I use FF8, but it does not matter.
>P.S.: I've checked again examples from Alex. He used additionally IRZMAT to remove coordinates, which are coupled with scan coordinate. The calculation runs fine if I remove two torsions 2-3-4-6 and 2-3-4-7.
Yes, I forgot to mention this. But the scan stops abnormally with extremely large gradient when 2-3-4-5 torsion equals 180, I don't know why.