My solution to your problem would be to sample fast (you bought a fast card so use it), and then decimate the sampled data array and only save every N points, where N is a time varying function defined by your logrithmic function.
IE at first you save all points, then save every other point, then every third point, then every fourth point etc.
It's piecewise, but I garantee you if you are doing any interesting math on the data, you'll be a lot happier with uniformly (piecewise)sampled data rather than trying to recast functions into non-uniform (or god help you, random) sampling space.
I don't know of any way other than stopping sampling with a DAQ card to change the sample rate. But I really really like uniformly sampled data, and haven't looked at other solutions. You could use an external trigger and feed it a logrithmic decaying pulse train.
Sheldon
Technical geek, engineer, research scientist, biodegradable...