The first stage is quite simply one of these http://www.technobotsonline.com/re-540- ... -11-1.html which is an easy way to get the motor speed down to a reasonable rate for the worm stage (which isn't well suited to the ~10,000rpms of the motor). With the como drills range the ratios ‰ 11:1 usually have larger MOD spur gears than 50:1 and above, hopefully better able to take the torque.
Then the worm stage is comprised of one of each:
http://uk.rs-online.com/web/search/sear ... &R=5216963
http://uk.rs-online.com/web/search/sear ... &R=5216890
for a reduction of 40:1. Worms allow a great reduction in a small space, at the price of lots of friction. You can see in the above picture roughly how I'll be housing this, of course with lots of bearings and additional shafts
The final stage is simply a chain stage (9T sprocket to 17T sprocket) which will be rigged on the side of the housing, I may get some bigger sprockets though as I'm dubious that 6mm chain will stand up to shocks on the arm.

This set-up, using the standard motor and ignoring inefficiencies would provide ~22rpm no load output at 14.8v with ~10kg/m stall torque. The motor I'll probably be using is rated at approx 40% greater rpm at ~3 times the stall torque (with an awful lot more current and heat!) http://www.robotmarketplace.com/product ... P5737.html