A small amount of random noise was added to synthetics to prevent zero-divide errors (in cases where the recfunk codes tries to compute an RF for an interval of x=0 data). In the update, this bit of noise was made smaller, and explicitly zero-mean. There were some low-frequency artifacts cropping up in synthetic RFs. The updated versions seem to avoid this problem:
anirec_synth.f
anirec_synth_circle.f
for info see link
1 comment:
Jeff - I've been combing through this code as you know, and I've found a region that could be prone to some instability.
This is for the current version, using matget() as opposed to matget_old()
there is never a call to the defective() routine - therefore the variable idfct (in respget) is never set or initialized.
There are three seperate chunks of code in respget that use this variable, which I think should be commented out - otherwise you risk mucking up the R/T coeffs if those memory locations contain random data. If this needs to be in there, let me know.
Post a Comment