How does the Hubble Space Telescope "Drift and Shift"?

Question

The Phys.org article Mystery of the universe's expansion rate widens with new Hubble data says:

Astronomers have been using Cepheid variables as cosmic yardsticks to gauge nearby intergalactic distances for more than a century. But trying to harvest a bunch of these stars was so time-consuming as to be nearly unachievable. So, the team employed a clever new method, called DASH (Drift And Shift), using Hubble as a "point-and-shoot" camera to snap quick images of the extremely bright pulsating stars, which eliminates the time-consuming need for precise pointing.

"When Hubble uses precise pointing by locking onto guide stars, it can only observe one Cepheid per each 90-minute Hubble orbit around Earth. So, it would be very costly for the telescope to observe each Cepheid," explained team member Stefano Casertano, also of STScI and Johns Hopkins. "Instead, we searched for groups of Cepheids close enough to each other that we could move between them without recalibrating the telescope pointing. These Cepheids are so bright, we only need to observe them for two seconds. This technique is allowing us to observe a dozen Cepheids for the duration of one orbit. So, we stay on gyroscope control and keep 'DASHing' around very fast."

I understand that the technique skips the use of guide stars, but I'm wondering exactly what does happen. What does the word recalibrating mean in "..move between them without recalibrating the telescope pointing..."

Do they just execute a blind change in direction using Hubble's attitude control system of gyros and reaction wheels, expose, and move on to the next field, and hope the target object was imaged?

For a little background see Hubble's one-gyro mode; how does it work for attitude control, stabilization, and slewing? and answer(s).

Answer 1

The "drift-and-shift" (DASH) mode was described in a paper by Momcheva et al. (2016), and works the way you supposed: a sequence of nearby images is taken using blind offsets without guide-star recalibration: move to the first field and take an image (possibly first acquiring a guide star), move to a nearby target field and take another image (no guide star acquisition), repeat...

Guide-star "calibration" means correcting the internal model of "where in the sky HST is pointing" by locating a bright star with well-known coordinates, tweaking the pointing of HST to put the guide star in the right position within the telescope's focal plane (so that the center of the focal plane is now pointing where the observing proposal specifies), and setting up a monitoring process where the position of the guide star is continually tracked during an observation to make sure the telescope pointing is not drifting (and correcting it when it is). The whole acquisition-and-setup process introduces an overhead of about six minutes; since a typical HST observing window is only about 50 minutes (before the spacecraft's orbit carries it around to the other side of the Earth and it can't see the target any more), trying to change target fields multiple times would rapidly eat up your actual observation time.

(In some [most?] cases, it might make sense to do guide-star acquisition for the very first field, to which HST has likely just made a very large slew, to ensure the initial pointing is accurate and that the (smaller) offsets to new fields will be reasonably accurate.)

Not having a guide star does impact the stability of the telescope pointing during an observation: guiding using just the gyros means a drift of about 0.001 to 0.002 arc seconds per second of time. So long exposures are not a good idea, because they'll be smeared by the telescope drift: an exposure of 100 seconds means a typical drift of several pixels during that time. But since Riess et al. were taking very short exposures (just a few seconds), this wasn't a problem for them.