This keyword is used in the declaration of fixed point types. Delta represents the absolute error bound of a fixed point type; that is, delta gives a hint to the compiler how to choose the smallest incremental value of the type, called the Small of the type. This Small value is accessible via the attribute T'Small.
For Ordinary fixed point types, it is by default a power of two not greater than the delta; it is specifyable by an attribute clause to be any value not greater than the delta.
For decimal fixed point types, the Small value is equal to the delta, which of course must be a power of ten.
type Fixed_Point is delta 0.01 digits 10; -- A decimal fixed point type My_Fixed_Point : Fixed_Point := 0.0; ... My_Fixed_Point := 0.1; -- Ok My_Fixed_Point := 0.02; -- Ok My_Fixed_Point := 0.001; -- Oops - this value is too small. -- It will be represented by zero.
Ada Reference Manual
Ada Quality and Style Guide