What number should I use for "Voice Frames per TX"?

It depends on what codec you choose and balance between bandwidth utilization and impact of packet loss. The bigger this value, the higher bandwidth utilization because more voice frames are packed into the payload field of a UDP/RTP packet and thus the network header overhead would be lower. However, the impact of a packet loss on perceived voice quality will be bigger.

For PCMU/PCMA, the default is 2 and max is 10

For G723, the default is 1 and max is 32

For G726-32, the default is 2 and max is 20

For G729, the default is 2 and max is 64