Renormalization of exponential sums
   Basic idea :
   Poisson formula
   
   leads to
   
   
   If  small, much less terms in the right hand side. So right hand side varies more 
   slowly with  One erases the details (Mendès-France, Berry, Goldberg). 
   Three questions:
    is there a simple analytic description of the correction terms in these formulae?
    are these correction terms really error terms with respect the exponential sums,
      i.e. are they smaller than the renormalized sum?
    what is the behavior of these terms for small values of ?
   To present the renormalization scheme, we concentrate on quadratic exponential 
   sums i.e. on the case