mirror of
https://github.com/saitohirga/WSJT-X.git
synced 2024-12-02 17:15:11 -05:00
96 lines
2.7 KiB
Fortran
96 lines
2.7 KiB
Fortran
|
subroutine decode65a(dd,npts,newdat,f0,nflip,mode65,sync2,a,dt, &
|
||
|
|
nbmkv,nhist,decoded)
|
||
|
|
|
||
|
|
! Apply AFC corrections to a candidate JT65 signal, then decode it.
|
||
|
|
|
||
|
|
parameter (NMAX=60*12000) !Samples per 60 s
|
||
|
|
real*4 dd(NMAX) !92 MB: raw data from Linrad timf2
|
||
|
|
complex cx(NMAX/8) !Data at 1378.125 samples/s
|
||
|
|
complex c5x(NMAX/32) !Data at 344.53125 Hz
|
||
|
|
complex c5a(512)
|
||
|
|
real s2(66,126)
|
||
|
|
real a(5)
|
||
|
|
logical first
|
||
|
|
character decoded*22
|
||
|
|
data first/.true./,jjjmin/1000/,jjjmax/-1000/
|
||
|
|
data nhz0/-9999999/
|
||
|
|
save
|
||
|
|
|
||
|
|
! Mix sync tone to baseband, low-pass filter, downsample to 1378.125 Hz
|
||
|
|
dt00=dt
|
||
|
|
call timer('filbig ',0)
|
||
|
|
call filbig(dd,npts,f0,newdat,cx,n5,sq0)
|
||
|
|
call timer('filbig ',1)
|
||
|
|
|
||
|
|
! NB: cx has sample rate 12000*77125/672000 = 1378.125 Hz
|
||
|
|
|
||
|
|
! Find best DF, f1, f2, and DT. Start by downsampling to 344.53125 Hz
|
||
|
|
call timer('fil6521 ',0)
|
||
|
|
! Add some zeros at start of c5 arrays -- empirical fix for negative DT's
|
||
|
|
nadd=1089
|
||
|
|
c5x(:nadd)=0.
|
||
|
|
call fil6521(cx,n5,c5x(nadd+1),n6)
|
||
|
|
n6=n6+nadd
|
||
|
|
call timer('fil6521 ',1)
|
||
|
|
|
||
|
|
fsample=1378.125/4.
|
||
|
|
a(5)=dt00
|
||
|
|
i0=nint((a(5)+0.5)*fsample) - 2 + nadd
|
||
|
|
if(i0.lt.1) then
|
||
|
|
! write(23,*) 'i0 too small in decode1a:',i0,f0,a(5),fsample,nadd
|
||
|
|
! flush(23)
|
||
|
|
i0=1
|
||
|
|
endif
|
||
|
|
nz=n6+1-i0
|
||
|
|
|
||
|
|
! We're looking only at sync tone here... so why not downsample by another
|
||
|
|
! factor of 1/8, say? Should be a significant execution speed-up.
|
||
|
|
call timer('afc65b ',0)
|
||
|
|
! Best fit for DF, f1, and f2
|
||
|
|
call afc65b(c5x(i0),nz,fsample,nflip,a,ccfbest,dtbest)
|
||
|
|
call timer('afc65b ',1)
|
||
|
|
|
||
|
|
sync2=3.7e-4*ccfbest/sq0 !Constant is empirical
|
||
|
|
|
||
|
|
! Apply AFC corrections to the time-domain signal
|
||
|
|
! Now we are back to using the 1378.125 Hz sample rate, enough to
|
||
|
|
! accommodate the full JT65C bandwidth.
|
||
|
|
|
||
|
|
call timer('twkfreq ',0)
|
||
|
|
call twkfreq65(cx,n5,a)
|
||
|
|
call timer('twkfreq ',1)
|
||
|
|
|
||
|
|
! Compute spectrum for each half symbol.
|
||
|
|
! Adding or subtracting a small number (e.g., 5) to j may make it decode.\
|
||
|
|
! NB: might want to try computing full-symbol spectra (nfft=512, even for
|
||
|
|
! submodes B and C).
|
||
|
|
|
||
|
|
nsym=126
|
||
|
|
nfft=512
|
||
|
|
j=(dt00+dtbest+2.685)*1378.125
|
||
|
|
if(j.lt.0) j=0
|
||
|
|
|
||
|
|
call timer('sh_ffts ',0)
|
||
|
|
do k=1,nsym
|
||
|
|
do i=1,nfft
|
||
|
|
j=j+1
|
||
|
|
c5a(i)=cx(j)
|
||
|
|
enddo
|
||
|
|
call four2a(c5a,nfft,1,1,1)
|
||
|
|
do i=1,66
|
||
|
|
jj=i
|
||
|
|
if(mode65.eq.2) jj=2*i-1
|
||
|
|
if(mode65.eq.4) jj=4*i-3
|
||
|
|
s2(i,k)=real(c5a(jj))**2 + aimag(c5a(jj))**2
|
||
|
|
enddo
|
||
|
|
enddo
|
||
|
|
call timer('sh_ffts ',1)
|
||
|
|
|
||
|
|
call timer('dec65b ',0)
|
||
|
|
call decode65b(s2,nflip,mode65,nbmkv,nhist,decoded)
|
||
|
|
dt=dt00 + dtbest + 1.7
|
||
|
|
call timer('dec65b ',1)
|
||
|
|
|
||
|
|
return
|
||
|
|
end subroutine decode65a
|