The mechanical structure of the spectrometer houses the optical components comprising a 9 cm double sided optical path difference Michelson interferometer and an infrared detector unit which accepts an incoming beam with a nominal diameter of 36.1 mm and a maximal divergence of 4.1 by 4.1∘. Out of the possible configurations for Fourier transform spectrometers (Carli et al., 1999), a single linear-slide two-port configuration with cube corners was selected for compactness, robustness, and optical (vignetting) considerations. This optical setup is not shear-compensated and requires alignment. Since the instrument is operated in various orientations relative to the direction of gravity, it has to be sufficiently stiff to preserve this static shear alignment during gimbal rotation. High stiffness is also needed to suppress vibrational shear that otherwise could be the cause of ghost contributions to the spectra.

GLORIA uses time-equidistant sampling with post processing to account for velocity variations (Brault, 1996). In the ideal case, this technique can cope with velocity variations approaching 100 %. In reality, the optical paths of the infrared radiation and of the reference will not be completely identical and residual errors will arise. Based on unpublished experiences with MIPAS-STR, the goal for the velocity stability of the optical path difference measured by the reference laser system is 5 % RMS which provides a reasonable trade-off between technical limitation and theoretical requirements.

In addition, the design of the spectrometer shall be compact and lightweight to allow its integration in the gimbal. Integrated local electronics are used on every gimbal frame as well as on the spectrometer itself in order to minimize the number of cables running through the gimbal axes. Furthermore, the pointing stabilization requires stiffness of the overall system – including the gimbal and the spectrometer – to enable pointing stability within 0.7 arcmin (1σ) in elevation during the recording of an interferogram.

The mechanical systems have to cope with low environmental temperatures (down to 200 K) and pressure changes from 1000 hPa at ground to 70 hPa at flight altitude. Also vibrations impact the instrument and the spectrometer optics. The vibration requirements are driven by the performance of the interferometer. The design of the optics was specified to perform measurements during cruise flights with vibrations fitting two requirements. The first one defines that acceleration spectral peaks shall remain below 10 % of the power spectral density (PSD) in DO 160C (Radio Technical Commission for Aeronautics (RTCA), 1989) “standard random vibration test curve for equipment installed in fixed wing aircraft with turbojet or turbofan engines, curve C”. This vibration limit is derived from measurements at different locations on the research aircraft Geophysica (MDB, 1996), where the envelope of the maximum PSD values are smaller than the 10 % DO-160 curve. The second requirement concerns the overall root mean square acceleration Grms which can be calculated by the square root of the integrated PSD curve. This value is given in the DO160C curve C with 4.12 g. The Grms values during cruise are assumed to be below 10 % of DO160C from experiences with MIPAS-STR on Geophysica and therefore smaller than 4 ms-2. This corresponds to PSD values below 1 % of DO160C.

These vibration values are given for the attachment points of the whole instrument at the carrier. However, these values are used as requirements for the spectrometer, too. In reality, the vibrations may be damped or increased by the structure of the gimbal. Additionally, vibrations can be caused by airflow.

Additional constraints are imposed by the flight certification process; static load strengths must be demonstrated by calculations and testing. The requirements are different for HALO (Wernsdorfer and Witte, 2008) and for Geophysica (MDB, 2002). In Geophysica the instrument is inside of the fuselage and has to be handled as a built-in component. The choice of materials shall conform to aviation safety regulations.

The optical components have to be cooled below 220 K in order to sufficiently reduce the background thermal radiation and thus the background photon noise. The temperature should be spatially uniform in the optic module to avoid thermomechanical misalignment and disturbances due to air density fluctuations in the spectrometer. The temperature must also be temporally constant, with a drift rate smaller than 2 Kh-1 to extend the time intervals between calibrations and to achieve a good radiometric accuracy (Kleinert et al., 2014).

For proper operation of the spectrometer in the harsh environment, care has to be taken to keep the cooled interferometer volume dry and clean. The outer surface should be kept free of condensate to reduce contamination and avoid disturbances at critical components such as on the entrance window or on electrical connectors. The requirements are summarized in Table 1.

The cooling system has to be located close to the interferometer optics – within the gimbal to avoid insulated or pressurized hoses from one gimbal level to the other, which may impact the pointing stabilization performances. Compressor or Stirling coolers for the required working temperature and environment are bulky, heavy and generate vibrations. Their integration in the system is difficult. Therefore, a non-electrical cooling system with a reservoir of coolant was considered. It allows direct contact between the coolant reservoir and the interferometer on a large area and therefore enables good spatial temperature uniformity. The working duration should achieve at least 24 h to cover flights with two legs and short stopovers without maintenance. The pitch agility of the gimbal, covering 112∘, makes the use of liquid coolants difficult.

Experiences from the precursor instruments MIPAS-B2 and MIPAS-STR have shown that operations during scientific measurement campaigns lead to additional requirements for the use of coolants: the coolant has to be refillable independently of the filling level or temperature of the instrument. This is required to handle delays in launch schedules. A coolant commonly available has to be chosen for operational considerations. The instrument's servicing ports have to be accessible through small hatches in the aircrafts fuselage to allow servicing without dismounting the cowling of the instrument bay. Finally, the capability to stabilize the temperature of the optic module during flight, on-ground and in the laboratory in a reproducible way is important for measurements and tests.

The GLORIA Instrument consists of the spectrometer in the three-axis gimbal, the blackbody calibration system, and the power electronics which are attached by a mounting frame to the hard-points in the fuselage of the aircraft by silicone vibration isolators. The central control computer can be positioned in the cabin – for flights on HALO – or in the bay – for flights on Geophysica.

Figure 1 shows the GLORIA instrument on the HALO aircraft. The gimbal has a shield with an opening in front of the entrance window of the spectrometer. An elongated opening in the belly pod allows observation of the horizon with different gimbal yaw angles.

The spectrometer dismounted from the gimbal is shown in Fig. 2. The total mass of the spectrometer is 48 kg of which the cooled fraction is about 21 kg. The outer dimensions are 670 by 500 by 420 mm.

The spectrometer structure can be split into the three distinct parts shown in Fig. 3: the housing, the optic module, and the cooling system. The housing is built upon the gimbal pitch plate and forms, with additional cover plates, a protection shell containing the insulation. Both the cover plates and the gimbal pitch plate also serve as a support to electronic units. The optic module is a sealed compartment containing all optical components, including the entrance window and the detector.

A cooling system based on dry ice has been chosen. Contamination of the interferometer volume with gaseous carbon dioxide has to be avoided because it is an atmospheric trace species measured by GLORIA. Therefore, the cooling system exhaust has to be diverted and the interferometer volume has to be properly sealed. For work in laboratories and during standby, an additional feedthrough with the possibility to inject cryogenic liquid nitrogen from a storage vessel for controlled cooling was added. This second coolant increases flexibility for ground operation. It is not suited for flights because the cooling system is not designed to store a liquid at 70 K.

Besides the cooling system, the use of vacuum insulation panels (VIP) is an important part of the thermal design. The highly efficient panels ensure low heat input and, therefore, allow a small coolant reservoir for the required holding time. With small insulation thickness due to the exceptionally low thermal conductivity, the volume required for the insulation is maintained small. The spatial temperature gradients in the optical system are reduced by using an all-aluminium structure with direct contact to the cooling tank and the use of the mentioned insulation. An insulation feedthrough assembly was designed to reduce local heat input along cables and cooling system tubes.

The entrance port for the incoming radiation breaks the thermal containment of the instrument. An assembly consisting of two air-spaced Germanium windows was preferred over a shutter system in order to meet the requirements in mass and size. This solution also enables the system to perform measurements in a humid environment at low altitudes or even on ground, supported by a heater for the outer entrance window to prevent condensation. Hygroscopic breathers allow pressure compensation between the closed cooled interferometer and the outside environment while protecting the optics from humidity and other contaminants.

The optical and electronic units are based on a modular design, which allows parallel maintenance of the subsystems and easy replacement during development or operation. It also facilitates improvement to specific subsystems. The reproducibility of the positioning for the optical components is ensured by dowel pin and hole or slot and key configurations. Electrical connections between interchangeable parts are realized by docking connectors.

Several PT1000 platinum resistance temperature sensors, four triaxial accelerometers, and two pressure sensors are mounted at key locations in and on the spectrometer. Together with the sensors on the gimbal and support structure, which also include a microphone, these sensors enable a detailed instrument characterization.

The base structure of the housing is the pitch level plate, working as a stiff chassis which holds further housings components as well as the optic module and the cooling system. The components are shown in Fig. 3. Both ends of the pitch level plate are mounted to the gimbals' pitch bearings and motor drive. The position of the pitch plate is such that the pitch rotation axis coincides with the spectrometer's center of gravity.

The housing cover consists of thin-walled plates with stiffening ribs machined out of high-strength aviation grade aluminium. The housing has only openings for the detector, for the entrance window, and for the insulation feedthrough assembly. The box protects the optic module and the VIP insulation.

The combined optic module and cooler are fixed to the pitch level plate by three glass-fibre reinforced plastic (GFRP) spacers. They allow a stiff connection while providing thermal insulation. A high degree of stiffness is important for this connection as it is critical for the line of sight stabilization performance. The position and orientation of the plates compensate different thermal expansion between optic module and pitch level plate.

The optic module comprises of the entrance window assembly, the optical components of the Michelson interferometer, the imaging lens system, and the detector. The optic module also holds a reference laser system for measuring the optical path difference. The optic module is described in detail below in this section and shown in an exploded view in Fig. 5.

In order to fulfil the thermal and logistical requirements described in Sect. 2.1 we decided to cool the optic module by a reservoir filled with dry ice. Charging solid CO2 into a coolant tank (Piesch et al., 1996; Shallman and Shallmann, 2006; Pint and Thom, 2001) requires a large opening which disrupts the insulation. Additionally handling, deliverance, and storage of dry ice are costly. We have found a way to avoid these disadvantages by in situ production of dry ice in the coolant tank by expansion of liquid CO2 (Büst, 2003, E.P. 1429093B1). With this method, it is possible to fill liquid CO2 via a thin non-insulated flexible transfer tube and to store the coolant easily as pressurized CO2 in gas cylinders. Carbon dioxide supplied in gas cylinders is commercially available. It is ideal for the application during campaigns even in remote areas as it has nearly unlimited storage time and poses little problem for transportation.

Cooling without dry-ice filling is also possible by injection of small amounts of liquid CO2 into the tank. In addition, it is possible to use liquid Nitrogen sprayed inside the coolant tank at controlled intervals. In flight, the cooling system operates with dry ice, whereas liquid N2 or CO2 is destined for laboratory and ground operations.

Figure 7 shows the coolant tank construction. A small polyethylene pipe with holes in different zones of the coolant tank is used for filling carbon dioxide. The positions of the holes are optimized to fill the tank volume homogenously with CO2 snow. A constantan heating wire inside the injection polyethylene pipe protects blocking caused by freezing during the filling procedure. It also allows opening frozen sections of the pipe for the recharging process. A PTFE tube with larger diameter and perforated with holes drains the sublimation gas. The injection pipe and its integrated heating wire are guided and protected inside the PTFE drainage tube.

The small holes in the injection pipe act as spray nozzles and in these sections the polyethylene pipe is led outside the PTFE drainage tube. The injection pipe path and size, as well as the number of nozzles, their orientation and their opening have been empirically optimized in several tests to maximize the filled CO2 snow mass.

The coolant tank has a total volume of 2.2 L and is machined out of a monolithic aluminium block by milling and spark-erosion. The coolant tank is designed for an overpressure of 10 bar to withstand the pressure forces exerted by the dry ice during filling and to operate the coolant tank at pressures up to 5 bar. The overpressure capability allows control of dry-ice temperatures from 173 K at 0.1 bar up to 216 K at 5 bar. The construction uses pillars to optimize the pressure resistance. The coolant tank is mounted directly below the optic module, which acts as cover for the tank; this configuration makes optimal use of conductive and convective cooling. The cooler is sealed by phenyl silicon O-rings (phenyl vinyl methyl quartz or PVMQ).

A separate inlet port on the bottom of the cooling tank enables access with an injection lance of a vacuum insulated transfer hose to spray liquid nitrogen in the tank. For safety reasons, two independent pressure relief valves and a bursting disc (not shown on Fig. 7) protect the cooler from overpressure.

A functional schematic drawing which shows the GLORIA thermal setup and cooling system with its different cooling possibilities is shown in Fig. 8. During dry-ice filling or for controlled cooling, the exhaust port at the pitch level plate hose has to be fully open to release the exhaust gas. After dry-ice charging, a smaller hose leading to the aircraft outlet is connected to the main exhaust port.

GLORIA has performed more than 100 flight hours on two different carriers. In this section results from three exemplary flights are presented: flight 1 on 11 December 2011 on the Geophysica in the polar winter and flights 8 and 19 on HALO at mid-latitudes on 28 August and on 25 September 2012.

Figure 10 shows temperature and altitude data of one flight with the Geophysica and one flight with HALO. The flight characteristics of both carriers differ in terms of altitude, duration, and velocity. The outside air temperature (OAT) during cruise varies from 190 to 230 K in function of the flight altitude and the meteorological conditions. Depending on the Mach number, the air flow at the aircraft warms up due to adiabatic compression; the so-called “ram rise” heating is about 17 K for Geophysica at Mach 0.66 and 28 K for HALO at Mach 0.81. The ram rise is included in the total air temperature (TAT). The air temperature measured in the Geophysica instrument bay reached 225 K (30 K above OAT) and in the HALO belly pod 253 K (33 K above OAT). The spectrometer housing with the pitch level plate is exposed to the ambient aircraft bay environment. During the first to second hour after takeoff, it cools down and stabilizes near the bay air temperature. The dry-ice cooled optic module operates at a temperature of about 210 K with a drift under 1 Kh-1 during the HALO flight. During the Geophysica flight, the drift was higher as the coolant tank was not charged to full capacity. The amount of coolant and the observed performances were sufficient for the typically short Geophysica flights. The optimized filling procedure with injection pipe heating and consequently a more complete fill level was later developed in preparation of the longer HALO flights.

Figure 11 displays the conditions observed during flight 8. Figure 11a illustrates flight altitude, the gimbal yaw angle, and the belly pod air temperature. The gimbal coordinate system is relative to the aircraft orientation; the gimbal yaw angle is varied during operation between -10 to 128∘, where at 0∘ the entrance window is pointing along the aircraft longitudinal axis. At -10 and 20∘ yaw angles, the window is respectively aligned with the cold and the warm blackbody for radiometric calibration. The CM measurements are carried out at around 86∘ gimbal yaw angle. During the DM measurements, which were performed between 11:00 to 13:30 UTC and between 16:00 to 16:20 UTC as shown in Fig. 11, the gimbal yaw angle changed rapidly: the yaw angle is varied every 3 s in 4∘ steps from 128 to 44∘ yaw angle to scan the horizon. This measurement mode can be clearly recognized by the variation of the gimbal yaw angle in Fig. 11a. The atmospheric measurements are periodically interrupted for internal calibration.

Figure 11b shows the outside static air pressure given by aircraft measurements, which goes down to 130 hPa. The pressure inside the belly pod generally follows the outside pressure with a slight fluctuation correlated with the gimbal yaw angle due to air streaming through the side opening. The differential pressure measurement shows the pressure difference between the inside of the optic module and the belly pod. The pressure fluctuations in the belly pod can also be seen in this differential pressure, causing variations down to -8 hPa. The plot also depicts the pressure variations caused by the breathing of the instrument during ascent, descent, and dives in the range of ±2 hPa. This is caused by flow resistance through the hygroscopic breathers.

Figure 11c presents the acoustic pressure in the belly pod and the vibrations measured at the pitch level plate. Shown are the RMS values for time periods of 1.2 s during DM measurements and 12 s during CM, respectively. Beside an obvious and expected correlation between the values and the gimbal yaw angle, the mean of these values depend clearly on the flight level. For lower cruise altitudes of 12.6 km at 10:30 UTC the acceleration mean level is about 2.3 ms-2 and goes down to 1.1 ms-2 at 15 km height at 15:00 UTC. The behavior of the acoustic pressure is similar. The dependency of the vibrations on the gimbal yaw angle, as well as on the flight altitude, and therefore air density, is explained by aerodynamic forces caused by the belly pod opening.

Figure 12 shows the spectrograms of the vibrations measured at the pitch level plate – the base of the spectrometer – and the acoustic pressure in the belly pod. The sampling rate of the sensors is 1 kHz. In addition to Fig. 11c, these plots depict the spectral contributions for different gimbal yaw angles. The microphone as well as the accelerometer clearly show the strongest vibration amplitudes between 200 and 250 Hz throughout the whole flight. This excitation probably originates aerodynamically at the belly pod side opening and appears at all gimbal yaw angles with varying amplitude. The highest value occurs at a gimbal yaw angle of 90∘. Analysis of different sensor data indicates that the aerodynamic excitation changes with the position of the gimbal's shield and its opening relative to the airflow.

In order to avoid the strong vibrations at 90∘, the CM measurements have been performed at a yaw angle of 86–87∘. Another clearly visible frequency band at 40–60 Hz in the accelerometer data is related to resonant frequencies of the gimbal and its shield. All other visible bands in both plots have considerably smaller amplitude.

A key indicator of the spectrometer function quality is the measurement of the moving mirror velocity using the reference laser system (Learner et al., 1996; Kimmig, 2001). It gives a good insight on the stiffness and damping characteristics of the system and on the impact of the vibrations on the IR measurement. The velocity measured with the reference laser system is directly proportional to the velocity of the moving cube corner and should ideally be constant. Unwanted velocity fluctuations are generated by fluctuations in the cube corner movement, but also by relative movements of the optical parts to each other. Such relative movements can e.g., be caused by independent vibrations of the optical components and the structure.

In order to quantify the stiffness of the spectrometer the inflight vibrations and the effects to the reference signal velocity are shown in Figs. 13 and 14. These pictures show the vibrations on the pitch level plate and the velocity signal recorded by the reference laser system over one interferogram at 15:18 UTC during flight 8 while the spectrometer is operated in CM. The environmental conditions remained stable at the time as shown in Fig. 11 in the previous section. The gimbal yaw angle varied only slightly in the range of 86.0 to 86.6∘ and the altitude was close to 15 km.

Figure 13 shows the power spectral density (PSD) of the vibration magnitude in the direction of scanner motion measured on the pitch level plate. This plot is derived from a single spectrum. Furthermore the integrated value of the PSD, the cumulative spectral power (CSP), is displayed. The peaks in the PSD and steps in the CSP curve show very clearly that the largest power contribution of the vibrations is found around 50 and 220 Hz. The red curve shows the PSD limit for single vibration peaks in operational conditions as described in Sect. 2.1 and summarized in Table 1. It can be noted that the excitation at 217 Hz exceeds this requirement. The integral of the vibration excitation, the so-called root mean square acceleration Grms, can be derived out of the CSP to 0.87 ms-2. This is lower than the requirement which was set to 4 ms-2.

The velocity signal detected by the reference laser for the same time interval shows a standard deviation of 9 %, exceeding the target value of 5 % given in Table 1. Figure 14 shows the amplitude spectrum of the velocity. The largest contribution to the velocity fluctuations is again around 217 Hz in accordance with the measured vibrations. Furthermore, there are sharp peaks which are found at multiples of 6.41 Hz. They are caused by the leadscrew of the scanner which has a rotation rate of 6.41 Hz. The second largest peak, which can be found at 128.2 Hz, is generated by the cogging torque of the twenty electrical coils in the stator of the drive.

The main part of the velocity fluctuations in the interferometer is caused by forced vibrations of the optical components induced by the strong acoustic and mechanical excitations around 220 Hz as shown in Fig. 12. The strong velocity variations around 220 Hz lead to so-called “ghost”-signatures in the spectra (Learner et al., 1996; Kimmig, 2001). The frequency distribution of the velocity variations is broad in comparison to individual lines, leading to a smearing of ghost lines, which makes them indiscernible from noise. Only broad band signatures like the ozone band produce visible ghost patterns. A ghost amplitude of 1–2 % is estimated from these broad patterns. Spectral regions contaminated with these broad band signatures are used with great care in the retrieval of atmospheric parameters. A precise quantification of the impact on temperature and trace gas profiles has not been performed yet.

The temperatures measured inside and outside the spectrometer and the flight altitude are shown in Fig. 15 for flight 19 on the HALO aircraft. The same flight was already discussed in Sect. 3.1. In Fig. 15b the temperatures inside the optic module and the coolant tank are shown. The sensor on the coolant tank, identified as “coolant tank” is the farthest away from the central body and shows the lowest temperature of the cooling system. The sensor “coolant tank at central body” is located at the interface between the central body and the coolant tank. It shows the highest temperature of the cooling system and the lowest temperature of the central block as expected. A third sensor, positioned in the IR objective, shows the highest measured temperatures inside the optic module with the exception of the entrance and detector windows. The temperature of the latter is influenced by the environmental pressure, window heating and indirect heating by electronic components. Both window temperatures are shown in Fig. 15d.

The cooler had been filled with dry ice approximately 2 h before takeoff. During flight preparation, visualized as a constant GPS height near 0 m in Fig. 15a, the coolant tank warmed up rapidly with about 4 Kh-1. During ascent, the coolant tank temperature drops again due to the ambient pressure decrease. In this flight phase, the increase rate of the optic module temperature becomes smaller before stabilizing at the typical inflight increase rate of approximately 1 Kh-1.

The inflight temperature of the optic module remains below 220 K. The temperature difference between the coldest and warmest components in the optic module is below 4 K while the coolant tank is at most 3 K colder than the optic module.

The belly pod and the cooling system pressure are shown in Fig. 15c. The pressure in the unpressurized coolant tank follows the aircrafts ambient pressure. Due to pressure decrease with increasing altitude the sublimation temperature of the dry ice drops. This leads to a higher sublimation rate and thus to a higher overpressure of 50 hPa. The overpressure diminished to approximately 22 hPa during flight. The higher overpressure observed in the cooling tank during flight compared to ground operation is also caused by the reduced gas density at flight level requiring a higher volume for the same mass flow of CO2.

The temperature drifts of the entrance and detector windows are required to be below 2 Kh-1, as for all optical components. For the windows, this requirement is technically challenging due to the exposed position at the boundary of the cold system with unavoidable gaps in the thermal insulation. As depicted in Fig. 15d, temperatures of outer and inner windows of the entrance window assembly are strongly influenced by the aircraft environment. The outer window is also strongly affected by its heating system which is needed to prevent condensation. The temperature gradient inside the entrance window assembly is approximately 35 K inflight without heating and 80 K on ground while heating.

The inner window temperature drift is with 1 Kh-1 similar to the other optical components and thus in accordance with the requirements. Stronger temperature drifts can be observed while the outer window heater is operating. These drifts do not affect the performance of the system as no scientific measurements are taken while the heater is active.

The outer window temperature is strongly affected by the environmental conditions and the outside airflow. This leads to temperature drifts that exceed the requirements. For instance between 10:00 and 11:00 UTC a drift of -10 Kh-1 is observed correlated with the belly pod temperature. This seems to indicate an atmospheric condition change. Other sudden changes in the outer window temperature exceeding 10 Kh-1 can also be observed. These are mostly correlated with the gimbal yaw angle changes, indicating an influence of the outer air flow around the window.

The detector window temperature shown in Fig. 15d varies smoothly without sudden changes during flight. The drift of 5 Kh-1 is significantly smaller than for the outer entrance window but still exceeds the requirement.