Franken said:
In Dread Ink, the Grave Hand of James Kuyper Did Inscribe:
Can you say a few more words about this? There's questions of precision,
but I can't quite see the geometry of it. ...
This is going way off-topic; I just wanted to give a concrete example
from my own personal experience of a situation where the sub-second
timing involving leap-seconds is important. In the course of doing so, I
made a mistake, one that nobody who's likely to be visiting this
newsgroup is likely to recognize (or care about): the mirror motion is
reported using relative times, rather than absolute times, and as a
result is not sensitive to leap-second issues.
What I should have referred to is the second most rapidly moving part of
the system, which is the satellite itself. It moves at 7500 m/s, so
missing a leap second would cause a systematic error in our ground
positions of roughly 7500m, or roughly 75 times our entire error budget.
That's still more than enough to make tracking leap-seconds important;
but it's not as impressive as the error that could have been caused if
it had been the case that the calculation of the mirror motion were
sensitive to leap-second issues.
To address your question about the geometry: our calculations work
essentially by tracing light rays backwards. We start with a line of
sight from the center of the detector toward the center of the lens, and
trace that line of sight backwards, reflecting it through the rotating
mirror, and then from there to the surface of the earth. In the course
of this calculation, we do about six different coordinate system
transformations, several of which involve coordinates systems what are
moving with respect to each other. Those transformations depend upon the
rotation of the mirror, the current attitude and position of the
satellite, and the current rotation of the Earth. For the level of
precision required, we need to take into consideration the facts that
the Earth rotates neither at a constant rate, nor around a fixed
rotation axis. We start by approximating the earth as an ellipsoid, and
then adjust that using a Digital Elevation Model of the earth with 30
arcsecond resolution.
... Are you talking of stereographic
Riemannian projection?
The stereographic projection projects from a point on the surface of a
sphere, through another point on the surface of the sphere, to a flat
plane tangent to the sphere at the point opposite to the projection
point. If you ignore the reflection through the mirror, there's a
similarity between our algorithm and the stereographic projection, but
also a lot of differences. We're using an ellipsoid, not a sphere. The
projection point is at an altitude of 705 km, not on the surface of
ellipsoid. The focal plane on which our detectors lie can, if you ignore
the mirror, be considered to be approximately parallel to the earth's
surface, and just a few centimeters higher than the projection point,
and therefore not likely to touch the surface of the earth at any point
(though it could if the pitch or roll angles were high enough, but that
only happens during maneuvers). Also, we don't treat the coordinates on
the focal plane as components of a complex number, so "Riemannian"
doesn't apply, either.