I recently find that the noise seed block needs negative seed for
gaussian and uniform distributions. The commit 0d6871b3 has provided
documentation for this and also changed the seed in noise seed from
positive value to 0 to avoid misleading. However, the seed value in
channel model block hasn’t been updated. And in some python scripts,
randint(0, 100000) as the seed. This is misleading and doesn’t
give what we want.
I attach a patch to address the issue.