|Chesapeake Times, Vol 5 | April 2021|
In SonarWiz 7.07.05 we will be introducing several changes to the sidescan and sub-bottom gain dialogs to make them easier to use and to support a new workflow for EGN. We’ve also removed an undocumented normalization step in gain processing that dates back to the 8-bit era of SonarWiz. Taken together, these changes may affect how your processed data looks, especially if you stack gains to achieve a particular effect. We hope that the changes make it easier to understand how gains are being applied to your data.
A bit of SonarWiz History
Until recently, most sidescan and sub-bottom customers were mainly interested in producing beautiful imagery. The specific intensity values were not interesting so long as the relative image was sharp and well defined. SonarWiz gain algorithms were originally developed to make relative changes to the ping intensity values. Gains could be stacked, and if too much gain was applied to a portion of a ping such that it overflowed, rather than clipping the data, the algorithms were designed to rescale or normalize the data so that the whole dynamic range was compressed to fit the display. It was possible to apply gains—even unreasonable gains—and yet still produce an image that looked ok.
The problem with SonarWiz normalizing the gains behind the scenes was that it made it difficult to trace an intensity value from raw to processed if an undocumented normalization step was applied to the data. In some pathological cases, it was possible to increase the gain on one line, expecting it to get brighter, but SonarWiz would rescale the whole data set such that all the other lines got darker instead!
Today, clients are paying increased attention to how the digitized intensity values are affected by signal processing, and the modern display architecture of SonarWiz no longer requires the dynamic range to be limited to 8-bits. So, we have converted most gain algorithms to work directly with floating-point intensity values and we have removed most of the hidden normalization steps. The user gain controls will now apply gains exactly as the user interface indicates. If you stack gains, expect the intensity values to increase accordingly.
User Gain Controls (UGC)
The User Gain Control settings have been enhanced to make it easier to use. You will notice that the graph axes are clearly labeled, that you can toggle the X-axis from 2-way travel time to sample range, and that the cursor position always indicates the gain value that will be applied to the raw intensity at that point in the ping. We’ve also reduced the vertical scale of the graphs so that more reasonable gain settings are offered.
You will find similar changes to UGC in the sub-bottom gain dialog. In addition, cursor tracking is applied between the gain graph and the SBP signal trace so that you can match up exactly where you want the gain applied:
Empirical Gain Normalization (EGN)
There are two new changes to the EGN user interface: 1) We have introduced a way to build and apply individual EGN tables for each trackline in the dataset, and 2) we have exposed a smoothing parameter to the EGN tables that gives the user more control over how sensitive the algorithm is to beam-to-beam changes in average intensity.
EGN was originally designed so that all files in a survey are added to a single EGN table and the user interface in SonarWiz was designed to support this. However, under some survey conditions, building a new EGN table for each track line produces better results. In this mode, you are using EGN much like beam angle correction. In prior versions of SonarWiz, creating a new table for each file was a tedious manual process. In SonarWiz 7.05.05, we’ve added a new check box to the table creation dialog that will automatically create and assign a new EGN table for every file in the project if desired.
The second new option for EGN is the surfacing of a table smoothing parameter. EGN correction tables are stored by beam angle and range. The correction for each sample is determined by interpolation on the EGN table. We found that in some cases, neighboring nodes in the table of corrections were sensitive to outliers and this would result in along-track stripes in the processed data. Smoothing the correction table reduces the artifact significantly. But, as with all things, sometimes smoothing introduces artifacts of its own. By allowing advanced users to control the smoothing or even to turn it off entirely, EGN will work well on more data sets.
If you are interested in learning more about SonarWiz gain algorithms, we recently gave an in depth webinar on each of the available algorithms. You can watch the webinar on YouTube: Cocoa with CTI: Sidescan Gain Settings
– David Finlayson, Chief Scientist