pre.cluster to denoise

Hello every one,
I’m dealing with 454 pyrosequencing analysis of 16S rRNA genes. To denoise, i try to use the “pre.cluster” in Mothur. The questions is that, how to set the opotion of “diffs”. Give an example here, based on that you understand the denoising principle of “pre.cluster”:
If the length of my 16S is 300bp, and you set “diffs=1” or “diffs=2” or “diffs=3”, and the OTUs No. won’t drop, comparing from the results as for no such a “pre.cluster” treatment. Because, “diffs=1”, the difference is 1/300=0.33%, having no such a treatment, these sequences should be calculated as an OTU.
So, how to set the opotion of “diffs” is very important, i think. Please tell me how you choose the number for it, considering the length of your 16S.

As we lay out in, we generally allow 1 diff per 100 bp.