Robust, spatially scanning, open-path TDLAS hygrometer using retro-reﬂective foils for fast tomographic 2-D water vapour concentration ﬁeld measurements

We have developed a fast, spatially direct scanning tunable diode laser absorption spectrometer (dTDLAS) that combines four polygon-mirror based scanning units with low-cost retro-reﬂective foils. With this instrument, tomographic measurements of absolute 2-D water vapour concentration proﬁles are possible without any calibration using 5 a reference gas. A spatial area of 0.8m × 0.8m was covered, which allows for application in soil physics, where greenhouse gas emission from certain soil structures shall be moni-tored. The whole concentration ﬁeld was measured with up to 2.5 Hz. In this paper, we present the setup and spectroscopic performance of the instrument regarding the 10 inﬂuence of the polygon rotation speed and mode on the absorption signal. Homogeneous H 2 O distributions were measured and compared to a single channel, bi-static reference TDLAS spectrometer for validation of the instrument. Good accuracy and precision with errors of less than 6 % of the absolute concentration and length and bandwidth normalized detection limits of up to 1.1 ppmv · m · p Hz − 1 were achieved. 15 The spectrometer is a robust and easy to set up instrument for tomographic reconstructions of 2-D-concentration ﬁelds that can be considered a good basis for future ﬁeld measurements in environmental research.


Introduction
In many applications in science and engineering, the spatial distribution of gas con-20 centrations is of high interest. A highly topical example comes from environmental science: as the permafrost in arctic regions is gradually thawing, gases that affect the climate effect, such as methane, carbon dioxide or water vapour, are emitted from the soil (Anisimov and Nelson, 1997;Nakano et al., 2000;O'Connor and Boucher, 2010;Schuur, 2011;Vonk et al., 2012;Zimov et al., 2006). This process is suspected to to certain soil structures. For instance, it is presumed that greenhouse gases emanate from subterranean reservoirs through cracks in the ground (Lopez et al., 2008;Schuur, 2011;Walter Anthony et al., 2012;Zimov et al., 2006). Taking e.g. a picture of the ground and comparing it to the spatial distribution of gas concentration in that area is a possibility for identifying most relevant structures in soil emission. 10 Any instrumentation that shall be used for such field applications needs to fulfil certain requirements: first of all, the equipment has to be kept as simple as possible to reduce transport issues. At the same time, spatial resolution needs to be suitable. Alignment effort should be minimized and robustness against harsh environmental conditions needs to be ensured. In the special case of permafrost monitoring, mea-15 surement areas of at least 0.5 m edge length to the range of several meters have to be covered. Moreover, fast 2-D-concentration field recording rates of at least 1 Hz are necessary for 2-D flux measurements, which could be obtained in combination with velocity measurement techniques.
Existing soil emission instruments have certain disadvantages if used for this type of 20 measurement. There are instruments for gas concentration measurements (Babilotte et al., 2010;Crosson, 2008;McManus et al., 2008) or even flux measurements (Mc-Dermitt et al., 2010;Hendriks et al., 2008;Tuzson et al., 2010) that are well-suited for field applications but that do not provide a 2-D-concentration distribution. Certain variants of direct TDLAS (tunable diode laser absorption spectroscopy ) are power- 25 ful techniques for field applications, as no reference gas measurement is needed for calibration (Buchholz et al., 2014) and it is based on robust, industrially available components. For receiving spatially resolved concentration fields from line-of-sight data, multiple measurement paths have to be combined to a tomographic setup. Introduction The most straight-forward way to do this is to arrange several laser sources around the field of measurement, while detectors are positioned at the opposite side. These setups allow for very fast measurements, since all channels can be recorded simultaneously and the field measurement rate equals the rate with that the concentrations along each projection are measured. However, such highly parallel, bi-static setups 5 either require large amounts of light-source-detector-pairs and the appertaining data acquisition channels (Wright et al., 2005;Terzija et al., 2008;Terzija and McCann, 2011;Ma et al., 2013 or are somewhat limited regarding spatial resolution (Ma et al., 2009;Deguchi et al., 2012). Additionally, a disadvantage of bi-static setups in field applications is the cumbersome alignment of many detectors. Another well-established method for tomographic absorption measurements is the step-wise movement of the measurement object, such that several line-integrals can be measured with a single bi-static channel (Liu et al., 2013;Song et al., 2013). This technique is of course not applicable for immobile measurement objects such as soil sections.
The most promising solutions are realized by sequential spatial scanning of the laser 15 beam. Kasyutich and Martin (2010) have developed a tunable laser absorption spectrometer with minimized amount of equipment and at the same time relatively high spatial resolution. Although there is only one measurement channel, multiple beams can be evaluated consecutively. This is facilitated by two mirrors for fan-beam scanning that are moved by carriages along a circular rail. However, the sequential scanning 20 measurement of the whole field takes about 160 s, so the instrument is only suited for stationary conditions. Wang et al. (2010), by contrast, present an instrument that continuously rotates laser beams at four different positions across the field. Thus it is very fast (100 ms field scan −1 ) and nevertheless has reduced complexity: due to retroreflecting mirrors on the opposite field side, only one detector per scanner is neces- their measurements in a way that the whole field is more or less covered with laser beams because they did obviously not expect any absorption outside of this target region. In contrast to this, in the environmental application the absorbing medium may be everywhere and small fan-beam angles therefore lead to small measurement field coverages.

5
In this paper, we present a tomographic absorption spectrometer that consists of four fan-beam scanners. Instrumentation effort is minimized with a setup that contains combined emitting and detecting units and very inexpensive retro-reflective foils as reflecting targets. The basic principle of a single-channel foil TDLAS instrument has been described in a previous publication (Seidel et al., 2012). This static instrument has now been expanded to a spatially scanning 2-D-spectrometer. Due to the retroreflecting foils, the instrument is very easy to align. Furthermore, many absorption paths of wide-angled fan beams are realized, attaining a good spatial coverage of the measurement field.
In the following, we provide a detailed description of the 2-D-TDLAS fan-beam in-15 strument and its performance compared to a static path-averaged measurement. In Sect. 2, we briefly explain the measurement method TDLAS that is used for the singlechannel concentration measurements. Direct TDLAS is based on Lambert-Beer's law, of which we use an extended version to deal with non-specific absorption and broadband light losses (Ebert and Wolfrum, 2011;Schulz et al., 2007): Here I 0 (λ) is the incident light emitted by a diode laser that is periodically wave-5 length tuned across a molecular absorption line. After passing the absorbing medium, where a wavelength-depending attenuation takes place, the intensity I(n, λ) is detected. The exponential relationship between detected and incident light depends on the temperature-dependent line-strength S(T ), a pressure-and temperature-dependent absorption line shape described by Φ(λ − λ 0 ), the absorber number density n and the 10 absorption path length l . In addition to that, background radiation E (t) and broadband transmission losses T r(t) are considered. With known absorption path length, dynamic wavelength tuning characteristics of the laser ∂λ ∂t , gas pressure p and gas temperature T , n can be derived from: The absorber volume concentration (expressed in [ppmv]) is then calculated via the ideal gas law, as temperature and pressure are known. This version of direct TDLAS has been presented for numerous applications, e.g. biological (Hunsmann et al., 2008;Wunderle et al., 2009), engineering (Ortwein et al., 2010Wagner et al., 2009;Teichert et al., 2003), or atmospheric research (Buchholz et al., 2012).

TDLAS tomography hardware
The presented tomographic spectrometer consists of four scanning units that are located at the edges of a 0.8 m × 0.8 m measurement field (view Fig. 1 field edges is covered with retro-reflective foil (Seidel et al., 2012), so that many reflected beams can be detected. The scanning units comprise the emitting as well as the detecting optics. All optical components are included in a closed box that can be purged with e.g. dry air or nitrogen and that protects the optics from the environment. Each of the units is connected to a 1-to-4 fibre-beam-splitter that receives light from 5 a 1370 nm DFB-laser. Inside the box, the light is emitted by a fibre-coupled grin lens with about 1 mm diameter that is placed in a hole drilled into the center of an off-axis parabolic mirror. Afterwards, the light is guided onto a rotating polygon mirror with five facets. It passes a wedged window and enters the measurement field, where it is retroreflected at the opposite side by the foil. It re-enters the box through the window, is reflected at the rotating polygon mirror and focussed onto a 1 mm diameter, uncooled InGaAs detector by the off-axis parabolic mirror. As the polygon is rotated continuously by a stepper motor, the beam is scanned horizontally across the measurement area.
Overall, the laser beams cover more than 75 % of the measurement field area. We compared the measurement quality of the 2-D scanning instrument to a sta-15 tionary, non-scanning reference TDLAS setup (view Fig. 2). For this purpose we used a bi-static setup: light from a fibre-coupled DFB laser at 1370 nm is collimated by a collimator. After passing an absorption path of 700 mm length, it is detected by an InGaAs detector. All other processing steps are identical to the previous tomografic system. Below, we shortly describe details regarding the retro-reflecting foils (Sect. 3.1), the 20 scanning system (Sect. 3.2) and the data acquisition (Sect. 3.3).

Retroreflecting foils
Instead of a costly system with many light sources, detectors and mirrors, we use a setup with retro-reflecting foil. This foil enables us to detect hundreds of beams with only four data acquisition channels by scanning the beam across the area. Usually, 25 such foils are used for safety and traffic applications, which makes them industrially available and low-cost. They reflect light like a retro-reflector even for long distances and large impact angles, such that incident beams at the edges of the fan-beam cone 12833 Introduction can also be used. An elaborate presentation of the foil reflection characteristics and static water vapour measurements is discussed in a previous paper (Seidel et al., 2012).

Polygon scanning unit
The laser beams are spatially scanned by rotating polygon mirrors. Four stepper motors 5 are rotated continuously and their positions are recorded by TTL signals of the motor drivers, such that the angles of the laser beams are known. At the same time, the laser is tuned with 5039.8 Hz repetition rate across the H 2 O absorption line. Depending on the polygon revolution speed, the beam moves by a certain fraction of a degree on the retro-reflecting foil during one concentration measurement. Different rotation 10 speeds have been realized and checked for suitability (view Sect. 4). Every scanner has a polygonal mirror with five facets, which results in a total scanning angle of 144 • . The laser beam at 18.5 • is used as a reference for the absolute position once in the beginning of the measurement. As long as the motors are supplied with voltage, the position does not have to be re-calibrated since the relative position is recorded during 15 the motor movement.

Data acquisition and laser light source
A fibre-coupled DFB-laser at 1370 nm with a current tuning range of about 2 cm with 100 S s −1 and 18 bit. All signals were triggered once by the master trigger coming from the signal generator and then measured continuously. For evaluation, the start of the TLDAS tuning ramp has to be known. As we measured continuously for speed reasons, we needed a time reference on the TDLAS signal itself since the time base precision of the PXI rack proved to be insufficient. Therefore, we added a pulse from 5 a pulse generator onto the TDLAS signal and split the data during post-processing by an edge-detection algorithm. For evaluation of the TDLAS measurements, we used an in-house LabVIEW software.

TDLAS measurements
By measuring the uniform H 2 O vapour concentration distribution of the surrounding 10 atmosphere and comparing it to a static reference channel, the performance of the instrument regarding accuracy and precision was examined. We tested two different motor velocities (30 rpm and 15 rpm) as well as a step-wise polygon movement (only Channel 0) and studied the influence of rotation on the TDLAS measurement. For the continuously moved motors, velocities correspond to measurement rates of the whole 15 field of 2.5 Hz and 1.25 Hz, respectively. The duration of the step-wise movement of the whole field was on the range of a few minutes. Overall, we evaluated more than 360 different beams during continuous scanning. In the step-wise case, the motor was moved in steps of 1.  Figure 3 displays a representative example of the measured and fitted absorption profile that was achieved while scanning continuously at 1.25 Hz field scan rate. No averaging was applied here, so the time resolution was approximately 0.2 ms. In this case, the overall standard deviation of the fit residual (1σ) that was attained at 12 038 ppmv H 2 O is 1.57 × 10 −2 ODe. The optical density is defined as The noise on the signal is dominated by electrical disturbances caused by the electric stray field of the stepper motors. A signal to noise of 108 was calculated by the ratio between the peak absorption and the standard deviation of the residual. With an absorption length of 1.98 m, the normalized detection limit is 3.1 ppmv · m · Hz −1 .

10
Depending on the position of the laser beam and the rotation speed of the motor, we achieved signal to noise ratios of up to 224 and hence detection limits of up to 1.1 ppmv · m · Hz −1 .
An overview of typical concentration measurements while continuously scanning is given in Fig. 4. In the left column, all four fan-beam units, represented by channels 15 0-3, are shown for a 2-D field measurement rate of 2.5 Hz, in the right column for 1.25 Hz. For all cases, the H 2 O concentration was plotted for all angles in a fan-beam, while the reference concentration is plotted as a solid line. Precision is enhanced when it is scanned slower, presumably because transmission and absorption length variation during one wavelength tune decrease. A full listing of scanning measurement 20 performance is shown in reference concentration by about 170 ppmv to 597 ppmv, which corresponds to 1.3 % to 4.8 % of the absolute concentration. Although this bias lies within the measurement uncertainty of the static reference instrument, the systematically negative deviation of the scanning measurement should be kept in mind and evaluated further. It is probably caused by inaccuracies of the assumed beam path lengths that were determined by 10 a measurement of a known uniform H 2 O concentration.
We also compared the continuous scanning mode to step-wise measurements, where the motor is sequentially moved but stopped for measuring the absorption signal. The overall time for scanning the whole field step-wise took more than four minutes. Results are shown in Fig. 5 and listed in detail in Table 1. Note that the reference con-15 centration is higher than during the scanning measurements, since the step-wise measurement took place a few days later. The standard deviation of the H 2 O-concentration is in the order of magnitude of the standard deviation that was achieved with the continuously scanning measurement. However, the systematic deviation from the reference measurement is slightly smaller. Since the beam is not moved during the absorption 20 measurement, there is no variation of absorption length and hardly any transmission variation during one tuning run. This might be a source of measurement error, as well as inaccuracies of absorption lengths of both reference and scanning channels. Preliminary numerical simulations were carried out, confirming that tomographic measurements with noise in the achieved order of magnitude are possible. 25 The accomplished field measurement rate of 2.5 Hz can be increased further by raising the laser tuning rate, e.g. by using a VCSEL instead of a DFB laser. With a VCSEL, tuning repetition rates of 30 kHz or more are realistic (Witzel et al., 2013), corresponding to up to 15 Hz field measurement rate.

Conclusions and outlook
We have presented a TDLAS spectrometer for spatially resolved 2-D-concentration measurements that may serve as a prototype for permafrost field measurements. Besides covering a relatively large measurement area of 0.8 m×0.8 m, it achieves rather fast field measurement rates of up to 2.5 Hz. A high number of usable laser beams 5 (> 350) and a relative field coverage of more than 75 % are favourable regarding tomographic spatial resolution. Measurements of homogeneous H 2 O distributions prove that accuracy and precision are satisfying regarding tomographic reconstructions. Relative standard deviations of the path-averaged concentrations amounted between 3 and 6 %, whereas the relative systematic deviations to a reference measurement amounted between 1.3 and 4.8 %.
In the future, the field measurement rate can be increased by raising the tuning rate of the laser, which requires replacement of the DFB laser by a VCSEL laser. We aim to carry out tomographic measurements and reconstructions with known concentration distributions for validation and later on measure first 2-D-concentration fields above 15 soils. Other molecules, such as CH 4 , shall also be quantified. As static measurements with distances between emitting side and retro-reflective foil of more than 5 m have already been successfully realized (Seidel et al., 2012), scanning experiments covering larger areas seem possible and are planned. Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version