1. Sound velocity Sound velocity (SV) errors bend the beam paths. The classic signature is a "smile" or "frown" across the swath: outer beams curve up or down relative to the nadir. Check this early, because it shows up in a cross-line or a flat seabed patch before anything else. Practical habits: Take SV profiles often enough for the water you are in. River mouths, thermoclines and tidal fronts change faster than the schedule says. Compare the surface SV probe at the transducer against the profile's top value. A mismatch means one of them is wrong. Never apply a profile without checking its depth range covers the deepest sounding. Keep a simple SV log: time, position, file name and which lines it was applied to. When a smile appears in processing a week later, that log is how you find the cause in minutes instead of hours. 2. Sensor alignment (patch test) The patch test solves for the residual angular and timing offsets between the multibeam head and the motion sensor. The usual order is latency, pitch, roll, then heading. The order matters because each error masks the next. Run it on a suitable slope or a distinct object, in conditions similar to the survey. Repeat it after any change to the installation, and after the vessel has been mobilised or remobilised. File the results with the raw data. A patch test nobody can find later is as good as none. Compare the new values with the previous patch test on the same installation. A large jump in roll or pitch is a reason to ask what changed on the mount, not just a number to type into the software. 3. Static offsets and draft Before any calibration, the lever arms between the reference point, the GNSS antennas, the motion sensor and the transducer have to be right. These come from the dimensional control survey of the vessel. A wrong offset will not always show up in the patch test, because the test can partly absorb it into an angle. Check the offsets against the drawings and the measured survey, and confirm the sign convention your acquisition software uses. Draft and squat are easy to forget. Read the draft marks at the start and end of the day, and remember that the transducer sits lower as fuel and water are used up. On small vessels, a change of a few centimetres is already a visible part of your vertical budget. 4. Motion and heading Heave, pitch, roll and heading come from the motion sensor and gyro/GNSS heading. Bad heave shows up as a regular ripple along track. Bad heading shows up as feature smearing at the swath edges. Watch the motion data quality indicators, not only the depth display. 5. Vertical reference: tide or GNSS height If you reduce to chart datum with tide, the gauge, the datum offset and the time base all have to be right. If you use RTK or PPK heights, the geoid or separation model and the antenna-to-waterline offsets matter just as much. A consistent vertical bias between lines is usually a reduction problem, not a sonar problem. 6. Positioning and timing A fixed time offset between position and ping data creates along-track shifts that grow with vessel speed. Check the time stamping chain (PPS, NMEA latency, network delay) rather than assuming it is fine. 7. Check lines and crossline analysis Cross-lines are the cheapest independent check you have. Run them early, compare against the main lines, and look at the statistics, not just the colours. Repeat when conditions change. 8. Data cleaning: what not to delete Cleaning is where good data gets damaged. Automatic filters are useful, but they cannot tell a boulder from a spike. Before you reject a cluster of soundings, look at it in more than one view and ask whether it could be a real object: debris, a pipeline, a wreck or a rock outcrop. On construction and route surveys, those "spikes" may be exactly what the client is paying you to find. Keep the rejected soundings flagged, not deleted, so someone else can review your decision. Write down which filters and settings you used. If two processors clean the same block differently, the difference should be explainable. A quick guide to error signatures Smile or frown across the swath: sound velocity. Outer beams tilted to one side on a flat seabed: roll offset. Features shifted forward or backward when lines are run in opposite directions: pitch offset or latency. Features rotated or smeared at the swath edges: heading offset. Regular ripple along track: heave or motion sensor problem. Steps between lines or between days: vertical reference, tide or draft. None of these is proof on its own, but each one tells you where to look first. What "good enough" means Fit the check to the specification. For IHO S-44 (6th edition) the allowable total vertical uncertainty is computed as sqrt(a² + (b × d)²) at 95% confidence, where d is depth. Exclusive Order uses a = 0.15 m and b = 0.0075, Special Order uses a = 0.25 m and b = 0.0075, Order 1a/1b uses a = 0.5 m and b = 0.013, and Order 2 uses a = 1.0 m and b = 0.023. Know which order your contract requires before you decide a residual is acceptable. As a quick example: at 20 m depth, Order 1a allows sqrt(0.5² + (0.013 × 20)²), which is about 0.56 m. Special Order at the same depth allows about 0.30 m. The difference is large enough to change how often you take SV profiles and how strictly you watch the motion data. A simple end-of-line habit Before leaving the line: check coverage, look at the swath in cross-section, compare to the nearest cross-line, and note anything odd in the log. Five minutes per line saves a reprocessing campaign. The same habit applies at the end of the day. Spend a few minutes reviewing the whole day's data as one surface. Problems that are invisible on a single line, such as a slow drift in heave or a tide step, often become obvious when you see all the lines together.