Application of NASA core Flight System to Telescope Control Software for 2017 Total Solar Eclipse Observation

The DICE instrument consists of imaging and mechanism parts to obtain scientific data between 1 and 14 R_Sun (0227 and 373) from the total solar eclipse observations (Figure 1). To increase the signal-to-noise ratio (SNR) of the corona during the total solar eclipse, two instruments were built to obtain as much data as possible.

2.1.1. Imaging Part

2.1.2. Mechanism Part

The DICE filter wheel system operates by rotating the bandpass filter wheel and polarizer at 90^O and 60^O in one direction, respectively. The filter wheel uses spur gears with a gear ratio of 9.75 or 9.85. The motor, MOOG Animatics' SM2316D-PLS2, which is a DC servo type with an encoder resolution, can perform 4000 counts per revolution and provide RS-232 communication. The filter wheel is equipped with a limit sensor for position initialization. The polarization filter wheel shows polarization brightness within a 1% deviation in the repeatability test (Cho et al. 2020).

Zoom In Zoom Out Reset image size

An observation system was built using a pointing system carrying a science payload and a power distribution unit (PDU) with batteries (Figure 2). The science payload mounts two DICE instruments and control computers, and a wireless network switches on the mechanical structure. The total payload weight is approximately 12 kg. An operator can adjust the sight of the science payload with the intended target of the pointing system and handle DICE instruments using a remote computer via a wireless network switch.

Zoom In Zoom Out Reset image size

We utilized a robotic equatorial mount for slewing and manual adjustment to the center of the images using the hand controller. Because of the large field of view of the DICE instrument, we can aim for the pointing system to find the Sun in the images. In addition, a tracking drift is acceptable during a short period of totality if the polar alignment of the pointing system is performed using GPS. The mount uses Bisque's Paramount MYT software, which can load a payload of a maximum of 45 kg and track an astronomical object with an accuracy of 1''. Our aim was to develop software for controlling the DICE instrument, software prototypes of the scientific balloon mission BITSE operated by Wallops Arc-Second Pointer (WASP), and an independent pointing system (Gopalswamy et al. 2021).

The control computer controls the DICE camera, filter wheels, and handles commands and data. The computer used was the ASUS UN65U model, which is equipped with an Intel 7th generation i5-7200U (2.5 GHz) processor, 4 GB SDRAM, and 256 GB SSD. The computer was connected to the camera and filter wheels through USB ports, using an RS232-USB converter for the filter wheel. The two computers were connected to the network hub so that the remote computer could communicate with them.

Because the observation of the total solar eclipse takes place outdoors, the PDU was built to provide stable power for electronics. The PDU can have a 9–20V DC power input port and 4 × 5V, 2 × 12V, and 2 × 48V power output ports. For the primary power source, the PDU used two 12V car batteries. The PDU provides 5V to the network hub, 12V to the cameras and control computers, and 48V to the filter wheels and mount.

Two dedicated embedded systems control each of the two instruments. In addition, they used two sets of PDUs and batteries. This fail-safe configuration requires more electrical power and is heavier than a single configuration for controlling the two instruments. The observation system remained safe for a short observation time in the total solar eclipse. The network router and pointing system are the only single points of failure in the observation system. The impact of pointing system failure is minimized because of the large field of view. We can mitigate the risks of network failures using a scheduler in embedded software to run an observation program automatically at the absolute time for totality. The system time of the embedded computers was synchronized using the network time protocol system software with a remote control computer. If a single computer is fault tolerant and has acceptable performance, it can handle multiple instruments simultaneously to reduce the weight load and electrical power. In this case, the embedded software should control the two instruments in multiple processes, multiple threads, or polled I/O.

The three critical elements of the cFS are the dynamic runtime environment, layered software architecture, and component-based design, which are suitable for most embedded software development. The cFS helps develop platform-independent software through the functional abstraction of hardware and operating systems and provides an API for its core services. McComas et al. (2016) show a layered software architecture that consists of stacked abstraction libraries, the core flight executive (cFE), and applications that are dynamic link libraries. The cFS can run applications with plug-and-play capabilities in parallel and support their communications on a software bus using the message queue. Developers can cooperate for development and test process on application basis using the cFS. Loosely coupled application modules have the advantage of being selectively applied to other missions.

cFS is based on codes that have been verified through past NASA space missions. This can be an additional advantage of telescope systems. Time-critical systems are easy to build using cFS, as they support real-time systems. Based on the C language, the cFS can have lightweight executable files, control most hardware drivers, and efficiently use computational resources; thus, it is suitable for building a portable observation environment with limited power. Because the operating principles of the telescope system on the ground and in space are similar, applications developed with the cFS can be reused when the ground telescope system is applied to space.

The operating system abstraction layer (OSAL) provides APIs that abstract operating system functions, such as multitasking, inter-process communication (IPC), interrupt and exception handling, and file systems. The platform support package (PSP) is a software library that defines all toolchain settings required when building the cFS in a specific operating system and hardware platform. The PSP is implemented using the board support package (BSP) of a hardware platform and contains cFS startup code.

The cFE provides five fundamental services for the operation of modular applications together with the abstraction layer. The cFE comprises a software bus (SB), executive service (ES), event service (EVS), time service (TIME), and table service (TBL). The cFE is middleware such that the applications can be reused in other running environments. SB uses a queue for messaging to support the publish/subscribe messaging patterns to communicate with each other. The message uses the Consultative Committee for Space Data Systems (CCSDS) standard space protocol (CCSDS 2020). The ES manages a task lifecycle in a multithreading environment. EVS handles the prioritized event logs of tasks. TIME provides functions for time calculation and handles external time tone and information to synchronize the system time. TBL manages the data blocks of the tasks in a logical unit called a table. It can dump the data blocks in the memory into a file and load the file into memory.

The application layer includes reusable open-source applications developed by NASA. NASA applications have been used in various space missions and their performance and reliability have been proven (Wilmot 2021). The cFS with NASA applications can be a baseline for building a software system that satisfies mission requirements because NASA applications have the advantage of essential functions commonly required by flight software. The most critical role of flight software is data processing (DS: data storage, HK: housekeeping, MD: memory dwell, CF: CCSDS file delivery protocol), automated operation (SC: stored command, LC: limit checker, SCH: scheduler), resource management (FM: file manager, MM: memory manager), and safety (CS: check sum, HS: health and safety).

NASA is dedicated to systematic software reuse for the cFS, which includes technical documents, such as requirements, designs, tests, and user guides, as well as reliable source codes. Technical issues regarding the cFS can be resolved through community effort. NASA continues to use the cFS for various space missions to ensure software quality. Most cFS applications include automated test scripts to perform regression tests according to application changes. However, the testing and operation can be performed by ASIST (Advanced Spacecraft Integration and System Test, http://sed.gsfc.nasa.gov/etd/583/tech/asist) and ITOS (Integrated Test and Operation System, http://itos.gsfc.nasa.gov), which are for US government purposes only. McComas (2018) distributed OpenSatKit along with COSMOS, an open-source software for testing and operating cFS, which is a great alternative if ASIST and ITOS are difficult to use.

The DICE-embedded software was developed based on the version 6.5a of the cFS. The cFS includes OSAL and PSP, which support x86 processor-based computers and Linux operating systems. An Ubuntu Server 16.04 LTS 32 bit was used as the operating system. The embedded software was set larger for both the maximum buffer size of the software bus and the block size of the executive service to utilize the memory of the control computer.

We built embedded software integrated with cFS mission-specific and reusable applications. The mission applications include camera control (CAM), filter wheel control (WHL), data acquisition and processing (DAP), and observation control (OBS) applications. The reusable applications consist of a scheduler (SCH) application that generates periodic commands in a unit of seconds, and the command ingest (CI) and telemetry output (TO) applications that process remote command and telemetry on the software bus through Ethernet. Figure 3 shows the activities of the applications for the data distribution and observation control of the embedded software.

Zoom In Zoom Out Reset image size

The applications are dynamically loaded by the cFE executive service and run on a per-thread basis. The software bus distributes commands and telemetry to the data pipes of the applications in a publish-subscribe messaging pattern. Applications can process messages queuing on a specific data pipe sequentially.

The CI and TO transfer commands and telemetry through the Ethernet network, so the DICE-embedded software can be remotely monitored and controlled over the network. The SCH generates a periodic command at 1 s intervals. This is necessary to perform time-synchronized functions of the applications and generate periodic status information, such as housekeeping telemetry.

The WHL and CAM can independently control the filter wheel and camera. The CAM stores image frames in memory space using the memory pool function of the cFE executive service. The OBS manipulates WHL and CAM using messages with the parameters of each observation sequence according to an observation program file, which consists of the filter wheel position, camera exposure time, and integration limit for image frames. The OBS receives telemetry of the current filter wheel status and camera status, constructs the header of the scientific data, and then generates science data by combining the header with an image frame. The DAP receives the science data, performs image processing, and writes the data as files for onboard storage. After DAP finishes writing the file, it returns the memory space where the image frame is stored.

3.2.1. Filter Wheel Control

3.2.2. Camera Control

3.2.3. Data Acquisition and Processing

DAP performs image processing, real-time image transfer, and science data-writing for local storage. These functions are selectively activated via remote commands. They are useful for aligning the optical system to a target object on a remote system. DAP processes messages containing image frames generated by CAM and releases a block from the memory pool after a science file has been stored. The memory pool can hold image frames as much as its own size; therefore, DAP and CAM can be synchronized to use memory resources without conflict. The output file format is memory-dumped data with a science header, including filter wheel positions, camera exposure time, and observation time. The Python script for post-processing is used to convert them to FITS files using Astropy (http://www.astropy.org), a community-developed core Python package for astronomy (The Astropy Collaboration et al. 2013, 2018).

Image processing can calculate image statistical information (pixel values of 1% of top and bottom and middle) and the optimized exposure time is calculated using the following equation:

$$\begin{eqnarray*}&&{T}_{\mathrm{opt}}={T}_{\exp }\displaystyle \frac{{I}_{\mathrm{opt}}}{{I}_{\exp }},\end{eqnarray*}$$

I_opt is the optimized target intensity and T_exp and I_exp are the exposure time and intensity of the source image, respectively. I_opt and I_exp generally use half of the maximum pixel value of a camera and the average intensity of a source image, respectively; however, I_exp uses the pixel value of the top 1% for solar observation. The real-time images were reduced for display on a remote computer of a suitable size by down-sampling the original image to 256 × 256 × 8 bits.

3.2.4. Observation Manager

The control software is implemented by referring to the ground system, which is a utility in the cFS software package. The control software is written in Python, which can run on various operating systems and has a graphical user interface (GUI) implemented with the QT framework. The control software can transmit commands by calling the command utility with a character user interface (CUI), another utility of the cFS software package compiled with C language in a remote software environment.

The control software consists of backends that communicate via an Ethernet network with the DICE-embedded software and frontends that interface with the users (Figure 4). One copy of the backends and two of the frontends were run on a remote computer to control each embedded software. The telemetry system receives telemetry via UDP and relays it to the telemetry router, which distributes telemetry receivers via TCP/IP. The telemetry receiver processes telemetry data to serialize the data objects used in the ground system by a telemetry factory, which creates the data objects in the definition of data formats. The ground system handles a GUI that monitors telemetry and generates remote commands by calling the command sender. The ground system communicates with an embedded computer that connects one camera and two filter wheels.

Zoom In Zoom Out Reset image size

Figure 5 shows an example of the user interface of the control software. It is designed to arrange the commands necessary for observing the total solar eclipse and determine the observation progress. It also includes functions of manual commanding and a log trigger for integration and test purposes. The screen layout is arranged for content related to commands to appear on the left and telemetry on the right. The functions of each part are summarized as follows:

• 1.  
  A: Network setting for connection with the DICE embedded software
• 2.  
  B: Manual command transmission
• 3.  
  C: Time synchronization
• 4.  
  D: Bandpass filter control
• 5.  
  E: Polarization filter control
• 6.  
  F: Camera control
• 7.  
  G: Observation control
• 8.  
  H: Telemetry transmission control
• 9.  
  I: Data processing and storage control and real-time image display
• 10.  
  J: Display event message
• 11.  
  K: Display housekeeping information
• 12.  
  L: Log control for performance measurement

Zoom In Zoom Out Reset image size

We performed functional tests using DICE software while integrating the observation system in the laboratory. The bandpass filter wheel was visually inspected to ensure that it transitioned to the expected position for each filter. The polarization filter wheel was verified by repeatedly measuring the polarization angles in rotation at a single point (Cho et al. 2020). The camera control function was verified using multiple image files. We executed a test observation program file to verify the observation automation function during normal operation for two hours. We then verified the observation functionality in a similar field configuration of the total solar eclipse using a DICE system, which was mounted on a pointing system, powered by an external power source, and controlled using a remote computer. The DICE system in the field configuration was verified by capturing images of the full Moon, which is known to have a similar corona brightness as that of a total solar eclipse.

A total solar eclipse occurred across the east and west of the United States on 2017 August 21. We observed the eclipse in Jackson, Wyoming, where the total duration was not too short, and it was forecasted to have low cloud cover. The observation site was located in the National Elk Refuge (43° 33' 3407 N, 110° 37' 3062 W), away from the roadway, to avoid dust and interruptions from other people. The maximum duration is 2 minutes 18 s. Table 1 lists the times and the Sun's positions in each phase of the total solar eclipse. Figure 6 shows a photograph of the DICE observation system deployed in the field.

Zoom In Zoom Out Reset image size

Table 1. Observation Time and Solar Positions during Phases of the Solar Eclipse

Event	UT	Local Time	Δt	Altitude	Azimuth
Partial phase start	2017-08-21 16:16:46	2017-08-21 10:16:46	-1:19:19	385	1132
Totality start	2017-08-21 17:34:56	2017-08-21 11:34:56	-0:01:09	502	1345
Totality max	2017-08-21 17:36:05	2017-08-21 11:36:05	0:00:00	504	1349
Totality end	2017-08-21 17:37:14	2017-08-21 11:37:14	0:01:09	505	1352
Partial phase end	2017-08-21 19:00:30	2017-08-21 13:00:30	1:24:25	578	1682

Download table as:  ASCIITypeset image

The optimal exposure time (T_e ) during the total solar eclipse for each filter is estimated for the DICE telescope aperture through the exposure time (T_t ) of the pinhole aperture during the pinhole observation. Assuming a coronal brightness of 10⁻⁶ disk brightness, T_e is given by:

$$\begin{eqnarray*}&&{T}_{e}={10}^{6}{T}_{t}{\left(\displaystyle \frac{{D}_{p}}{{D}_{o}}\right)}^{2},\end{eqnarray*}$$

where the pinhole diameter (D_p ) is 0.2 mm, the DICE telescope aperture (D_o ) is 50 mm. T_t is derived from the target intensities (I_t ) through the intensities (I_p ) and exposure times (T_p ) from the pinhole observation:

$$\begin{eqnarray*}&&{T}_{t}={T}_{p}\displaystyle \frac{{I}_{t}}{{I}_{p}},\end{eqnarray*}$$

I_t is set to 20,000 and 40,000, which is approximately 1/3 and 2/3 of the CCD dynamic range, respectively, considering the expected spectral change of the corona in comparison with the solar disk. Table 2 shows the intensities and exposure times in the pinhole observation and calculation of exposure times for target intensities during the solar eclipse.

The observation program was composed of flat and dark imaging, pinhole observation for partial and total solar eclipse observations. We aimed the pointing system at the open sky away from the Sun and operated the filter wheel and the camera manually so that the flat imaging shot a consistent distribution of light from the sky. Dark imaging was performed before and after the partial solar eclipse. Image processing was paused during the total solar eclipse to capture as many images as possible within a short period of time.

The total solar eclipse observation was automated using the observation program of the DICE software. Three program files were created respectively for dark imaging, pinhole observations, and eclipse observations. The observation files were uploaded to the DICE control computer using FTP. The DICE software controlled the number of frame exposures in each sequence of the observation program during a limited time of the total solar eclipse observation and a limited frame count for the dark and the pinhole images.

We successfully performed the eclipse observation using one of the DICE telescopes. The other failed during the total eclipse because of frame transfer error of the camera. The pinhole imaging, every 10 minutes, acquired a total of 120 images in 10 repeats, with a total of 12 images for four bandpass filters and three polarization angles. Dark imaging yielded a total of 130 images for each exposure time of up to 5 s. The total solar eclipse observations yielded 161 images. The polarization filter was rotated every 60° to obtain an image set for polarization angles of 0°, 60°, and 120° per bandpass filter. We obtained two complete sets of observational images for the total solar eclipse, except for oversaturated images, owing to the sudden change in solar brightness at the start and end of the total eclipse. Figure 7 shows a set of observational images of the total solar eclipse. Each bandpass image was obtained with the expected signals using the exposure times calculated through the pinhole observation shown in Table 2.

Zoom In Zoom Out Reset image size

We created software procedures for data processing and plotting using interactive data language (IDL). All total eclipse images were preprocessed to remove dark current images. We attempted to measure the flat field images of each filter on the day of the total solar eclipse, but we could not obtain useful flat field images because of the sky's polarization pattern. The flat images are important to correct the systematic bias of the two different systems; therefore, we planned to take flats in a laboratory after the day of the observation. Because only one instrument worked in capturing the total eclipse, we continued data processing without flat-field correction. We assumed that the pixel-to-pixel sensitivity variations and the effect of dust or scratches on the optics and CCD window were negligible across a few pixels of tracking drift during the total eclipse.

As shown in Figure 7, the three polarization angles of each filter resulted in different coronal structures. For their co-alignments, we used the following procedure: (1) We aligned the normalized images with a specific polarization degree of each filter to its first-time image by using a cross-correlation technique. (2) We aligned the images with 60 and 120 polarization degrees of each filter to its image with zero polarization degree using the position of the solar disk. (3) We aligned the 3990, 4025, and 4249 Å images to the 3934 Å images. (4) By comparing our four filter images and the SDO/AIA 304 Å observed at 17:36 UT on 2017 August 21 we checked the co-alignment. After the co-alignment, we stacked images with the same polarization angles for each filter to improve the signal-to-noise ratio of the coronal structures.

Using three polarization angle images of each filter, we estimated the polarized brightness (pB), total brightness (tB), degree of linear polarization (dP), and angle of linear polarization (aP). Here, pB, tB, dP, and aP are defined as follows (Billings 1966 Lu et al):

$$\begin{eqnarray*}\begin{array}{rcl}{pB} & = & \displaystyle \frac{4}{3}\left[{({I}_{0^\circ }+{I}_{60^\circ }+{I}_{120^\circ })}^{2}\right.\\ & & {\left.-\,3({I}_{0^\circ }{I}_{60^\circ }+{I}_{60^\circ }{I}_{120^\circ }+{I}_{0^\circ }{I}_{120^\circ })\right]}^{\tfrac{1}{2}},\end{array}\end{eqnarray*}$$

$$\begin{eqnarray*}&&{tB}=\displaystyle \frac{2}{3}({I}_{0^\circ }+{I}_{60^\circ }+{I}_{120^\circ }),\end{eqnarray*}$$

$$\begin{eqnarray*}&&{dP}=\displaystyle \frac{{pB}}{{tB}},\end{eqnarray*}$$

$$\begin{eqnarray*}&&{aP}=\displaystyle \frac{1}{2}\mathrm{arccot}\displaystyle \frac{2{I}_{0^\circ }-{I}_{60^\circ }-{I}_{120^\circ }}{\sqrt{3}({I}_{60^\circ }-{I}_{120^\circ })}.\end{eqnarray*}$$

It should be noted that the field of view (FOV) of the DICE ranged from the solar disk center up to 14 solar radii. When we carefully investigated the coronal structures using the exposure times (192, 236, 288, and 69 ms) for each filter (3934, 4025, 3990, and 4249 Å), the coronal structures could be well observed until approximately two solar radii. From this, we cropped all the images within the FOV from the solar disk center to a solar radius of 2.5.

Figure 8 shows examples of pB, tB, dP, and aP images for each filter observed in the 2017 total eclipse. From this, we find the following characteristics: (1) coronal structures such as East, West, North-West streamers are well observed as shown in pB and tB. Based on a careful investigation if the tB image of the 3934 Å filter, we found that the radial leakage patterns (gray arrows) were caused by the prominences (orange arrows) that contaminated the neighboring pixels due to the strong scattering light. (2) The dP of the coronal streamers generally increased with the increase in the heliocentric distance until a specific solar radius and then decreased. The maximum dP were located at a height of 1.2 solar radius on the eastern streamer, as shown by black arrows in dP. The values of dP are similar to one another with 0.3 (3990 Å), 0.31 (4249 Å), 0.31 (3934 Å), and 0.33 (4025 Å). The mean value of dP at the sky background are consistent with 0.1 except for 0.08 for 3934 Å. (3) The aP of the coronal structures were roughly 90° on the left and right and roughly 0° on above and below, similar to those obtained using the Thomson scattering theory. The aP of the sky background had the same values as 75.1, (3990 Å), 73.7 (4249 Å), and 76.5 (3934 Å), except for 60.0 (3934 Å). Our results therefore demonstrated that DICE is highly effective for the observation of coronal structures such as coronal morphology and polarization degree.

Zoom In Zoom Out Reset image size

Vorobiev et al. (2020) observed the 2017 total solar eclipse in Madras, Oregon, using a polarization imaging camera equipped with Bessel B and R filters. They found that the maximum dP was 0.47 near the height of 1.5 solar radius in the eastern streamer. The sky foreground at the Bessel B and R filters showed 0.16 and 0.08 of the mean values of dP and a difference of ∼11° of the mean values of aP was observed.

Our findings of the maximum dP of the solar corona and the dP and aP of the sky background, however, differ from their results. These discrepancies might be due to the different filter characteristics and geographic and weather conditions at the observation sites. The lower value of dP obtained from DICE might be caused by the strong scattering of light from prominence pixels, which leads to an increase in the tB of each filter. The 3934 Å filter, especially, transmitted more light from the prominent pixels. As our bandpass filters had narrower bandwidths than theirs, no significant differences in dP and aP were observed between the bandpass filters.

While we describe the polarization characteristics of solar corona and sky background, Cho et al. (2020), on the other hand, studied the general data processing and results for the uncertainty of the DICE observation data and the solar corona temperature. Unfortunately, they were not able to obtain the velocity of the solar corona, as it could not be calculated owing to the low SNR caused by the insufficient observation data collected during the short total eclipse observation time. It is interesting to note that polarization cameras can obtain multiple polarization angles simultaneously by attaching a micropolarizer array to an imaging sensor, making it possible to obtain high SNR data by image stacking (Reginald et al. 2017, Vorobiev et al. 2018). Gopalswamy et al. (2021) developed a solar coronagraph equipped with a polarization camera to observe the temperature and velocity of a solar corona at an altitude of 40 km.