double precision problem

Question

Question from a beginner. I used doubles in a program to calculate cable sag. The formula itself is pretty sophisticated lots of math.tanh()'s and math.sqrt()'s. After analyzing the results i found out there's a slight precision problem. Doubles don't seem precise enough for this particulair problem. Is there a way to solve this? ---Posted from iPhone App---

Vulpes · Answer

Doubles have a precision of 15 or 16 digits which is enough for most engineering applications. Also all the trignonometic functions in the Math class take double arguments and return a double result so there's no way you could use the decimal type (precise to 28 or 29 digits) instead. Internally floating point operations are carried out to a higher precision (19 digits) and then rounded off to 'double' or 'float' precision. This means that you can sometimes improve overall precision by performing intermediate operations in a different order. I'd also try to isolate any intermediate operations where int or long could be used instead of double as these operations will always (unless they overflow or underflow) produce exact results.