How (the heck) can thrusters alone stabilize New Horizons well enough to take high magnification 30 second long exposures in order to see +21 m objs?

Question

Wikipedia's Long Range Reconnaissance Imager says that LORRI is a:

...telescope aboard the New Horizons spacecraft for imaging. LORRI has been used to image Jupiter, its moons, Pluto and its moons, and Arrokoth since its launch in 2006. LORRI is a reflecting telescope of Ritchey-Chrétien design, and it has a main mirror diameter of 20.8 cm (8.2 inches) across. LORRI has a narrow field of view, less than a third of a degree. Images are taken with a CCD capturing data with 1024 × 1024 pixels. LORRI is a telescopic panchromatic camera integrated with the New Horizons spacecraft, and it is one of seven major science instruments on the probe. LORRI does not have any moving parts and is pointed by moving the entire New Horizons spacecraft.

Later in design it says:

The design can take images at very low light levels required for the mission, including light levels 1/900 those of Earth when it is at Pluto.[4] For the Arrokoth encounter the longest exposure time (up to ten seconds for the Pluto flyby) was increased. This was accomplished after the Pluto flyby by the team, to support taking images in even lower light levels.

After the Pluto flyby, exposure times of at least 30 seconds were made possible, which was also useful for taking reconnaissance images and enabling imaging down to a magnitude of 21.

LORRI is pointed by moving the entire spacecraft, which limits the exposure time. The spacecraft does not have reaction wheels and is stabilized by thrusters.

Seeing a +21 magnitude star with a 20.8 cm telescope will require a long exposure time as mentioned. I am surprised that thrusters alone can stabilize this spacecraft and therefore this telescope so well without the use of any other attitude control system.

Unlike the push-broom cameras popular on some deep space spacecraft that use slow spacecraft rotation to scan their 1D CDD across a field for imaging purposes, LORRI is a "normal" 2D CCD camera and so requires steady pointing during an exposure.

Question: How (the heck) can thrusters alone stabilize New Horizons well enough to take high magnification 30 second long exposures in order to see +21 mag. objects?

Are there any special tricks to this? Perhaps firing opposite thrusters for very slightly different times rather than trying to fire only one for a very short time? Are there any other sources of torque that are used to fine-null any rotation during the up to 30 second long exposures?

From Wikipedia's [New Horizons; Propulsion and attitude control] (https://en.m.wikipedia.org/wiki/New_Horizons):

There are 16 thrusters on New Horizons: four 4.4 N (1.0 lbf) and twelve 0.9 N (0.2 lbf) plumbed into redundant branches. The larger thrusters are used primarily for trajectory corrections, and the small ones ... are used primarily for attitude control.

Answer 1

The LORRI Wikipedia article cites the following two sources for this:

• LONG-RANGE RECONNAISSANCE IMAGER ON NEW HORIZONS, Cheng et al. (2008)
• Emily Lakdawalla Planetary Society article

The latter being the likely origin:

The New Horizons team has now enabled LORRI to take images with exposure times of 30 seconds. [...]. Before, the dimmest objects they could reliably detect were at about a visual magnitude of 20, but the longer exposure pushes that down to 21.

LONG-RANGE RECONNAISSANCE IMAGER ON NEW HORIZONS, Cheng et al. (2008) says:

LORRI has a 4×4 pixel binning mode, for which its limiting magnitude requirement is V>17 in a single exposure of 9.9s. This 4×4 pixel-binning mode will be used to search for the target KBO and to perform optical navigation on approach. A special spacecraft guidance mode is available for the KBO search in which the spacecraft will hold the target within the 4×4 pixel pointing tolerance for 10 second exposures. [emphasis added]

• article describing pixel binning

The paper later mentions:

In general, exposures of 50 to 200 milliseconds are typical for LORRI, although the maximum exposure time is 29.9 seconds. [emphasis added]

So perhaps the capability was always there? Just now it has been "enabled", i.e., turned on?

Either way, the same paper also gives the pixel field-of-view (IFOV) as 4.94 micro radians (vertically/horizontally, not diagonally). This means the 4x4 binned 'super-pixel' has an IFOV of ~20 micro radians. The regular/normal 3-axis attitude control system specifications are (per The New Horizons Spacecraft, Fountain et al.):

The control algorithms must maintain the spacecraft attitude to within ±24 µrad (3σ) and the spacecraft rotational rate to within ±34 µrad/s (3σ). [emphasis added]

This looks something like this:

Where the grid represents the individual pixels, the red square the 4x4 binned pixel, and the blue circles the normal (3-axis) allowable spacecraft pointing error.

This is evidently not good enough for imaging dim objects, hence the aforementioned special mode, which I suspect is just lowering the dead bands in the attitude controller. However, there is also problems with too accurate a spacecraft attitude; another point of note from Cheng et al.:

For optical navigation, LORRI is required to be able to image a star of visual magnitude V=11.5 at SNR>7 in a single 100 ms exposure, with full width at half maximum (FWHM) >1 pixel. It is not desirable for too great a fraction of the energy from a point source to be imaged onto a single pixel, because stellar images become too undersampled. [emphasis added]

(for what its worth the section in between these quotes is the first quoted passage from Cheng et al. above)

At 40 AU from the Sun, LORRI is predicted to be able to detect a 50 km diameter object, of albedo 0.04 and at phase angle 25°, from a distance of 0.35 AU, more than 40 days before the object would be encountered. [emphasis added]

I do not know if/how the undersampling concern applies in the 4x4 pixel binning mode, but the test Kuiper Belt Object described in the latter quote has an angular diameter of ~1 micro radian, less than a single pixel:

So how is the increase in pointing accuracy only now achieved?

I think it comes from, somewhat counter-intuitively, the reduced efficiency of the propulsion system.

New Horizons uses a blowdown propulsion system which is characterized by a decrease in feed/tank pressure as propellant is used, seen in figure 3 from THE USE OF THE AEROJET MR-103H THRUSTER ON THE NEW HORIZONS MISSION TO PLUTO, Stratton (2004) (thanks comment section):

This corresponds to a decrease in the minimum impulse bit (Figure 6, Stratton (2004)):

And with reference to the angular velocity change from a single thruster pair pulse (min impulse bit), Cheng et al. says (thanks comments):

These numbers improve (decrease) as the feed pressure of the spacecraft propulsion system decreases.

Finally, the Planetary Society article (from Jan 2018) indicates fuel levels:

Of the 75 kilograms of propellant that they launched with, 21 kilograms remain. Another 12 or so will be spent completing the extended mission, leaving about 9 for a possible second mission extension.

Answer:

The ~30 % decrease in minimum impulse bit (from propellant use) lowers the spacecraft's minimum angular velocity change to a level that permits 30 second exposures in the "special spacecraft guidance mode" combined w/ 4x4 pixel binning.