Lightweight diode laser spectrometer „CHILD“ for balloon-borne measurements
of water vapor and methane
Wolfgang Gurlit, John P. Burrows (*), Rainer Zimmermann, Ulrich Platt (***), Carsten Giesemann,
Jürgen Wolfrum, Volker Ebert (**)
(*) Institute of Environmental Physics, University of Bremen, Otto-Hahn-Alle 1, 28201 Bremen
(**) Institute of Physical Chemistry, University of Heidelberg, Im Neuenheimer Feld 253, 69120
Heidelberg
(***) Institute of Environmental Physics, University of Heidelberg, Im Neuenheimer Feld 229, 69120
Heidelberg
A new lightweight near-infrared tunable diode laser spectrometer CHILD (Compact High-altitude In-situ
Laser Diode spectrometer) was developed for flights to the stratosphere as an additional in-situ sensor on
existing ballon borne payloads. Free-air absorption measurements in the near infrared are made with an
open-path Herriott-cell with new design features. It offers two individual absorption path lengths
optimised for CH4 with 74 m (136-pass) and H2O with 36 m (66-pass), respectively. New electronic
features include a real-time gain control loop that provides an auto-calibration function. In flight-ready
configuration the instrument mass is about 20 kg including batteries. It successfully measured
stratospheric CH4 and H2O profiles on high-altitude balloons on four balloon campaigns (ENVISATvalidation) between October 2001 and June 2003. On these first flights, in situ spectra were recorded
from ground level to 32,000 m altitude with a sensitivity of 0.1 ppm (ground) to 0.4 ppm (32,000 m) for
methane and 0.15 – 0.5 ppm for water.
OCIS codes: 120.6200, 300.6260, 010.1280
1. Motivation & Introduction
Water vapor and methane are two of the most important greenhouse gases in the atmosphere. Water
vapor accounts for 62 % of the natural greenhouse effect and plays an important role in many
atmospheric gas-phase reactions. Although less than 1% of the atmosphere’s total water content resides
in the stratosphere, changes of the stratospheric water-mixing ratio are considered to strongly influence
radiative forcing of the climate. The concentration of stratospheric water vapor in the "overworld" above
the  = 380 K potential temperature surface is determined by dehydration of air entering the stratosphere
through the tropical tropopause, methane oxidation in the stratosphere, transport by the poleward-anddownward (Brewer-Dobson) mean circulation [1] and other processes, which are only partly understood.
The CHILD Spectrometer page 1 of 32
06.02.2016
Methane on the other hand plays an important role in both radiative transport and photochemistry of the
atmosphere. Since the 18th century, atmospheric methane concentrations have doubled [2, 3] from 0.9 to
1.8 ppm. Both species are linked together by the fact that the oxidization of methane in the stratosphere is
the major source of stratospheric water. However, the observed 2 ppmv increase of stratospheric water
since the 1950s can only partly be explained by the increase of tropospheric methane (0.55 ppmv,
corresponding to 1.1 ppmv H2O) during the same period. This increase could be leading to the increased
formation of polar stratospheric clouds (PSC), an enhanced concentration of HO X radicals, and to lower
temperatures in the stratosphere. Due to the coupling of methane and water in the stratosphere, it is an
important task to monitor both species simultaneously.
While in situ measurements with high spatial resolution of important atmospheric trace gases by aircraft
or balloon missions are considered extremely important for the understanding of transport processes and
the radiative balance in the upper troposphere and lower stratosphere (UTLS), they suffer from the high
cost of these missions leading to sparse measurements and snapshot-like results. One way to achieve a
higher number of these measurements is to develop small, automated light-weight instruments, that can
be used piggyback on any mission that provides adequate flight trajectories. Stratospheric balloon
payloads typically launched e.g. by CNES (Centre National d’Études Spatiales, France) are in the range
from 100 kg to 500 kg payload. On such payloads, an additional compact instrument with a total mass
around 20 kg in flight-ready configuration can be integrated in many cases. TDL instruments used in the
past years on high-altitude ballons still require their own gondola and balloon [4, 5]. The new TDL
spectrometer CHILD described here is explicitely designed to use existing platforms and to substantially
reduce the costs and increase the frequency of highly resolved profile data acquisition. For this reason, a
lot of emphasis was put into a very low mass construction and reliable automation features. We
demonstrate the operational capabilities of the CHILD instrument by putting it piggyback as an
additional sensor on four flights of the TRIPLE gondola, which carried a set of instruments for the in-situ
detection of water vapor and other tracers. Due to the low mass of CHILD, the flights to the stratosphere
were possible with no additional cost for the launch of TRIPLE. As a result of the potential sensitivity to
local contamination by outgassing of water vapor, CHILD, like all other instruments installed inside the
large, open frame of TRIPLE, had to avoid any materials that could lead to corruption of the in-situ
measurements by outgassing of water or other substances. During stratospheric float at approx. 32 km
altitude, vertical oscillations of the whole vehicle (balloon envelope, typ. 150.000 m3 volume and
payload installed 150 meters below) occur during the adjustment of the ballon´s valve by telecommand.
In this situation, and partly also during ascent, high water spikes that originate from outgassing of the
The CHILD Spectrometer page 2 of 32
06.02.2016
balloon can be seen in the data of all water-sensing instruments. Best probing situation is then on the
descent, when the air is first probed by the payload before it comes into contact with the balloon
envelope.
2. Experimental
Tunable diode laser absorption spectroscopy (TDLAS) in the infrared (IR) [5-7] and near infrared (NIR)
[4, 8-10] spectral range is one of the most sensitive methods for stratospheric in-situ humidity
measurements [1]. Further methods for high-resolution and precise measurements of H2O in the
stratosphere are A) frost point (FP) hygrometers (FPH) [11], and B) Lyman-alpha photo-fragmentation
fluorometers (LAF) [12-14].
The FPH relies on creating a constant amount of frost on a temperature-stabilized, chilled optical mirror.
The optical losses are measured with a light source and are kept constant by adjusting the mirror´s
temperature. The measured mirror temperature is therefore directly related to an inherent property of
water – to the frost point (a similar concept is possible with the dew point). This makes these systems
stable and reproducible, with a lower detection limit for the best devices of 15 ppbV corresponding to a
FP of –100°C [15]. However due to the high thermal mass of the mirror it offers only a very limited time
resolution of 10 sec. and up to 120 sec. at very low frost point temperatures.
Another important technique, LAF, [12-14] is nearly as sensitive but offers a much higher time resolution
of up to 1 second. LAF uses a narrow-band Lyman alpha (L) light source to cause H2O-photofragmentation. The products of the fragmentation process are excited OH molecules that can be
selectively detected via their fluorescence in the 280 to 330 nm spectral range. This technique requires
sampling of the gas into the photolysis cell, which might lead to problems with contamination, adsorption
and limited exchange times of the flow system. Several earlier LAF instruments have shown false signals
due to ambient solar radiation [12]. The measurement is depending on the precise knowledge of the
photo fragmentation process, the OH yield, the quantum efficiency of the fluorescence and the selective
absorption of the L-radiation by the water molecules only. LAF hygrometers have to be carefully
calibrated and traced to a secondary standard in order to yield the desired absolute water concentration
[12,13,14]. Measurements of CH4 in the stratosphere are performed using IR-photometers, gas
chromatography (GC), or cryogenic whole air samplers [13,16].
As described above, it is very interesting to monitor H2O and CH4 in parallel. This is possible by the third
technique used for humidity measurements, which is laser-based absorption spectrometry. In combination
with tunable single-mode diode lasers it offers excellent sensitivity and selectivity. The concept is very
The CHILD Spectrometer page 3 of 32
06.02.2016
flexible and can be used for all IR-active molecules. Several TDL-based hygrometers have been deployed
successfully for airborne H2O measurements in the near-infrared [4, 8-10]. By using several diode lasers
at suitable wavelengths even a simultaneous detection of multiple species [5] is feasible, making this a
nearly universal tool for atmospheric chemistry studies [4, 5, 8-10]. The sensitivity can be adapted in a
wide range via a suitable change in absorption path length. Furthermore gas sampling can be avoided by
use of open-air absorption paths. This allows high gas exchange rates, thus high temporal resolution and
at the same time minimizes adsorption effects and the risk of contamination.
The instrument described here uses two independently tunable diode lasers that emit around 1393 nm and
1648 nm to detect water and methane, respectively. The measurement principle of absorption
spectroscopy is based upon a spectrally resolved measurement of the attenuation of the laser beam
propagating through the measurement volume. The attenuation as a function of the optical frequency  of
the laser, can be described by Beer’s law
I() = I0() exp {-S(T) g(L}
where I0() is the initial laser intensity, I() is the measured intensity after passage of an absorbing
medium of thickness L and a number density N of absorbers. The molecular absorption coefficient is
characterized by S(T), the temperature dependent line strength, and a function g() centered at
frequency 0 describing the shape of an absorption line.
The line profile is measured by a continuous scan of the laser wavelength across the absorption line. By
varying the laser temperature, the emission wavelength can be coarsely tuned onto the center of the
absorption line. A fast (Hz to kHz) periodic modulation of the injection current (saw-tooth or triangle
shape) then allows a fine tuning across the line. One side effect of this current modulation is an undesired
intensity modulation I0() of the laser light, which has to be accounted for. The dynamic wavelength
tuning allows a calibration free measurement of absolute number densities if the spectroscopic line
parameters are known [17, 18]. The availability of narrow bandwidth low-noise laser diodes has made
possible minimum detectable absorptions (MDAs) around 10-5 in the field and 10-6 in the laboratory,
while the theoretical MDA defined by the laser´s shot noise is typically in the 10 -8 range.
3. Instrument Description
The CHILD instrument, which was developed to fulfil the requirements of low weight, high sensitivity
and automated operation consists (see figure 1) of a new dual-species open-path multi-pass Herriott-cell
[19, 20] with a compact flange mounted optical head and a 19-inch electronic box, containing the laser
The CHILD Spectrometer page 4 of 32
06.02.2016
controllers, power supply, current amplifiers, modulation electronics and the lithium batteries for the
flight. The mass of the spectrometer optics (excluding electronic components) is only 6.6 kg. The overall
dimensions of the optical unit are 25 cm in diameter and 75 cm in length.
For the measurement of water vapor and methane, two distributed feedback (DFB) lasers were used.
These lasers contain a holgraphic grating that assures single mode operation, while at the same time a
limited tunability by variation of temperature and current is possible. Both lasers were characterised and
selected from a number of devices. Technical data are shown briefly in table 1.
Table 1
The two lasers are free-beam devices that are housed in small metal packages of approx. 7 mm diameter.
The free-beam design enabled a more flexible and compact design of the optical head, as well as a high
optical efficiency that was needed for the large number of reflections in the miniaturised Herriott-cell.
Further, it was chosen for the much lower cost of the free-beam laser devices.
Figure 1: TDL-Spectrometer with electronic box
Optical Head with 3-beam design
The compact optical head (visible on the right side in figure 1, and in modified version on the left side of
figure 5) contains all optical components needed in addition to the multi-pass cell itself (see figure 2).
The diode lasers are mounted in modified commercial housings, which contain the thermo-electric
coolers (TEC), temperature sensors for wavelength stabilisation and the collimating optics. The
collimating lens is adjustable along the optical axis (z) and the laser in x-y-direction. For each of the two
laser diodes, a three-beam layout is implemented. By using uncoated wedged beam splitters, each of the
two collimated laser beams is divided into a main beam for the absorption measurement and two
additional low-power beams (4% of the initial power each). These two additional beams are directed 1)
to wavelength reference cells and 2) to background (I0) signal detectors. The background signals are
needed mainly to compensate the effects of amplitude modulation which is caused by the periodic rampshaped variation of the laser current used for wavelength tuning. The reference cells contain a defined
concentration of H2O and CH4, respectively. They serve as wavelength references for the line-lock
The CHILD Spectrometer page 5 of 32
06.02.2016
scheme. To keep the vapor pressure high enough the water reference cell is stabilized at a temperature of
+5°C.
The main beams are coupled from the optical head into the Herriott-cell via folding mirrors. After
passage through the absorber, the beams exiting the cell are focused on the signal detectors (InGaAs). All
steering mirrors are mounted on lightweight, custom-made titanium solid-state hinges to prevent
misalignment. The base plate of the optical setup is equipped with small electrical heaters to minimize
water adsorption.
Figure 2: optical layout inside the optical head
Multi-pass Cell
The main part of the optical setup is the dual-species multi-pass Herriott-type absorption cell shown in
figure 1. It is open to the atmosphere to minimize adsorption effects of water, a major source of
uncertainty for many airborne water sensors. In order to simplify the integration of the instrument into
existing balloon payloads mass and size had to be minimized. The desired relatively small mirror spacing
(0.55 m) and the demand for a long absorption path resulted in a high number of optical passes. By use of
two independent coupling holes for the two species the absorption path length is independently adapted
to the relative magnitude of the expected absorptions for both species. This was achieved by designing a
new Herriott cell that allows 136 passes for methane (resulting in 74 m path length) and 66 passes for
water (36 m path length) at the same time by using spherical aberation. This concept eliminates the need
for line switching and increases the dynamic range. First priority for the optimisation process was the
136-pass pattern for methane, which is more critical than the 66-pass pattern for water. Figure 3 shows
the spot pattern of the two laser beams on the rear mirror, recorded with a near-IR sensitive vidicon
camera. Extensive model studies and raytracing calculations were made before the design of the cell was
finalised. These calculations, which are described in detail in [21] include simulations under all
environmental and mechanical conditions that the cell may encounter during launch and flight.
Figure 3: Spot pattern
The cell consists of two gold-coated spherical 5 inch Zerodur mirrors with identical focal lengths of 262
mm. A broad temperature range (30°C to -90°C) has to be expected during the flight. In order to avoid
condensation of tropospheric water vapor during balloon launch the mirror temperatures can be
controlled with separately attached heaters. As a consequence of the small spot distance of the 136-pass
absorption path the correct function depends critically on the exact spacing of the two Herriott mirrors. In
The CHILD Spectrometer page 6 of 32
06.02.2016
case of a longitudinal misalignment of the cell the laser beams will miss their output aperture and
recirculate between the mirrors. The total length tolerance over the full temperature range is below 100
µm. For this reason, three FeNi-steel tubes with low thermal expansion were used as the main structural
elements of the Herriott-cell. The remaining effects of mechanical expansion or contraction are further
reduced by using a passive thermal shift compensation system (PTSC). Figure 4 shows the principle of
PTSC:
Figure 4: back mirror support with PTSC
The back mirror of the Herriott cell is bedded on three pistils made of aluminum alloy which are mounted
on the end of the FeNi-tubes. The length of the tubes and the pistils is chosen to provide an identical
absolute thermal expansion by the aluminum pistils and the FeNi tubes. The condition is fulfilled when
the ratio of the lengths is inversely proportional to the ratio of the thermal expansion coefficients. The
FeNi-tubes and the aluminum pistils will expand and contract in opposite directions and both effects will
compensate each other. In order to optimise the exchange of the air inside the Herriott-cell, the whole
assembly is mounted horizontally in the gondola. Care is taken not to have other components of the
payload close to the cell, and to make sure an unobstructed airflow through the cell. During a hard
landing with strong groundwinds in fall 2002, the CHILD cell, like all other components of the TRIPLE
payload, suffered mechanical damage and needed to be refurbished. For the next flight (March 2003) a
protective "birdcage" was constructed from lightweight bicycle parts. The Herriott-cell with its protector
ready for launch with TRIPLE can be seen in figure 5.
Figure 5
Signal processing and inherent calibration
The signal processing chain was designed to maintain the sensitivity and calibration accuracy over time
under the harsh conditions of stratospheric balloon measurements. During its flight to the stratosphere,
the Herriott-cell may experience a large temperature variation, mechanical stress and possible
contamination, resulting in fluctuations of its optical transmission. A special signal processing approach
was developed to electronically compensate the resulting fluctuations of the cell´s (or other component´s)
optical transmission not originating from absorption measurements. The technical solution is based on a
dual-channel automatic gain control (AGC) technique. The basic idea is that the unwanted fluctuations
happen on a slower timescale (e.g. several seconds or even minutes) than the scanning ramp which is
The CHILD Spectrometer page 7 of 32
06.02.2016
periodically produced at 220 Hz to record the absorption signal. The amplifiers used for the absorption
and background signals are fast (gain-bandwidth product 20 MHz), but the AGC adjustment by a
feedback loop is set to slow time constant of several seconds.
The signal variations from beginning to end of each single scan ramp or the absorption itself are therefore
too fast to substantially influence the AGC. Thus, the two outputs (background / absorption) are balanced
by automatic gain adjustment to the same signal level of 1230 millivolts at the center of each scan.
(controlled by averaging over a whole scan) To achieve this, each of the two preamplifiers has its own
and independent automatic gain control (AGC) loop, referenced to the same stabilised bandgapreference. The output signal level as specified of 1230 mV at the center of the scanning ramp (with no
absorption present) can be maintained over a dynamic range of 3 orders of magnitude at the inputs of the
AGC circuits, which are able to produce gain factors from 0.5 to 500. We refer to this principle as a
"dynamic differential amplifier" [25], as only shorttime-differential signals between the two inputs are
amplified, while longer term signals are balanced and cancelled out as "common mode" in the following
subtraction stage (figure 6).
Figure 6
The analogical subtraction (absorption-background) is then done by a fixed-gain differential amplifier
stage. The resulting absorption signal (figure 6, signal 3) is background-free and referenced to an I0 of
1230 millivolts. Thus, the depth of the absorption line is inherently calibrated to this I0, e.g. each percent
of absorption is equal to 12.3 millivolts of line-depth. An additional (fixed-gain) amplification is then
applied to adapt the absorption signal to the input of the A/D converter.
The AGC function automatically compensates for any longer-term changes in signal intensity, which can
be caused by variation of laser power, optical transmission (e.g. misalignment or Herriott cell mirror
contamination), detector or laser degradation or electronic drift. The reference cell signal, IREF, is
processed in the same manner in order to safely and conveniently extract the line position for the linelocking algorithm.
The output voltages of the AGC amplifiers US (absorption signal) and UI0 (background) as a function of
optical frequency  can be described by the following equations:
The CHILD Spectrometer page 8 of 32
06.02.2016
U S ( )  AS VS PL ( ) [1  a( )]
U I 0 ( )  AI 0VI 0 PL ( )
Here, PL() is the optical laser power, and a() is the molecular absorption depth in the signal path. AS
and AI0 are factors describing the different light transmissivities of the signal and reference light paths in
the absence of molecular absorption, as well as detector sensitivities and pre-amplifier characteristics; VS
and VI0 are the amplifications of the analog AGC amplifiers. Due to the effects described above, A S and
AI0 may vary on a slow time scale, but can be assumed to be constant within each collected spectrum.
Because of the normalisation and under the condition that molecular absorption is small (i.e., a()<<1
over the entire scan range), it can be assumed that:
ASVS = AI0VI0
The normalised main and background signals still carry a strong triangular variation due to the laser
intensity modulation. This is very effectively removed by an analog subtration circuit, which also
includes a gain of G = 25.7 ± 1% to amplify the small absorption signal.
U D ( )  G  U I 0 ( )  U S ( )
The difference signal can now be readily converted to the instantaneous relative absorption a()
according to:
a( ) 
U D ( )
.
G  U I 0 ( )
This difference signal is schematically shown in figure 6. As the AGC scheme provides optimum
balancing of both input signals, the output signal that results from the subtraction is now backgroundfree. All dynamic range of the following 16 bit A/D conversion can then be used for the absorption
signal instead for a large background.
The CHILD Spectrometer page 9 of 32
06.02.2016
The advantage of a true zero-background normally can only be achieved by the use of derivative
modulation techniques like frequency modulation (FM), two-tone frequency modulation (TTFM) or
wavelength modulation (WM). All these techniques however result in a derivative-like line shape that
strongly depends on instrumental parameters, which makes the calibration of these instruments and the
precise conversion from measured signals to gas concentrations more difficult [22-24].
The difference signal UD, the background signal UI0 and also the reference cell signal UREF (which is also
normalised to 1230 mV mean level) and the laser current ILD are then sampled by a 16-bit A/D converter
at a sampling rate of 95.238 kHz. The difference signal is additionally passed through a 2nd order Besseltype low-pass filter to remove any high-frequency noise beyond the Nyquist frequency.
An important point to note is that the absorption information determined by this scheme is inherently
calibrated (auto-calibration, described in [25]). Since both voltages are sampled by the same mulitplexed
A/D converter and both AGC amplifiers use the same reference voltage, the only source of systematic
error in the absorption measurement is the amplification G, which can be relatively easily controlled.
Onboard Computer and Software
The complete instrument is controlled by a single industry standard PC-104 format computer board with
an AMD 5x86 processor running at 133 MHz. A 16-bit ADC board with 8 differential input channels and
100 kS/s total sampling rate is used to collect the analog scan signals. A second multi-I/O board
including a 16-bit, 32 channel ADC, a 4-channel, 12-bit DAC, and digital I/O, is used for collection of
housekeeping signals and control of various instrument functions. A 224 MB solid-state flash disk serves
as the storage medium for software and the collected data.
The PC-104 board software, written in the C programming language, is operated under a Real Time
Linux (RTLinux) Kernel version 2.0.37.RT.1.1 [26]. Certain time-critical software components are
implemented as real-time tasks running with kernel privileges.
A set of various housekeeping signals is collected once per second through the 32-channel ADC board.
These include temperatures at critical points of the optical head, absorption cell and electronics, the air
temperature and the air pressure measured by a temperature-stabilised Baratron pressure sensor. These
The CHILD Spectrometer page 10 of 32
06.02.2016
values, along with status variables of the software control loops, are logged also to the on-board flash
disk and transmitted through the telemetry downlink for inspection during flight. Heaters of the Herriott
cell mirrors, optical head, H2O reference cell, electronics and Baratron pressure sensor are controlled by
MOSFET switches. A software-implemented pulse width modulation (PWM) allows a proportional
control of heater power. The temperatures of the H2O line lock cell and the Baratron are stabilised by a
software-driven PID control loop.
The laser temperatures are stabilised by a P-type (proportional) analog control loop. Also, to correct for
residual temperature deviation as well as long-term drift of the laser diode, and to keep the wavelength
scan range centered around the absorption line, a second real-time software-driven I-type control loop is
implemented which stabilises the absorption line position seen in the reference cell signal. After each
averaging cycle, the line position is extracted from the reference cell spectrum, whose integrated
deviation is used to correct the analog temperature set-point via a 12-bit DAC. To extract the line
position, a specially developed center-of-mass algorithm is used, which requires neither a fit routine nor
floating-point arithmetic. The line position can therefore be calculated within a short, predictable time
interval. It is also very stable against noise or spikes. Under typical stratospheric flight conditions, the
line position was found to be stable within ± 100 MHz for H2O and ± 200 MHz for CH4.
For each species, four ADC channels are sampled at a rate of 95.238 ksamples/sec each: (1) the ramp
signal modulating the laser current Ilas(t), calibrated at 10 mV/mA; (2) the normalized background
detector signal UI0(t), (3) the amplified differential absorption signal UD(t), and (4) the normalized
reference cell signal ULL(t). Each spectrum comprises 216 data points for each channel, thus covering one
half-period of the ramp signal. Between 5 and 100 individual spectra are co-added immediately after
collection, the actual number varying between channels and the different flight phases. After averaging,
all spectra are stored on the on-board flash disk for post-flight data analysis, a part of the spectra is also
transmitted through the telemetry downlink for pre-analysis during flight.
4. Data Analysis
The raw spectra collected by the on-board software are represented in a 32-bit integer format. They are
first re-scaled to represent ADC channel input voltages; also, the calibrated modulation ramp signal is
used to re-scale the independent axis in terms of relative optical frequency, . =0 represents the
nominal absorption line position.
The CHILD Spectrometer page 11 of 32
06.02.2016
From a theoretical point of view, the absorption depth a() can be described as a Voigt line shape with
superimposed error contributions due to electronic background, b(), detector and electronic noise, n(),
and interference fringing, f():
  0  l 
a( )  A0 V 
,   b( )  n( ) 
 d d 
f ( )
Here, A0 is the integrated absorption, d the Doppler line width due to thermal broadening, l the Lorentz
line width due to collisional broadening, and 0 is the frequency of the line center. The Voigt line shape
function V(x,y) is defined as:
V ( x, y ) 
2 ln 2 y
 3/ 2  d


0
e t
dt
y 2  ( x  t)2

2
with
 V ( x, y ) dx  1.

In the case of methane, the single Voigt line shape function V(x,y) has to be replaced by a superposition
of four Voigt functions, as a group of four absorption lines (6067.1 cm-1) is used for the methane
measurement.
Interference fringes are caused by unwanted, Etalon-like behaviour of optical windows, laser collimation
lenses and the Herriott-cell. f() is, theoretically, a superposition of sine-shaped functions of various
periods and amplitudes, whose phase may vary strongly between individual spectra. It can be separated
into a long-period part (period >> d) and a short-period part (period << d). In practice, the contributions
of both the electronic background b() and the long-period part of f() can be represented by a
polynomial background function. The overall short-period fringe amplitude was found to be
f short ( )  10 4
2
f short ( )  10 4
2
f short ( )  10 4 . The noise term n() was found to be
2
negligible in all our measurements.
Though the optical head, which contains all laser, beamsteering and detector components, is open to the
atmosphere and flushed by a fan, the water vapour measurements can be influenced by a certain amount
of residual water vapour inside the optical head. This contamination is corrected for by an additional
analysis of the H2O background beam signal, which is only influenced by water inside the optical head.
Because the optical pathways are precisely known, the background signal (which by default should
The CHILD Spectrometer page 12 of 32
06.02.2016
contain no absorption) offers the possibility to measure the amount of residual water vapor there. This
measurement is then used to correct the absorption signal of the Herriott-cell for additional water
absorption on the (relatively short) optical path inside the optical head. The resulting spectrum, scaled by
a path length factor of CH2O = 2.3 ± 0.1, is then subtracted from the atmospheric spectrum. However, this
correction scheme is based on the assumption of an equal distribution of the residual water inside the
optical head.
The resulting absorption spectrum a() is initially approximated by a Levenberg-Marquardt fit algorithm
which varies the parameters A0, l, 0, and the background polynomial coefficients. The residual
spectrum resulting from the fit process is dominated by the short-period part of f(), which is also the
main source of statistical uncertainty in A0.
The Doppler line width, caused by thermal broadening, is determined by [27]
d 
 0 8k BT ln 2
c
m
where 0 is the central frequency of the line, c the speed of light, kB Boltzmann’s constant and m the
molecular mass. The Lorentz line width (or pressure broadening parameter) l depends on ambient air
pressure and temperature by [28]:
T 
 l  pair  cbr,air   ref 
 T 
n
(Tref = 296 K)
Spectroscopic parameters of the absorption lines used by CHILD are listed in Table 2. The air
broadening coefficient cbr,air is a physical property of the given absorption line. For H2O, we use the value
given by [29]. For the CH4 methane absorption lines used by our instrument, no accurate determinations
of this parameter have been published to our knowledge (the HITRAN database gives only estimates), it
is determined from the data itself in a special Levenberg-Marquardt fit analysis run in the pressure range
100...1000 hPa with l as a variable parameter.
Table 2
In the main analysis run on the data set, l is then calculated using cbr,air based on ambient pressure and
termperature and held fixed for the Levenberg-Marquardt fit, which then produces the final integrated
absorption (A0). Using predetermined values of cbr,air , rather than leaving l as a variable fit parameter,
greatly reduces scatter in A0.
The CHILD Spectrometer page 13 of 32
06.02.2016
Finally, the integral absorption A0 can then be converted to a volume mixing ratio (VMR) by:
VMR  A0
kT
,
pL S
where k is Boltzmann’s constant, T the absolute air temperature, p the air pressure, L the absorption path
length and S the absorption cross-section.
Figure 7
Figure 7 shows typical measured spectra at a pressure level with relatively low absorption (50 hPa). It
shows the averaged spectrum after subtraction of the polynomial background (Data-BG), the fitted line
shape (Fit), the polynomial background (scaled by a factor of 0.2 in the CH4 plot), and the residual scaled
by a factor of 5.
Accuracy
Apart from the statistical uncertainty, which is mainly caused by interference fringes in our instrument,
possible sources of systematic errors in the calculated VMR are (1) the pressure measurement error, p/p
 1% except for very small pressures below 20 hPa; (2) the temperature measurement error, T  4K; (3)
the amplification error of the subtractor circuit and ADC sensitivity error, with a total of 1.1%, and (4)
the uncertainty of the absorption cross-section, S. Because of wavelength scaling errors, we must also
consider (5) the calibration error of the tuning ramp signal, 1% and the uncertainty of the laser tuning
coefficient, estimated to be  2% in total. All errors are given as 2 values.
The cross-sections S are taken from the HITRAN database [28, 30] for CH4 and from [31] and [32] for
H2O, with a specified uncertainty of 4% in both cases. Except for the H2O measurement at pressure levels
p < 40 hPa, where the uncertainty in the contamination correction CH2O dominates, the total systematic
uncertainty of our data is almost constant at  6%, with S being the leading systematic error term.
However, since S is a physical constant, more accurate determinations of S may improve the accuracy of
our measurements in the future.
The CHILD Spectrometer page 14 of 32
06.02.2016
5. Results and discussion
Figure 8
Figure 8 shows the profiles of methane (scaled by a factor of 2) and water vapor measured during the
flight of our instrument on Sept. 24, 2002 at Aire-sur-l’Adour (southern France, 44°N). Flight parameters
were matched to an overpass of the ENVISAT satellite in order to use the acquired profile data for the
validation of the SCIAMACHY spectrometer. Depending on altitude, the data were averaged over
intervals of 5 to 200 seconds. Error bars include both statistical and systematic errors. The data shown
here were collected during the controlled descent to avoid H2O contamination caused by outgassing from
the balloon. Residual water inside the optical head has been corrected for as described above. The
potential water concentration [H2O]p = 2 [CH4] + [H2O] is also shown. Leaving aside the region below
100 hPa, where [H2O]p is influenced by extratropical troposphere-stratosphere exchange, it is found to be
independent of altitude, as expected. Its mean value above 100 hPa was determined to be 7.9 ± 0.4 ppm,
in good agreement with previous observations [13, 33] and an assumed rate of stratospheric H2O increase
of about 50 ppb/a. The CHILD instrument is designed for stratospheric measurements. In the troposphere
below 100 hPa, the water vapor absorption signal may be "clipped" electronically, because it is getting
too large. However, from the slopes of the water vapor absorption line, a retrieval is still possible from
approx. 600 hPa above. At even lower tropospheric layers, in addition to the electronic clipping, the
absorption signal gets saturated (completely absorbed) in the Herriott-cell. Both effects can be avoided
by changing the laser´s emission wavelength to a weaker absorption during part of the flight by
telecommand. However, this feature was only used during ground testing and not on the flight, as the
mission of our flights were stratospheric measurements.
Especially for the water vapor in-situ measurements, we found that the sensitivity is more dominated by
the danger of local contamination than by instrumental factors like signal to noise ratio. At stratospheric
float, when the contol valve of the ballon is openend by telecommand in order to stabilise the vertical
position, some altitude oscillations of a few hundreds of meters can occur. Though the payload is
installed 150 meters below the balloon, these oscillations produce large water peaks in the measurements
(at least factor 10 above the stratospheric values) that originate from the large surface of the balloon
The CHILD Spectrometer page 15 of 32
06.02.2016
envelope (typ. more than 10.000 m2). Even an (at least short-time) contamination of the instruments is
possible. For this reason , an outgassing phase of at least 15 minutes at stabilised float position is always
done, before the best measurements are then made during descent, when the air is first probed by the insitu measurement before it is possibly contaminated by the balloon.
6. Conclusion
The newly developed lightweight CHILD spectrometer was flown four times piggy-back as an additional
sensor on the larger payload ”TRIPLE” with a high-altitude balloon from Aire sur l’Adour (France) and
Kiruna (Sweden). Profiles of water vapor and methane were obtained by in-situ measurements. The TDL
spectrometer has new features that are intended to make routine operation easier, more reliable and more
cost-efficient. These new features are 1) a small size Herriott-cell with two different optimized optical
pathlengths for the two absorbers, and 2) an electronic autocalibration and autobalancing circuit that
ensures long-term stability and inherent calibration of the measurements. The data collected during these
first flights show the CHILD´s high sensitivity and reliability. All technical innovations of the instrument
worked successfully. For the year 2004 a tropical campaign from Teresina (Brasil) is planned.
Detailed comparisons of the CHILD data with measurements of the SCIAMACHY spectrometer on
ENVISAT are underway as part of the official ENVISAT validation programme.
7. References
1) SPARC Assessment of upper tropospheric and stratospheric water vapor, WCRP-113, WMO/TD Mo.
1043, SPARC Report No. 2 (2000)
2) IPCC, Climate Change 2001: The Scientific Basis. Contribution of Working Group I to the Third
Assessment Report of the Intergovernmental Panel on Climate Change (IPCC). Cambridge University
Press (2001)
3) E. J. Dlugokencky, K. A. Masarie, P. M. Lang und P.P.Tans, “Continuing decline in the growth rate
of the atmospheric methane burden” Nature 393, 447–450 (1998)
The CHILD Spectrometer page 16 of 32
06.02.2016
4) G. Durry, G. Megie, “Atmospheric CH4 and H2O monitoring with near-infrared InGaAs laser diodes
by the SDLA, a balloonborne spectrometer for tropospheric and stratospheric in situ measurements”,
Applied Optics, 38, No. 36, 7342-7354 (1999)
5) G. Moreau, M. Pirre, F. Goffinont-Taupin, C. Robert, C. Camy-Peyret, S. Payan, R. Ramaroson,
“Results and goals of SPIRALE after the first flight from Gap in June 1999”, 15 th ESA Symposium on
European rocket and balloon programmes and related research, SP-471, 309-314 (2001)
6) P. Werle, ”Spectroscopic Trace Gas Analysis Using Semiconductor Diode Lasers”, Spectrochimica
Acta A 52 805-822 (1996)
7) David S. Bomse, Alan C. Stanton, Joel A. Silver, “Frequency modulation and wavelength modulation
spectroscopies: comparison experimental methods using a lead-salt diode laser”, Applied Optics 31, 718731 (1992)
8) R. D. May, “Open-path, near-infrared tunable diode laser spectrometer for atmospheric measurements
of H2O”, J. Geophys. Res., 103, 19,161-19,172 (1998)
9) D.M. Sonnenfroh , W.J. Kessler , J.C. Magill , B.L. Upschulte , M.G. Allen , J.D.W. Barrick, “In-situ
sensing of tropospheric water vapor using an airborne near-IR diode laser hygrometer”, Applied Physics
B Lasers and Optics Volume 67, Issue 3, 275-282 (1998)
10) J. A. Silver and D. C. Hovde, “Near-Infrared Diode Laser Airborne Hygrometer”, Rev. Sci. Instrum.
65, 1691-1694 (1994).
11) A. W. Brewer, B. Cwilong, and G. M. B. Dobson ,“Measurement of absolute humidity in extremely
dry air”, Proc. Phys. Soc., 60, 52-70 (1946)
12) M. Zöger, A. Afchine, N. Eicke, M.-T. Gerhard, D. S. McKenna, U. Mörschel, U. Schmidt, V. Tan,
F. Tuitjer, T. Woyke, and C. Schiller, “FISH: A novel family of balloon borne and airborne Lyman-a
photofragment fluorescence hygrometers”, J. of Geophys. Res., Vol. 104, 1807-1816 (1999)
The CHILD Spectrometer page 17 of 32
06.02.2016
13) M. Zöger, A. Engel, D. S. McKenna, C. Schiller, U. Schmidt, T. Woyke, “Balloon borne in-situ
measurements of stratospheric H2O, CH4 and H2 at mid-latitudes” , J. of Geophys. Res., Vol. 104, 18171825 (1999)
14) E. J. Hintsa, E. M. Weinstock, J. G. Anderson, R. D. May and D. F. Hurst, “On the accuracy of in
situ water vapor measurements in the troposphere and lower stratosphere with the Harvard Lyman-
hygrometer”, J. Geophys. Res., 104, 8183-8189 (1999)
15) Buck Research Inc. Specifications, Model CR-2 Cryocooled Hygrometer, BUCK RESEARCH
INSTRUMENTS, LLC, 5375 Western Avenue, Boulder, CO 80301
16) U. Schmidt, G. Kulessa, E. Klein, E.-P. Röth, P. Fabian, and R. Borchers, “Intercomparison of
balloon-borne cryogenic whole air samplers during the Map/Globus 1983 Campain”, Planet. Space Sci.,
35, 647-656 (1987)
17) H. Teichert, T. Fernholz, V. Ebert, “Simultaneous In Situ Measurement of CO, H2O and Gas
Temperature in a Full-Sized Coal-Fired Power-Plant Using Near Infrared Diode Lasers”, Applied Optics
42, No 12, 2043-2051 (2003)
18) V. Ebert, T. Fernholz, C. Giesemann, H. Pitz, H. Teichert, J. Wolfrum and H. Jaritz, “Simultaneous
Diode-Laser-Based In-situ-Detection of Multiple Species and Temperature in a Gas-fired Power-Plant”,
Proc. Comb. Inst. 28, 423-430 (2000)
19) D.R. Herriott and H.J. Schulte, “Folded optical delay lines”, Appl. Opt. 4, 883-889 (1965)
20) D.R. Herriott, H. Kogelnik, and R. Kompfer, “Off-axis path in spherical mirror interferometers”,
Appl. Opt. 3, 523-526 (1964)
21) C. Giesemann, "Entwicklung und Einsatz eines Diodenlaserspektrometers zum quantitativen in-situNachweis von Methan und Wasser in der Stratosphäre", Dissertation, Univ. Heidelberg, (2003, in
German)
The CHILD Spectrometer page 18 of 32
06.02.2016
22 ) R. D. May and C. R. Webster, “Data processing and calibration for tunable diode laser harmonic
absorption spectrometers”, J. Quant. Spectrosc. Radiat. Transfer, 49, 335-347, (1993)
23) G. C. Bjorklund, “Frequency modulation spectroscopy: a new method for measuring weak
absorptions and dispersions”, Opt. Letters 5, 15-17 (1980)
24) G. C. Bjorklund, M. D. Levenson, W. Lenth, C. Ortiz, “Frequency modulation spectroscopy: theory
of lineshapes and signal to noise analysis”, Applied Physics B 32, 145-152 (1983)
25) W. Gurlit, “Diodenlaser für spektroskopische Anwendungen: Neue Aspekte für die Konstruktion
hochempfindlicher Feldinstrumente“, Cuvillier Verlag Göttingen, ISBN 3-89588-786-2, (1997, in
German)
26) M. Barabanov, “A Linux-based Real-Time Operating System”, Master Thesis, New Mexico Institute
of Mining and Technology, (1997)
27) W. Demtröder, “Laserspektroskopie“, Springer-Verlag, ISBN 3-540-52601-3, (1991)
28) L. S. Rothman, C. P. Rinsland, A. Goldman, S. T. Massie, D. P. Edwards, J-M. Flaud, A. Perrin, C.
Camy-Peyret, V. Dana, J.-Y. Mandin, J. Schroeder, A. McCann, R. R. Gamache, R. B. Wattson, K.
Yoshino, K. V. Chance, K. W. Jucks, L. R. Brown, V. Nemtchinov and P. Varanasi, “The Hitran
Molecular Spectroscopic Database And Hawks (Hitran Atmospheric Workstation): 1996 Edition“,
Journal of Quantitative Spectroscopy and Radiative Transfer, 60, 665-710 (1998)
29) L. Moretti, A. Sasso, L. Gianfrani und R. Ciurylo, “Collisional-broadened and Dicke narrowed
lineshapes of H216O and H218O transitions at 1.39µm”. J. Mol. Spectroscopy 205, 20 (2001).
30) J. S. Margolis, “Measured line positions and strengths of methane between 5500 and 6180 cm-1”,
Applied Optics 27 (19), 4038– (1988).
31) B. Parvitte, V. Zéninari, I. Pouchet and G. Durry, “Diode laser spectroscopy of H2O in the 7165–
7185 cm-1 range for atmospheric applications”, Journal of Quantitative Spectroscopy and Radiative
Transfer 75 (4), 481–505 (2002).
The CHILD Spectrometer page 19 of 32
06.02.2016
32) R. A. Toth, “Extensive measurements of H216O line frequencies and strengths: 5750 to 7965 cm-1”,
Applied Optics-LP 33 (21), 4851–4867 (1994).
33) D. F. Hurst, G. S. Dutton, P. A. Romashkin, P. R. Wamsley, F. L. Moore, J. W. Elkins, E. J. Hintsa,
E. M. Weinstock, R. L. Herman, E. J. Moyer, D. C. Scott, R. D. May und C. R. Webster, “Closure of the
total hydrogen budget of the northern extratropical lower stratosphere”, J. Geophys. Res. 104 (D7),
8191–8200 (1999).
Acknowledgement
The authors would like to acknowledge the support of Cornelius Schiller and Fred Stroh at
Forschungszentrum Jülich, and Andreas Engel from University of Frankfurt, Germany, who made
possible the first flights of our instument piggy-back on the TRIPLE-Gondola, and who continue to give
us important support. The development of “CHILD” was funded by the German Government (BMBF)
under contract number 01 LA 9835.
The CHILD Spectrometer page 20 of 32
06.02.2016
Figure 1
The CHILD Spectrometer page 21 of 32
06.02.2016
Figure 2
The CHILD Spectrometer page 22 of 32
06.02.2016
Figure3
The CHILD Spectrometer page 23 of 32
06.02.2016
Adjustment Screw
Thread Sleeve
Invar-tube
Herriott-Mirror
Tension Spring
Mirror Heater
Piezo actuator
(optional)
Aluminiumbolt
Figure 4
The CHILD Spectrometer page 24 of 32
06.02.2016
Figure 5
The CHILD Spectrometer page 25 of 32
06.02.2016
1230 mV at
scan-center
abs.
signal
0
abs.
signal
scan
AGC 1
+
reference
1230 mV
back
ground
signal
subtractor
abs.line
referenced
to I0 = 1230 mV
-
AGC 2
0
back
ground
signal
0
1230 mV at
scan-center
scan
Figure 6
The CHILD Spectrometer page 26 of 32
06.02.2016
absorption depth
0.02
H2O 7181.2 cm
p = 50 hPa
VMR = 5.8 ppm
0.01
Peak = 1.610
Fit
Data-BG
Residual 5
-1
Background
-2
RMS = 5.210
-4
0.00
-0.01
-0.02
-0.03
0.003
Fit
Data-BG
Residual 5
-1
CH4 6067.1 cm
p = 50 hPa
VMR = 1.23 ppm
0.002
Backgrd 0.2
Peak = 7.110
absorption depth
-4
RMS = 9.810
-5
0.001
0.000
-0.001
-0.002
-0.003
-8000 -6000 -4000 -2000
0
2000 4000 6000 8000
 (MHz)
Figure 7
The CHILD Spectrometer page 27 of 32
06.02.2016
2 [CH4]
10
2 [CH4] + [H2O]
7.9  0.4 ppm
3.6 ppm
pressure level (hPa)
[H2O]
100
194hPa
TP
0
2
4
6
8
volume mixing ratio (ppm)
Figure 8
The CHILD Spectrometer page 28 of 32
06.02.2016
10
12
Figure Captions
Figure 1: TDL-Spectrometer with electronic box before integration into the TRIPLE gondola. Including
all packaging and the necessary power supply by batteries, the total weight is less than 25 kg. To enable
optimum contamination-free flow of air through the cell, it is installed horizontally in the gondola. An
additional advantage is that with this installation, protective covers (with the additional risk of opening
by telecommand) to avoid contamination of the mirrors during ascent in the troposphere are not needed.
Figure 2: All laser, detector and beamsteering components are installed inside the compact optical head,
that also contains the heatable reference-cells. From each of the two collimated laser beams, two lowintensity reflections are coupled out by a beamsplitter. These low-intensity beams are directed to the
reference-cell (used for the linelock) and to the I0 (background) detector. The main beam is brought into
the Herriot-cell (Z-direction in this figure) by a folding-mirror. The whole optical unit has a mass of only
6.6 Kg.
Figure 3: Spot pattern on the Herriott-cell´s back mirror, imaged with an IR-sensitive vidicon camera.
The outer pattern with the closer spacing belongs to the laser for methane, the inner pattern with the
wider spacing is the pattern of the water vapor absorption path.
Figure 4: Section of thermal shift compensation. Misalignment of the Herriott-cell by thermal shift in the
operating range from +40° to –60° is avoided with this compensation mechanism. Remaining shift of the
invar tubes, that form the structure of the cell, is compensated by small bolts from aluminum shifting the
mirror to the opposite direction. Thermal expansion coefficients of the two materials have to be inversely
proportional to the length of the elements.
Figure 5: Ready for launch from Esrange (Kiruna, 67°N) in march 2003. The protective "birdcage" made
from lightweight bicycle parts was added to the Herriott-cell after it had suffered serious structural
The CHILD Spectrometer page 29 of 32
06.02.2016
damage during a hard landing in southern France on a previous campaign in fall 2002. It also helps to
better distribute the mechanical load on the Herriott-cell during launch and flight.
Figure 6: Signal flow in autocalibration scheme. Both background and absorption signal are scaled to a
fixed average value of 1230 mV, but the reaction time for gain changes is 3 orders of magnitude slower
than the scan frequency. Thus, all information in the signals is fully preserved, but slower effects caused
by possible shift of optical transmission, or by contamination of optical parts, are completely
compensated by the automatic gain control. As a result, the absorption signal is autocalibrated. For
details see text.
Figure 7: Typical observed spectra at a low pressure of 50 hPa for water vapor (above) and methane
(below). Absorption path length is 36 meters (66-pass) for water and 74 meters (136-pass) for methane.
Temporal resolution (averaging time) at middle to high altitudes like shown here is typically 1.5 to 10
seconds. The fit residuals have been magnified by a factor of 5 in the plots. The minimum detectable
absorption for methane is limited by optical fringes in the Herriott-cell.
Figure 8: Methane and water vapor profiles from balloonborne TDL measurement (France, fall 2002).
Methane is plotted here as 2[CH4] to illustrate the concept of the potential hydrogen 2 [CH4] + H2O.
Time and place of the measurement were coordinated with an overpass of ENVISAT, in order to use the
CHILD measurement for validation of the SCIAMACHY spectrometer.
The CHILD Spectrometer page 30 of 32
06.02.2016
Target Species
CH4
H2O
laser type
STH910002Z-5
SU1393-DFB-CD
maker
Laser Components
Sensors Unlimited
center of emission
1648.2 nm
1392.53 nm
scan range
+- 8.5 GHz
+- 7.5 GHz
linewidth
< 30 MHz
< 30 MHz
max. opt. power
9.2 mW @ 90 mA
7.6 mW @ 80 mA
tunability by current
1.6 x 10-2 nm/mA
7.09 x 10-3 nm/mA
tunability by temperature
9.72 x 10-2 nm/K
8.68 x 10-2 nm/K
absorption path length
74.0 m
35.9 m
MDA (minimum
4-8 x 10-5
4 x 104
detectable absorbance)
(limit: optical fringes)
(limit: possible local water vapor)
Table 1: Laserdiode technical data
The CHILD Spectrometer page 31 of 32
06.02.2016
Abs. line or line
S [cm-1/(molec cm-
cbr,air [MHz/hPa]
cbr,air [MHz/hPa]
group
2)]
(HITRAN estimate) (value used)
n
(HITRAN value)
(value used)
CH4 23: R(5) group
3.06810-21  4%
(6067.1 cm-1)
[27, 29]
H2O 1+3: P(3)
1.4210-20  4%
303  202
[30, 31]
1.95
1.85  0.10
(our determination)
3.05
3.20  0.035
[28]
(7181.17 cm-1)
Values of S and cbr,air are given for the reference temperature Tref = 296 K.
Table 2: Spectroscopic parameters used for CHILD data processing
The CHILD Spectrometer page 32 of 32
0.75
06.02.2016
0.73