that can help a little, it is simply the size of the coil that allows you get some performance in a test, depending upon the size of the container that you might use with the 3" your describe as "magnetic gravel and black sand."
If you are trying to get a true-case evaluation of a detector and coil I suggest the sample material either be a typically encountered 'object,' such as a dirt clod or rock, such as my "Homestead Rock" from a Central Oregon homestead site where they was a decayed structure and some building walls and chimney rocks were scattered about. Those rocks are usually flattish, so as to stack easier, and not round, and my Homestead Rock measures about 6" wide X 9" long X 2" thick at the center.
When I want to compare detector/coil performance for challenging round, I never like to use a smaller plastic or wooden or cardboard box with sample dirt. Instead, the soil make-up is either in its natural state covering a natural area of ground, or a mixture I put in a larger-size plastic tub that is at least 2'X3', and the test sample target is located under the direct center, not close to an edge. Then, when the sample soil is covering a larger-size area, it eliminates the cheater-performance of a larger-size and/or DD coil.
Just some thoughts, not knowing what you had in mind for a sample soil container. As a rule, when dealing with an absolute natural condition of really bad ground, it eliminates possible errors and requires that the make and model detector, when used in the motion-based Discriminate mode be adjustable to a TRUE All Metal accept or Zero Disc. setting. That would be models such as the Compadres, Eldorado