The best approach to take is to start with a recommended amount of primers, i.e. a final concentration of 40 nM. However, for your particular application this may produce some pirmer-dimers, non-specific DNA amplification or no amplification. What I suggest is that you set up a standard PCR reaction, run the reaction and analyse the results. If the results are not satisfactory, repeat the reaction while keeping all the variables constant, except the primer concentration. You could set up multiple PCR reactions with increasing primer concentrations, starting at 20 nm to 1 µM. The resulting PCR reactions would give you an idea of the best primer concentration, or if the primer concentration was the problem in the first place.
I have found more often with PCR reactions that the amount of DNA in the reaction is critical for successful PCR. Make sure that you stay more or less within 15 - 30 ng or DNA per reaction.
Remember that optimizing PCR reactions are no easy or quick task. Take your time and do not take any shortcuts. Lucky for you 16S PCR using 27F and 1492R is well documented.
I agree with what Roelof has written on PCR optimization. You can use 5 pmoles of each primer and 0.5 U of Taq polymerase per 20 µl of final reaction volume as start conditions. Most probably it is not primers concentration which is problematic though. Usually if standard conditions don't work its a matter of template DNA - make sure that the quality of your preparation is good enough (A260/280 ~1.8 and A260/230 > 2. If you have problems with isolation, try bead beating + flocculation kits (e.g. Qiagen's DNeasy PowerSoil) even with pure strains - they effectively lyse even very though bacs and get rid of many problematic contaminants.