That depends on the equipment being used. In a Harrick 2.5" Plasma Chamber I use 40 seconds on high RF power (18 watts I think). I don't use O2 in the chamber but I do crack the needle valve to let just enough air in to make the plasma a bit brighter. I also like to use RF corona to bond PDMS-PDMS. I actually have higher yields with corona than vacuum plasma. In the end, you will need to optimize your process in your lab settings with your equipment. I have noticed that what works in some labs does not work in others and lab conditions seem to play a role in bond strength.

What Ron Roisterer said it very true. I have used multiple plasma chambers in different labs and you have to test them all out. I would suggest starting with 0.5 mbar (air not O2) at the highest RF power (maybe adjust the needle valve to maximise brightness). Try various times from 20 s to 60s. Place the two bonded pieces together for 30min on a 60C hotplate. Then compare your results for the different times. 

I have also found that you can also over expose the PDMS to the plasma resulting in poor bonding