Open Access
Add to BibSonomyAbstract
Light spectrometers are highly versatile state-of-the-art measurement devices. However, using these systems, e.g., in semiconductor device characterization, creates challenging obstacles with respect to measurement time. We present a new, flexible and accurate approach to either characterize optical properties of arbitrary photosensitive devices or examine the spectral components of light reliably. Using a spatial light modulator (SLM) in combination with frequency division multiplexing methods, it is possible to significantly improve signal-to-noise ratios and decrease measurement times. Moreover, the use of SLM ensures a greater reliability of the setup because conventional moving parts are replaced. The feasibility and experimental setup are described in detail. The setup has been validated for various applications by comparative measurements.
© 2015 Optical Society of America
1. Introduction
Currently, grating spectrometers are state-of-the-art instruments widely used for the analysis of irradiation spectra or characterization of a certain light receiver, e.g., photodiodes, solar cell devices or pyroelectric devices. In most applications, however, the measurement becomes challenging owing to low light intensities and, therefore, small signal amplitudes. This constraint calls for special methods that allow signals of interest to be reliably separated from background noise and other disturbing influences. A widely used approach is lock-in technology [1]. Amplifiers equipped with this technique can recover the amplitude of a measurement signal previously modulated at different frequencies. This allows the wanted signal to be isolated from background noise, even in case with dominating background noise levels. The operation of lock-in amplifiers relies on the orthogonality of sinusoidal functions. Considering the modulation frequency, basic trigonometry identities can be used to demodulate the measurement signal to recover amplitudes. The technical implementation is typically based on either analog control loops [1] or, more recently, digital signal processing technology. Optical modulation usually relies on opto-mechanical light modulators, e.g., chopper wheels. To improve speed and reliability, these can be replaced by nonmoving spatial light modulating (SLM) elements such as liquid crystal modulators or digital micromirror devices (DMD). The nature of most SLMs enables the possibility of modulating different signal components, e.g., different parts of a dispersed spectrum, by different frequencies and/or phases. In this way, spectral parts can be coded. Fourier transform methods or mixer-based demodulation algorithms then enable the superposition of various frequencies to be decoded and, thus, amplitudes and wavelengths to be assigned.
Approaches involving spectrometers making use of SLMs have been presented for different end applications [2–5]. E. P. Wagner et al. were among the first to report on the use of a DMD in visible spectroscopy. The DMD was used to select certain parts of a dispersed spectrum, which subsequently have been redirected onto a photomultiplier tube. It was reported that the setup outperformed state-of-the-art spectrometers based on charge-coupled devices in speed and sensitivity [2]. An enhanced approach was presented by J. D. Batchelor et al. using a DMD in analytical atomic spectroscopy. The frequency modulation of certain parts of the DMD allowed for using the DMD as a scanning slit while increasing the SNR significantly [3]. Utilizing the multiplexing advantage in a DMD-based spectrometer using up to four different modulation carrier frequencies simultaneously was later reported by T. M. Spudich et al. [5]. The implementation was based on a slightly modified nView projection system which was controlled by the video output of a computer. The DMD active mirror area was sectioned in either four quadrants or four vertical stripes and frequency modulated at 14.92, 20.00, 25.00 or 34.48 Hz. By different experiments with a non-dispersed light source and a photodiode detector, it was shown that a DMD can potentially be used in multiplex spectroscopy.
In contrast to conventional spectroscopy applications, SLMs are hardly used for the characterization of optical and electrical properties of solar cell devices or light sensors. In the majority of existing advanced multiplex approaches, a fast controllable source formed by light-emitting diodes (LED) [6–12] or semiconductor lasers [13,14] is used for generating quasi-monochromatic irradiance modulated at different frequencies.
The use of DMD technology opens a variety of possibilities to improve state-of-the-art measurement technology with respect to speed and noise behavior. Previous LED multiplexing approaches can be surpassed in bandwidth, resolution and irradiance. In this paper, we describe a new approach implementing the multiplex advantage in spectroscopy using a DMD-based spectrometer setup which can modulate a dispersed spectrum at up to 256 different frequencies. In contrast to previous broadband DMD spectrometer systems such as those presented in [4] and [15], our design does not require any customized refractive optics. Moreover, using standard reflective optical elements allows a larger useable spectral bandwidth ranging from UV to NIR. It is, therefore, particularly suitable for conventional medium-resolving (>1 nm) broadband spectroscopy in which an incoherent source of irradiance or a certain sample is to be examined. Possible fields of application are high-speed solar cell device characterization, low-noise/high-reliability outdoor spectroscopy and spectroscopy in harsh environments.
2. New spectrometer setup
The new measurement system is based on two diffraction gratings and an SLM. The basic working principle is noted in Fig. 1.
Fig. 1 Functional principle of the spatial light modulator (SLM)-based spectrometer setup. Light diffracted from grating G1 is imaged onto the SLM, which modulates the spectral components at different frequencies. Diffraction grating G2 recombines the quasi-monochromatic light to polychromatic white light imaged to a detector. E refers to irradiance, I refers to electric currents, λ refers to wavelengths and f to frequencies used for modulation.
The SLM modulates with diverse frequencies the light intensities of different spectral components diffracted by diffraction grating G1. This allows each quasi-monochromatic wavelength to be coded with a different frequency. The entire modulated light spectrum is subsequently combined into polychromatic white light using a second diffraction grating G2 and finally projected onto the surface of an appropriate detector, e.g., a photodiode. Conversely, instead of a detector, a certain device under test (DUT) such as a solar cell device can be illuminated to determine electrical and optical device properties. The electrical current generated by the detector or DUT consists of various spectral components. The identification of each component’s frequency and amplitude allows the recognition of the corresponding light wavelengths. This can be accomplished by the use of digital signal processing methods such as well-known Fourier transform operations. Moreover, the modulation frequency bandwidth can be halved using modulation schemes that use both signal frequency and signal phase, e.g., analog quadrature amplitude modulation.
2.1 Spectrometer optical setup
The core component of the experimental setup is a digital micromirror device (DMD, DLP® 0.7 XGA 2xLVDS Type A DLP7000) from Texas Instruments, which serves as an SLM. Compared to polarization-based light modulating devices such as liquid crystal on silicon (LCOS), DMDs can be used without crossed polarizers. Therefore, the available light intensity can be used to a greater extent. The most important factor is, however, the high switching frequency of single DMD mirrors, which enable pattern modulation frequencies on the order of 10 Hz to 30 kHz [16].
The used DMD consists of a rectangular array of 1024 × 768 (XGA) square mirrors attached to the surface of an integrated circuit chip. Each mirror is mounted on a torsion hinge that runs between the diagonally opposite corners. The alignment of the mirrors is adjusted through the influence of electrostatic fields. The maximum modulation frequency in the present case is limited to 23 kHz by the data processing hardware DLP V-7000 from Vialux GmbH [16].
Figure 2 illustrates the configuration of the optical system setup, which consists basically of two mirrored Czerny-Turner spectrometers in a typical 4f configuration [17].
Fig. 2 Simplified illustration of the optical layout. The light path is exemplarily represented by two wavelengths. For simplification, all optical components (Mx = mirrors, Gx = diffraction gratings, Sx = apertures/slits) are represented by transparent (lens) optics with identical focal lengths f.
To simplify the illustration, all components are represented by transparent (lens) optics. The optical layout is generally similar to so-called 4-f zero dispersion compressor setups used in femtosecond pulse-shaping applications [18,19]. A short rightward-pointing arrow represents the incoming light, which is initially limited by aperture stop (slit) S1 and then collimated through collimating mirror M1, diffracted by diffraction grating G1 and focused through focusing mirror M2 onto the SLM surface. Once the diffracted spectrum has passed the SLM, the now-modulated quasi-monochromatic light components are recombined into polychromatic white light in the second Czerny-Turner stage. By analogy with the first stage, the second stage is represented by collimating mirror M3, diffraction grating G2 and output focusing mirror M4. The exit aperture stop S2 reduces stray light and absorbs parasitic light caused by higher grating diffraction orders. The real implementation is based on reflective optical elements. In this case, entrance slit, collimating mirror, diffraction grating, focusing mirror and output focal plane are arranged in a W-layout [17]. Most standard requirements can be very well covered by the use of off-the-shelf spectrometers. However, using a DMD places special demands on the spectrometer layout and its components, particularly through the diagonal orientation of its tiltable micromirrors [16]. This necessitates a custom design. Remarkably good results concerning bandwidth, optical throughput and resolution can already be achieved by applying basic spectrometer calculations (e.g., presented in [20] or textbooks [21]) with appropriate initial values and the use of off-the-shelf optics. Using focal lengths in the range of 300 mm and consistent 2 inch active diameters (NA ≈0.08) for the mirrors turned out to be an acceptable tradeoff between aberrations, light throughput and, moreover, price and availability of optical components. In this manner, the grating inclusion angle Φ is defined by the triangle running between the centers of mirror M1, grating G1 and mirror M2 (in Fig. 2 the grating inclusion angle can be considered as a break at the position of G1 and G2). Much smaller focal lengths, however, would lead to larger angles and increase unwanted influences of astigmatism. Larger focal lengths would reduce coma and astigmatism and thus allow improved optical resolution and SNR at the expense of light throughput [22]. Another important constraint in the DMD spectrometer design is the definition of the spectral operating range. We decided to use the system with four pairs of plane-ruled diffraction gratings with different blaze angles. Each grating is dedicated to a fixed angular position dispersing approximately 450 nm. This leads to a total spectral bandwidth of approximately 1800 nm starting at approximately 280 nm subdivided into four sequential measurement steps. Using the fundamental grating equation given by
with the light incidence angle α, the diffraction angle β, the grating groove density G and the grating diffraction order n at a certain wavelength λ [22] allows us to calculate the angular dispersion for the central wavelength λc dispersed by a particular diffraction grating. This can be accomplished by taking the system symmetry into account, consideringTransposing Eq. (1) into a product form by applying the sum-to-product trigonometric identity and substituting incidence angle α by Eq. (2), the diffraction angle β can be expressed by:Differentiating the grating equation with respect to β and keeping α constant allows calculating the reciprocal angular dispersion. Introducing the effective focal length Lf of the focusing mirror enables calculating the linear dispersion on a focal plane [22]:With consideration of constant tilting angles of single mirrors in the XGA DMD of ± 12° the entire DMD has to be tilted. By introducing the inclination angle γ of the focal plane the effective focal length as well as the spectral compression of the output field can be corrected by the cosine of γ. This allows a more accurate approximation for the reciprocal linear dispersion:Influences on the spectral plane caused by the off-axis illumination of spherical mirrors are neglected in this calculation. Thus, it is possible to calculate the extent to which a particular spectral interval is spread out across the DMD surface, respective of the focal field of the spectrometer with good approximation. The best practicable approximation for groove density G is, therefore, given by diffraction gratings with 85 to 100 grooves per millimeter, assuming that the dispersed spectrum is projected onto the entirely available mirror space of an XGA DMD (lateral extension of 14 mm). Figure 3 shows the efficiency curves of the four diffraction gratings G1 + G2 to G7 + G8 implemented in our setup. The plotted vertical lines group the entire spectral range into four bandwidth regions at 712 nm, 1130 nm, 1544 nm and 2030 nm. To enable gapless measurements, all four regions have 10 nm overlap. The elongated efficiency maximum of G7 + G8 provides additional headroom in the infrared regime and thus allows a (theoretical) fifth bandwidth region. To switch over between wavelength-ranges a grating turret is used. As this mechanism is based on a single rotating axis, the effective focal length Lf varies slightly in the interval of 301.1 to 303.6 mm according to the change of the light incidence angle α in the interval of 1.6° to 5.0°. This results in small differences between spectral bandwidths (BW) diffracted by the first three grating-sets with equal groove density. The difference in the blaze wavelength (BZ) is caused by the fact that every grating was chosen to have its (approximately) highest diffraction efficiency within its assigned spectral bandwidth.
Fig. 3 Optical efficiency curves of the four used diffraction gratings within their assigned wavelength regions including or close to blaze wavelength (BZ). The useable bandwidth (BW) varies slightly owing to different light incidence angles.
The following experimental descriptions and measurements will focus on the first grating stage, which is highlighted in Fig. 3.
Figure 4 illustrates the final geometric arrangement of the SLM-based spectrometer. The illumination optical path is represented by solid lines. After modulation through the DMD, the optical path is represented by dashed lines. As seen, the diffraction gratings G1 and G2 as well as the DMD are tilted by an angle of 45°. This allows the diffracted light spectrum to be imaged on the entire DMD surface, which must be illuminated with diagonal incidence [16]. Moreover, aberrations are greatly reduced using the angular diffraction of the first diffraction order to achieve the appropriate light incidence angle on the DMD active surface.
Fig. 4 Sketch of the optical setup based on two Czerny-Turner spectrometers in 4f configuration. Mirrors are marked with Mx and diffraction gratings with Gx.
With respect to optical properties, the system approximates to the greatest possible extent a series connection of two symmetrical Czerny-Turner spectrometers [17]. Thus, aberrations caused by the off-axis illumination of the first mirror M1 (respective of M3) will be largely reversed by the second mirror M2 (respective of M4).
Figure 5 shows a photograph of the optical setup equipped with two identical ruled 100 grooves/mm diffraction gratings (1) and (3) with a blaze-wavelength of 450 nm on the right and the DMD (2) on the left. The entire available active mirror area of the DMD is illuminated with the dispersed light spectrum.
Fig. 5 Photograph of the optical setup used within this work: diffraction gratings (1), (3) and DMD (2) mounted with a 45° tilt angle.
2.2 Electrical measurement setup
To measure the detector’s current spectrum, the signal chain illustrated in Fig. 6 was used. A First Sensor PC1-6 1 mm2 silicon PIN-diode was applied as a detector and directly connected to a femtoampere bias Texas Instruments LMP7721 operational amplifier (OPA) in transimpedance configuration (I/V-converter). Owing to the low luminous flux of the preliminarily used halogen light source, the use of an ultra-low bias OPA configured with closed loop gain is ideally suited.
To avoid aliasing effects, for smoothing of the quasi-continuous signals modulated by the DMD and rejection of unwanted DMD crossover switching disturbances, the OPA’s output signal initially passes a tenth-order Butterworth lowpass filter (–3 dB passband at 2.9 kHz; –100 dB stopband at 10 kHz) before signal digitization. The filter output voltage signal is subsequently quantized using a 104 dB SNR 20-bit Linear Technology LTC2378 analog-to-digital successive approximation (SAR) converter (evaluation board DC2135A combined with DC890B data acquisition board), which is eventually connected to a personal computer.
To enable measurements with repeatable accuracy, a discrete trigger signal was used to always start quantization at the same DMD frequency-phase relation during all measurement iterations.
2.3 Signal processing
Both the DMD-controller and quantization hardware are controlled by different C + + based computer programs. The controlling and timing of the DMD are frame based [23]. Each frame contains a previously defined number of patterns representing the different frequency carriers. These patterns, in turn, consist of multiple columns of micromirrors whereas the number of on-state micromirror rows corresponds to a certain amplitude value calculated from the amplitude values of sinusoidal functions with appropriate frequencies. In the case of the XGA DMD used here, amplitudes are modulated in 768 substeps approximating a roughly 9 bit sinusoidal intensity modulation. Compared to coarse square wave modulation, the system is able to achieve lower total harmonic distortion (THD) values and, therefore, higher SNR.
From the perspective of data processing, computed amplitude values are temporarily stored in look-up tables and subsequently used to determine maps of mirror states. These maps are then transferred via USB to the DMD controller.
The core steps of data processing of previously quantized measurement data are a high-resolution fast Fourier transform followed by a line search-based pitch detection algorithm. This allows the exact frequencies and amplitudes to be found, which vary slightly owing to the nonideal transfer behavior of all involved electronic components. By means of pitch detection, the absolute amplitude accuracy was improved significantly. All signal processing functions have been realized by implementing diverse MATLAB/Octave functions.
3. Measurement results and discussion
3.1 Optical calibration and characterization
Relative signal throughput is directly related to the entrance slit width, which determines the optical resolution of a spectrometer system. Therefore, a compromise must be reached between light throughout and optical resolution. In the particular case of the DMD spectrometer, the DMD pattern resolution must be considered additionally. Here it is regarded as an additional intermediate exit slit. The system’s resolution is, therefore, dominated by the width of the entrance slit or the physical DMD pattern width, whichever is smaller.
With regard to the common bandpass (resolution) formula [22], the slit width wslit is either given by the entrance slit width or the DMD pattern width:
The resolution of the DMD spectrometer experimental setup was examined with 256 patterns and accordingly groups of 4 adjacent columns of mirrors. Theoretically, this results in approximately 1.7 nm DMD resolution. The entrance slit was opened stepwise until the width no longer dominated the system resolution. The DMD was then controlled as an adjustable bandpass. This means that four micromirror columns (according to Eq. (6)) were tilted in on-state while the remaining micromirrors have been left in off-state. After data acquisition the subsequent next four columns have been activated and so on. An exemplary section of a 430 nm full bandwidth measurement episode is shown in Fig. 7.
Fig. 7 Detail section of an episode measurement. Four mirror columns were switched on simultaneously. The maximum resolution was limited by the benchtop spectrometer used for examination.
The experiment was limited, however, by the resolution of the used benchtop spectrometer to approximately 4.5 nm FWHM. The peak-to-peak difference between two consecutive measurement steps (wavelength sub-ranges) was in any case below 1.5 nm. In the course of future development, it is foreseen to characterize the optical properties with a higher resolving spectrometer. For simplification, further measurements have been taken with reduced optical resolution and, therefore, a lower amount of frequency patterns. This, however, results in improved noise behavior supported through the increased signal intensity. Considering the integer constraint for the number of mirrors, 32 columns of mirrors per pattern have been selected. In this way, the total mirror plane was subdivided into 32 columns. Owing to the arbitrary spectral characteristic of the measured light from the halogen lamp weighted by the two gratings and other optical components, benchmarking FWHM exclusively is not sufficient. For this reason, the centroid of every DMD selected wavelength range was evaluated. The results are illustrated in Fig. 8.
Fig. 8 Measured centroids of sequentially activated 32 mirror width pattern represented by squares. Deviations from the ideal linear slope are represented by circles.
Ideally, all centroids should form a perfect linear slope. Deviations from this slope are plotted as single circle dots and scaled by the right side of the y-axis. The maximum deviation ranges between ± 0.6%. Relative to the mean FWHM of 13.8 nm, this results in ± 0.1 nm rounded. The standard deviation here is σ = 0.27%. Besides optical aberrations, alignment errors and the trenches between single micromirrors as well as the partial reflecting torsion hinge of off-state mirrors, the deviations can chiefly be attributed to distortion caused by the off-axis illumination of the collimating and focusing mirrors. As shown in Fig. 5 the DMDs mounting plate and the diffraction grating mechanics limit the smallest possible inclusion angle Φ in our setup. This problem will be addressed in future optical setups with rotating diffraction grating mounts and optimized DMD and slit optomechanic systems. Inaccuracies caused by the used benchtop spectrometer cannot be ruled out.
3.2 Frequency behavior
To determine the amplitude accuracy of the modulation, signal-to-noise and total harmonic distortion measurements have been carried out. For this purpose, a 4 cm2 silicon reference solar cell (primary calibration by the Physikalisch-Technische Bundesanstalt) [24] was selected as a detector to simulate realistic conditions of a feasible light receiver. However, in true spectrometer applications, the use of a dark current optimized photodiode such as the PC1-6 would be preferred. In solar cell device characterization, large cells are quite common. For comparability, each single measurement was sampled at 100 kS/s within a measurement interval of 100 signal-periods and carried out three times.
Figure 9 visualizes the current amplitude deviations at different modulating frequencies from the current amplitude of the reference solar cell under continuous illumination. Therefore, the entire active DMD area was modulated with a sinusoidal signal at various frequencies in the interval from 104 to 950 Hz. Lower frequencies are possible in principle, but the noise behavior will be increasingly influenced by 1/f noise from the electronic components. Considering the nonoverlapping condition for symmetric rectangular or sinusoidal signals as stated in Eq. (7), fmin = 104 Hz is an appropriate choice to avoid interference among harmonics. Frequency spacing is, therefore, equal to df = 6.5 Hz using 32 patterns.
Fig. 9 Current amplitude deviations of the reference solar cell under modulated illumination from its mean value plotted by squares. Included Standard deviations are always below 300 ppm.
The squares in Fig. 9 show that frequencies in the range from 104 Hz to 500 Hz vary within a tolerance band of ± 0.3% in amplitude compared to the mean value. This value improves to ± 0.2% if the sampling takes place within a specific period of time (here 1 second) independent from the signal period, as it appears in the final application. Higher frequencies, however, tend toward a noticeable attenuation of the measured amplitude. The reason for this lies in the capacitive behavior of the nonbiased detector solar cell, which dominates the response time at higher frequencies. However, it cannot be ruled out that the lowpass filter component tolerances have a certain influence on the attenuation in the neighborhood of 1 kHz. Vertical error bars indicating the standard deviations over the three measurements are included in Fig. 9; however, they are below 300 ppm and therefore are largely smaller than the symbols. The same holds true for the (electrical) SNR. For lower frequencies, SNRs on the order of 78 dB are realistic, whereas they slightly degrade for frequencies above 600 Hz. It should be noted, however, that the effective SNR value strongly depends on the assumed computational parameters. Here, all SNR and THD values have been computed from the single-sided power spectral density (PSD) of the Hanning window-weighted [25] and gain-corrected signal. The signal itself consists of values sampled at 100 kS/s. For the SNR determination, the power of the first ten harmonics was excluded from the noise power computation. As the DMD generated sinusoidal signals approximated by 768 digital steps, higher harmonics of the fundamental signal are not treated as noise. The PSD was, in addition, limited to 2500 Hz.
The THD plotted in Fig. 10 benefits largely from the pseudo-sinusoidal waveform approximated by the DMD. Compared to amplitude accuracy behavior, the opposite effect can be observed as the THD improves (lowers) its value at higher frequencies.
Fig. 10 Measurement results of the total harmonic distortion (THD) measurement for 100 signal periods of diverse frequencies.
This effect is even stronger if the sampling takes place within a specific time period, as the number of sampled signal cycles increases with higher frequencies. This enables a better reconstruction of the fundamental sine waves. In case of the signal period defined sampling time, the capacitive behavior of the detector and the non-ideal electronic parts have an additional frequency-dependent smoothing influence on the waveform. Additionally, harmonics of higher frequencies are stronger suppressed by the anti-aliasing filter (see paragraph 2.2). This virtually improves the result.
3.3 Frequency multiplex measurements
To demonstrate the performance of the presented DMD spectrometer approach, a frequency-doubled neodymium-doped yttrium aluminum garnet (Nd:YAG) diode-pumped solid-state laser was used as a narrowband light source. All examinations have been performed by measuring the visible laser emission at 532 nm. To emulate weak illumination conditions, the laser intensity was reduced to a level of picoampere photocurrents generated by the First Sensor PC1-6 photodiode. Figure 11 plots the results of four measurement cycles within 300 milliseconds with 32, 64, 128 and 256 patterns in ascending order. Patterns are formed by equally modulated groups of mirrors according to Table 1. Corresponding modulation frequencies are noted in parentheses. For all measurements, the entrance slit of the system was adjusted slightly narrower than its theoretical value derived from Eq. (6). Thus, the overall resolution was always dominated by the DMD. To achieve appropriate optical signal-to-noise values, for both 128 and 256 patterns, measurements of the light intensity were approximately doubled.
Fig. 11 Measurement series of Nd:YAG laser emission at 532 nm examined within 300 milliseconds with different DMD resolutions of 256, 128, 64 and 32 patterns.
Table 1. Summary of the Nd:YAG laser measurement results
The results in Fig. 11 show a slightly positive or negative shift from the center wavelength of 532 nm. This is due to the discrete wavelength steps defined by the respective pattern number; however, the used laser shows slight spectral changes (jitter). The effect of the discrete wavelength steps becomes particularly apparent in the case of the 32 pattern measurement in which the main peak is represented by the 535 nm bin. The neighboring bins are 521 nm below and 548 nm above. Owing to the spline interpolation that was applied to all four graphs, the interpolated peak wavelength is in the interval of 534 ± 2 nm. Table 1 summarizes all peak and FWHM values. The measured FWHM is compared to the calculated FWHM referring to Eq. (6) for every pattern resolution. The deviation ΔFWHM varies in the interval from 0.4 to 1.2 nm. However, the FWHM of 2.5 nm measured with 256 patterns is already in the range of the laser emission FWHM. For the remaining measurements with 32, 64 and 128 patterns, FWHM is also affected by the fixed grid of the discrete patterns.
The optical signal-to-noise ratio S/Nopt was computed according to Eq. (8), where S/Nopt treats higher diffraction orders as RMSnoise and S/N-hopt excludes higher diffraction orders from the calculations. The peak amplitude value is considered by Apeak:
Generally, the S/Nopt values serve as relative indications because the laser light source is inherently noisy. The degraded S/Nopt for the measurement with highest resolution of 256 patterns (Fig. 11, top) is most influenced by the reduced light intensity owing to the narrowed slit and, therefore, is the intensity distribution represented by the more distinctive sidelobes. Their shifted spectral position is misleading; the real cause is diffraction at the entrance slit, which was imaged to the DMD spectral plane. This generally unwanted effect becomes visible owing to the coherent characteristic of laser emission. Nevertheless, it impressively highlights the system’s performance regarding resolution, sensitivity and aberrations. The difference between the amplitudes of the right and left side lobes in Fig. 11 - top of approximately 15% suggests increased distortion. This was most likely caused by the non-corrected optical elements in conjunction with the broader diffraction pattern of the slit image. To achieve higher resolution measurements with coherent light sources further measures like compensating refractive optics (e.g. presented by T. C. Wood et al. [15]) must be considered.With regard to the measurement time and data processing effort, higher modulation frequencies can be used to achieve improvements. In the 832 to 2489.5 Hz frequency interval, measurements within 100 ms are already feasible without substantial loss of accuracy and resolution.
4. Conclusion
We present a new SLM-based frequency division multiplex spectrometer approach. The use of Texas Instruments digital micromirror technology allows us to improve medium-resolving light spectroscopy with respect to speed and signal quality. The signal quality improvement relies on the same basic principles as commonly used for lock-in technology. Moreover, the new DMD spectrometer is able to benefit extensively from the multiplex advantage coming from time parallel modulation of various light spectral components. This is possible without using any susceptible turning parts such as optical chopper wheels. It has been demonstrated that an easy setup made of off-the-shelf optical components can be used to resolve the emission spectrum of a low-intensity laser reliably within 100 to 300 ms measurements. By extrapolating the conducted examinations to the total available optical bandwidth of approximately 1800 nm, spectral measurements with SNRs on the order of 76 dB can be conducted within 4 s. By omitting any moving parts, this technology is particularly suitable for medium-resolving light spectroscopy in harsh environments.
Our future work aims toward improvements of the optical properties and using the setup for high-speed solar cell device characterization.
Acknowledgments
This work has been partly supported by the Federal Ministry for Economic Affairs and Energy (BMWi) under the HeKMod4 project, contract number 0325750. The authors are responsible for the content of this paper.
Thomas Missbach acknowledges the fellowship support from the Karlsruhe School of Optics & Photonics (KSOP).
References and links
1. J. H. Scofield, “Frequency-domain description of a lock-in amplifier,” AJP 62, 129–133 (1994).
2. E. P. Wagner II, B. W. Smith, S. Madden, J. D. Winefordner, and M. Mignardi, “Construction and evaluation of a visible spectrometer using digital micromirror spatial light modulation,” Appl. Spectrosc. 49(11), 1715–1719 (1995). [CrossRef]
3. J. D. Batchelor and B. T. Jones, “Development of a digital micromirror spectrometer for analytical atomic spectrometry,” Anal. Chem. 70(23), 4907–4914 (1998). [CrossRef] [PubMed]
4. A. Y. Fong, “Application of digital micromirror devices for spectral-response characterization of solar cells and photovoltaics,” Proc. SPIE 7596, 75960I (2010).
5. T. M. Spudich, C. K. Utz, J. M. Kuntz, R. A. Deverse, R. M. Hammaker, and D. L. McCurdy, “Potential for using a digital micromirror device as a signal multiplexer in visible spectroscopy,” Appl. Spectrosc. 57(7), 733–736 (2003). [CrossRef] [PubMed]
6. S. Tulloch, “The Design and Use of LED Reference Light Sources for Q.E. Calibration,” Instrument Science Group Royal Greenwich Observatory, RGO TECHNICAL NOTE 107 (1996).
7. D. L. Young, B. Egaas, S. Pinegar, and P. Stradins, “A new real-time quantum efficiency measurement systeM,” in Proceedings of IEEE Photovoltaic Specialists Conference (IEEE, 2008), 1–3. [CrossRef]
8. J. Schmidt, L. A. Vasilyev, J. E. Hudson, G. S. Horner, E. A. Good, and M. Dybiec, “Characterization and gauge study of a high speed quantum efficiency apparatus,” in Proceedings of IEEE Photovoltaic Specialists Conference (IEEE, 2010), 1710–1714. [CrossRef]
9. G. S. Horner, L. A. Vasilyev, J. Schmidt, J. E. Hudson, M. Arbore, M. Dybiec, and E. A. Good, “Flash quantum efficiency measurements for multi-junction solar cells,” Proceedings of IEEE Photovoltaic Specialists Conference (IEEE, 2011), 1716–1720. [CrossRef]
10. W. Reetz, D. Erdweg, W. Hilgers, P. Kaienburg, A. Gerber, B. E. Pieters, and U. Rau, “A novel high speed spectral response measurement system based on LED light sources,” Proc. 26th EU PVSEC, 113–116 (2011).
11. J. A. Rodriguez, M. Vetter, M. Fortes, and C. Alberte, “Development of a very fast spectral response measurement system for silicon thin film modules,” 2013 Spanish Conference on Electron Devices (CDE), 369 – 372 (2013). [CrossRef]
12. B. H. Hamadani, J. Roller, B. Dougherty, and H. W. Yoon, “Versatile light-emitting-diode-based spectral response measurement system for photovoltaic device characterization,” Appl. Opt. 51(19), 4469–4476 (2012). [CrossRef] [PubMed]
13. W. Warta, J. Sutter, B. F. Wagner, and R. Schindler, “Impact of diffusion length distributions on the performance of mc-silicon solar cells,“ 2nd WCPEC, Vienna, 1650-1653 (1998).
14. Th. Pernau, P. Fath, and E. Bucher, “Phase-sensitive LBIC analysis,” in Proceedings of IEEE Photovoltaic Specialists Conference (IEEE, 2002), 442–445.
15. T. C. Wood and D. S. Elson, “A tunable supercontinuum laser using a digital micromirror device,” Meas. Sci. Technol. 23(10), 105204 (2012). [CrossRef]
16. Texas Instruments Incorporated, “DLP 0.7 XGA 2xLVDS Type A DMD (Rev. B),” Datasheet DLPS026B, Texas (2013).
17. M. Czerny and A. F. Turner, “Über den Astigmatismus bei Spiegelspektrometern,” Z. Phys. 61(11-12), 792–797 (1930). [CrossRef]
18. M. Hacker, G. Stobrawa, R. Sauerbrey, T. Buckup, M. Motzkus, M. Wildenhain, and A. Gehner, “Micromirror SLM for femtosecond pulse shaping in the ultraviolet,” Appl. Phys. B 76(6), 711–714 (2003). [CrossRef]
19. A. M. Weiner, “Femtosecond pulse shaping using spatial light modulators,” Rev. Sci. Instrum. 71(5), 1929–1960 (2000). [CrossRef]
20. J. M. Lerner, “Imaging spectrometer fundamentals for researchers in the biosciences–a tutorial,” Cytometry Part A, Spectral Imaging 69A(8), 712–734 (2006). [CrossRef]
21. J. F. James, Spectrograph Design Fundamentals (Cambridge University Press, 2007).
22. Ch. Palmer and E. Loewen, Diffraction Grating Handbook, 6th edition (Newport Corporation, 2005).
23. R. Hoefling and E. Ahl, “ALP: universal DMD controller for metrology and testing,” Liquid Crystal Materials, Devices, and Applications X and Projection Displays X, Proc. SPIE 322, 5289 (2004).
24. J. Metzdorf, “Calibration of solar cells. 1: The differential spectral responsivity method,” Appl. Opt. 26(9), 1701–1708 (1987). [CrossRef] [PubMed]
25. F. J. Harris, “On the use of windows for harmonic analysis with the discrete Fourier transform,” Proceedings of the IEEE (IEEE, 1978), 51–83. [CrossRef]
