10. Design investigations: sweeps, tolerances, noise and diagnosis
Previous · Contents · Next: SPICE in TinyCAD
The examples in this chapter are bundled with TinyCAD. Choose Open Example → SPICE tutorial, select the named example, then choose Open Drawing. No download is needed.
A useful simulation report records the question, circuit/model version, analysis, stimulus, measurement window and acceptance criterion. A convincing plot without those details is difficult to reproduce.
Parameter sweeps and tolerances
Return to rc-ac. In Tune / sweep, run the prepared R1 values 1 kΩ, 2.2 kΩ and 4.7 kΩ. Their expected cutoffs are 159.15, 72.34 and 33.86 Hz. This is a parameter sweep: it asks how deliberate component choices affect response. It differs from the DC sweep in chapter 2, which changes an independent source’s DC value during one analysis.
For tolerances, use Analysis lab → Tolerance studies. Define R1 and C1 bindings, choose a distribution and seed, and add a cutoff requirement. With independent bounded ±5% values, the extreme RC corners give approximately 144.36–176.35 Hz. A Monte Carlo sample does not necessarily reach either corner. A normal distribution’s standard deviation is not a hard ± limit. Record which convention you selected and whether variables are correlated.
TinyCAD can retain test-case conditions and measurement/requirement definitions. Use Apply only when you want temporary tuning to become a real schematic edit. Save a separate circuit revision when changing topology or model text; an acquisition hash is an identifier, not a recoverable schematic revision.
Noise: a different kind of output
Open rc-noise. It requests output noise at OUT, referred to independent input source V1, from 1 Hz to 100 kHz at 27°C. The ideal source is noiseless; R1 contributes thermal noise filtered by C1.

The output-noise total is approximately 64.2 nV RMS over that band. For an
ideal RC circuit integrated from zero to infinite frequency, the capacitor’s
mean-square noise is kT/C, giving about 64.4 nV RMS at 300.15 K and 1 µF.
The finite simulation band explains the small difference. Spectral density
has units V/√Hz; its integral is formed from the squared density and frequency
bandwidth, then square-rooted. It is not the arithmetic sum of the displayed
density samples. The input-referred total is a different result.
Transfer functions and poles
Open rc-tf. Its DC small-signal transfer result is
1 V/V, output resistance is 1 kΩ, and input resistance is effectively infinite
because the capacitor is open at DC. A very large printed input resistance
such as 1e20 represents a numerical open-circuit result, not a measured
physical 100 exaohm resistor.
Open rc-pz. The pole is approximately −1000 rad/s,
matching s = −1/(RC). Divide its magnitude by 2π to obtain 159.15 Hz. A pole
in the left half of the complex s-plane represents decay in this linear system.
The absence of a finite zero is expected for this first-order low-pass transfer.
These analyses operate on the linearised model. They do not replace transient testing of clipping, switching or startup. Arbitrary output/input gain is also not the same thing as feedback-loop gain; stability measurements need a deliberate loop test configuration.
A practical measurement menu
| Task | Suitable quantities |
|---|---|
| Sensor conditioning | Gain, offset, noise, cutoff, settling time |
| Pulse interface | Rise/fall time, delay, pulse width, duty cycle |
| Amplifier | Bias, gain/phase, bandwidth, RMS output, power, clipping |
| Power stage | Ripple, peak current, mean power, energy, efficiency |
| Component variation | Sweep overlays, worst observed values, requirement results |
TinyCAD also offers a virtual oscilloscope, multimeter, function generator and current-limited bench supply under Lab Equipment. Their impedances and limits matter: for example, the function generator normally has 50 Ω output resistance. They provide another workflow for the same electrical questions.
Diagnose a failed or implausible run
Check ground, unintended shorts and floating nodes first. Confirm component pin mapping and model names. For AC, verify the independent source’s AC amplitude; for transients, verify duration, maximum step and finite stimulus edges. Read the solver log and confirm the run reached the requested end time. A partial waveform is not a completed result.
Try a simpler operating-point or isolated-stage test before changing numerical settings. Document any added leakage resistance, source resistance or parasitic capacitance: it changes the circuit as well as helping convergence. Export CSV when a specialised calculation is needed, and retain units and complex columns.
Further reading: the ngspice control-language tutorial
shows command-line automation. TinyCAD’s document model intentionally does not
execute arbitrary .control scripts; use its supported analysis and study
controls, or export a netlist for a separate ngspice workflow.
Download the lesson circuits
The same examples are already available in Open Example. For an independent copy, download a .dzn drawing and use File → Open in TinyCAD. The .cir companion is SPICE text for inspection or a compatible external simulator, not a TinyCAD drawing. XSPICE circuits require Desktop for fresh runs.
| Example | Drawing | Netlist |
|---|---|---|
| rc-ac | TinyCAD drawing | SPICE text |
| dc-sweep | TinyCAD drawing | SPICE text |
| rc-noise | TinyCAD drawing | SPICE text |
| rc-tf | TinyCAD drawing | SPICE text |
| rc-pz | TinyCAD drawing | SPICE text |