Greatest Software Ever Written 4225
 |
On Fri, 18 Aug 2006 10:59:34 -0700, Eugene Miya I think we're saying the same thing, but to pick nits, I still say there's a...
On Thu, 17 Aug 2006 17:04:56 -0700, Eugene Miya
See the following, esp. the last paragraph. I think Miles has a typo; the architecture is definitely right, but the model number is wrong. By the 1980's most of the machines had been converted to 1100-80's, and by the mid 1990's when this report was written, some were 2200's.
From the 1993-1996 IAU Commission 4 Report:
3. Numerical Ephemerides of Planets (Standish)
3.1 CALTECH JET PROPULSION LABORATORY; PASADENA, CALIFORNIA
JPL's latest planetary and lunar ephemeris, DE403 (Standish et al., 1995) was integrated from 3000 BC to 3000 AD and has been made available via anonymous FTP navigator.jpl.nasa.gov - 128.149.23.82 and in limited quanbreasty on CD-ROM.
The ephemeris was fit to nearly 40,000 observations, including optical transits, astrolabe, astrometry, occultation timings, radio astronomy, radar ranging, spacecraft, frame-tie measurements, and lunar laser ranging. DE403 is referenced directly onto the International Celestial Reference Frame (ICRF) by the inclusion of VLBI measurements of spacecraft in orbit around Venus (Magellan) and Mars (Phobos) and by the inclusion into the fits of the frame-tie determination (Folkner et al., 1995).
The ephemeris is still integrated on the Univac 1108, because that computer has a 60-bit mantissa in double precision. We find that modern workstations with only 53-bit mantissas do not have the precision necessary for integrating the motion of the moon at the sub-centimeter level, necessary for reducing and adjusting to modern Lunar Laser Ranging data. Quadruple precision is about 100 times slower -- too slow when integrating many centuries of motion. A new formulation by Fukushima (1996) gives an error of only a few centimeters after 25 years, as opposed to a few meters when using Cartesian coordinates. Further alternatives are being explored.