After FEC Packet loss
be sure you read this mode-0-vs-mode-1
f0(n,m,p), a helper function for calculation; assume network packet loss is p uniformly, and you are sending n packets, the chance of >=m packets arrives out of n:
f(x,y,p), the after-fec packet loss; assume network packet loss is p uniformly, and for each x orignal packets you send y redundant packets, the change of >=y packet get lost:
f(x,y,p) = 1 - f0(x+y,y,p)
f(20,10,10%)=0.0089% means if -f20:10 is used, if the real-packet-loss is 10%, then the after-fec-packet-loss will be 0.0089%
Assumption is still: network packet loss is p uniformly, and for each x orignal packets you send y redundant packets.
g(x,y,p), the after-fec packet loss for mode 1; a given packet itself is lost, and we also lost >=y-1 packets from the rest x+y-1 packets:
g(x,y,p) = p * (1 - f0(x+y-1,y-1,p))
after-fec packet loss of mode 1 is stricting better than both:
- orignal packet loss
- mode 0's after-fec packet loss
(assuming packet loss p is fixed and uniform)