Operating Limits of a Two Axis Beam Type Airborne Gravity Meter Abstract: John Halpenny Airborne gravity measurements require a system that is accurate over a wide range of aircraft motions. The Geodetic Survey Division of Geomatics Canada operates a Lacoste & Romberg air/sea meter which was originally optimized for marine operations, but which has been used in airborne surveys on a number of different aircraft. This paper examines the limits to vertical and horizontal motion which can be tolerated by this meter and still keep within its operating range. Two cases are examined here. Tests in a Cessna Caravan by Sander Geophysics Ltd., described by Brunton et al. at this conference demonstrate hand flying operations at slow speed in turbulent conditions, while flights in a NRC Convair illustrate much faster and smoother flight with an autopilot. Both flight regimes create conditions where the meter can go off level or move the sensor mass against its stops. This paper discusses techniques for minimizing the problems and lays out the ultimate and practical limits of motion during a survey. This work makes use of data extracted from various test programs and is part of GSD s ongoing study of airborne gravity. Introduction The Lacoste & Romberg air-sea meter was developed in the 1960 s to measure gravity from a moving platform. Over 100 meters were built and were used primarily on ship surveys, although several were tested in aircraft. Our organization, then the Dominion Observatory, purchased S-56 in 1970 and used it for marine surveys for many years. In 1980, the sensor was replaced with a new linear sensor, SL-1, and much of the analog control system was replaced by a computer. In 1989 the DEC LSI-11 control computer was replaced by a desktop PC running QNX, resulting in much more reliability and less cost. Since then we have made several upgrades to the computer system and electronics system, but we have kept the original platform, gyros and accelerometers. In addition to marine surveys, the meter has been operated in airborne tests with Sander Geophysics Ltd. (SGL), Intera Kenting Ltd., and the US Naval Research Laboratory. We have performed one production airborne survey over the Arctic ocean with the Institute for Aeronautical Research (IAR), part of the Canadian National Research Council. Theory of Operation Operation of the linear sensor is similar to operation of Lacoste s older beam type sensors. A proof mass, usually called the beam, is supported by a spring and a measuring screw, and the screw is adjusted by a motor until the spring force just equals gravity. A feedback system moves the screw to keep the beam centered, and a system of dampers keeps it from hitting the stops during rapid acceleration. When the beam is stationary, the screw position represents the gravity reading, but under most conditions the beam is in motion and its velocity is used to correct the gravity. The screw reading is very accurate and stable, but the beam velocity correction is less accurate when there is a lot of motion.. The limits of the beam motion are set in the factory to provide adequate motion range and sensitivity. On our system, the beam moves from center to stop after an acceleration of 30,000 mgal for 7.5 seconds, or a velocity change of about 2 m/sec. In practice, the allowed velocity change would be smaller since the beam may not be centered and it should not operate near the stop. The sensor is held level in a gimbaled platform supported by bungee cord shock mounts. Two Honeywell GG-49 gyros measure the rotation about the two horizontal axes and torque motors keep the sensor steady. 556
A pair of Systron Donner accelerometers measure horizontal accelerations and the sensor is slowly rotated until there is no horizontal gravity component. The platform leveling period may be adjusted but is usually set to 4.75 minutes for aircraft operation. This will keep it level on most lines but will cause it to tip over when the aircraft turns, and it will take several minutes to come back to level. During this time a level correction may be calculated by comparing the accelerometers and GPS horizontal accelerations, but it is not as accurate as flying with a truly level platform. The platform will stay level during a turn if the platform period is set to a large value, which effectively turns off the accelerometer inputs. An operator or navigation system switch is set when a turn occurs and reset when it is finished so the leveling process can continue. The meter originally had a third axis which rotated the entire platform and could be used as a complete Schuler tuned navigation system, but it was mechanically large and complex and took a long time to align. The third axis was later removed. Operating limits The performance of the beam and level systems will be illustrated by three different flights. The flight on April 25, 0 in SGL s Cessna Caravan shows slow flight during typical daytime low level turbulence. Speed is less than 50 m/sec., the air is rough and an autopilot is not used. A typical smooth flight was taken from the Arctic survey in the IAR Convair on Sept 28, 1998 and is flown at about 110 m/sec. using an autopilot. A very smooth flight was made during a test by the Convair in the Ottawa area on Feb 7, 1. This used an autopilot in smooth early morning conditions and was flown at about 85 m/sec. 560 520 480 440 400 0 120 240 360 480 15:30:00 Apr 21 Seconds 15:40:00 height raw beam Figure 1 : Height and Beam Position On April 25 Figure 1 is a ten minute segment taken from the SGL s Cessna Caravan on April 25. The upper trace shows height changes of up to 40 m. and vertical velocities of up to 1.5 m/sec. The lower trace is the beam position, with the beam stops being at the limits of the chart. The beam motion does not hit the limit, but it comes quite close. If the beam does hit the stop, the reading will be wrong and there will be a data gap the length of the filter. It can also be seen that if the beam is not centered to start with, the motion will cause it 557
to hit a stop. Unfortunately SL-1, unlike the older Lacoste meters, does not have a clamp which the operator can use to bring the beam to the center position at the start of a line, but must rely on the spring tension adjustment to move the beam. 400 360 320 280 240 0 120 240 360 480 13:30:00 Sep 28 Seconds 13:40:00 raw beam gps ht Figure 2: Height and Beam on September 28 By comparison, Figure 2 shows a flight over the arctic ocean in NRC s Convair with an autopilot. Although the largest bump gives a vertical motion of nearly 1 m/sec, the height variations are less than 5 meters and the response of the meter is much more subdued. There is no problem with the beam approaching the stops and the accelerations are much smaller. 558
560 520 480 440 400 0 120 240 360 480 12:40:00 Jan 6 Seconds 12:50:00 height raw beam Figure 3: Height and Beam on Feb 7 Figure 3 shows the meter operation under very smooth conditions. The beam is centered and there is little correction to the reading. These are the conditions which perhaps give the best gravity traces. Platform Level When there is no horizontal acceleration, the meter platform will level itself and stay level. The motion is a damped oscillation with a period of 4.75 minutes, which was set to allow a shipborne meter to settle down in a reasonable time but be insensitive to wave motions. The period is adjustable, and can be set to any value, including 84 minutes for a Schuler tuned platform, or infinite, entered as 999 minutes, which means that the accelerometer inputs are ignored and the platform is fixed in space. Longer periods increase the setup time and thus are only useful if the system is already level. On any reasonably straight line the meter levels itself within a few minutes and stay that way. During a turn, the sustained horizontal acceleration is read as an error, the platform tilts to compensate and only becomes level again after several minutes. It is possible to detect a turn using the accelerometers and a navigation input and ignore the sideways acceleration. However, because there are only 2 axes, the platform turns with the aircraft and changes the earth rate component in each axis, so it will still tip after the turn is finished. 559
1000 1000 0 480 960 1440 1920 2400 15:40:00 Apr 25 Seconds 16:20:00 c acc l acc fa5m faofff Figure 4: Accelerations on rough flight In Figure 4, taken in a test area flight, the accelerations are shown during a flight segment that includes a turn. The cross acceleration, in dark red, is larger than the long acceleration and is quite large during the turn. Also notice that it does not return to zero immediately after the turn, but takes some time to settle out. The aircraft bank angle was kept to 10 degrees to avoid losing any GPS satellites, and the acceleration is small so there is not a clear distinction between lines and turns. The meter tries to use the GPS cross acceleration to distinguish turns from lines, but it is just as noisy as the accelerometers 560
1000 1000 0 480 960 1440 1920 2400 13:50:00 Sep 28 Seconds 14:30:00 c acc l acc fa5m faoff Figure 5: Accelerations on Sept. 28 In the smooth flight and turn of Figure 5, the sideways accelerations are much smaller during the line and the level corrections are insignificant. However, notice that the cross acceleration goes to saturation during the turn. The accelerometer is limited to 0.25 g, which is equivalent to a 15 degree turn, and most aircraft maneuvers are larger than that. The gravity does not change very much during the turn but the off level correction, based on the saturated accelerometer, goes off the scale. When the line begins, the meter is only slightly off level and is much more easily corrected. In this case, the automatic turn correction has shut off the platform and has operated reasonably well due to the fact that there is a very clear distinction between turns and lines. 561
1000 1000 0 480 960 1440 1920 2400 11:45:00 Seconds 12:25:00 c acc Feb 7 l acc fa5m faofff Figure 6: Accelerations on Feb.7 Figure 6 shows a line that is not only smooth, but the turn has been limited to 10 degrees and the accelerometers do not saturate. The turn cutoff has worked well and the platform is almost level after the turn. The turn produces a lump in the (green) gravity trace and there is an off level correction that unfortunately does not improve the performance very much. 1000 Figure 7: Improved level correction 1000 0 480 960 1440 1920 2400 11:45:00 Seconds 12:25:00 c acc Feb 7 l acc fa5m faofff 562
The level correction has been greatly improved in figure 7 by adjusting the offset of the long accelerometer. However, this offset is slightly different from the one found during the platform calibration, which may indicate that it drifts, probably with temperature changes. Conclusions The performance of our gravity system is affected by aircraft motion. During smooth flight, the gravity trace is stable and few corrections are needed. The system can operate during daytime turbulence with height excursions of up to 20 meters and vertical accelerations of 1.5 m/sec, but there is the possibility of hitting a stop and losing data, and great care is needed in calibrating the accelerometers. Rough flights also cause more noise in the GPS position and require much more care in matching the meter and GPS times. The meter is also supported by fairly soft bungee cords and the sensor may move significantly with respect to the aircraft. The meter data is recorded at 10 hertz and the GPS data for the more recent flights is collected at the same rate. However, the gravity processing is done at 1 hertz, and there may be distortion when the vertical accelerations and gravity are lined up on rough days. I plan to reprocess some flights at the higher rate and I expect some improvement. The accelerometers are the oldest part of the system and are not temperature controlled. They are adequate for smooth flights, but could do better in rough air if they were moved into the temperature controlled meter case, or perhaps replaced entirely. 563