Validation
Whether a calculator can be trusted is decided by what stands behind its answer, not by the answer itself. These reports work each calculation by hand from a published definition and set the result against what the tool returns, with every step shown.
What a report contains
Each report states what it covers, derives the expected values from primary sources, names the acceptance criterion, shows the working in full, and ends with what the exercise does not establish. The expected values are held apart from the calculation engine and are written from the sources rather than computed from the tool's own constants, because a tool compared against itself will always agree.
Coverage is counted rather than asserted. A report states how many of the things the tool offers it has actually derived, the build compares that count against the tool, and a mismatch fails the build. Adding a capability without validating it is therefore not possible.
Reports
7 of 7 calculators has a published report. The rest are listed so that the gap is visible; the work is proceeding one tool at a time and this page is the record of where it has reached.
-
Unit converter
Every unit the converter offers is derived by hand from its primary definition, and every conversion within a quantity is checked against those hand-derived factors.
44 units · 276 ordered conversions · Report dated 2026-09-17 · Read the validation report
-
Decay and half-life
The decay law is checked as exact mathematics across half-lives spanning seventeen orders of magnitude, and the half-lives themselves are set against an independent evaluation that publishes uncertainties.
Report dated 2026-09-17 · Read the validation report
-
Gamma dose rate and shielding
The unit chain is derived by hand, the interpolation is checked against every row of the attenuation tables, and the emission intensities are derived from an independent evaluation rather than compared with published gamma constants, which turn out to be a different quantity.
Report dated 2026-09-18 · Read the validation report
-
Mass and activity
Derived by hand from the definitions of the mole and the becquerel, with every molar mass set against the evaluated atomic masses. The report found that the calculator had been using mass numbers in place of molar masses, states the size of the error it caused, and records its removal.
Report dated 2026-09-18 · Read the validation report
-
Detection limits (MDA / MDC)
The three worked examples MARSSIM prints in its own text are reproduced by this engine, which is a stronger check than a derivation because it fails when the algebra is right and the answer is not. Where the calculator departs from the standard, the size and the direction of the departure are measured.
Report dated 2026-09-19 · Read the validation report
-
Beta dose rate and shielding
The infinite-medium dose rate is derived exactly from conservation of energy; the three empirical fits behind the shielding mode are checked for form and domain and set against NIST electron data for the same quantity, not a similar one. What could not be established is listed rather than glossed.
Report dated 2026-09-19 · Read the validation report
-
ALARA and job planning
A calculator with no data to compare against, so every expression is derived instead. The point-source assumption is turned into a distance criterion by solving the finite-source case exactly, and all nine inputs that have no answer are enumerated — every one of them was returning a number.
Report dated 2026-09-19 · Read the validation report
Scope of the exercise
Validating a conversion is a different task from validating a physical model. The unit converter rests entirely on definitions, so its report can be complete: there is a finite list of units and every one of them is derived. The calculators that use nuclear data and physical models cannot be complete in that sense, and their reports will have to say what was compared, against which published values, and how far apart the two were. Where a figure on this site is a computed derivation rather than a copied table value, the derivation is described on the methods page.
None of this makes RadCalc a substitute for an accredited calibration or test, and the disclaimer says so in the terms that matter.