I think your 5x5x5 unit cell could be the smallest possible, which is great!
Actually, my original design sketches were for a 4X5X5 cell, one tile narrower than this design, but that design didn't have any provision for access stairs from below. The dwarves would hace had to walk all the way to the side of the map and up stairs going around the array, then down from overhead walkways to work on it. 5X5X5 let me give each cell its own access stairway, and also made drainage of excess water from the temporary register easier.
I think it's a shame though that it needs to be built on the surface. Certainly waterwheels could be substituted with some minor modifications, which would make it significantly bigger.
I considered a waterwheel-based design. Waterwheels would have given me 60 more units of power in each cell, which would have required 12 more dummy load gears in each cell if I used the same type of mechanical logic. Could have been done, but with a lot more parts required. Ultimately I chose the simplest and easiest to build design.
Have you considered any less space efficient designs that don't use floodgates and bridges? Or that require fewer steps to run? Or fluid conserving designs? Since there are so many z-levels in every map now, space is rarely in short supply, so speedier designs could be more optimal. I believe water flows to be the biggest FPS hog, rather than items.
One thing I could do to speed the mechanism up is perform the overpopulation check and underpopulation check at the same time. My design uses the same windmill/gear/pump assembly to perform both checks, with the bridges and mode switch gear reconfiguring the device between steps. If both tests were performed at once the machine would run faster, but would take about twice the space and require about twice as many parts per cell. I also considered a fluid-conserving design, where water was shunted back and forth between two storage cells to denote if a particular cell in the grid was 'alive' or 'dead'. Again, this design was chosen for simplicity of construction. In a 8X16 grid I have to build 128 individual cells, so making each cell as simple as possible was the most important design goal.
Just in the process of building this I came up with several ways the design could be improved, but this was such an insanely time-consuming megaproject to build that I'm not going to attempt anything similar anytime soon.
Oh god, I want this save. Or at least a video. Either way this needs to be shared in the name of !!science!!.
Soon. A little bit of final testing, a few finishing touches to put on it, and I will upload a save, map archive, and movies.