Title: Improved Geolocation Accuracy For POES Imagery 11 December 2002
1Improved Geo-location Accuracy For POES
Imagery11 December 2002
NOAA/NESDIS Office of Satellite
Operations Office of Systems Development
Peter Phillips Cynthia Hampton James
Valenti
2Topics
- Basics of POES Geo-location and Clock Management
- The Problem
- Legacy Architecture and Findings
- Corrective Actions
- New System Details
- Calibration and Validation
- Operational Implementation
- User Impact
3Geo-location and Clock Management
- Timing information (Day of Year and Millisecond
of Day) is embedded in POES imagery data - Imagery users geolocate pixels by projecting this
time onto an ephemeris-based map of spacecraft
location - POES spacecraft have no internal means to know
what time it is--time is set by the ground, and
an on-board crystal oscillator provides pulses to
advance the clock - The on-board oscillator drifts relative to true
time, requiring the NOAA/NESDIS Office of
Satellite Operations (OSO) to measure the
difference between the spacecraft time and a true
timing reference--called a clock delta - OSO clock delta measurements are used in two
ways - Provided to users to correct timing information
prior to geolocation - Used to correct the spacecraft clock to true
periodically
4The Problem
OSO-measured NOAA-15 and -16 clock deltas did not
match imagery In example, OSO measured a clock
delta of -400 ms, but imagery showed it was 1200
ms This caused a geolocation error of over 10 km!
5Legacy Findings
- Though synchronized to Global Positioning System
(GPS) time, clock delta measurement system only
processed spacecraft and ground reference times
to nearest 100 milliseconds - Expected clock delta measurement error of 100
milliseconds could not explain magnitude of
problem - OSO and users had image navigation software which
performed best fit of pixels using coastlines - For 40-day period in 2001 where OSO clock delta
measurement for NOAA-16 remained constant at -600
milliseconds, average clock delta from navigation
software was 440 milliseconds - Navigation results normally distributed, with
standard deviation of 600 milliseconds - Results consistent between OSO and users
6Legacy Findings(continued)
- In 1999, NOAA-15 on-orbit tests to set spacecraft
clocks resulted in time being 500 milliseconds
off target value on 5 of 8 attempts - Clock delta measurement system extracted time
code information from different spacecraft data
stream than users - Used TIROS Information Processor (TIP)
data--spacecraft time code only available once
every 32 seconds - Users extract time code from High Resolution
Picture Transmission (HRPT) frames, which is
available 6 times per second
7Corrective Actions
- NASA engineers discovered bug in NOAA-15 and
NOAA-16 on-board clock-setting software - Used 1 Hz instead of 2 Hz reference to determine
hardware cycle - Corrected via flight software patch in 2000
- In mid-2001, NOAA/NESDIS Office of Systems
Development (OSD) engineers discovered incorrect
Polar Frame Synchronizer (PFS) TIP data blocking
setting for KLM spacecraft - Caused clock delta error of -900 milliseconds
- Reconciling brought clock deltas to within 1? of
navigation results - OSD procured new clock delta measurement system
in 2002 - Part of PFS upgrades to Wallops and Fairbanks
Command and Data Acquisition Stations (CDAS) - Goal was to keep spacecraft clocks within 75
milliseconds of GPS reference--equal to dimension
of 1 high-resolution pixel
8New System Details
- Uses HRPT data stream
- PFS receives ground timing reference from GPS
receiver and performs internal Time Stamping as
follows - Extracts spacecraft time from header of every
third frame of HRPT - Latches GPS-based Ground Receipt Time (GRT) to
end of frame - Passes spacecraft time/GRT data pair to main
ground system computer - Since Time Stamping is internal to the PFS,
blocking and transmission delays no longer impact
clock deltas - Spacecraft time and GRT processed with
1-millisecond precision
9New System Calibration
- New system includes settable GRT offset
- Accounts for delay from spacecraft time
extraction to GRT latch - Must be correct for system to produce accurate
clock deltas - Testing at WCDAS compared new to legacy system
- Legacy system had proper PFS blocking factor for
TIP data - New deltas highly consistent within and between
contacts - New deltas differed from legacy by 1000
milliseconds in all cases - PFS vendor found 640 millisecond delay in GRT
output - 360 milliseconds of difference remained between
new and legacy - WCDAS tests in January 2002 of MIT/Lincoln Labs
clock delta measurement system recorded similar
differences with legacy - Inspection of legacy code showed incorrect block
transfer delay term--value of -400 milliseconds,
but should be -750 milliseconds
10New System Validation
- Timing system calibration validated by Aerospace
Corporation and MIT/Lincoln Labs - Post-installation testing at WCDAS and FCDAS with
corrected GRT offset showed clock deltas
consistent with initial test results and
identical between stations - Navigation of imagery with spacecraft time set to
within 75 milliseconds of true validated by OSO
HRPT ingest system - Navigation also validated in Local Area Coverage
(LAC) data by Air Force Weather Agency (AFWA)
11Operational Implementation
- New system clock deltas first used operationally
for NOAA-17 - Spacecraft time set to within 2 milliseconds of
true on day following launch - Users report excellent geolocation of NOAA-17
imagery - Clock deltas form basis for daily clock
corrections to compensate for on-board oscillator
drift - Archived clock delta information used to
determine rate, in milliseconds per day, of drift
relative to GPS reference - Daily 24-hour clock decrement term in spacecraft
stored command table modified to include drift
rate correction - OSO maintains spacecraft clock deltas to 75
milliseconds - By August 2002, OSO using new system for clock
management of NOAA-14, 15, 16, and 17 spacecraft
12Operational Implementation(continued)
13User Impact
Direct Data users can now use POES imagery
directly, without any need for post-ingest
navigation to correct for timing errors!
14Backup Slides
15Navigation Histogram
16Legacy Architecture
17New Architecture
18PFS Blocking Cycle
19Legacy Clock Delta Equation
20New Clock Delta Equation
21Calibration Details
- PFS software includes settable GRT offset
- Accounts for delay from spacecraft time
extraction to GRT latch - Must be correct for system to produce accurate
clock deltas - Initial testing at WCDAS compared new to legacy
system - Legacy system had proper PFS blocking factor for
TIP data - GRT offset was -173 milliseconds--length of 1
HRPT frame plus average link transit time from
spacecraft to ground - New deltas highly consistent within and between
contacts - New deltas differed from legacy by 1000
milliseconds in all cases - PFS vendor notified of test results and reviewed
design - Found delay in output of ground time code
following receipt of IRIG-B clocking signal - Delay was 640 milliseconds, causing GRT to be
less than expected at time of latch
22Calibration Details(continued)
- 360 milliseconds of difference remained between
new and legacy - WCDAS tests in January 2002 of MIT/Lincoln Labs
clock delta measurement system recorded similar
differences with legacy - Inspection of legacy clock delta calculation code
showed incorrect offset term to account for block
transfer delay--value of -400 milliseconds, but
should be -750 milliseconds - Findings explained differences, justified change
of GDP 225WA GRT offset from -173 to 467
milliseconds - Post-installation testing at WCDAS and FCDAS with
new GRT offset showed clock deltas consistent
with initial test results and identical between
stations - Calibration validated by Aerospace Corporation
Geolocation Study and MIT/Lincoln Labs