Underflow is a condition in floating-point arithmetic that occurs when the result of a calculation is smaller (closer to zero) than the smallest positive value representable by the data format. It is the opposite of overflow, where a number exceeds the maximum representable value.
In standard floating-point formats (such as 32-bit single-precision), there is a limit to the smallest non-zero number that can be stored. If a calculation falls below this limit, the system typically replaces the result with zero (known as "flush to zero"). While often acceptable, this leads to a loss of precision. In high-precision measurement and control programming, engineers must be aware of underflow risks to ensure data integrity when processing extremely small values.