Float Epsilon Calculator

Calculator Form

Main value

Comparison value

Precision mode

Custom significant binary bits

Absolute tolerance

Relative tolerance

Output decimal places

Example Data Table

Value	Precision Mode	Expected Machine Epsilon	Common Meaning
1	Double precision style	2.220446049250E-16	Smallest step near one for many standard numeric systems.
1	Single precision style	1.192092895508E-7	Typical step near one for compact floating storage.
1000000	Double precision style	ULP grows with value scale	Large numbers have wider nearby spacing.
0.000001	Double precision style	ULP shrinks with value scale	Small normal values can have tiny spacing.

Formula Used

Machine epsilon: ε = 2^{1 - p}

Unit roundoff: u = ε / 2

ULP near x: ULP(x) ≈ 2^{floor(log₂(|x|)) - (p - 1)}

Next value up: x + ULP(x)

Next value down: x - ULP(x)

Relative spacing: ULP(x) / |x|

Recommended comparison tolerance: max(absolute tolerance, relative tolerance × max(|a|, |b|, 1))

Close comparison rule: |a - b| ≤ recommended tolerance

How To Use This Calculator

Enter the main number you want to inspect.
Enter another number for comparison.
Choose double, single, or custom precision.
Enter custom binary bits only when custom mode is selected.
Add absolute and relative tolerance values.
Select output decimal places.
Press the calculate button.
Review epsilon, ULP, next values, and tolerance status.
Use CSV or PDF export for records.

What Is Float Epsilon?

Float epsilon is the smallest meaningful step between nearby floating point numbers at a chosen scale. Computers store many decimals in binary form. That storage is fast, but it is not exact for every value. A float epsilon calculator helps you see this spacing before rounding creates surprises.

Why Precision Gaps Matter

Two decimal values can look equal on screen, yet differ in memory. This happens in finance, engineering, statistics, graphics, simulations, and scientific scripts. The gap changes as numbers grow. A value near one has a small spacing. A very large value has a larger spacing. That is why a fixed decimal rule is often weak.

What This Calculator Shows

This tool estimates machine epsilon, unit roundoff, ULP, next representable values, and equality tolerance. It also compares two numbers. You can see the absolute difference, relative difference, and how many ULP steps separate the values. These results help you decide whether two computed answers are close enough.

Choosing a Precision Mode

Double precision uses 53 significant binary bits. Single precision uses 24 bits. A custom mode lets you study other formats or lessons about floating arithmetic. Higher precision means smaller gaps and more reliable decimal digits. Lower precision is faster in some systems, but it needs wider tolerance checks.

Using Epsilon Safely

Never use epsilon as a magic constant for every comparison. Scale it by the magnitude of the numbers. Combine absolute tolerance for values near zero with relative tolerance for larger values. This mixed rule is common because it handles tiny and huge values better.

Practical Use Cases

Use this calculator before validating algorithms, storing sensor data, comparing imported spreadsheets, testing simulations, or building numeric APIs. It can also support classroom examples. The table and export buttons make it simple to document assumptions. Clear precision notes reduce debugging time and improve trust in technical results.

Reading The Result

The ULP value is the key output. It tells you the spacing around the selected number. The recommended tolerance is not a law. Treat it as a starting point. Increase it when input data is noisy. Decrease it when your method requires strict control and the source values are stable during repeatable tests and later audits.

FAQs

What is float epsilon?

Float epsilon is the smallest step that changes a floating point value near one. It helps explain why some decimal calculations produce tiny differences.

Is epsilon the same for every number?

No. Machine epsilon describes spacing near one. ULP spacing changes with number size, so larger numbers usually have larger gaps.

What does ULP mean?

ULP means unit in the last place. It estimates the gap between adjacent representable floating point values around a selected number.

Should I compare floats with exact equality?

Usually no. Use a tolerance rule instead. Exact equality can fail when calculations include rounding, conversion, or repeated operations.

What is unit roundoff?

Unit roundoff is half of machine epsilon. It estimates the maximum relative rounding error for many normalized rounding operations.

Why add absolute tolerance?

Absolute tolerance helps compare values near zero. Relative tolerance alone can become too strict when both numbers are very small.

Why add relative tolerance?

Relative tolerance scales the allowed error with number size. It is useful when comparing large values or results from complex calculations.

Can this calculator replace numerical testing?

No. It supports precision planning and comparison checks. Always test real algorithms with realistic inputs, edge cases, and accepted error limits.