WSJT-X/lib/ftrsd/ftrsd_paper/bodide.f90
Steven Franke ebccd2267a Computes theoretical probability of error for a bounded-distance decoder.
git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@6302 ab8295b8-cf94-4d9e-aec4-7959e3be5d79
2015-12-22 00:08:31 +00:00

55 lines
1.5 KiB
Fortran

program bodide
! Compute probability of word error for a bounded distance decoder.
! Hardwired for non-coherent 64-FSK and the JT65 RS (63,12) code on GF(64).
!
! Let ps be symbol error probability.
! The probability of getting an error pattern with e symbol errors is:
! ps^e * (1-ps)*(n-e)
! The number of error patterns with e errors is binomial(63,e)
! Overall probability of getting a word with e errors is:
! P(e)= binomial(63,e)* ps^e * (1-ps)*(n-e)
! Probability that word is correct is P(0 to 25 errors) = sum{e=0}^{25} P(e)
! Probability that word is wrong is 1-P(0 to 25 errors)
! P_word_error=1-( sum_{e=0}^{t} P(e) )
!
implicit real*16 (a-h,o-z)
integer*8 binomial
integer x,s,XX,NN,M
character arg*8
nargs=iargc()
if(nargs.ne.1) then
print*,'Probability of word error for noncoherent 64-FSK with bounded distance decoding'
print*,'Usage: bounded_distance D'
print*,'Example: bounded_distance 25'
go to 999
endif
call getarg(1,arg)
read(arg,*) nt
M=64
write(*,1012)
1012 format('Es/No P(word error)'/ &
'----------------------')
do isnr=0,40
esno=10**(isnr/2.0/10.0)
hsum=0.d0
do k=1,M-1
h=binomial(M-1,k)
h=h*((-1)**(k+1))/(k+1)
h=h*exp(-esno*k/(k+1))
hsum=hsum + h
enddo
ps=hsum
hsum=0.d0
do i=0,nt
h=binomial(63,i)
h=h*ps**i
h=h*(1-ps)**(63-i)
hsum=hsum+h
enddo
pw=1-hsum
write(*,'(f4.1,4x,e10.4,4x,e10.4)') isnr/2.0, ps, pw
enddo
999 end program bodide