lights are assembled from several colored (red, yellow, or green) LEDs housed in a single 8 or 12” diameter casing. For example, depending on the design and requirements, single traffic lights produced by ZGSM Technology Co., Ltd (China) incorporate from 64 to 208 individually packaged 5-mm LEDs with round lenses providing view angles of up to 60° (Figure 10).The round lens would also serve as a good light collector should the LED is used in the photodetector mode. If individual addressing is applied to each LED, the traffic light would
12. Lighting Research Center. NLPIP, Volume , Issue 5, 2005. http://www.lrc.rpi.edu
make a very good two-dimensional photodetector array that can be used for position sensing13 or simple imaging14 applications. For example pattern recognition software based on artificial neural networks, similar to those used by IMS previously for multifunctional sensors, will be employed for object identification, or event classification purposes. We will also test the array, imaging accuracy, time dependence, sensitivity, and depth of field.
13. Anssi Mgkynen, Timo Rahkonen, Juha Kostamovaara. A binary photodetector array for position sensing. Sensors and Actuators, A 65 ( 1998) 45-53.
14. V. P. Fedosov. Analysis of the Photodetector Array Spectrum Using the Influence Function. Russian Microelectronics, Vol. 30, No. 4, 2001, pp. 269–273. Translated from Mikroelektronika, Vol. 30, No. 4, 2001, pp. 314-319.