There are a few things that may result in an abrupt offset from one profile to the next.
1) The probe orientation must be completely vertical. A non-vertical orientation results in the geometry being such that the same horizontal component of flow is not measured at each bin in the profile (so you typically end up with a sloped profile), so the profiling area is not located where you would expect it to be. Also, make sure that there are no changes in tilt or heading while moving the probe vertically.
2) Insufficient data collection times are used. Local variations in the flow need to be averaged out over the collection period to produce consistent velocities. The flow conditions must also be checked to ensure that they’re the same each time the probe is moved (e.g. if there is a long time between measurements, it’s possible that the actual flow may have changed. This comes down to “knowing” how your flume operates.)
3) The raw data MUST be properly filtered and analyzed to make sure that it is “good data”. Including bad or noisy data typically pulls the averaged velocities for a bin towards zero.
4) The signal-to-noise ratio is not the same for all bins in the profile. The geometry of the overlap region between the active and passive transducers results in the SNR dropping towards the end points of the profile. If there are insufficient carriers in the water column to provide a high-quality signal over the full profiling range, the range may have to be restricted to only those bins that exhibit good SNR. Check also the correlation values.
5) Look for weak spots caused by acoustic interference. When you use the same ping timing for a large range of stacked profiles, it is pretty much inevitable that weak spots will be encountered. These also tend to appear as bins in which the velocities appear to be pulled towards zero with higher-than-expected SNR/correlation values. These can be removed or reduced by using the “max ping” algorithm and adjusting the velocity range appropriately. The “Adaptive once” ping timing can also be used, but the current algorithm tends to quickly resort to extended velocity timing, which requires better SNR to operate correctly.
6) If none of these things apply to your data, please contact Nortek support. Your Vectrino Profiler may need recalibrating.