there is a big difference between planning a data network from scracth, or upgrading an exisitng network (your case).
So in your case, you want to know how to increase voice capacity, by decreasing capacity for PS.
FPDCH or SPDCH are static timeslots that cannot be preempted by voice. Usually, you can choose a value between 2 to 4. Selecting SPDCH rather than FPDCH should decrease your GPH load.
Next step is to maximize usage of each PS timeslot (pdch) :
pile up several TBF on the same PDCH
authorize EGPRS on all ts
authorize maximum throughput (mcs9)
reduce the inactivity timers, so that idle TBF are released and make room for new TBF
Then OPDCH .. it is an interesting parameter, but should not affect your voice capacity. All PDCH in the OPDCH zone can & will be preempted by voice whenever it is needed.
You can put a high value, but avoiding to stress your GPH board too much.
In your study you can also measure how many TCH are used AT MOST in your cells, and calculate how many PDCH are left. Would that be enough to provide a good-enough experience to the user ?
If number of remaingin ts is too low at the busy hour, you can then “force” Half rate to be triggered early (CSPRIO or something like this). It’s probably the most efficient parameter to restrain the voice to take too much space in the cell.