I'm the above-mentioned person using Magellan GPS underwater. (You may have seen my
recent article in Underwater Magazine). A free eBook is available at:
http://groups.msn.com/divergps *(a non-commericial site...nothing for sale).
I use one of 3 possible deploment methods, depending on depth and conditions. Lately, I've been "towing" a clear housing (Otterbox 9000) on the surface...the housing is only pulled below briefly to "mark" a point, or to gain awareness of the boat's position.
Here's the set up... The housing floats on the surface with an attached nose cone (to reduce drag)...this is followed (on the surface)--by about 25 ft. of bouyant poly-line....which is attached to a unique "no drag" dive flag. I simply tug lightly (to get the housing to move forward to the desired position)--then, pull down in a hand-over-hand position...gain data...drop the housing (which returns to the surface).
I've been using different variations of this method for about 6 years with good results.
As stated...the above-mentioned web site will provide further details, and further methods of deployment. *Be sure to see the photo section. **To answer a frequently asked question: It's often asked if this method allows the housing to drift to a distant point (rendering a mark innacurate)...not a problem, in this applicaton...it's where the housing is pulled below (losing signal) that matters. This is easily controlled by the diver. Most divers are not in the water enough to bother, but researchers and the most avid divers can have plenty of fun with this! You can test the basic theory on land: Take a Magellan gps and set up a land simulation. You simply cover the antenna with alluminum foil (to simulate periods of signal loss underwater). In real world applications (after significant periods of signal loss)...it takes about 60 seconds to get a fresh signal. However, in the above described surface-tow method, the wait is avoided, and data is available in mere seconds. Constant signal?---towed antenna??...that's fine if you want to go to the trouble, but I'm trying to use the most basic method possible.