Floating-point math can get ugly little rounding errors. You're seeing them here.

No need to set any precision, simply make sure the magnitude of the values are above some $eps (epsilon) or set them to zero.

Like so:  $pdl = $pdl * ( (abs $pdl) > $eps);