Engineering Calculations

Give calculations another engineer can audit. This skill sets a calculation method, not a formula bank, a topology, a component value or a part choice.

Boundaries

• Name the quantity and the physical model before choosing equations. Use a formula only inside its assumptions (for example ideal converter, steady state, CCM, small-signal).
• Keep given, derived, assumed and datasheet values apart. Every rating, parasitic, material parameter or test value needs a source; ask for it or mark it NOT_VERIFIED.
• Use SI units throughout, convert prefixes explicitly where mistakes are likely, and say whether a value is peak, RMS, average, instantaneous or rated.
• Absolute maximum ratings are limits, not operating points. Include tolerances, derating, temperature, transients and margins when they change the answer.
• When an assumption changes the answer, give a range or a conditional result, not one falsely precise number.
• A calculated result is an estimate, not hardware verification. Claims of compliance, safety, thermal adequacy or design closure need source data and independent validation.
• Read KiCad files with targeted, read-only queries. Write files only when the user asks.
• Cite only sources you actually read.

Evidence baseline

• Label each significant claim VERIFIED (named evidence: CAD data, a calculation or an official source), ASSUMED (assumption stated), ESTIMATED (method and margin stated) or NOT_VERIFIED (evidence missing or check not run).
• Never open or read a PDF. Datasheet values come from the user, official manufacturer web pages, or NotebookLM answers with cited text, used only after the user approves uploading the file. With none of these, the value is NOT_VERIFIED.
• KiCad design and library files are read-only. Propose changes; do not apply them.
• If the project has .claude/rules/engineering.md, follow it; it is the full policy.

Workflow

1. Define the target. The quantity, operating mode, constraints, available inputs and needed accuracy. Ask only for a fact without which the answer would mislead; otherwise state a bounded assumption.
2. Choose and justify the model. State the equations and the conditions under which they hold. Add non-ideal effects where they matter.
3. Normalize inputs. Show units, prefixes and operating ranges, and mark each value as nominal, worst case or manufacturer-specified.
4. Calculate reproducibly. Substitute values and check dimensions. Use a calculator or local Python for error-prone numbers, sweeps, optimization, matrices or several operating points. Report a number from code only after the code has run.
5. Check the result. Limiting cases, signs, dimensions, conservation laws and physical sense. Cross-check with an independent method when an error would be costly.
6. Check sensitivity. Min, nominal and max conditions, tolerances, temperature, losses, derating and parameter uncertainty, where they apply. Find the worst case; do not assume it sits at nominal.
7. Report evidence and limits. Keep math correctness, model validity and measured performance apart. Mark missing evidence NOT_VERIFIED.

Optional companions

• design-assist: calculations compare architectures; a formula alone does not choose the architecture.
• component-select: turn calculated limits into device requirements. One matching parameter does not pick an MPN.
• stock-select: sourcing comes after technical fit.
• component-check: actual datasheet ratings, tolerances, operating conditions and package limits.
• sch-inspect: targeted schematic facts; inferred connectivity stays unverified.

These help if installed. If one is missing, do the step directly under the same evidence rules, or mark the dependent check NOT_VERIFIED.

Output

Match the size of the answer to the request. For a real calculation:

1. Inputs and assumptions: values with units and the operating range.
2. Model and equations: formulas, conditions and limits.
3. Results: values with sensible precision and units; a compact table for several operating points. Label each result ESTIMATED and state the method and margin.
4. Checks and implications: sanity checks, margins, sensitivity, and the next evidence the design decision needs. Mark critical unknowns NOT_VERIFIED.

For simple arithmetic, give the result and the key assumptions only. When the user supplies an equation to review, check its derivation and conditions before substituting numbers.

Example

Inputs:  R = 10 kOhm +/-1%, C = 100 nF +/-10% (given); first-order RC low-pass, no load (ASSUMED)
Model:   fc = 1 / (2 pi R C)
Results: fc nominal 159 Hz | ESTIMATED | exact formula, nominal values
         fc range 143 to 179 Hz | ESTIMATED | R and C at tolerance extremes
Checks:  C tolerance dominates; C temperature drift NOT_VERIFIED (dielectric unknown).