The Design and Engineering of Curiosity

by Emily Lakdawalla

  MARDI was not required to operate after landing, so was never tested on Earth for survival through Mars day/night temperature cycles. It has no heaters. However, it also has no moving parts, so there is no reason to expect it to suffer from Martian conditions any more than MAHLI and the Mastcams do. Since landing, MARDI has been used to image the ground beneath the rover, documenting the rock fragments and outcrops along rover traverses.

  Curiosity’s MARDI was built by Malin Space Science Systems, who had also built MARDI instruments for Mars Polar Lander (which crashed) and 2001 Mars Surveyor (which was canceled). The principal investigator is Michael Malin. The Surveyor MARDI later flew to Mars on Phoenix, but was not actually used during the landing because of late-appearing concerns about the spacecraft computer’s interface with the instrument. Although Curiosity’s MARDI bears the same name as these predecessors, it is a wholly different instrument. It has a successor instrument already in space, the JunoCam aboard NASA’s Juno orbiter mission to Jupiter. (Interestingly, JunoCam launched a few months before MARDI on Curiosity.)

  7.3.1 How MARDI works

  MARDI is mounted to the left front side of the rover, pointed straight downward (Figure 7.9).9 It uses the same detector, electronics, and software as MAHLI and the Mastcams (see section 7.2.1), with much simpler optics. It is in focus at any distance beyond 2 meters. It obtains color 1600-by-1200-pixel images over a wide field of view of 70-by-55°. Images taken after landing, from an elevation 66 centimeters above the ground, are slightly out of focus, and MARDI’s ability to resolve ground features since landing is the same as it was from a height of 2 meters. On the ground, the MARDI field of view covers an area about 92 centimeters across the driving direction and 66 centimeters along the driving direction. The fixed MARDI view includes part of the left front wheel and the area immediately behind it and next to it, underneath the rover. Further MARDI facts are summarized in Table 7.4. Figure 7.10 shows examples of MARDI images taken under different conditions. They have been projected to correct for the distortion caused by its wide-angle lens.

  Figure 7.9. Location of the MARDI instrument on the rover. Inset photo taken at Malin Space Science Systems; large photo taken during final assembly at Kennedy Space Center, showing MARDI’s point of view beneath the rover in descent configuration. NASA/JPL-Caltech/MSSS.

  Table 7.4. MARDI Facts.

  2 km elevation

  2 m elevation

  66 cm elevation

  depth of field

  2m - infinity

  angular resolution

  0.76 mrad/pixel

  field of view (FOV)

  70° x 55°

  spatial resolution

  1.5 m/pixel

  1.5 mm/pixel

  0.5 mm/pixel at image center (out of focus)

  7.3.2 Using MARDI

  Landing video.

  Figure 7.10. Examples of MARDI images. Top: View of the heat shield during descent. Middle: sol 45 image under daylight conditions, showing reduced contrast. Bottom: sol 738 image under twilight conditions, which improves contrast. Images 0000MD0000000000100035C00, 0045MD0000300000101520E01, and 0738MD0003120000102267E01. NASA/JPL-Caltech/MSSS.

  On landing day, MARDI switched on about 6 seconds before heat shield separation and took 1504 images at an average 3.88 frames per second with exposures of 0.9 milliseconds. The first 25 were black, seeing the inside of the heat shield; the next 622 chronicled the final 2.5 minutes of landing, from heat shield separation to touchdown; and the final 857 were taken on the surface. The MARDI descent images not only captured the dynamics of the spacecraft’s descent, they also documented large gravel being propelled toward the rover as the rockets impinged on the surface, quickly providing an explanation of how the rover deck came to be salted with gravel.10 Unfortunately, the front element of the MARDI lens was coated with dust during the landing. The dust coating scatters sunlight, which has the effect of reducing the contrast of MARDI images taken since landing.

  Clast surveys. At parking spots between traverses, MARDI shoots a photo to document the sizes and shapes of rock fragments on the surface. Since sol 310, MARDI clast survey images have mostly been taken at twilight (around 18:30 local true solar time), which reduces the light scattering off of the dusty front window, producing much higher quality images (see Figure 7.10).11 Beginning on sol 488, MARDI has also sometimes been used before and after short “bumps” to obtain 4 or 5 overlapping images for analysis of the three-dimensional shapes of rock clasts.12

  Sidewalk mode. MARDI can take movies during drives, acquiring mosaics along drive paths.13 In sidewalk mode, MARDI takes an image every 3 seconds, but only saves the image if onboard software determines that the new image is significantly different from the previous one. The saved images have more than 75% overlap. Returning every third image to Earth allows the construction of a mosaic, but if all images are returned, the team can generate a digital elevation model in addition to the mosaic. Sidewalk mode was tested on sol 651 and has since been used to document terrain thought to be hazardous to rover wheels as well as science stops like Pahrump Hills. A complete list of sidewalk mode observations to sol 1647 is in Table 7.5.Table 7.5. MARDI Sidewalk Mode sequences to sol 1800.

  Sol

  Description

  651

  Test of sidewalk mode

  691

  Characterize terrain thought to be hazardous to rover wheels

  739

  Characterize terrain thought to be hazardous to rover wheels

  780

  Stratigraphy of Pahrump Hills outcrop

  785

  Stratigraphy approaching Book Cliffs

  787

  Stratigraphy between Book Cliffs and Gilbert Peak

  790

  Stratigraphy to Alexander Hills

  792

  Stratigraphy to Chinle outcrop

  794

  Stratigraphy to Whale Rock

  797

  Stratigraphy between Whale Rock and scuff test site

  1181

  Document sediment interaction with the wheels at Bagnold Dunes

  1281

  Document drive across knobbly Stimson contact

  7.4 MAHLI: MARS HAND LENS IMAGER

  7.4.1 Introduction

  The Mars Hand Lens Imager (MAHLI, pronounced “Molly”) functions as it is named: it works like the hand lens that any field geologist carries in order to examine the grain size and structure of rocks. But it’s capable of many other tricks. Located on the turret on the end of the robotic arm, MAHLI can work very close to targets, taking photos with microscopic detail, like its predecessor, the Microscopic Imager (MI) on the Mars Exploration Rover mission. Unlike MI, MAHLI is focusable, from 2.04 centimeters to infinity. When held at its minimum working distance, MAHLI can take images with resolutions of 13.9 microns per pixel, enough to resolve the finest grains of sand. But its wide (~35°) field of view makes it a useful tool for imaging large and distant objects, too. It can acquire detailed mosaics of interesting outcrops with arm motions, and obtains 3D information from stereo pairs or its focal depth.14 Figure 7.11 shows parts of the MAHLI camera, and Table 7.6 summarizes MAHLI facts.

  Its position on the end of the robotic arm allows MAHLI to examine Curiosity itself, so the mission often commands it to document the condition of the wheels, remote sensing mast, and instruments. And it’s taken some of the most iconic photos of the whole mission: self-portraits of the rover at drill sites, such as the one on the cover of this book. MAHLI was built and is operated by Malin Space Science Systems, San Diego, California. The principal investigator is Ken Edgett of Malin Space Science Systems.

  Figure 7.11. Upper left: MAHLI camera head with 89-millimeter-long pocket knife for scale. Upper right: MAHLI camera head with dust cover open. From Edgett et al, 2012 . Bottom: MAHLI on Mars as viewed from the right Mastcam, Curiosity sol 128. Image 0128ML0006510000103943E01. NASA/JPL-Caltech/MSSS/Emily Lakdawalla.

  Table 7.6. MAHLI Facts.

  Target l
ocated 25 cm away from front window

  Target at infinity

  depth of field

  1 mm

  –

  field of view (FOV, diagonal)

  34°

  38.5°

  FOV (horizontal 1600 pixels)

  26.8°

  31.1°

  FOV (vertical 1200 pixels)

  20.1°

  23.3°

  instantaneous field of view (IFOV)

  402 μrad

  346 μrad

  focal ratio

  f/9.8

  f/8.5

  effective focal length

  18.4 mm

  21.4 mm

  7.4.2 How MAHLI works

  MAHLI components include a camera head mounted on the turret at the end of the rover arm, an electronics assembly located inside the body of the rover, and a calibration target mounted on the “shoulder” of the rover arm, its azimuth actuator. The camera head detector and other electronics and internal electronics assembly are identical to those of Mastcam (see section 7.2.1). Its optics, calibration target, and usage are different.

  7.4.2.1 Camera head and electronics

  MAHLI’s optics include a sapphire window in front of a group of stationary elements (refractive lenses). Behind the sapphire window and stationary elements is a movable group of three elements, operated by a single motor.15 Because the sapphire window has a long wave cutoff filter and the lens elements filter out ultraviolet light, only light with wavelengths ranging from 394 to 670 nanometers reaches the detector.

  Various design elements help MAHLI survive extreme temperatures, exposure to dust, and sampling-related vibration. MAHLI was designed to operate between –40°C and +40°C, but because it is outside the rover it must be able to survive far more extreme temperatures when it is powered off. It has a movable lens cover to keep the dust out. The dust cover includes a transparent Lexan lens window in order to permit imaging with the cover closed. Unfortunately, as with MARDI, a thin film of dust covered the window during landing, and photos taken with the cover closed have very low contrast and an orange-colored cast. A contact sensor assembly with two probes protects the camera against contact with the surface, causing the forward motion of the arm to stop when it has brought MAHLI’s sapphire window within 17 millimeters of a hard surface. A vibration isolation platform, which connects MAHLI to the turret through a set of three springy wire rope assemblies, isolates MAHLI somewhat from the intense vibration of the drill and CHIMRA.

  MAHLI carries six light-emitting diodes (LEDs) on a ring around the outside of the front lens element in order to illuminate science targets from different directions and with ultraviolet light. Windows on the dust cover allow light from the LEDs to be visible even with the cover closed. Four of the LEDs emit white light, positioned in two pairs on either side of the lens. The white-light LEDs can be commanded to operate together or independently (with either one side or the other sides or both sides lit), to simulate the way a geologist tilts a rock while examining it with a hand lens to catch glints of sunlight off of crystal facets. Nighttime white-light LED imaging also allows scientists to directly compare the colors of rock targets at different locations without having to account for differences in solar illumination.

  Two ultraviolet LEDs that emit light at a wavelength of 365 nanometers are intended for identifying minerals that are fluorescent or phosphorescent, and are usually used overnight. They do leak some light in short visible wavelengths, so white surfaces appear blue in MAHLI nighttime ultraviolet images. So far there has been no unambiguous detection of fluorescent or phosphorescent minerals. It would have helped to equip MAHLI with shorter-wavelength LEDs, but no such LEDs were available when MAHLI was being designed (no flight-qualified ones, anyway).

  MAHLI’s camera head is connected to the electronics through an astonishing 12.7 meters of cable harness. As with Mastcam, MAHLI images are usually stored onboard the instrument in raw 8-bit form without compression or Bayer interpolation. Full MAHLI frames can be acquired at a maximum rate of about 1 frame per second, slower than the maximum rate for Mastcam. MAHLI has video capability because of its shared electronic design with Mastcam and MARDI, but it has rarely used this ability. Table 7.7 lists the few video observations with MAHLI. A rare, spectacular MAHLI video observation happened on sol 687, when MAHLI took video of the target Nova as ChemCam lasered it. MAHLI was able to see the flash of the plasma excited by ChemCam, and saw motion of dust as ChemCam blasted the target.Table 7.7. Summary of all MAHLI video mode observations to sol 1800. “ATLO” = Assembly, test, and launch operations, i.e. images taken before launch.

  Date or Sol

  Target Name

  Day/Night

  30 Nov 2010

  rover deck

  ATLO

  2 Dec 2010

  MAHLI cal target

  ATLO

  3 Feb 2011

  infinity focus position

  ATLO

  26 May 2011

  ATLO

  ATLO

  165

  Sayunei

  Night

  166

  Sayunei

  Day

  282

  CheMin inlet

  Night

  687

  Nova

  Day

  7.4.2.2 Focusing MAHLI

  MAHLI’s focusing ability means it can point at targets from working distances as close as 2.04 centimeters from the sapphire window, and also focus at infinity. MAHLI can focus on targets with the cover in either the open or closed position. MAHLI focuses using the same motor that opens and closes the dust cover. It is a stepper motor with up to 16100 step positions (Figure 7.12). From steps 0 to 5100, the cover is closed. Then the motor engages the dust cover. Beyond a motor count of 12000, the cover is fully open. From motor steps 0 to 1480, MAHLI is in focus at its minimum working distance of 2.04 centimeters with the cover closed. With the cover still closed, it is in focus at infinity at a motor count of about 4523.

  To work with the cover open, the MAHLI team commands it to fully open to a motor count of 15504, then steps to the desired focus position. MAHLI is in focus at infinity at a motor count of 12552, and in focus at its minimum distance of 2.04 centimeters at a motor count of 15600. The stepper motor moves at 150 motor steps per second, so it takes less than 2 minutes to open the cover and focus the camera. The dust cover sweeps through a space about 5 centimeters beyond the camera. To be sure of ample space to operate the dust cover safely, it is never articulated with the camera head less than 10 centimeters from a target.

  Figure 7.12. Relationship between MAHLI motor count, behavior of the lens cover, and working distance at which the target is in focus. By Emily Lakdawalla after Edgett et al. ( 2015 ).

  Like the Mastcams, MAHLI can either be manually focused (commanded to a specific motor count position) or autofocused. The MAHLI team manually focuses images when the scene will contain a wide depth range, such as during rover wheel imaging or wide views of outcrops. Autofocus is more commonly used for hand-lens imaging where it is crucial because of MAHLI’s shallow depth of field. The depth of field ranges from about a millimeter at a working distance of 2 centimeters, to about 8 millimeters at a working distance of 12 centimeters, and continues increasing toward infinity. In fact, MAHLI focus is so sensitive to distance from the target that MAHLI autofocus distance is a tactically useful measure of the distance between a commanded turret position and a target. Ordinarily, rover planners nudge a target with the APXS contact sensor to precisely determine its range from the rover (see section 9.3). But when the APXS contact sensor isn’t available for ranging (when a surface is especially rough or when it is composed of loose materials), MAHLI autofocus distance works as a measure of the distance to the target.

  7.4.2.3 Z-stacks

  MAHLI can obtain “z-stack” images in order to compensate for its shallow depth of field. To obtain in-focus images over a greater distance range, MAHLI can be commanded to acquire several (typically 8 or
16) images at different motor count positions. Later, using a different set of commands, onboard software will locate the best-focus parts of each image, and merge them into a single data product, called a focus-merge image product or a z-stack. As a byproduct of the focus merge process, MAHLI also creates another data product called a range map – essentially, a digital elevation model of the target. Returning a single, color-interpolated z-stack and its associated grayscale range map requires less data volume than returning the 8 images used to produce them, so it functions as a kind of data compression.

  MAHLI z-stacks are usually not made until later in the sol, or even one or many sols after the original images were taken. The resulting focus-merge data products have file names and time stamps that reflect the date and time that they were produced, not the date and time of the original data, which can make them difficult to track down. Recognizing this issue, the MAHLI team has released an open-source “Principal Investigator’s Notebook” for each of the public releases of MAHLI data, and information about which sol’s data contributed to z-stack images is included in metadata delivered to the PDS.16

  7.4.2.4 Figuring out MAHLI image scale

  Both the field of view and the pixel scale depend upon the working distance between MAHLI and a target. To the MAHLI team, working distance is measured from the front of the sapphire window to the target. For rover planners, the zero point is located at the minimum distance they can safely command MAHLI to take a photo, 19 millimeters in front of the sapphire window. This is usually – but not always – called the “toolframe distance”, “standoff distance” or “RP [rover planner] distance,” which distinguishes it from MAHLI instrument “working distance.” Unfortunately, rover planners sometimes refer to this as the “working distance,” which gets confusing. To convert rover-planner distances into MAHLI instrument working distances, add 19 millimeters. If you are not certain of the convention being used, you can unambiguously determine the range to the in-focus parts of an image using the motor count.