The following table shows how one axis' sensors are blocked/unblocked as the stick is moved. "B" is a blocked sensor, "O" is an open (unblocked) one.
<<< << < center > >> >>>
OOO BOO BBO BBB OBB OOB OOO
Notice the ends: both ends are the same: all sensors unblocked. How does the game know which side it's on? The "side" pin mentioned earlier tracks this on the joystick. The the side pin is added to the one axis table:
<<< << < center > >> >>>
OOOO OBOO OBBO OBBB BOBB BOOB BOOO
I stole that from Urebel's website.Using a combo of NOT and AND chips, or just a series of NAND chips, you could set up logic that would detect ALL movement off center (Left=OxxO, Right=BOxx).
It is also possible to set it to detect only second and third stage movement (Left=OxOO, Right=BOOx), or only third level movement (Left=OOOO, Right=BOOO).
For your sticks you will need to invert the output of the axis pins, and then run everything into the AND gates.
Unfortunately, I don't have the time to figure out all the pinouts to connect the two.
The biggest problem may be the connection to the JAMMA harness though.
I'm not sure how the 49-way is connected to the JAMMA harness, so you may need to force the output to different pins on the harness to hook up the two different types of games.
My fear is that the 49-way may be sending its data down the normal stick/button connections, and those signals would have to be interrupted during 8-way games, and the proper filtered signals inserted in their place.