Here's a simple little routine that will help you validate real numbers.   It is written in Fortran 90/95.   You'll want k8 to be equal to 8 for most double precision variables on most PC computers.   Make it 4 for single precision on most PC computers.   This code is tested and used on software which I've written.   If you're using Fortran 77,  you'll need to do some  alterations.   The intent of the code below should be sufficient for you to figure out how to alter it appropriately.

Dan    
www.dtware.com


        SUBROUTINE  Validate_Real_k8 ( StrNum, value, valid )
        !***********************************************
        ! Check to see that a character string represents a valid real number
        ! Valid real numbers have the following attributes:
        !
        !   They are representable by the fortran compiler used with this code.
        !   ie. No error is flagged when attempting to use a READ statement on
        !   the data, <StrNum>.
        !   They do NOT begin or end with the any of the following letters:
        !     eEdDqQ
        !   They do not contain embedded blank characters.
        !   If <StrNum> is all blanks, that is still considered a valid number.
        !
        ! Input:
        !   StrNum = Character string which is supposed to represent a valid
        !       real number. StrNum is not altered by this routine.
        !       StrNum may be up to 50 character digits (including decimal, etc)
        ! Output:
        !   valid = .true.  if num is a valid real number or is all spaces.
        !         = .false. otherwise.
        !   value = Numeric value from StrNum if valid is .true.
        !         = 0.0 if the value from StrNum if valid = .false.
        !         = A variable with a Kind Type of "K8"
        ! Uses:
        !   KindTypes = module holding valid kind type parameters
        !*************************************************
        USE KindTypes
        IMPLICIT NONE
        Character(*),     intent(in)      :: StrNum
        Real(k8),            intent(inout) :: value
        Logical,              intent(out)   :: valid
        Character(Len(StrNum))        :: num
        Integer                              :: io
        Character(1)                      :: fd, ld

        num = ADJUSTL ( StrNum )
        Read ( num, '(F50.0)', iostat=io ) value
        IF ( io /= 0 ) THEN
            ! the entered value is not a valid 'real' number
            valid = .false.
            value = 0.0
            RETURN
        END IF

        io = Len_Trim ( num )
        fd = num(1:1)         ! first digit
        ld = num(io:io)       ! last  digit
        IF ( fd=='e' .OR. fd== 'E' .OR. ld=='e' .OR. ld== 'E' .OR. &
             fd=='d' .OR. fd== 'D' .OR. ld=='d' .OR. ld== 'D' .OR. &
             fd=='q' .OR. fd== 'Q' .OR. ld=='q' .OR. ld== 'Q' .OR. &
             INDEX(num(1:io),' ') /= 0                                 ) THEN
            ! Some compilers will accept a real number starting or ending
            ! with just a 'd' or an 'e'.  I don't consider this a valid entry
            ! LF95 ver 5.5 does this.  Also,  LF95,5.0 allows q or Q to be
            ! used to denote exponentiation just as E and D are used.
            valid = .false.
            value = 0.0
        ELSE
            valid = .true.
        END IF
        END SUBROUTINE  Validate_Real_k8