Behind the result
How we check calculations
A result should be possible to question, reproduce, and understand.
These calculators and guides are published by Eucalyptus One LLC. We check the mathematics and the browser interface separately. Each tool states its model and limits; a successful calculation does not establish that the model fits a particular design.
Begin with an answer we can check
Our automated tests compare selected examples with analytical answers or independently published values. An equal-resistor divider should halve an unloaded input. Integrating x³ from 0 to 1 should give 1/4. The impedance example is compared with NXP’s published microstrip geometry.
The worked guides spell out inputs and rounded results so you can repeat those examples. Passing one example does not establish accuracy for every input.
Check the boundaries as well as the example
Tests include invalid and missing inputs, supported limits, unit conversions, numerical overflow, and relevant convergence or residual checks. Exact matrix tools compare rational values. Floating-point checks allow a stated numerical tolerance; they do not assert exact equality for approximate methods.
For numerical integration, we compare rules and refinements on known functions. For electrical calculations, we check unit conversions and relationships such as capacitance increasing in proportion to load power. Rejected inputs should produce a useful error and disable an obsolete download.
Exercise the page people actually use
Browser tests run in Chromium, Firefox, and WebKit. They exercise calculations, validation, reset, downloads, navigation, and automated accessibility checks. We also inspect screenshots at desktop, laptop, tablet, and phone sizes, including equations and diagrams.
Those checks complement hands-on testing; they do not certify accessibility or replace testing with assistive technology and physical devices.
Know what the checks leave open
Software checks do not validate component tolerances, a manufactured PCB, a converter’s transient behavior, or a fabricator’s finished stack-up. The impedance estimate has no guaranteed physical error bound. Use manufacturer data, appropriate simulation, and measurement when the application needs them.
Display rounding is separate from calculation precision. Saving more digits does not improve the accuracy of the physical model. Inputs and results stay in your browser; the site does not upload them for verification.
Found a result that looks wrong?
Contact support with the tool name, inputs and units, expected result, and the reference you used. Include only information you are comfortable sharing. A small, reproducible example helps us distinguish an input problem from a model limitation or calculation defect.