I believe that I could recast the problem into a different scheme. Are you suggesting that I implement a finite difference method to solve the equations? I will think about this approach a little more. I suppose that it would not be too difficult to write a simple Euler or Runge Kutta where I could explicitly specify a dt. NDSolve has a StepMonitor option, but am I correct in assuming that you can not fix the step size because it is adaptive.