EUROLAS Workshop

Detecting and eliminating errors in the EUROLAS network

Hosted at Herstmonceux, UK, by the Natural Environment Research Council (NERC) Space Geodesy Facility (SGF) - 2002 March 11-13

Fourteen representatives from 10 EUROLAS stations, one representative from the NASA network and one from the ILRS central bureau attended the workshop. A list of attendees is given at Appendix 2. Several people brought Stanford timers so that direct intercomparisons could be made between them and the Herstmonceux timers during the course of the workshop.

The meeting was held to address issues of common interest to the EUROLAS cluster of stations in order, for example, to exploit more fully its unique capability for very precise, simultaneous tracking. Such capability is seriously compromised if one or more stations generate corrupt data from locally induced observational errors: it is particularly destructive if these errors vary with time or the circumstances of the pass. This capability should also be exploited in order to improve data quality from EUROLAS, which of course will be of benefit to the mission of the whole ILRS.

The discussion, in four half-day sessions, dealt with specific elements of the SLR technique: Detectors; Reduction/Post Reduction errors; Counters and Timers and Miscellaneous items. The following summary picks out key points of the discussion and highlights recommendations. It is hoped that some more detailed descriptions of good practice for getting best results in specific areas can be posted on the ILRS website in due course.

Introduction:

Van Husson set the scene by emphasising the keys for stations to achieve consistent millimetre accuracy:

Aggressively pursue and eliminate all site biases;

Improve the quality of data available to analysts;

Comply fully with format and data centre standards.

He pointed out that one of the prime science drivers is the desire to determine the radial component of an orbit (LAGEOS and altimetry satellites) at the millimetre level. His definition of 1mm accuracy was that the mean range bias for any pass should be £ 1mm and that the normal point scatter should be at the same level. This is not the same as saying that the single shot RMS needs to be 1mm! Thus each station's performance goal should be 1mm accuracy and repeatability in both the short-term (over a pass of, say, 50 minutes) and in the long-term (pass-to-pass, day-to-day, month-to-month).

General Principles that were accepted from the outset:

All stations should be aiming at 1 mm accuracy on a pass-by-pass basis;

Stations should have primary responsibility for data integrity - the goal must be for the user community to have 100% confidence in the data from the stations;

Stations should routinely reject suspect data instead of sending them to data centres for (possible) rejection by others;

At the same time, stations and analysts need to continue to improve their monitoring and analysis techniques to enable stations to identify and solve subtle performance problems;

Daylight ranging is essential;

Measurements should be traceable through data and calibration records.

Session 1: Detectors

What operational conditions, procedures and monitoring lead to the best results?

C-SPAD:

The C-SPAD employs a cooled detector chip and has two output channels - one "raw" output pulse and the other compensated for the effects of walk with return energy. Benefits of use and features to be treated with care were fully discussed.

Multi-photon return levels:

Uncompensated C-SPAD's outputs display time-walk at higher return energy levels. Even for the compensated outputs there is still a residual 'walk' for high signal strength returns from satellites with extended arrays - including LAGEOS, ETALON, and Topex/Poseidon. If left uncorrected, this time walk will bias the measurement results. Similar results are found in PMT's.

• C-SPAD should be tuned for zero-walk with respect to return signal strength for calibration ranging - tuning is effective only for a single pulse length (and the pulse length must be one available to the tuner - Graz);
• Stations should confirm the tuning of their C-SPADs by conducting calibration ranging experiments at many different return energies; (see, for example, Appleby and Gibbs, Proc 9^th Int. Workshop on Laser Ranging Instrumentation, Canberra, pp 92-102)
• If signal strength can be measured, shot-by-shot correction for this walk is theoretically possible;
• If signal strength cannot be measured directly, return rate, if correctly computed (see Appendix 1), can be used as proxy of signal strength, either to apply corrections or to maintain a particular regime of return energy;
• C-SPAD stations should routinely compute return rates, at the very least in order to monitor its variability and to give an indication of potential range bias;
• C-SPAD stations should perform high-low signal strength experiments during a test calibration and on a LAGEOS and ERS-2/Stella/Starlette pass (again, see Proc 9^th Int. Workshop on Laser Ranging Instrumentation, Canberra, pp 92-102); NERC SGF to assist or, if required, carry out the processing of the data from these experiments - contact Graham.Appleby@nerc.ac.uk

All C-SPAD users

• Pre-arming - arm the C-SPAD well before the satellite 'track' to avoid variable walk (>50ns for the uncompensated channel; >100ns for the compensated channel):

Stations should conduct C-SPAD arming tests to verify local practice;

• Optimising alignment of the detector minimises range bias effects and jitter;

Stations should conduct alignment tests to quantify the effects;

• Temperature - C-SPAD electronics are partially compensated for temperature change, but there is a residual walk with temperature change.

Stations should calibrate frequently when the ambient temperature is changing rapidly.

PMT:

For photo multiplier/constant fraction discriminator systems stations should:

keep returns well after (>20ns) the CFD gate;

control return rate to minimise 'walk' effects (<30%) and/or measure return pulse energy to correct times of flight with previously determined time walk model.

Session 2: Reduction/Post Reduction Errors

What can be done at stations after the pass to maximise data quality?

Ä = Send as a recommendation to an appropriate ILRS Working Group.

Stations should:

Compute useful parameters for each satellite pass in order to assess data quality. For example, NERC SGF compute (using DISTRIB subroutine, Sinclair 1995, available through EDC at ftp://ftp.dgfi.badw-muenchen.de/pub/laser/software/DISTRIB) and archive useful parameters (e.g. RMS, skew, kurtosis, LEHM-peak, LEHM = Leading Edge Half Maximum) from each reduction and flag values that fall outside the normal range for each satellite;

Be very careful about accepting pass segments widely spaced in time, especially segments with few returns, since fitting programs are flexible enough to accommodate noise;

Limits normal points to a minimum number of returns:

in daytime 6;

at night 3,

unless, for purely practical reasons, these are too high (e.g. low-repetition rate systems; some LEO satellites with very compact arrays); Ä

Compute (and control for better consistency) the return rates (keeping to the same regime) for both calibration and satellite ranging, and again at reduction time (to reject data outside return rate/energy limits);

Use a formalised algorithm for return rate calculation (outlined above and to be distributed as subroutine) Ä

Long and Short-arc quality checks are carried out daily by NERC SGF and show up problems at levels of 20mm for some stations and for some passes.

For feed back on recent passes, stations should regularly consult http://nercslr.nmt.ac.uk/slrweb/auto_analysis.html

Data Centres should:

Reintroduce full rate data archiving in MERIT II format; Ä

Formalise a procedure for withdrawal (i.e. complete removal from data centre) of bad-pass data: (stations and analysts to agree!). Ä

Session 3: Counters and Timers.

How can stations get the best out of their timers? Most of the discussion centred on the Stanford SR620 time interval counters.

Ideally timers should be located in a temperature stabilised environment and should never be switched off;

If they are only switched on for operational use, recommend that they be given a warm-up period of at least one hour before being used for data taking;

Input threshold settings on counters should be set by software and not via the front panel controls;

The outline procedure and results of counter tests carried out during the workshop are given at http://nercslr.nmt.ac.uk/current_rnd/workshop_sr.html

Stations should apply the results from their SR620 counter tests to their routine data as soon as possible;

Graz/Herstmonceux agreed, as a matter of routine, to archive and analyse counter differences on many real passes so as to monitor long-term stability and range dependencies;

Matera agreed to make direct shot-by-shot comparisons between their event timer and their SR620 to characterise the range dependency against a linear device;

Agreed that, wherever possible, the 'Herstmonceux' counter tests should be repeated at other sites;

Short-arc quality control could be used to analyse "raw" and "counter-corrected" LAGEOS passes from all stations on a test basis - recommend that stations send relevant pass data to NERC SGF for analysis.

Proposal from Ivan Prochazka

The Technical University of Prague has developed and demonstrated the portable standard concept through its P-PET system.

EUROLAS Stations agree to:

collaborate on a proposal to the EU to procure a P-PET European portable standard; and

support P-PET comparisons at as many stations as possible once a European standard is available.

Session 4: Miscellaneous

Other issues relating to establishment of Europe-wide good practice were discussed.

General:

Encourage station personnel to visit each other for the purposes of information exchange, collaboration, specialised advice and assistance and the spread of 'good practice' in all matters relating to data acquisition and processing.

Meteorological Measurements:

Blunders in meteorological measurements are not uncommon and in particular, errors in atmospheric pressure can have a very significant impact on data accuracy.

Stations should conduct regular comparisons of local measurements with duplicate sensors or nearby meteorological stations; 

NERC SGF will prepare its second Herstmonceux Druck digital pressure sensor for on-site comparisons at as many of the EUROLAS sites as possible, designing the experiments and posting the device to the nearest station for a round Europe trip.

Epoch Measurements:

Most stations use a local source of epoch driven either by a 10MHz output from GPS (synchronised by 1pps) or by a local time service that contributes to and is monitored against UTC. There appear to be no problems at any station in achieving the requisite sub-microsecond level of accuracy for epoch. However since timing errors of this magnitude would be difficult to detect, stations should:

Check their clocks routinely against an independent source of UT.

Laser Stability:

NERC SGF has carried out extensive tests on the effects on system precision and accuracy of laser temperature, dye-strength and repetition rate and found that keeping each within strict limits improves consistency - outside these limits, the observer is prevented from operating the system.  Building from this and similar experiences in the network, stations are urged to:

Carefully monitor system parameters as a health check on their operation.

Calibration techniques/station practices:

Changes in calibration value indicate that something has happened at the station that may bias the range data. Stations should:

Investigate (and understand) any significant changes in calibration value (there must be a reason for it);

Calibrate at least once per hour (unless they are doing real-time calibration), even if this means interrupting (long) satellite passes;

Perform calibration ranging in the same signal strength regime as used for satellite ranging; 

Perform calibration and satellite ranging through the same optical and electronic paths.

EUROLAS near-real-time display:

AIUB has established a near real-time display of station tracking status to: expedite recovery of manoeuvred satellites, check on 'best' IRV set and time-bias to use, dynamically prioritise satellite passes (adjust to the activities of other stations in the cluster, and exchange messages. For information on this utility, see SLRMail 372, "Near-real-time status exchange", by W. Gurtner, 13^th July 1999.

EUROLAS recommends that:

Other stations/networks join; and

The current refresh rate of display be changed from 30 seconds to 15 seconds to aid capture of fast LEO satellites;

Quality of predictions:

It was agreed for reasons of flexibility, that when available, several options for prediction sets should be accessible by the stations.  For information, NERC SGF uses:

• GFZ IRVs (for ERS-2); IRVs + drag functions (for CHAMP, GRACE);
• CODE/NERC IRVs (for GPS, GLONASS);
• MCC IRVs (for REFLECTOR);
• HTSI daily IRVs (for all other satellites);
• (NERC SGF - daily/weekly - for backup).

Time bias corrections:

A proposal by Werner Gurtner to set up a server to compute instantaneous time bias values was warmly welcomed:

Time biases will be evaluated in near real-time using the time-bias functions routinely computed and frequently updated by NERC SGF;

Values will be computed for all satellites for all available prediction sets, subdaily, daily, weekly and monthly; and will use GFZ drag functions as appropriate;

Service will be run from AIUB (in a similar form to the EUROLAS near-real-time display) with NERC time bias functions updated every 15 minutes.

NB - already implemented! - See SLRMAIL 921 for details.

Collocation results:

An analysis by Van Husson of the ITRF2000 coordinates of the reference points for the various geodetic instruments located at SLR stations showed that nearly every site has problems (at the few mm level) in at least one coordinate of the local tie vectors that need to be resolved. In the laborious task of resolving these issues:

Every significant ILRS sites should be collocated with a GPS receiver contributing to the IGS global network

Appendices

Computation of return rates:

Count the number of laser shots, s, in a given time interval - say 15 seconds;

For each of these shots, check whether a noise event has been detected 'before' the true return (i.e. count noise below the 'track') - each of these reduces by one the effective number of laser shots - get s_corrected;

Count the number of 'true' returns, r, during the same time interval;

Then the 'true' return rate  = (r /s_corrected)   ´ 100  (%).

Participants - Herstmonceux - 2002 March 11-13

Responsible Government Official: Carey Noll
NASA's Privacy Policy and Important Notices

Send us your comments

Last modified date: Friday, November 05, 2004
Author: Carey Noll
Maintained by: Carey Noll