I did a little work modifying my autopilot a few months ago but was hesitant to post details. I've basically violated its warranty and made it a little weaker in order to achieve a tighter turning radius for the boat. This was a tradeoff that I wasn't even aware I was making back when I had the autopilot installed.
A little background. If you're considering buying an autopilot for your wheel steered boat (tillers are a whole different thing), you should be aware that Raymarine, Simrad and possibly other manufacturers, limit the angle the rudder is able to deflect to each side to around 35 degrees. In talking with the Raymarine folks, they felt that if the rudder deflected much more than 35 degrees that the forces that could be imposed on the rudder could overwhelm the autopilot and reduce its lifespan. So, they require that when you install the autopilot on the boat that you restrict the maximum rudder deflection to each side to 35 degrees.
I believe my Pacific Seacraft originally had a 55 degree maximum rudder deflection - so when the autopilot was added, I lost 20 degrees of rudder travel to each side. This is almost a 40% loss of rudder deflection. This means that the turns I used to be able to make were tighter than they were after installing the autopilot. This doesn't matter when you are out in open water sailing around. But when in close quarters maneuvering the boat to dock or move around a marina - being able to turn tightly is nice.
After realizing this limitation I started thinking about how I could compromise the system to achieve a tighter turning radius. Basically, if you end up with something that allows the rudder to deflect more than 35 degrees, you violate the warranty. In the end, I decided that I wanted to be able to maneuver the boat more easily and have modified my setup.
A few pictures to clarify how this all works.
The autopilot computer controls a linear drive which can push or pull a rod attached to a tiller arm which is attached to my rudder post. This is how the autopilot turns the rudder.
The bronze arm is the tiller, the autopilot arm on the left is attached to the tiller and the tiller is attached to the rudder post. At the top of the rudder post is the quadrant, which is how the steering cables attach to the rudder.
The steering cables wrap around the quadrant, so as you turn the wheel you pull on the cables which rotate the quadrant and the rudder turns. But clearly you don't want to be able to turn the rudder around until it hits the hull, so manufacturers build a support bar which crosses above the quadrant, and then there is a post on the forward part of the quadrant. As the quadrant rotates, the post (seen with a short hose chafe sleeve above) will hit the bar and stop the wheel from turning the rudder any further.
Originally there were no pieces of wood on the support bar, which would give the maximum deflection of the rudder - it could rotate on one side until the center post hit one side of the support bar, and then rotate to the other side until the center post hit the other side.
Without the pieces of wood in the above picture, you can get 55 degrees of travel in each direction.
After my autopilot was installed, there were two pieces of starboard where the wood is now. The starboard stops were 1 3/4" wide, which limited the travel to around 39 degrees to each side.
My modification was twofold. I moved the attachment point of the autopilot from being 10" out along the tiller to 8". This gave a larger deflection for a given amount of travel of the autopilot rod. The Simrad unit uses the same tiller arm and attaches at 8" so the tiller arm is strong enough to accommodate this new attachment point. I then found the maximum limits of travel for the autopilot and created the wood stops to have the rudder stop travel before pulling or pushing the autopilot rod out of/into the linear drive stops.
I ended up with the two pieces of wood being different sizes and my rudder now travels out to 50 degrees when turning to port and to 43 degrees when turning to starboard. I gained a little to starboard and 11 degrees to port, roughly a 30% improvement to port. The boat doesn't turn as well to port as starboard, as to starboard you can use prop walk in reverse to tighten a turn, this doesn't work to port.
After realizing the limits my autopilot imposed on me and going through this process, I started to think about my choice of autopilot. There are different autopilots made which are mounted above deck and which drive the wheel directly. These autopilots may not be as elegant, but they wouldn't impose any limit on the rudder travel. See the CPT autopilot for example. I would be tempted to go something like this route if I was to do it all over...