General Overview
Online Documents
Access to Data
Daily Observation History
Photo Gallery
Related Sites
Rules of the Road
TIDI Web Site
The Final Word
HRDI Instrument Operations

HRDI Instrument Operations

Theory of Operation

A typical horizontal wind speed of 10 m/s for the middle atmosphere causes a Doppler shift of ~2 x 10-5 nm (4 x 10-4 cm-1). This small shift requires a high-resolution spectrometer with high sensitivity and very good stability. The Fabry-Perot was chosen because of its large light gathering power and high resolution (P. Jacquinot, J. Opt. Soc. Am. 44, 761-765, 1954). Since complete descriptions of the Fabry-Perot interferometer have been given elsewhere (G. Hernandez, Cambridge University Press, 1986; J.M. Vaugh, Hilgar Press, 1989) only an outline is provided here. The Fabry-Perot interferometer is a deceptively simple device that consists of two transparent plates which are extremely flat (up to lambda/200) and parallel (tilt < 2 x 10-11 rad). The inside surfaces are coated with a relatively high-reflectivity coating (~80 to ~95%) that is usually a multilayer dielectric stack. The device is a resonating cavity which forms Haidinger fringes at an infinite distance from the etalon. The fringes are brought back to some finite distance by the use of imaging optics. The transmittance for an ideal Fabry-Perot etalon is given by the Airy function

Eq. 1 (1)
where R is the plate reflectivity, T is the transmittance, M is the order of interference which in turn is given by
Eq. 2 (2)
where is the index of refraction of the gap (here =1), t is the plate separation, is the wavenumber of the light and is the angle from normal incidence. The peak transmittance of such an ideal etalon is unity and occurs whenever the order of interference is an integer. The passbands can occur for infinite combinations of µ, t, theta, and nu. If we vary only the wavenumber, the separation between peaks at normal incidence is the free spectral range:
Eq. 3 (3)
If M is an integer for the conditions µ=1 (vacuum), t=t0, theta=theta0=0, and nu=nu0, then in order for M to remain an integer the following condition must be satisfied:
Eq. 4 (4)
This equation illustrates that for space flight applications there are two practical methods for scanning a Fabry-Perot interferometer in wavenumber: 1) by looking at different angles through the interferometer, which amounts to viewing different spatial locations at the focal plane of the imaging optics, and 2) by varying the thickness of the etalon gap. Both of these techniques are utilized by HRDI.

Fig. 3. Effect of multiple etalons on the instrument passband. (a) HRE and filter only, (b) HRE, MRE and filter, (c) HRE, MRE, LRE, and filter.

The Airy function describes a periodic response, as shown in Fig. 3(a). The instrument response is adequate to observe narrow spectral features with a very small background, so that light will be transmitted only through one order. This requirement is satisfied in the upper atmosphere where widely spaced emission lines are viewed against a dim background. The line of interest is isolated by adding a low resolution interference filter. In the lower atmosphere (below the stratopause), where HRDI observes absorption lines in scattered sunlight, a single etalon would transmit light through all orders and the signal would be overwhelmed by this parasitic light. It is necessary to use a multiple-etalon system in order to reduce this unwanted light (J.E. Mack et al. Appl. Opt. 2, 873, 1963; D.P. McNutt, J. Opt. Soc. Amer. 55, 288, 1965). If the thicknesses of the etalons are properly chosen (W.R. Skinner et al., Appl. Opt. 26, 2817, 1987), then it is possible to reduce the "parasitic" light of the instrument to an acceptable level. The effect of adding a second and third etalon with an interference filter to the system is shown in Figs. 3(b) and 3(c). The HRDI instrument employs one etalon with a fixed gap of 1.0 cm which is scanned by observing different angles through the etalon using a multi-channel detector. The gaps of the second and third etalons (0.186 and 0.025 cm thickness) can be varied slightly (~1000 nm) with piezoelectric materials so that they may be adjusted into resonance with any order of the fixed, high-resolution etalon. This allows the instrument to scan in wavelength over a range which is limited only by the interference filter.

Instrument Tuning

Equation (2) is the basis for developing the equations to properly tune the instrument. It is obviously necessary that all three etalons be in resonance at the same wavelength for the instrument to operate properly. The HRE has fixed posts, but, since more than a free spectral range is projected on the detector, all wavelengths are present there. The MRE and LRE have piezoelectric posts that allow their gaps to be adjusted. The tuning algorithm adjusts the etalons from a known reference position that is close to the desired wavelength. This is carried out in two steps: first, the angle through the HRE which will cause the desired line to be in resonance is determined. Second, given the wavelength and angle, the voltages that must be applied to each of the posts of the MRE and LRE which bring these etalons into alignment are found. In this section the necessary steps are outlined. The actual algorithm is somewhat more complicated since thermal effects and phase dispersion of the coatings must be included. Since the desired lines for HRDI are always close to a reference line (the system was designed for that to be so), and since the HRDI instrument operates in a vacuum (mu=1), the order of interference can be written

Eq. 5 (5)
where the subscript o indicates values of t and nu that cause m to be an integer when theta=0. In the development that follows a subscript r represents the conditions that match a reference condition. These conditions are determined from calibration data. No subscript indicates the state of the instrument that is necessary for tuning to wavenumber nu. The HRE has no thickness adjustment capability, so the difference in orders from the desired to reference condition is
Eq. 6 (6)
In order to find theta-squared it is important to note that a change of an integer number of orders results in the same instrument state, and so the first step is to find the change in orders, M, assuming for the moment that :
Eq. 7 (7)
where the free spectral range is that of the HRE. This is permissible since the change in order due to the angle change is always less than one order. The integer number of orders are removed and Eq. (7) inverted to find theta-squared:
Eq. 8 (8)
where int() denotes the closest integer. This sets the condition for resonance for the HRE. For the MRE and LRE the procedure is similar, except now t is found given nu and theta-squared. For these etalons values for Delta M are defined as
Eq. 9 (9)
where the free spectral range now refers to either the MRE or LRE. This equation is inverted to give
Eq. 10 (10)
The etalons are controlled by applying voltage to the posts to change their length. Extensive calibrations have determined that a second order polynomial is sufficient to represent thickness changes:
Eq. 11 (11)
where V is a command voltage and alpha1 and alpha2 are constants that are specific to each post. Equation (10) then becomes
Eq. 12 (12)
This quadratic function is readily solved for V.

These equations show that there are several important parameters needed to properly tune the interferometer. First, the reference conditions must be well known. This amounts to six voltages and one angle for each reference condition. Usually, the instrument is adjusted to maximize the signal from an atomic emission line of a spectral lamp. These lines are quite stable and since lamps are part of the on-board calibration equipment they can be used to recalibrate these parameters in flight. Second, the free spectral range and the voltage constants alpha1 and alpha2 are required. In addition, the thermal drift terms (which have been excluded in this simplified analysis) must be determined. It has been necessary to redetermine the reference conditions several times since launch. Some of these were required to compensate for the MRE post failures which are described below. During the first major solar array problem the instrument temperature became quite cold (<-20° C), and after the instrument was turned back on a noticeable shift was observed. The other constants have remained stable and have required no adjustment.

Table 1. Summary of HRDI parameters.

Primary diameter 17.78 cm
Field of view 0.12° x 1.369° (wide), 0.12° x 0.12° (narrow)
Transmittance ~25%
Number of channels 32
Integration period 0.096 s
Off period 0.032 s
Repetition period 0.128 s
Effective quantum efficiency 0.053 (631 nm), 0.043 (691 nm)
  0.037 (725 nm), 0.030 (768 nm)
Dark signal 0.63 counts/0.096 s/channel @ -10°C
  1.47 counts/0.096 s/channel @ -5°C
Plate diameter 13.2 cm
Coated diameter 9.4 cm
Plate material Spectrosil-B
Coating materials ZnS-ThF4
Post type fixed (HRE), piezoelectric (MRE and HRE)
Post material zerodur (HRE), PZT-5H + quartz (MRE and LRE)
Spectral range 630-770 nm
Number of filters 13
Filter bandwidth 0.8 - 1.5 nm
Spectral resolution (HWHH) ~0.020 cm-1
Continuum sensitivity ~1 x 10-4 counts/0.096 s/R/cm-1
Thermal drift -23.0 m/s/°C (762 nm), -36.5 m/s/°C (690 nm), -26.5 m/s/°C (630 nm)

Instrument Performance

Instrument Problems

The HRDI instrument has been collecting scientific data since November 2, 1991. There have been several periods when the instrument was either off or not in a condition to make useful measurements. These are summarized in Table 2. There have been two occasions when there have been hardware problems that have temporarily prevented scientific data collection. Fortunately, both had work-arounds that allowed complete recovery. Both problems occurred in the electronics that operate the piezoelectric posts on the medium resolution etalon (MRE). The first failure occurred in the drive circuitry. In order to minimize the time needed to adjust the etalon gap there are two circuits used to drive the PZT; one to lengthen the post, and the other to shorten it. One of these failed, leaving positive control in only one direction. In order to move the post in the other direction sufficient time (~4 s) must be allowed for the post to relax to the desired position. The solution to this problem was to examine the applied voltages for each of the lines in the scientific mode and order them such that sequencing through desired lines always has positive control. The voltage was reset during the slew from front to back or back to front viewing. The slew takes approximately six seconds which is sufficient for the post to relax. The second problem occurred with MRE post 1. A digital-to-analog converter (D-A) which converts the computer calculated voltage to the analog value had a one-bit failure. The D-A is a 16-bit converter with 12 bits varying the voltage by a factor of 2 for each bit and 4 bits which select a voltage range. One of these latter bits failed. This was fortunate, since it allows the full resolution of the device to be retained. The instrument was designed to have a dynamic range of 2 x 103 nm (~6 orders of motion), which is far more than necessary. In principle, only 1 order of motion would be necessary. The additional orders were included to allow for paralleling the etalon and to compensate for temperature variations. The quality of the constructed etalon was very good, so virtually none of the additional dynamic range was required. This could be used to shift etalon movement from the dead region on the D-A to one that was still active. The instrument software was modified to do this and full scientific capability was restored. Neither of these work-arounds would have been possible without the flexibility provided by the instrument computer.

Table 2. HRDI timeline.


September 12, 1991 1 UARS launch
October 1, 1991 20 HRDI activation
November 2, 1991 52 Scientific observations began
March 30-April 26, 1992 201-228 Instrument data of limited value due to MRE post 2 partial failure
June 2-July 22, 1992 265-301 Instrument off due to spacecraft solar array problem
February 2-12, 1993 510-520 Instrument data of limited value due to MRE post 1 partial failure
February 15, 1993 523 Calibrations performed on nightside only
June 14, 1993 642 Begin temperature feedback
August 2, 1993 691 Improved yaw offset calibration
August 2, 1993 691 Begin improved temperature feedback
August 4-9, 1993 694-698 Instrument off due to spacecraft solar array problem
September 17-23, 1993 737-743 Instrument off due to spacecraft solar array problem
October 2-7, 1993 752-757 Instrument off due to spacecraft solar array problem
October 22-26,1993 772-776 Instrument off due to spacecraft solar array problem
December 8-9, 1993 819-820 Instrument off due to low battery power
March 15, 1994 916 Reduced calibration frequency began
March 24-April 6, 1995 1290-1303 Instrument off due to spacecraft solar array problem
April 17, 1995 to present 1314- Instrument off due to spacecraft solar array problem

Thermal control

Initially the temperature of HRDI was controlled by trying to hold the radiator to a specific temperature. This setpoint is adjustable in steps of 2.5° C. If the radiator temperature fell below the radiator setpoint, then heaters on the radiator were turned on and the power supplied to the heaters was proportional to the difference between the desired and actual temperature. Once the setpoint was exceeded they would be turned off. The amplitude of the temperature variation of the interferometer was damped by the long path from the radiator. Examination of the interferometer temperature data showed that the high frequency cycling of the radiator was also effectively damped. As the angle between the orbit plane and the sun (the beta angle) varied during the spacecraft's monthly precession, heat load into the interferometer changed. This resulted in a low frequency variation of the instrument temperature (Fig. 4) of greater than 1.5° C. The interferometer drifts at a rate of ~30 m/s/° C, this amounts to about a 45 m/s variation. In addition, as the radiator ages the emissivity of the radiator drops, and it becomes increasingly more difficult to keep the instrument cold. This becomes significant during the solstice periods when the spacecraft can be in continuous daylight for several days and the solar beta angle can approach 80°. In this condition the heaters are off, and the interferometer temperature rises by about 1° C. After about 500 days of operation it was realized that the use of the instrument computer could significantly improve the temperature stability. The temperature of the etalons was monitored and when they rose above the desired temperature the radiator setpoint was lowered. This forced the radiator to lower its temperature and consequently cool the interferometer. The radiator setpoint was then raised when the etalon temperature fell below the desired temperature. This method of temperature control was used from June 14 to August 2, 1993 (642 to 691 days after launch). This was not an optimum setup and a positive feedback resulted that caused the interferometer to oscillate at high frequency. On August 2, 1993 (day 691), the algorithm was modified slightly so that the temperature of the optical bench (see Fig. 2b) was controlled instead of that of the interferometer. This is an intermediate point between the radiator and the interferometer and effectively eliminates the high frequency components present in the first control attempt, and significantly reduces the low frequency components that were present when no computer control was attempted. The temperature is held to within a few tenth's of a degree, except during solstice periods when the temperature rises about 1° C.

Fig. 4. Temperature of the high resolution etalon during the first 1000 days of UARS operation.

In-flight calibrations

There are four different types of in-flight calibrations: 1) high frequency calibration checks which are performed every 1-2 days; 2) low frequency periodic detailed calibrations (approximately every 36 days); 3) instrument pointing; and 4) one of a kind or infrequent special calibrations. Here, the discussion will be limited to the calibration checks, which provide most of the information on the interferometer performance, and the instrument pointing.

The frequent calibrations were initially performed daily. Nine different calibration sequences were defined that took from five to ten minutes to complete. They were cycled through every third terminator crossing. Terminator crossings were chosen because the HRDI data is most difficult to interpret when the solar zenith angle is near 90°. Therefore it was acceptable to give up some of that data. Every third terminator crossing was selected since this would give alternating night-to-day and day-to-night crossings, preventing any bias, in both scientific and calibration data. After about 500 days it was realized that calibrations performed in the night-to-day portions were causing a small, but significant, loss in daytime scientific data. Since there was no noticeable difference between the night-to-day and day-to-night calibrations, it was decided to perform the calibrations every second day-to-night terminator crossing starting February 15, 1993 (523 days after launch). At the same time, some calibrations that were providing little useful information were eliminated and some new ones were added. These new calibrations monitor the performance of the damaged MRE etalon. By 900 days after launch it was realized that one of the spectral lamps containing a mixture of helium, argon, and krypton (HAK) was significantly aging (it was only coming on a fraction of the time it was commanded) and the instrument was changing slowly enough so that the calibration frequency could be reduced. Calibrations were then performed every fourth day-to-night crossing so that it takes about two days to complete a cycle. Table 3 describes the calibrations performed in this last period.

Table 3. Frequent HRDI Calibrations.

Segment Lamp Purpose of Test

all none detector dark count
1 incandescent interferometer sensitivity
2 neon instrument drift in the 14427 cm-1 (B band) region
3 HAK instrument drift in the 13151 cm-1 (A band) region
4 neon instrument drift in the 15856 cm-1 (gamma band) region
5 neon MRE, LRE finesses HRE & lamp width
6 none PLT post response test
7 none MRE electronics test

HAK = helium, argon, krypton gas mixture

LRE and MRE finesse

The finesse of the LRE and MRE can be determined by scanning the etalons over the spectral lines. Figure 5 shows the finesses for these etalons measured at 692.9 nm for the first 1000 days of operations. The LRE shows a slow drift towards higher values, starting at about 11.5 and is about 12.25 by day 1000. In contrast, the MRE has decreased from about 12.5 to approximately 12. The low values of the MRE finesse correspond to occasions after the MRE post problems had arisen and before the temperature control was implemented.

Fig. 5. Etalon finesse as a function of time as determined from observations of the neon line at 629 nm. (a) LRE and (b) MRE.

HRE width

The spectrum of a spectral lamp as viewed by the IPD provides some information on the width of the HRE etalon. Since the line width of the lamp is a significant fraction of the etalon width, only a relative change in the etalon passband can be determined. Figure 6 shows the width of the neon line at 692.9 nm as a function of time. An oscillation in the width of ~0.02 cm-1 is quite clear until approximately 600 days after launch. This corresponds to the time that the interferometer temperature control was improved and indicates that the oscillation was a temperature effect.

Fig. 6. Spectral width of combined HRE and neon line at 692.9 nm for the first 1000 days of UARS operation.

Instrument sensitivity

An incandescent lamp is used to look for changes in the sensitivity or transmittance of the interferometer. The lamp is located in the calibration deck and therefore does not determine if any changes are occurring in the telescope or transfer optics (see Figure 1a). The interferometer is tuned to a wavelength that is located in the center of the filter passband. Figure 7 shows the counts from channel 1 when the instrument is tuned to the center of a filter located at 692.3 nm. The signal varies by less than 10% throughout the mission, but shows a slight shift upward at approximately 300 days after launch. This was caused by a change in the MRE and LRE reference values at this time which was required by the cold period the interferometer encountered when the instrument was turned off (see Table 2). The signal continued to increase until about 700 days after launch when a decrease occurred. The reason for this is currently unclear, although it is probably due to either the LRE or MRE becoming slightly misaligned with the HRE.

Reference location

The position of a reference line on the detector provides information on the instrument drift and zero velocity information and is particularly important for accurate wind recovery (Burrage et al., Geophys. Res. Lett. 20, 1259, 1993; Burrage et al., SPIE 2266, 294, 1994). The location of the spectral lamp line in each of the three main spectral regions (630, 690, and 762 nm) is regularly monitored for these purposes. Figure 8 shows the position for the neon line at 629.9 nm as a function of time. In figure 8a the raw data is presented. Oscillations of 0.1 channels are readily observed. This corresponds to a shift of approximately 40 m/s. Notice after the improved temperature control was implemented the oscillation becomes much less and only a slow, long term drift remains. In Figure 8(b) the temporal effects have been removed, leaving just the instrument long term drift. The instrument drifted quickly at first, which is probably a result of instrument outgassing. A shift occurred during the time of the first solar array problem, probably because of the very low temperatures the instrument encountered. In Figure 8(c) the temporal drift has been removed and the result is a straight line with a scatter that is equivalent to an rms value of ~2 m/s. Notice the increase in scatter during the time of the first temperature control scheme, between days 642 and 691.

Fig. 7. Signal from incandescent lamp for the first 1000 days of UARS operation.

Fig. 8. a) Location in channel numbers of the neon line at 629.9 nm as a function of time: (a) raw data, (b) thermal effects removed, and (c) thermal and temporal effects removed.

HRDI Home Page