The approach to take depends on what exactly is "minimum error". For the lower frequencies it may be more appropriate to use the reciprocal counting technique, while the higher frequencies can be better counted directly. Where the breakpoint between 'high' and 'low' frequency is depends on factors such as desired resolution, counting speed and available counter resources.

Peter Danegger has a frequency meter project on his webpage: http://www.specs.de/users/dan.../index.htm
That should give you some ideas.


Rob.