I went through a very crude dimensional analysis here, and I think that the height comes out to be about:
the cube root of (the flow rate*the water viscosity/the mass density*the gravitational acceleration*the length)
This was assuming laminar flow.
Plugging in numbers, I got that only about 1mm height was needed.
However, that's on the high side (large Reynold's number) for laminar flow. So take this as a very approximate first estimate.
Mike W.
(published on 10/22/2007)