Hi Joe,
r3644 compiled and running. Decoding is fine.
Problem with Flatten option which is selected on startup. On my Mac OSX 10.9
the spectrum on the Wide Graph panel is a straight green line with no peaks.
The waterfall display is blank. With Flatten deselected, the green spectrum
appears. Although the waterfall display is now active the waterfall background
is now a green 'smear' from left to right indicative of too much gain. Gain
on the lower panel is set to 30dB.
Using Qt5.1.1
Diagnosis: if I turn on the write statements in flat3 I find:
fort.21: write(21,1010) f,y0(i),yfit(i),y0(i)-yfit(i) :: f is
just i*df; but y0(i), yfit(i) are zero always:
worse, when I change the format from f to e, the output shows that y0 and yfit
are mal-formed variables - the output is garbage
The problem is this statement: implicit real*8 (a-h,o-z) coupled with
this: y0(i)=db(s0(i))
The function db is deemed to be double precision by virtue of the implicit
statement but db is actually a single precision function. Consequently y0(i)
is filled with garbage since db only returns a 32-bit number.
The implicit statement is needed for the definitions of the arguments to polfit
so real*4 db is needed.
While trying to get to the bottom of this I was puzzled by symspec:
call flat3(s,iz,nfa,nfb,3,1.0,s)
call flat3(savg,iz,nfa,nfb,3,1.0,savg)
where an argument is repeated. The second argument is only used once at the
end of flat3 as:
s(i)=10.0**(0.1*y4)
Is there a reason why the duplication is needed? Drop the final argument in
symspec calls and flat3 definition, no need for real*4 s(iz), and write:
s0(i)=10.0**(0.1*y4) at the end of flat3.
Finally: the waterfall display remains a greenish background from left to
right when flatten is selected. Much better without flatten - am I missing a
tuning item?
--- John G4KLA_______________________________________________
Wsjt-devel mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/wsjt-devel
